1

Journal of Materials Processing Technology 211 (2011) 1898–1906

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

A fast algorithm for strain prediction in tube hydroforming based on one-step

inverse approach
M.S. Chebbah a,∗ , H. Naceur b , A. Gakwaya c
a
University of Biskra, P.O. Box 145, Biskra 07000, Algeria
b
University of Valenciennes, Lab. LAMIH, 59313 Valenciennes, France
c
Laval University, 1065 Avenue de la Médecine, G1V 0A6, Québec, Canada

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents recent developments of a simplified finite element method called the inverse
Received 5 October 2010 approach (IA) for the estimation of large elastoplastic strains and thickness distribution in tube hydro-
Received in revised form 21 April 2011 forming. The basic formulation of the IA, proposed by Guo et al. (1990), has been modified and adapted
Accepted 14 June 2011
for the modeling of three-dimensional tube hydroforming problems in which the initial geometry is a
Available online 23 June 2011
circular tube expanded by internal pressure and submitted to axial feed at the tube ends. The application
of the IA is illustrated through the analyses of numerical applications concerning the hydroforming of
Keywords:
axisymmetric bulge, made from aluminum alloy 6061-T6 tubing, the hydroforming of square section
Tube hydroforming
Shell element
hollow component and the hydroforming of a free Tee extrusion from welded low carbon steel LCS-1008
Inverse approach tubing. Verifications of the obtained results have been carried out using experimental results together
Large elastoplastic strains with the classical explicit dynamic incremental approach using ABAQUS® commercial code to show the
effectiveness of our approach.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction is not carefully adjusted, defects such as bursting or wrinkling may

Nowadays, tube hydroforming (THF) is still being one of the
most important manufacturing techniques used in sheet metal
forming. The increasing application of hydroforming techniques in
the automotive and the aircraft industries as well as manufactur-
ing components for sanitary use is due to its advantages compared
to classical processes as stamping or welding. Tube hydroforming
is a perfect technique for manufacturing tubes of complex shapes
with high level of repeatability and offers an effective integration of
structural components manufactured using a minimal space. Tube
hydroforming has many other advantages compared to conven-
tional stamping processes, according to Nikhare et al. (2010), these
advantages include weight reduction, strength improvement and
higher geometry accuracy of the final manufactured part. Histori-
cal overviews and discussion of future trends in tube hydroforming
can be found in Ahmed and Hashmi (1997), Koç and Altan (2001),
Singh (2003) or Hartl (2005).
In THF the main parameters that may affect the general feasi-
bility of the final product are: tube material and dimensions, tools
geometry, as well as the magnitude and the loading path of the
hydraulic pressure and the axial feeding. If one of these parameters
∗ Corresponding author. Tel.: +213 7 74 02 81 42.
E-mail addresses: ms.chebbah@univ-biskra.dz, chebbah mss@yahoo.fr
(M.S. Chebbah).
appear according to Langa et al. (2009), Di Lorenzo et al. (2010).
Since the recent past, the feasibility of a product was mainly car-
ried out by engineers using trial-and-error iterative procedures that
are expensive and error prone. Over the last decade, the numeri-
cal simulation of tube hydroforming using the finite element (FE)
method has received significant attention as an alternative to the
trial-and-error methods; however, there is still lack of a fast and
robust control of hydroforming process parameters at the early
design stage; where the classical based dynamic explicit method
is not convenient to use due to large amount of data needed to
carry out a simulation see Hosford and Caddell (2011).
Ahmetoglu et al. (2000) provided fundamental issues related
to material and lubrication requirements, material shaping capa-
bilities, tool design and process control in tube hydroforming of
low carbon steel and aluminum alloy 6061-T6 tubes. It has been
achieved through the establishment of a consortium between part
manufacturers, and material and equipment suppliers in the Ohio
State in the US.
Koç et al. (2000) used a design of experiments technique in
conjunction with FEM to facilitate the economical prediction and
optimization of the height as a function of geometrical parameters
subject to thinning of the wall thickness at the protrusion region.
Results in their study suggest that comprehensive and detailed
investigations of tribology in hydroforming should be conducted.
Mac Donald and Hashmi (2001) used LS-DYNA 3D to compare
bulge with a solid medium against a hydraulic medium. In their

0924-0136/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2011.06.012
 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906
Nomenclature
0 initial
ı virtual
h thickness
P intensity of pressure
p, q point on the mid-surface, in distance z
s curvilinear coordinate
z distance of point to the mid surface
u, v, w displacements in local coordinate
U, V, W displacements in global coordinate
up displacements of p
x, dx position vector, its variation
n, t normal and tangent vectors in the mid surface
N, M membrane forces, bending moments
Fint , Fext internal and external forces in global coordinate
[Bm ], [Bf ] membrane strain, bending strain
[Kt ] tangent matrix
[T] transformation local-global matrix
Greek symbols
r, ϕ, Z cylindrical coordinate
1 , 2 , 3 principal stretches
, e curvature, membrane strains
ε, ε̄ strain, equivalent strain
, ¯ stress, equivalent stress
investigation, they examined the effect of varying friction between
the bulging medium and the tube and the history of development
of the bulge and stress conditions in the formed component. They
concluded that the use of a solid bulging medium allows for greater
branch height, less thinning of the branch top and less stress in the
formed component when compared to the hydraulic bulging pro-
cess. Kim et al. (2002) introduced a backward tracing technique to
predict an appropriate pre-formed configuration and determine the
initial tube dimensions from the desired final shape. The developed
program was applied to a hydroforming process of a box expansion
in order to get the uniform wall thickness after hydroforming, and
the conceptual application has been proved to be successful on its
effectiveness and feasibility.
Kridli et al. (2003) investigated corner filling by 2D simula-
tions, using ABAQUS/Standard, and experiments. They examined
the effects of the strain-hardening exponent, initial tube wall thick-
ness and die corner radii on corner filling and thickness distribution.
They concluded that thickness distribution is a function of die
corner and strain-hardening behavior and the initial tube wall
thickness affects the pressure while maintaining the same thin-
ning. Kwan and Lin (2003) used the FE program DEFORM-3D to
investigate the cold hydroforming process of a T-shape tube. They
examined the influences of the process parameters such as the
internal pressure, the fillet radius, and counterforce on the mini-
mum wall thickness of formed tube. They found a suitable range
of the process parameters for producing an acceptable T-shape
tube that fulfills the industrial demand. Jain et al. (2004) intro-
duced “dual hydroforming” where the counter pressure as a new
process parameter to achieve favorable tri-axial stress state dur-
ing deformation process. They observed that the counter pressure
provides back support to the tube material and excessive thinning
and premature wrinkling could be prevented and thus, larger tube
expansion could be achieved. Ray and Mac Donald (2004) used a
fuzzy logic control algorithm in conjunction with LS-DYNA finite
element code for simulation and optimization of the forming load
path to avoid the failure of the tube. They sustain that by means
of a minor modifications in the strain limit setting in the load con-
1899
trol algorithm; their method can be used to determine the optimal
and feasible forming load paths for asymmetric or axisymmetric
components with relatively complex geometries. Abrantes et al.
(2005), used FEA program LS-DYNA to establish a basic understand-
ing of both free bulge and calibration against closed dies in tube
hydroforming process. They observed that formation of winkles,
resulting in folding and unfolding of tube regions, causes a dif-
ferenced springback effect along the longitudinal axle of the tube.
From the equivalence curves it was possible for them, to program
the parameters in order to obtain balanced axial displacements for
the punch strokes.
Kashani Zadeh and Mashhadi (2006) used ABAQUS FE code
to quantify the effects of coefficient of friction, strain hardening
exponent, and fillet radius on the parameters, protrusion height,
thickness distribution, and clamping an axial forces. Yuan et al.
(2006) investigated via FEM and experiments the hydroforming
of automotive rectangular section structural components. They
explored the effect of loading path on the failures and thickness
distribution and the reasons were analyzed for the failures, such
as bursting and folding. Mohammadi and Mosavi Mashadi (2009)
determined the loading paths for copper joint hydroforming via
FEM and a fuzzy controller. They determined rules to increase axial
feeding by introducing the calibration indicator, in order to have
a product with appropriate mechanical properties. The presented
method can enable automatic determination of optimal loading
paths for the hydroforming of complex structural parts in a short
time. Alaswad et al. (2011) conducted a finite element study along
with response surface methodology for design of experiment to
construct models for three responses (bulge height, thickness, and
wrinkle height) for X shape bi-layered tube hydroforming. They
found that, tube geometry has an important influence on the shape
of the hydroformed junction. They concluded that, the usage of a
larger die corner radius leads to higher bulges and smaller wrin-
kles. However, critical thickness reduction can be avoided for large
tube diameters if a big thickness is assigned for both layers.
In all of the previous cited research works dedicated for the
design and the analysis of tube hydroforming process parameters,
generally the classical incremental approach based on dynamic
explicit formulation is used by means of commercial codes such as
LS-DYNA® or ABAQUS® . Although, the classical incremental meth-
ods can provide accurate solutions for complex forming problems,
the simulations using these methods are very expensive in terms
of solving CPU-time and also for engineer’s time to set up and run
the problem: complex die CAD meshing, material data, initial tube
mesh, etc. see Numisheet (2008). Therefore, since the last twenty
years, significant research has been devoted to the development of
alternate approaches allowing fast solutions of the forming prob-
lems. These methods have become valuable tools in the preliminary
design stage of components or structures mainly used in the auto-
motive industry. The one-step methods are based on the general
assumption of knowledge of the final geometry of the 3D part and
the total deformation theory of plasticity. The unknowns are mate-
rial positions of points on the initial geometry as well as strains and
thickness variation.
The inverse approach developed by Guo, Batoz and their coau-
thors since 1990 (Guo et al., 1990), is very attractive since authors
showed that the IA can estimate large strains in deep drawing with
a very good accuracy compared to incremental analysis or to the
experiments (Batoz et al., 1998; Guo et al., 2000). More recently
Naceur et al. (2006, 2008) introduced new enhancements on the IA
to take into account the loading path in deep drawing simulation
in order to improve the stress state obtained at the end of forming.
While the IA formulation has gained a great success and
attracted many research groups for the fast simulation of sheet
metal forming, unluckily its development has been only limited to
the deep drawing simulation. Based on our knowledge only a very
1900 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906

The deformation gradient tensors at points q0 and q referenced

in C can be defined by Eqs. (2) and (3).

dx0q = [F 0 ]−1 dx = [xp,x − up,x ; xp,y − up,y ; n0 /3 ]dx (2)

dxq = [F z ]−1 dx = [xp,x + zn,x ; xp,y + zn,y ; n]dx (3)

We calculate the deformation gradient inverse tensor describ-

ing the movement between positions q0 and q referenced in C as
follows:

dxq0 = [F]−1 dxq (4)

The inverse Cauchy-Green left tensor can be obtained by:

[B]−1 = [F]−T [F]−1 (5)

Eigenvalues of the left Cauchy-Green tensor [B] and the assump-

tion of incompressibility give the three principal stretches 1 , 2 ,
3 and their direction transformation matrix [M].
Finally, the logarithmic strains are obtained:
Fig. 1. 3D shell kinematics in tube hydroforming.
   
εx εxy 0 ln 1 0 0
[ε] = εxy εy 0 = [M] 0 ln 2 0 [M]T (6)
few works have been done to the extension of the IA to the mod-
0 0 εz 0 0 ln 3
eling and simulation of tube hydroforming. In a simple application
limited to the round tubes, Nguyen et al. (2003a,b) used the IA to
analyze cylindrical tube hydroforming, where the final configura- 2.2. Constitutive equations
tion is axisymmetric. They used a constant strain triangular (CST)
membrane element together with a criterion based on a forming In the present Inverse Approach the elasto-plastic deformation
limit diagram to predict the critical sections where local necking is assumed to be independent of the loading path. The constitu-
can occur. More recently, Fu et al. (2009) proposed a study on one- tive relations remain in the framework of the deformation theory
step simulation for the bending process of extruded profiles. They of plasticity in which Hill’s anisotropic criterion (Hill, 1984) is
showed that, the stretch bending process of rectangular aluminum employed to describe the plastic flow. A planar anisotropic sheet
extrusions and three-point bending of stainless steel extrusions is considered and the small elastic strains are assumed to have the
can be simulated successfully. Chebbah et al. (2010) proposed a same anisotropic directions as the plastic strains. The total consti-
specific methodology based on the coupling between the IA and a tutive relations can be written by:
Response Surface Method based on diffuse approximation using an
¯
axisymmetric membrane/bending shell element.  = Es [P]−1 ε, Es = (7)
ε̄
The originality of the present work consists of the introduction
of a new modified version of the inverse approach proposed ini- where , ¯ ε̄ are the equivalent yield stress and equivalent plastic
tially by Guo et al., 1990, for the deep-drawing. New enhancements strain, respectively, Es denotes the secant modulus of the uniaxial
have been introduced to the algorithm in order to deal with hydro- stress–strain curve.
forming of tubes of general shapes (not only those of axisymmetric The matrix [P] is expressed in terms of the mean planar
shape). Starting from a 3D final tube mesh, the algorithm is able to anisotropic coefficient r̄ defined by the Lankford coefficients as
determine material positions of points in the initial tube. In the 3D following:
case, the final shape of the desired tube is discretized by triangular ⎡ ⎤
shell elements DKT12 (Guo et al., 1990). r̄
1 − 0
The application of the inverse approach is illustrated through ⎢ 1 + r̄ ⎥
⎢ r̄ ⎥
the analyses of some numerical applications. Verifications of the [P] = ⎢ − 1 0 ⎥,
obtained results have been carried out using experimental results
⎣ 1 + r̄ ⎦
2(1 + 2r̄)
together with the classical incremental dynamic explicit approach 0 0
1 + r̄
by means of ABAQUS® commercial code, to show the effectiveness
of our approach. 1 εP
(˛+900 )
r̄ = (r0 + 2r45 + r90 ) with r˛ = (8)
4 εPz
2. Formulation of the inverse approach
The 00 direction is defined as the rolling direction.
2.1. Kinematics of a large transformation
2.3. DKT12 shell formulation
In the inverse approach only the configurations of the initial
tube of cylindrical form C0 and the final 3D workpiece C are consid-
The known three-dimensional workpiece is discretized by flat
ered. Using the Kirchhoff assumption, the initial and final position
triangular shell elements called DKT12 presented by Batoz and
vectors of a material point can be expressed on C (Fig. 1):
Dhatt (1992) of constant thickness having three corner nodes and
x0q = x0p + z 0 n0 = xp − up + z 0 n0 three mid-side nodes (Fig. 2). This element is obtained by assembly
(1) of the element of membrane CST (u, v at corner nodes) with the dis-
xq = xp + zn
crete Kirchhoff triangular plate element DKT6 (w at corner nodes
where x = (x, y, z) is the position vector at the mid-surface. and s at mid-side nodes).
 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906 1901

Fig. 2. 3D mesh mapping onto the initial cylindrical tube surface.

Fig. 3. Positions of a point p in the initial configuration during iterations.

The virtual strains of membrane are expressed in terms of virtual e is the internal virtual work and W e is the external
where Wint ext
displacement components ıu and ıv (linear approximations) along virtual work related to the tool actions.
the local coordinates x, y as follows: 
e
Wint = ıεT  dv (12)
ıeT = ıu,x ıv,y ıu,y + ıv,x ; ıe = [Bm ]ıunm (9) ve
with
with uTnm = . . . ui vi . . .; i = 1, 2, 3
  ıε = ([Bm ] + z[Bf ])ıun (13)
y23 0 y31 0 y12 0
1 The global components of the internal force vector are defined by:
[Bm ] = 0 x32 0 x31 0 x21 ,
2A
x32 y23 x13 y31 x21 y12 e
Wint = ıU n F eint , ıU n = ıUi ıVi ıWi , i = 1, 2, 3 (14)
2A = (y31 x21 − x31 y21 ), xij = xi − xj T T T
with F eint
= [T ] ([Bm ] N + [Bf ] M)A
The element external forces are due to the hydraulic pressure
and can be computed using the equilibrium conditions. The tool
For the virtual curvatures of the DKT6 element, the rotation
actions are replaced by a normal pressure force because in tube free
components ıˇx and ıˇy of the normal (rotations from z to x and
hydroforming, the deformed configuration is coaxial with the initial
y) are linearly expressed with semi-C0 approximations in terms of
round tube. The nodal external force is specified through a normal
the rotations at the mid-nodes:
pressure of intensity P. At each node there are three equilibrium
ıT = ıˇx,x ıˇy,y ıˇx,y + ıˇy,x ; ı = [Bf ]ıunf (10) equations and four unknowns: three displacements U, V, W and
the force intensity P (Fig. 3).
with uTnf = . . . wi . . . i . . .; i = 1, 2, 3 kT
Fint k
= Fx(ext) k
Fy(ext) k
Fz(ext)  = Pk nkx nky nkz  (15)
∗ = ı ∗ = −ıw ∗
The virtual rotations ıˇnx are zero since the
sk ,nk
rotations at mid-nodes are known. We have only two dof per node. where nkx nky nkz  is the average normal in the wall of final tube to
We finally obtain the matrix [Bf ] as (see Batoz and Dhatt, 1992, the node i.
Section 6.4.3 for details): Then we can give the pressure nodal Pk normal with the wall of
final tube for each node:
⎡ . .
⎤ T
S4 C4 − S6 C6
.
. S5 C5 − S4 C4
.
. S6 C6 − S5 C5 Pk = nkx nky nkz  · Fx(int)
k k
Fy(int) k
Fz(int)  (16)
1⎢ ⎥
[Bf ] = ⎢ ⎥
. .
A⎣ ⎦
. . But in our case of tube, the radial displacement U, in the cylindri-
−S4 C4 + S6 C6 . −S5 C5 + S4 C4 . −S6 C6 + S5 C5
. . cal coordinates (r, ϕ, Z) for each node is known. As a consequence,
. .
−C42 + S42 + C62 − S62 . −C52 + S52 + C42 − S42 . −C62 + S62 + C52 − S52 we can use this assumption to reduce the number of displacements
unknowns to two displacements for each node (ϕ and W). And
where A is the area of triangular element, Ck = xji /Lk ; Sk = Yji /Lk the displacements in the Cartesian coordinate are related with the
(k = 4, 5, 6 for ji = 21, 32, 13) and Lk2 = xji2 + yji2 . displacements in the cylindrical coordinate at any iteration i as
The principal of virtual work express on the known final 3D follows:
configuration leads to: Upi = Ur cos ϕ − Upi Upi = R0 (cos(ϕ − ϕi ) − cos ϕ)
with (17)

Vpi = Ur sin ϕ + Vpi Vpi = R0 (− sin(ϕ − ϕi ) + sin ϕ)
e e
W= (Wint − Wext ) =0 (11) The nonlinear system of equations with only two degree of free-
elt dom (ϕ and W) per node is solved using a Newton–Raphson
1902 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906
Fig. 4. The computational procedure using the one-step inverse approach.
procedure. An approximation of the solution at iteration i, leads
to:
[K iT ]U = F ext (U i ) − F int (U i ) = −R(U i )
where [K iT ] i
= [∂R (U)/∂U]
In the system (18) the force corresponding to the rotation ϕ is
assumed to be the moment: M = Ft · r and r is the first coordinate
of node in cylindrical system in the final configuration.
Uki = . . . ϕki Wki . . .
where i is the iteration and k is the number of node.
Two criteria can be used in the IA. The first one is the relative
displacement norm criterion given by:
U{U}
err =
U{U}
where {U} is the displacement increment vector during one iter-
ation, {U} is the total displacement vector. A second convergence
criterion can also be used in the IA, it is based on relative residual
norm, given by
Ri {Ri }
err =
R0 {R0 }
where {Ri } is the residual vector at the ith iteration, {R0 } is the
initial residual vector i.e. at iteration 0. During all our application
the first convergence criterion has been used with a tolerance of
10−6 .
2.4. Computational procedure
The computational procedure is as follows (Fig. 4). First, the
= 2700 kg/m3 .
final configuration depicting the geometry of the workpiece is dis-
cretized in DKT12 membrane-bending elements. In the analyses
presented in this work, the incremental approach by means of the
ABAQUS code is first used, and the deformed configuration obtained
at a given instant is considered as the final configuration for the
inverse analysis (only for validation and comparison between the
two methods). The initial guess representing the guessed nodal
Fig. 5. Part geometries (axisymmetric bulge).
positions is first obtained by a radial projection of the nodes in the
final configuration onto the initial round tube.
The initial guess allows starting the inverse process using Eq.
(6) to estimate the strains. Using the uniaxial stress–strain curve,
we determine the equivalent stress and the secant modulus. After
that, the stresses in each element are computed using the constitu-
tive relation (7). After computing the internal and external forces
and assembling the element tangent stiffness matrix, a first esti-
mate of displacement increment is obtained by the resolution of
the equilibrium equations (system (18)). After transformation of
(18) displacements to the Cartesian coordinate, the initial positions can
be updated to start a new iteration. The iterative process stops when
the displacement norm is below or equal to a desired precision.
3. Numerical results
In this section we will present the simulation results of final tube
wall thickness distribution for calibration forming of axisymmetric
bulge, of square section hollow component and free T-branch form-
ing for different branch height. These results are compared with the
results of incremental explicit method of ABAQUS commercial code
and discussed. The loading path used in incremental analyses for
(19) all applications is the optimal in terms of thickness distribution. For
all following applications, we have considered the residual forces
norm of 10−6 for the convergence criteria. To reduce the simula-
tion time in incremental method case, the forming time used in the
simulation was sped up by a factor of 1000 with apparently no ill
effect for all our applications cases (see also Koç et al., 2000).
(20) 3.1. Hydroforming of axisymmetric bulge
The analyses of aluminum alloy tubes under calibration hydro-
forming conditions are carried out, and we consider the same
tube as used in the work of Ahmetoglu et al. (2000). This tube
is of 177.8 mm length and has a 63.6 mm outside diameter and
1.65 mm initial wall thickness; the geometric parameters of the
profile of the Die are given in Fig. 5. The axisymmetric initial tube
is made of an Aluminum alloy AA6061-T6 with a Young’s modulus
E = 70665 MPa, a Poisson’s ratio = 0.3 and density
The equivalent uniaxial stress–strain curve follows the Holloman
model: ¯ = 431.01ε−0.1126 .
Experiments with the aluminum alloy 6061-T6 were run with
an axial feed of 12.7 mm and a variable pressure given in Ahmetoglu
et al. (2000). The axial feeding force is applied to the upper side only,
while the other side is completely clamped. During the experimen-
tal process, pressure reaches its maximum value of 57 MPa.
 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906
Fig. 6. Comparison of thickness variations along the tube profile.
In this application because the tube is axisymmetric and the
axial loading is asymmetric, only a quarter of the tube is modeled by
2000 S3R shell elements (50 elements along the meridian direction)
in ABAQUS code. The deformed configuration obtained at the end of
the incremental analysis is considered as the final configuration for
the inverse analysis and it is modeled using DKT12 shell element.
Fig. 6 shows the thickness variation along the meridian pro-
file. We can observe that our results are in good agreement with
the results obtained using ABAQUS Dynamic Explicit code. One can
observe also, that the thickness curve obtained by the incremental
explicit-method is closer to the experimental result especially at
the die inner radius at 80 mm, than the one obtained using the IA.
This can be explained by the fact, that in the IA the deformation is
independent of the loading path.
Thickness distribution onto the final hydroformed tube is also
given in Fig. 7. We can observe a good agreement between the two
solutions. More quantitatively summary of thinning results for both
two approaches are given in Table 1.
As we can see the maximum thinning is estimated with very
good accuracy compared to the result we have obtained using
ABAQUS Explicit.
= 7800 kg/m3 .
Fig. 7. Comparison of thickness distribution.
1903
Table 1
Comparison of thickness variations.
hmin [%] hmax (%) CPU times
ABAQUS-Explicit −13.7 1.15 1 min 9 s
Inverse Approach −13.76 0.18 2.5 s
Experiment −14.56 1.21
Fig. 8. Principal strain in mid of each element along the curvilinear length.
In Fig. 8, we show the logarithmic strains along the curvilinear
length of the tube. These strains are given for the case of DKT12
shell element by the relation 6. We can remark that our results are
in agreement with those obtained with ABAQUS code, except soft
perturbation between the results of ε22 principal strain.
3.2. Hydroforming of square section hollow component
Fig. 9 shows the shape and dimensions of a square section hol-
low component to be formed. The side length of the square section
is 45 mm and the transition radius is 6 mm. A tube with 40 mm in
diameter and 2 mm in thickness was used in this simulation. The
material of the tube is a stainless steel, its mechanical properties are
as following: The material model of the tube blank is ¯ = 628 ε−0.32 ,
yielding stress y = 368 MPa, Young’s modulus of 208 GPa (Yuan
et al., 2006), Poisson’s ratio = 0.3 and density
The total axial displacement is 8 mm in each side of the tube and
the maximum calibration pressure is 120 MPa, the relation between
stroke and pressure is considered as linear.
For symmetry reasons of geometry and loading, only 1/8 of the
tube is modeled by 750 S3R shell elements (25 elements along the
meridian direction) in ABAQUS code. The same mesh is used also
in IA using the DKT12 shell elements for comparison purpose.
The results we obtained with IA in terms of thickness distri-
bution are −16.08% of thinning and +4.62% of thickness increase
(Fig. 10). These results are in good agreement with the results we
Fig. 9. Shape and dimensions of part.
1904 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906

Fig. 12. Principal strain in mid of each element along the Arc ADB.

Fig. 10. Comparison of thickness distribution. 3.3. Hydroforming of T-shaped part

T-shaped part is one of the cases where more of material feed-

Table 2 ing, in the circumferential direction of the tube in expansion zone,
Comparison of thickness variations.
is necessary to carry out the part. The essential objective of this
hmin [%] hmax (%) CPU times application is to demonstrate the importance of verifying the equi-
ABAQUS-Explicit −19.1 8.35 45 s
librium en the circumferential direction of the final configuration
Inverse Approach −16.08 4.62 1s in the IA. In this test a cylindrical low carbon steel (LCS 1008) tube
of 45 mm (outer diameter), 169.12 mm length and 2 mm thickness
was used as the blank. The diameter of the T-branches was equal to
that of the main tube as presented in Fig. 13. A power law plastic-
have obtained using ABAQUS Explicit. Our model well captured the ity model ¯ = 484 ε−0.19 was used for the simulation and the other
variation in thickness along a section of the tube that caused the material properties as: Young’s modulus of 200 GPa, Poisson’s ratio
problem of calibration in a Die of square section. This variation can of 0.3 and density of 7800 kg/m3 .
be explained by a rotation of positions of points in the initial con-
figuration around the tube axis. Instead, in axisymmetric Die we do
not see this variation (Fig. 7). The principal results we have obtained
are summarized in Table 2.
Refer to Fig. 11, from the wall thickness plots it was observed
that the simulation using the IA and the incremental method wall
thickness distributions and variation trend were in good agree-
ment for both curvilinear length (ADB and CD), the maximum error
between the two methods is 6.35%.
In Fig. 12, we show the logarithmic strains along the Arc ADB of
the tube. We can remark that our results are in accord with those
obtained with ABAQUS code for principal strain ε22 , except soft
perturbation between the results of principal strain ε11 .

Fig. 13. Geometrical parameters for a Tee-shaped part.

Fig. 11. Comparison of thickness variation (Arcs: ADB and CD). Fig. 14. Comparison of thickness distribution (stroke of 27.65 mm).
 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906 1905

Table 3
Comparison of thickness variations.

hmin (%) hmax (%) CPU times

ABAQUS-Explicit −0.9 45.95 13 min 5 s

Inverse Approach −1.39 55.89 15 s

The forming load was internal hydraulic pressure along with

incremental axial compressive load (tube end axial feed); the lead-
ing path is as presented in the work of Koç et al. (2000). Only the
so-called free forming is considered, where the internal pressure
reaches a maximum of 44 MPa. The opposite axial punches move
toward each other with the same velocity fixed to 3 mm/s and the
maximum stroke of 34.56 mm.
The interface between the tube and the die was modeled with
an automatic surface-to-surface contact algorithm for the case of
incremental method. By taking advantage of the symmetry, it was Fig. 15. Comparison of thickness distribution (Arc AB).
possible to model 1/4th of the T-branch. In the analyses presented
in this work, the incremental approach by means of the ABAQUS
code is first used (using 4000 S3R triangular shell element, 50
elements along the meridian direction), and the deformed con-
figuration obtained at a given instant is considered as the final
configuration for the inverse analysis (using 4000 DKT12 shell ele-
ment).
Fig. 14 shows the comparison of thickness distribution for stroke
of 27.65 mm, we can observe a good results correlation between the
two methods especially in the top side of the branch.
In Table 3 we summarized the thinning results for both two
approaches. We can see that the maximum thinning is estimated
with very good accuracy compared to the result we have obtained
using ABAQUS Explicit. These results are obtained with very small
CPU time using the inverse approach.
For more details we present in Fig. 15 the distribution of
thickness on the arc AB for different case of stroke (6.91, 13.82,
20.74 and 27.65 mm). For each stroke the protrusion height (Hp ) Fig. 16. Principal strain in mid of each element along the Arc AB (stroke of
obtained by the incremental approach is 28.69, 35.37, 42.82 and 27.65 mm).
49.81 mm, respectively. Although the error of the maximum thick-
ening between the two methods is increase with the augmentation
of stroke, in the case of 27.65 mm, this error rich only 6.8%. This convergence criteria (the residual forces norm less to 10−6 ) is note
maximum error obtained by the IA where the elements are under a verified in our model.
big geometric distortion (blue areas, Fig. 14) (for interpretation of In Fig. 16, we show the logarithmic strains along the Arc AB of
the references to color in the text, the reader is referred to the web the tube for the stroke of 27.65 mm.
version of the article.). This is certainly related to local effects that The deformed configuration obtained for the stroke of 27.65 mm
are due to the history of bending/unbending moment through the considered as the final configuration for the inverse analysis. The
radius are not taken into account in our model, let us remember mesh obtained by the inverse approach is similar (structured mesh)
that is based on a one step.We have not present the cases obtained to the initial mesh used for the incremental method, as presented
with more than the stroke of 27.65 mm because after this later the in Fig. 17.

Fig. 17. Deformed configuration used and initial configuration obtained by IA (stroke of 27.65 mm).
1906 M.S. Chebbah et al. / Journal of Materials Processing Technology 211 (2011) 1898–1906
4. Conclusions
This work focuses on the simulation of tube hydroforming pro-
cess in which the initial geometry is a cylindrical tube submitted to
an internal pressure and axial feeding at the tube ends.
A 3D membrane/bending shell element formulation based on
the inverse approach for tube hydroforming have been developed
in this paper. The present model can be used to estimate the
strain and thickness distributions in the workpieces hydroformed
from an initial round tube. The formulation considers the initial
and final configurations that are coaxial. For validation purposes,
experimental results based on literature review are used while the
incremental explicit dynamic approach was conducted for the sim-
ulation of all applications using ABAQUS software.
Application of tube hydroforming process simulation has been
performed to validate the proposed 3D model, and the obtained
results are in a good agreement with both experimental results
and results obtained using classical incremental dynamic explicit
approach by ABAQUS. The general CPU computing time using
the proposed FE inverse model is very small (even negligible) in
comparison with the computation time of incremental explicit
dynamic approach using ABAQUS. Globally, the results of numer-
ical applications obtained confirm that the model based on IA for
the hydroforming simulation is efficient and particularly suited as
tool to verifying the feasibility of product.
References
Abrantes, J.P., Szabo-Ponce, A., Batalha, G.F., 2005. Experimental and numerical sim-
ulation of tube hydroforming (THF). Journal of Materials Processing Technology
164–165, 1140–1147.
Ahmed, M., Hashmi, M.S.J., 1997. Estimation of machine parameters for hydraulic
bulge forming of tubular components. Journal of Materials Processing Technol-
ogy 64 (1–3), 9–23.
Ahmetoglu, M., Sutter, K., Li, X.J., Altan, T., 2000. Tube hydroforming: current
research, applications and need for training. Journal of Materials Processing
Technology 98 (2), 224–231.
Alaswad, A., Olabi, A.G., Benyounis, K.Y., 2011. Integration of finite element analysis
and design of experiments to analyse the geometrical factors in bi-layered tube
hydroforming. Materials and Design 32, 838–850.
Batoz, J.L., Dhatt, G., 1992. Modélisation des structures par éléments finis, vol. 3.
Editions Hermès, Paris, pp. 387–393 (in French).
Batoz, J.L., Guo, Y.Q., Mercier, F., 1998. The inverse approach with simple trian-
gular shell elements for large strain predictions of sheet metal forming parts.
Engineering Computations 15 (7), 864–892.
Chebbah, M.S., Naceur, H., Hecini, M., 2010. Rapid coupling optimization method
for tube hydroforming process. Journal of Engineering Manufacture 224 (2),
245–256.
Di Lorenzo, R., Ingarao, G., Chinesta, F., 2010. Integration of gradient based and
response surface methods to develop a cascade optimisation strategy for Y-
shaped tube hydroforming process design. Advances in Engineering Software
41, 336–348.
Fu, L., Dong, X., Wang, P., 2009. Study on one-step simulation for the bending process
of extruded profiles. International Journal of Advanced Manufacturing Technol-
ogy 43 (11–12), 1069–1080.
Guo, Y.Q., Batoz, J.L., Detraux, J.M., Duroux, P., 1990. Finite element procedures
for strain estimations of sheet metal forming parts. International Journal for
Numerical Methods in Engineering 30, 1385–1401.
Guo, Y.Q., Batoz, J.L., Naceur, H., Bouabdallah, S., Mercier, F., Barlet, O., 2000. Recent
developments on the analysis and optimum design of sheet metal forming parts
using a simplified inverse approach. Computers & Structures 78 (1–3), 133–148.
Hartl, Ch., 2005. Research and advances in fundamentals and industrial appli-
cations of hydroforming. Journal of Materials Processing Technology 167,
383–392.
Hill, R., 1984. A theory of the yielding and plastic flow of anisotropic met-
als. Proceedings of the Royal Society of London, Series A 193 (1033),
p.281–p.297.
Hosford, W.F., Caddell, R., 2011. Metal Forming: Mechanics and Metallurgy. Edition
Cambridge University Press, 331 pages, ISBN: 978-1-107-00452-8.
Jain, N., Wang, J., Alexander, R., 2004. Finite element analysis of dual hydroforming
processes. Journal of Materials Processing Technology 145 (1), 59–65.
Kashani Zadeh, H., Mashhadi, M.M., 2006. Finite element simulation and experi-
ment in tube hydroforming of unequal T shapes. Journal of Materials Processing
Technology 177 (1–3), 684–687.
Kim, J., Kang, B.-S., Hwang, S.-M., 2002. Preform design in hydroforming by the
three-dimensional backward tracing scheme of the FEM. Journal of Materials
Processing Technology 130-131, 100–106.
Koç, M., Allen, T., Jiratheranat, S., Altan, T., 2000. The use of FEA and design of experi-
ments to establish design guidelines for simple hydroformed parts. International
Journal of Machine Tools & Manufacture 40 (15), 2249–2266.
Koç, M., Altan, T., 2001. An overall review of the tube hydroforming (THF) technology.
Journal of Materials Processing Technology 108 (3), 384–393.
Kridli, G.T., Bao, L., Mallick, P.K., Tian, Y., 2003. Investigation of thickness varia-
tion and corner filling in tube hydroforming. Journal of Materials Processing
Technology 133 (3), 287–296.
Kwan, C.T., Lin, F.C., 2003. Investigation of T-shape tube hydroforming with finite
element method. International Journal of Advanced Manufacturing Technology
21 (6), 420–425.
Langa, L., Li, H., Yuanb, S., Danckert, J., Nielsenc, K.B., 2009. Investigation into the pre-
forming’s effect during multi-stages of tube hydroforming of aluminum alloy
tube by using useful wrinkles. Journal of Materials Processing Technology 209,
2553–2563.
Mac Donald, B.J., Hashmi, M.S.J., 2001. Three-dimensional finite element simulation
of bulge forming using a solid bulging medium. Finite Element in Analysis and
Design 37 (2), 107–116.
Mohammadi, F., Mosavi Mashadi, M., 2009. Determination of the loading path for
tube hydroforming process of a copper joint using a fuzzy controller. Interna-
tional Journal of Advanced Manufacturing Technology 43, 1–10.
Naceur, H., Guo, Y.Q., Ben-Elechi, S., 2006. Response surface methodology for design
of sheet forming parameters to control springback effects. Computers & Struc-
tures 84 (26–27), 1651–1663.
Naceur, H., Ben-Elechi, S., Batoz, J.L., Knopf-Lenoir, C., 2008. Response surface
methodology for the rapid design of aluminum sheet metal forming parameters.
Materials and Design 29, 781–790.
Nikhare, C., Weiss, M., Hodgson, P.D., 2010. Die closing force in low pressure
tube hydroforming. Journal of Materials Processing Technology 210 (15),
2238–2244.
Nguyen, B.N., Johnson, K.I., Khaleel, M.A., 2003a. Analysis of tube free hydroforming
using an inverse approach with FLD-based adjustment of process parameters.
Journal of Manufacturing Science and Engineering 125, 135–140.
Nguyen, B.N., Johnson, K.I., Khaleel, M.A., 2003b. Analysis of tube hydroforming by
means of an inverse approach. Journal of Manufacturing Science and Engineering
125, 369–377.
Numisheet, 2008. In: Hora, P. (Ed.) Part A: Proceedings of the 7th International Con-
ference and Workshop on Numerical Simulation of 3D Sheet Metal Forming
Processes, ETH Zurich, September 1–5, 2008. Interlaken, Switzerland.
Ray, P., Mac Donald, B.J., 2004. Determination of the optimal load path for tube
hydroforming processes using a fuzzy load control algorithm and finite element
analysis. Finite Elements in Analysis and Design 41, 173–192.
Singh, H., 2003. Fundamentals of Hydroforming. Society of Manufacturing Engineers,
Dearborn, MI, USA, 219 pages, ISBN-13: 9780872636620.
Yuan, S.J., Han, C., Wang, X.S., 2006. Hydroforming of automotive structural com-
ponents with rectangular sections. International Journal of Machine Tools &
Manufacture 46 (11), 1201–1206.
 3
Journal of Materials Processing Technology 211 (2011) 2012–2021

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

A theoretical and experimental study on forming limit diagram for a seamed

tube hydroforming
Xianfeng Chen a , Zhongqi Yu a,∗ , Bo Hou a , Shuhui Li a,b , Zhongqin Lin a,b
a
Shanghai Key Laboratory of Digital Autobody Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
b
State Key Laboratory of Mechanical System and Vibration, Shanghai 200240, China

a r t i c l e i n f o a b s t r a c t

Article history: The purpose of this work is to establish the forming limit diagram (FLD) for a seamed tube hydroforming.
Received 24 February 2011 A new theoretical model is developed to predict the FLD for a seamed tube hydroforming. Based on this
Received in revised form 21 June 2011 theoretical model, the FLD for a seamed tube made of QSTE340 sheet metal is calculated by using the
Accepted 26 June 2011
Hosford yield criterion. Some forming limit experiments are performed. A classical free hydroforming
Available online 1 July 2011
tool set is used for obtaining the left hand side forming limit strains, and a novel hydroforming tool set is
designed for the right hand side of FLD. The novel device required the simultaneous application of lateral
Keywords:
compression force and internal pressure to control the material flow under tension–tension strain states.
Forming limit diagram (FLD)
Seamed tube
Furthermore, the suitable loading paths for the left hand side of FLD by theoretical formulas and for the
Tube hydroforming right hand side of FLD by finite element (FE) simulations are calculated. Finally, a comparison between
Theoretical model the theoretical results and experimental data is performed. The theoretical predicting results show good
Loading paths agreement with the experimental results.
© 2011 Elsevier B.V. All rights reserved.

1. Introduction AA6061 tube based on the free-expansion tube hydroforming.

In recent years, the automotive industry has shown a growing
interest in tube hydroforming. The process has some consider-
able advantages compared to conventional forming processes like
stamping. It allows weight reduction, part consolidation, reduced
tooling cost, fewer operations and improved structural stiffness of
the hydroformed part. Due to its low cost, high productivity, high
quality and varieties in products, seamed tubes have been widely
used to manufacture hydroformed parts. However, bursting for
seamless and seamed tube is different under free hydroforming,
as illustrated in Fig. 1. The cracking position is located at the base
metal near the weld zone for a seamed tube. The main cause of
such defects is the constraint of the weld zone in a seamed tube
hydroforming.
Bursting is generally classified as an irrecoverable failure mode.
Hence in order to obtain the sound hydroformed parts, it is nec-
essary to predict the bursting behavior and to analyze the effects
of process parameters and weld metal on this failure condition in
hydroforming processes. The forming limit diagram (FLD) can be
used as a measure of the maximum formability of the material.
Some experimental and theoretical methods have been applied
to study the FLD for tube hydroforming. Davies et al. (2000) pro-
posed a tooling and experimental apparatus to establish FLD for
∗ Corresponding author. Tel.: +86 21 34206785; fax: +86 21 34204542.
E-mail address: yuzhq@sjtu.edu.cn (Z. Yu).
These experimental results were compared to the theoretical FLD
calculated via the Marciniak–Kuczynski (M–K) method. Varma et al.
(2007) and Varma and Narasimhan (2008) analyzed the local-
ized necking in aluminum tubes free hydroforming using the M–K
method along with an anisotropic version of the Gurson model.
They also studies the influence of loading conditions, such as pre-
scribed fluid pressure or volume flow rate in conjunction with
axial end feed. Kulkarni et al. (2004) studied the localized neck-
ing initiation in aluminum alloy tubes during hydroforming by
an experimental and numerical method. In addition, approximate
analytical approach has been adopted to obtain the peak internal
pressure and M–K analyses have been conducted to predict the
limit strains corresponding to onset of necking of the tube wall.
Hashemi et al. (2009) obtained a theoretical forming limit stress
diagram (FLSD) to predict necking initiation in tube hydroform-
ing via the M–K model. Hwang et al. (2009) used bugle tests to
establish the FLD of tubular material AA6011. These experimen-
tal results were compared to the theoretical FLD calculated via the
Swift’s diffused necking criterion and Hill’s localized necking. In
these studies, only the left hand side (ˇ < 0) of FLD could be obtained
from experiments for the difficulty to clamp both ends of tube in
free hydroforming for the right hand side (ˇ > 0) of FLD. The dif-
ference between the theoretical methods for sheet forming and
for tube hydroforming could not be considered. So it is question-
able whether a direct application of the concept of FLD obtained in
sheet metal forming to tube hydroforming is a physically possible
solution.

0924-0136/$ – see front matter © 2011 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2011.06.023
 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021 2013

Nomenclature

Pw “w” zone internal pressure

Pb “b” zone internal pressure
Pi instantaneous internal pressure
F axial force
Fmixed total load on the mixed tensile specimen
A cross-sectional area
t0 initial tube wall thickness
ti instantaneous tube wall thickness
d0 initial tube diameter
di instantaneous tube diameter
Fig. 1. Cracking position for seamless and seamed tubes in hydroforming.
¯ equivalent stress
1 circumferential stress
2 longitudinal stress of FLD for tube hydroforming from the perspective various combi-
FLD0 limit strain under plane strain state nations of loading parameters based on plastic instability. Kim et al.
ε0 initial equivalent strain (2004a) predicted the bursting failure based on the Hill’ quadratic
ε̄ equivalent strain plastic potential under combining internal pressure and axial feed-
ε1 circumferential strain ing. Finally, the hydroforming strain and stress limit diagrams were
ε2 longitudinal strain obtained from the above approach. Kim et al. (2009) investigated
ε3 radial strain the theoretical FLSD based on the local necking criterion to predict
e1 engineering circumferential strain bursting failure in tube hydroforming. Then the proposed analyti-
e2 engineering longitudinal strain cal approach based on the implementation of the FLSD was verified
dε1 circumferential strain increment with a series of bulge tests.
dε2 longitudinal strain increment In the above studies, all bursting failure predictions did not study
u1 circumferential micro-displacement the effects of the weld line of seamed tube. Kim et al. (2004b)
u2 longitudinal micro-displacement investigated the forming limit and bursting pressure level for a
u3 radial micro-displacement seamed tube that comprised weld metal, heat affected zone (HAZ)
0 initial tube radius and base material parts by means of the finite element method
1 instantaneous circumferential radius (FEM) combined with Oyane’s ductile fracture criterion based on
2 instantaneous longitudinal radius the Hill’s quadratic plastic potential. Song et al. (2010) performed
L0 initial tube length some experiments for seamed tube hydroforming. But the theoret-
w free tube length ical prediction for tube did not study the effects of the weld line. A
rd die profile radius seamed tube used in hydroforming is generally produced by high
s stroke of punch frequency electric resistance welding (HF-ERW) after a roll forming
h stroke of pressing block operation. For a steel seamed tube, the base metal has significantly
c width of pressing block lower yield strength and higher ductility than the weld metal. The
l0 initial contact length quality of the weld in a tubular blank is critical for a successful
K strength coefficient forming operation. However, no previous theoretical study on the
E elastic modulus influence of the weld metal properties on the forming limit of bulge
 Poisson’s ratio forming has been presented.
n strain hardening exponent As discussed all above, the FLD for tube hydroforming is now
m stress exponent of crystal structure receiving increasing attention. But current researches on the FLD
R anisotropy coefficient for a seamed tube hydroforming are restricted in the application
ϕ ratio of the equivalent stress and the major stress due to the experimental apparatus and the theoretical model. So it
˛ stress ratio must be studied urgently that a special apparatus for the right hand
ˇ strain ratio
f constant of failure criterion
 circumference ratio
rin instantaneous tube inner radius
rout instantaneous tube outer radius
 coefficient of friction

Some research works had shown that the FLD for tube hydro-
forming may be different from that for sheet metal forming (Chu
and Xu, 2004a,b). Xing and Makinouchi (2001) concluded that the
forming zone for tube hydroforming is narrower than for conven-
tional sheet stamping. In the biaxial tensile stress zone, the forming
limit for tubular hydroforming is significantly lower than that for
sheet forming; while in the compression-tension stress zone, it
is between those of the diffusion necking and the local necking
for sheet forming. Nefussi and Combescure (2002) highlighted two
Swift’s diffuse criteria for sheet forming and for tube hydroforming
to predict necking. Chu and Xu (2008) investigated the prediction Fig. 2. Theoretical analysis model of FLD for a seamed tube.
2014 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021

Fig. 4. The stresses acting on an element at the middle of the tube.

tube, the following equilibrium equation can be written (Woo and

Hawkes, 1968):
Fig. 3. Theoretical simplified analysis model of FLD for a seamed tube.
 2

1
Pi = ti + (3)
1 2
side of FLD and a theoretical model for a seamed tube hydroforming,
which reflects the effect of the weld zone on the forming limit. where Pi is the instantaneous internal pressure, ti is the instan-
In this study, it is aimed to develop a new theoretical model taneous tube wall thickness,  1 and  2 are stress components in
to predict the FLD for a seamed tube hydroforming. Based on this the circumferential and the axial directions, and 1 and 2 are the
theoretical model, the FLD of seamed tube made of QSTE340 is cal- instantaneous circumferential and the longitudinal radii.
culated by using the Hosford yield criterion. A novel experimental And
device is designed for the right hand side of the FLD. Good cor-
1 = 0 eε1 (4)
relation between the theoretical results and experimental data is
indicated. In Section 2, a new theoretical model is developed, and ti = t0 eε3 (5)
the tubular material properties are tested to predict the FLD for a
seamed tube hydroforming. In Section 3, three type experiments where ε1 and ε3 are strain components in the circumferential and
are performed and the strain paths on tube hydroforming FLD are the radial directions, 0 is the initial tube radius, and t0 is the initial
investigated. In Section 4, the theoretical results are compared with tube wall thickness.
the experimental data and discussed. In the end the conclusions are Assume now that the expansion occurs in the profile shown in
drawn in Section 5. Fig. 5. This assumption means that the longitudinal radius of cur-
vature 2 < ∞. During free forming, tube free bulging area can be
approximated as elliptic profile (Hwang and Lin, 2006). Therefore,
2. Theoretical model of FLD
using elliptic function to describe the longitudinal profile of bulging
area, as shown in Fig. 5 can be a tube of plastic deformation of the
2.1. Developing theoretical model
geometry of the instantaneous. In Fig. 5, L0 is the initial tube length,
w is the free expansion length, rd is the die profile radius, s is the
From a macroscopic point of view, a seamed tube is composed
stroke of punch and l0 is the initial contact length.
of weld metal, HAZ and base metal, as illustrated in Fig. 2, in which
In the elliptical cross-section, so the elliptic function can be
the hardness of weld metal and HAZ is higher than that of base
denoted as
metal. The plasticity and toughness of weld metal are poor, and
it is difficult to deform. Due to the constraints of the weld zone, x2 y2
base metal near HAZ shows “abnormal” deformation behavior in + 2 =1 (6)
a2 1
a seamed tube hydroforming. Fig. 1 shows the cracking position
which is always located at the interface of the base metal and the where a and 1 are the lengths of semimajor and semiminor axes
weld zone for a seamed tube. So we need to develop a new theoret- of the elliptical cross-section.
ical model to reflect the constraint effect of the weld zone on the Considering the point of contact with the die fillet: (rd + w/2, 0 )
base metal. is on the elliptical arc, one obtains
Theoretical analysis model of FLD for a seamed tube is shown  2r + w  1
in Fig. 2, where 1-direction is the major strain direction and 2- d
a= (7)
direction is the minor strain direction. Weld metal and HAZ are 2 (12 − 02 )
1/2

called “w” zone, and base metal is named “b” zone. The “w” zone
will only transmit the load when the external force acts on the tube
because it is high hardness and difficult to deform. So Fig. 2 can be
simplified to Fig. 3. The force equilibrium equation can be consid-
ered by applying the internal pressure at the “w” and “b” zone. So
the force equilibrium equation can be written as follows:

Pw = Pb (1)

The compatibility requirement assumes that the minor strain

increment is equal in both zones as follows:

dε2w = dε2b (2)

Consider a tube which is subject to internal pressure Pi and com-

pressive axial force F in Fig. 4. For an element at the middle of this Fig. 5. The deforming mode of tube hydroforming (2 < ∞).
 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021 2015

And
 2r + w 2  1

d
a= (8)
2 12 − 02

Substituting Eqs. (4), (5) and (8) into Eq. (3) leads to
 
t0 2 (12 − 02 )
Pi = exp(ε3 − ε1 ) 1 + 2
(9)
0 (rd + w/2)

Then the internal pressures of “w” and “b” zone can be written
as follows:
 2 − 2 )

t0w 2w (1w 0w
Pw = exp(ε3w − ε1w ) 1w + 2
0w (rd + w/2) Fig. 6. Measuring positions for micro-hardness test.
 2 2 )
 (10)
t0b 2b (1b − 0b
Pb = exp(ε3b − ε1b ) 1b + 2
0b (rd + w/2)

Substituting Eq. (10) into Eq. (1) leads to

 2 − 2 )

t0w 2w (1w 0w
exp(ε3w − ε1w ) 1w + 2
0w (rd + w/2)
 2 − 2 )

t0b 2b (1b 0b
= exp(ε3b − ε1b ) 1b + 2
(11)
0b (rd + w/2)

¯ = K(ε0 + ε̄)n (12)

Assume that the tube material obeys the Swift’s hardening

law:where K is the strength coefficient, ¯ is the equivalent stress,
ε0 is the initial equivalent strain, ε̄ is the equivalent strain and n is
the strain hardening exponent.
Then the equivalent stresses of “w” and “b” zone can be written
Fig. 7. Micro-hardness profile of a seamed tube.
as follows:
¯ w = Kw (ε0w + ε̄w )nw
(13) Levy–Mises flow rule yields (assuming volume constancy)
¯ b = Kb (ε0b + ε̄b )nb
˛(m−1) − R(1 − ˛)m−1
Defining ϕ and ˛ as the ratio between the equivalent stress and ˇ= (19)
1 + R(1 − ˛)(m−1)
the major stress and between the minor and major stresses.
Finally, one can predict the FLD for a seamed tube hydroforming
¯ 2
ϕ= , ˛= (14) by solving the non-linear equation math problems of Eq. (15).
1 1
Combining Eqs. (11), (13) and (14), one can write 2.2. Testing material properties of a seamed tube
 2 2 )

t0w 1 ˛w (1w − 0w
· · Kw (ε0w + ε̄w )nw · exp(ε3w − ε1w ) 1 + 2
The magnitude of Kw , nw , Rw , Kb , nb and Rb must be ascertained
0w ϕw (rd + w/2) firstly. So the determination of an appropriate method for measur-

2 − 2 )
 ing the width of the “w” zone is crucial and thus micro-hardness
t0b 1 ˛b (1b
= · · Kb (ε0b + ε̄b )nb · exp(ε3b − ε1b ) 1 + 0b
2
profile is used in this work. Fig. 6 depicts measuring positions along
0b ϕb (rd + w/2) the circumference direction. Micro-hardness profile of a seamed
(15) tube with 2.5 mm thickness and 58 mm outer diameter is plotted
in Fig. 7. This profile indicates that the weld metal and HAZ widths
are approximately 4 and 12 mm, respectively. So the width of “w”
Defining ˇ as the strain increment ratio zone is 16 mm.
ε2 Hence, the assumption of iso-strain in both “w” zone and “b”
ˇ= (16) zone used by Saunders and Wagoner (1996) is applied in this study.
ε1
Two schematic tensile test specimens are illustrated in Figs. 8 and 9.
Hosford yield criterion can be written as,

|1 |m + |2 |m + R|1 − 2 |m = (R + 1)¯ m (17)

where m is the stress exponent of crystal structure and R is the

anisotropy coefficient.
Combining Eqs. (14) and (17), one can write
 1/m
1 + ˛m + R(1 − ˛)m
ϕ= (18)
R+1 Fig. 8. Specimen sizes for tensile test (unit: mm).
2016 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021

Fig. 9. Tensile test specimens cut from a seamed tube.

According to Fig. 8, four specimens of base metal alone and four

specimens of mixed metal containing “w” zone and “b” zone metal,
which are cut from the seamed tubes, are prepared as shown in
Fig. 9.
For the iso-strain method based on the rule of mixtures longitu-
dinal strain is assumed to be constant across the specimen which
is cut from the seamed tube, such that
ε2w = ε2b (20)
where the subscript w and b refer to “w” zone metal and “b” zone
metal respectively.
The “w” zone metal and “b” zone metal material data sets are fit
to Swift’s hardening law (Eq. (13)).
In the tensile test using the specimen of mixed metal, the total
load on the mixed specimen Fmixed is represented as
Fmixed = ¯ w Aw + ¯ b Ab = ¯ mixed Amixed (21)
where Amixed is the cross-sectional area of the mixed specimen, Aw
and Ab are the cross-sectional areas of “w” zone metal and “b” zone
metal in the section of mixed specimen.
And
¯ mixed = Kmixed (ε0mixed + ε̄mixed )nmixed (22)
Combining Eqs. (13), (21) and (22), one can write
Kmixed (ε0mixed + ε̄mixed )nmixed Amixed − Kb (ε0b + ε̄b )nb Ab
¯ w = (23)
Aw
Fig. 11. The algorithm flowchart of predicting FLD for a seamed tube hydroforming.
As shown in Eqs. (22) and (23), the stress–strain relation for
the “w” zone is calculated. Fig. 10 shows the stress–strain relations
for the “w” zone and “b” zone and the two data are fitted to the tube and forming instability. The force equilibrium equation can be
flow curve as ¯ = K(ε0 + ε̄)n . Finally, material properties for each established based on the theoretical model and material properties
material are obtained as listed in Table 1. are obtained from tensile tests. Then one can predict the FLD for a
seamed tube hydroforming by solving the non-linear equation. The
2.3. Predicting FLD principle is as follows.
At a strain state, a strain increment dε1w is set, and then the
The method for developing a theoretical model of the FLD for strain increment dε1b can be obtained from the force equilibrium
a seamed tube is based on the structural characteristics of seamed equation. If the ratio of dε1b /dε1w is less than f (where f is the con-
stant of failure criterion), then the tube does not reach the bursting
state and the program goes through to the next strain increment
dε1w . The calculation is iterated until the constant of failure crite-
rion f is reached. Then the value of (ε2b , ε1b ) is the strain limit of a
seamed tube in bursting state. Accordingly, the algorithm flowchart
of predicting FLD for a seamed tube hydroforming is given in Fig. 11.

3. Experiments for FLD

Fig. 10. True stress–strain relation for weld and base zone materials.
In previous researches, the left hand side of FLD for tube hydro-
forming could be easily obtained by free expansion. But for the
right hand side of FLD, the difficulty is to clamp both tube ends
in tube free hydroforming to achieve the tension–tension strain
states. In this study, three type experiments were performed for
the whole FLD on the hydroforming machine designed in Auto
Body Technology Center of Shanghai Jiaotong University. The max-
imum allowable working pressure of the machine is 200 MPa and
the maximum allowable axial force is 1000 kN. The tubes used in
the experiments were straight seamed tubes with 58 mm outside
diameter, 2.5 mm wall thickness and 245 mm long. The tube mate-
 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021 2017

Table 1
Material properties of the “b” zone metal and “w” zone metal.

Material Strength coefficient Hardening exponent Initial equivalent strain Anisotropy coefficient
K (MPa) n ε0 r

“b” zone metal 680 0.145 0.040 1.17

“w” zone metal 695 0.123 0.021 1.00

rial was a QSTE340 steel. Additionally, for the tubes used for the In Eq. (25), the magnitude of the normal stress  N must be ascer-
forming limit experiments, circular grids with a diameter of 2.5 mm tained. Therefore, the normal stress  N will be addressed in more
are electrochemically etched on the tube surface before the exper- detail in the following.
iments. For an element in the guiding zone, the stress state can be
described as shown in Fig. 13. In Fig. 13, rout is the initial outer
3.1. Experiments for the left hand side of FLD tube radius, rin is the initial inner tube radius,  is the Poisson’s
ratio, and E is the elastic modulus. A plane strain state condition
A classical free hydroforming tool set is used in this study for is assumed, and the micro-displacement in all directions are u1 , u2
obtaining the left hand side of FLD, shown in Fig. 12. During free and u3 . Assume that u3 is a function of r, and u1 = u2 = 0.
forming, the tube is subject to axial force F and an internal pressure The strains can be written as
Pi . u3 ∂u3
ε1 = , ε2 = 0, ε3 = (27)
The forming limit is determined by controlling the material flow r ∂r
along linear strain paths. So in Fig. 4, the strain ratio of an element In the guiding zone, elastic deformation only occurs. The rela-
at the pole of the forming tube has to be kept as a constant value. tions between the strain and the stress can be written as
Asnafi (1999) and Asnafi and Skogsgardh (2000) constructed ana-
E
lytical models to determine the loading paths for isotropic material. 1 = [(1 − )ε1 + ε3 ]
(1 + )(1 − 2)
They are
E
 2  n 2 = (ε1 + ε3 ) = (1 + 2 ) (28)
¯ (1 + )(1 − 2)
Pi = t 1 − ˛ + ˛2 ε1
1/2 i 2−˛
(1 − ˛ + ˛2 ) E
3 = [(1 − )ε3 + ε1 ]
 2 ˛
 (1 + )(1 − 2)
× + (24)
di − ti 2 The following equilibrium equation can be written
1 ∂ 1
  (r3 ) − 1 = 0 (29)
2˛ r ∂r r
F = Fforming + Fsealing + Ffriction = Pi 12 1−
1 + ˛1 /2 Combining Eqs. (27) and (28), one can write
(d − 2t )
2  
E u3 ∂u3
+ Pi 
0 0
+ N d0 (l0 − s) (25) 1 = (1 − ) +
2 (1 + )(1 − 2) r ∂r
  (30)
E ∂u3 u3
For anisotropic material which is fit to Swift’s hardening law 3 = (1 − ) +
(1 + )(1 − 2) ∂r r
¯ = K(ε0 + ε̄)n , Eq. (24) can be written as
Substituting Eq. (30) into Eq. (29) leads to
   n  
Kti 1+R 2R E(1 − ) ∂2 u3 1 ∂u3 u3
Pi = ε0 + √ 1+ ˇ + ˇ2 ε1 + − 2 =0 (31)
[1 − ˛(2R/(R + 1)) + ˛2 ]
1/2
1 + 2R R+1 (1 + )(1 − 2) ∂2 r r ∂r r
  Then
2 ˛
×
di − ti
+
2
(26)  
∂ 1 ∂
(ru3 ) = 0 (32)
∂r r ∂r

Integrating it gives
1 C2
u3 = C1 r + (33)
2 r

Fig. 12. Free hydroforming dies for the left hand side of FLD. Fig. 13. Stresses of an element in the guiding zone.
2018 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021

Combining Eqs. (27) and (33), one can write

1 C2 1 C2
ε1 = C1 + 2 , ε3 = C1 − 2 (34)
2 r 2 r
Substituting Eq. (30) into Eq. (29), we obtain
E
1 C2

1 = C1 + (1 − 2)
(1 + )(1 − 2) 2 r2
E
2 = C1 (35)
(1 + )(1 − 2)
E
1 C2

3 = C1 − (1 − 2)
(1 + )(1 − 2) 2 r2
The boundary condition can be written as

3 (r = rin ) = −Pi , 3 (r = rout ) = −N (36)

Substituting Eq. (36) into Eq. (35), we obtain

Fig. 14. The axial force versus the internal pressure for different strain paths.
1 C2 (1 + )(1 − 2)
C1 − (1 − 2) 2 = − Pi
2 r E
in
(37)
1 C2 (1 + )(1 − 2)
C1 − (1 − 2) 2 = − N
2 rout E

Solving for C1 and C2 , Eq. (37) can be rearranged to yield

2 2
(1 + )(1 − 2)2 (Pi rin − N rout )
C1 = 2 − r2
E rout in
(38)
2 2
(1 + ) rin rout
C2 = 2 2
(Pi − N )
E rout − rin

Substituting Eq. (38) into Eq. (33), we obtain

2 r2
 2 −  r2 )
 Fig. 15. Hydroformed parts at different pressure levels under plane strain state.
(1 + )rin out (Pi − N ) (Pi rin N out
u3 = 2 − r2 )
+ (1 − 2) 2 r2
r (39)
E(rout r rout the free expansion region of the tube must be enough length. The
in in
result of the hydroformed product under plane strain state is shown
Substituting the boundary condition u3 (r = rout ) = 0 into Eq. (38),
in Fig. 15 (the right bursting product).
we obtain
In order to verify the material flow along plane strain path,
2P
2(1 − )rin i hydroformed parts at different pressure levels under plane strain
N = (40)
2
rin 2
+ (1 − 2)rout state are performed, as illustrated in Fig. 15. A three dimensional
image processing system, in Fig. 16, is used to measure the major
Substituting Eq. (40) into (25), we obtain strains e1 and minor strains e2 of the deformed grids on the tube
 2˛
 (d − 2t )
2 surface. And the values of the (e2 , e1 ) are used to construct the
0 0
F = pi 12 1 − + pi  experimental points in Fig. 17. The engineering FLD0 is 21.9%.
1 + ˛1 /2 2
2 p d (l − s)
2(1 − )rin i 0 0
+ 2 2
(41)
rin + (1 − 2)rout

In order to generate constant strain ratio paths, the relationships

between the internal pressure and the axial force are calculated
using Eqs. (26) and (41). Based on these formulas, the analytical
models for linear strain paths are designed.
Fig. 14 is based on Eqs. (26) and (41) and displays the yield
limit and the loading characteristics in hydroforming along dif-
ferent strain paths. Also in this case, a rapid increase in the axial
force can be observed at the end of the free forming (Fig. 14). This
behavior is caused by instability. The free forming process becomes
unstable when the internal pressure reaches a maximum.
Five different strain paths experiments with accompanying
strain ratios of ˇ = −0.08 to ˇ = −0.20 are performed, and the results
of the hydroformed product for different strain paths are shown in
Section 4.

3.2. Experiments for FLD0

The free hydroforming tool set in Fig. 12 is still used in this study
for obtaining the FLD0 . The different is that both ends of tube are
fixed. In order to control the material flow along plane strain path, Fig. 16. Calibration of the strains.
 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021 2019

Fig. 20. Finite element meshes for FE simulation.

Fig. 17. Validation of linearity for the strain path under plane strain state. 4. Results and discussion

The results of the hydroformed products after free hydroforming

with axial feeding and internal pressure for different strain ratios
are shown in Fig. 23. It is known that the bursting lines occur at the
base material near HAZ and the maximum bulge height increases
with the absolute value of the strain ratio ˇ. Fig. 24 displays the
hydroformed products with lateral compression force and internal
pressure.
A three dimensional image processing system, in Fig. 16, is
used to measure the major strains e1 and minor strains e2 of the
deformed grids on the tube surface. And the values of the true strain
(ε2 , ε1 ) are transformed from (e2 , e1 ) and used to construct the
experimental forming limit points in Fig. 25.

Fig. 18. Hydroforming dies for the right hand side of FLD.

3.3. Experiments for the right hand side of FLD

At tension–tension strain states, it is difficult to clamp both ends

of tube in free hydroforming and the requirements for hydroform-
ing equipment are higher than normal. So a novel hydroforming
tool set is designed for the right hand side of FLD, as illustrated
in Fig. 18. The principle is to control the material flow under
tension–tension strain states by applying lateral compression force
and internal pressure, as shown in Fig. 19. In order to achieve dif-
ferent loading paths, three pressing blocks are used, and the width
c is 50 mm, 75 mm and 100 mm. The stroke of the pressing block is
h.
The linear strain paths are determined by FE simulation. The
finite element meshes are shown in Fig. 20 and Fig. 21 shows the
simulation results for constant strain ratio paths of the node with
the lowest thickness value. The loading paths for different double
tension strain states are shown in Fig. 22.

Fig. 19. Principle schematic for the right hand side of FLD. Fig. 21. Simulation results of ˇ = 0.1 and ˇ = 0.2.
2020 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021
Fig. 22. Loading paths for different strain paths under double tension strain states.
Fig. 23. Hydroformed parts for ˇ = −0.2–0.
The current work assumes that the values of the stress exponent
of crystal structure mw = mb = 6 for BCC steel material and the con-
stant of failure criterion f = 10. The theoretical forming limit curve
can be obtained by solving Eq. (15) with the help of an algorithm
which was implemented using VC++ programming environment.
Fig. 25 displays the experimental and theoretical FLD to be dis-
cussed in the following section.
Fig. 25 shows a comparison between the theoretical FLD and
experimental data for QSTE340 seamed tube. The experimental
data are slightly larger than the theoretical values. The reasonable
correlation shows that the theoretical model can reflect the form-
Fig. 24. Hydroformed parts for ˇ = 0.1–0.4.
Fig. 25. Theoretical FLD of a seamed tube and experimental points.
ing limit behavior of a seamed tube. The FLD for a seamed tube
hydroforming can be obtained based on the theoretical model and
calculation method.
5. Conclusions
A new theoretical model is developed, which helps to predict the
FLD for a seamed tube free hydroforming. It is demonstrated that
the theoretical model is feasible and effective for a seamed tube
hydroforming. The predicted FLD matches well with the experi-
mental data, which demonstrates the accuracy and validity of the
theoretical model.
A novel experimental tool set is designed for the right side data
in the FLD, which can be used to control the material flow under
tension–tension strain states by applying the lateral compression
force and the internal pressure. Three type experiments are per-
formed for the FLD of a seamed tube hydroforming. The tool sets
are simple, appropriate for industrial use. Using the novel tool set
and the classical free hydroforming tool set, engineers can obtain
the whole FLD of tube hydroforming.
In the future, the sensitivity studies will be carried out to
evaluate the effect of some factors on the FLD, such as tubular mate-
rial properties, the parameters of the dies, the tubular geometric
parameters and different yield criteria.
Acknowledgements
The authors gratefully acknowledge the financial support
of National High-Tech Research and Development Plan (No.
2007AA04Z145) and Baosteel Group the supply of test materials.
References
Asnafi, N., 1999. Analytical modeling of tube hydroforming. Thin-Walled Struct. 34,
295–330.
Asnafi, N., Skogsgardh, A., 2000. Theoretical and experimental analysis of stroke-
controlled tube hydroforming. Mater. Sci. Eng. A 279, 95–110.
Chu, E., Xu, Y., 2004a. Hydroforming of aluminum extrusion tubes for automotive
applications. Part I: Buckling, wrinkling and bursting analyses of aluminum
tubes. Int. J. Mech. Sci. 46, 263–283.
Chu, E., Xu, Y., 2004b. Hydroforming of aluminum extrusion tubes for automotive
applications. Part II: Process window diagram. Int. J. Mech. Sci. 46, 285–297.
Chu, E., Xu, Y., 2008. Influences of generalized loading parameters on FLD predictions
for aluminum tube hydroforming. J. Mater. Process. Technol. 196, 1–9.
Davies, R., Grant, G., Herling, D., Smith, M., Evert, B., Nykerk, S., Shoup, J., 2000. Forma-
bility investigation of aluminum extrusions under hydroforming conditions. SAE
Technical Paper Series, 2000-01-2675.
Hashemi, R., Assempour, A., Abad, E.M.K., 2009. Implementation of the forming limit
stress diagram to obtain suitable load path in tube hydroforming considering
M–K model. Mater. Des. 47, 1498–1518.
 X. Chen et al. / Journal of Materials Processing Technology 211 (2011) 2012–2021
Hwang, Y.M., Lin, Y.K., 2006. Analysis of tube bulge forming in an open die consid-
ering anisotropic effects of the tubular material. Int. J. Mach. Tools Manuf. 46,
1921–1928.
Hwang, Y.M., Lin, Y.K., Chuang, H.C., 2009. Forming limit diagrams of tubular mate-
rials by bulge tests. J. Mater. Process. Technol. 209, 5024–5034.
Kim, J., Kim, S.W., Song, W.J., Kang, B.S., 2004a. Analytical approach to bursting
in tube hydroforming using diffuse plastic instability. Int. J. Mech. Sci. 46,
1535–1547.
Kim, J., Kim, Y.W., Kang, B.S., Hwang, S.M., 2004b. Finite element analysis for bursting
failure prediction in bulge forming of a seamed tube. Finite Elem. Anal. Des. 40,
953–966.
Kim, S.W., Song, W.J., Kang, B.S., Kim, J., 2009. Bursting failure prediction in tube
hydroforming using FLSD. Int. J. Adv. Manuf. Technol. 41, 311–322.
Kulkarni, A., Biswas, P., Narasimhan, R., Luo, A.A., Mishra, R.K., Stoughton, T.B.,
Sachdev, A.K., 2004. An experimental and numerical study of necking initiation
in aluminium alloy tubes during hydroforming. Int. J. Mech. Sci. 46, 1727–1746.
2021
Nefussi, G., Combescure, A., 2002. Coupled buckling and plastic instability for tube
hydroforming. Int. J. Mech. Sci. 44, 899–914.
Saunders, F.I., Wagoner, R.H., 1996. Forming of tailor welded blanks. Metall. Mater.
Trans. A 27, 2605–2616.
Song, W.J., Heo, S.C., Ku, T.W., Kim, J., Kang, B.S., 2010. Evaluation of effect of flow
stress characteristics of tubular material on forming limit in tube hydroforming
process. Int. J. Mach. Tool. Manuf. 50, 753–764.
Varma, N.S.P., Narasimhan, R., Luo, A.A., Sachdev, A.K., 2007. An analysis of local-
ized necking in aluminium alloy tubes during hydroforming using a continuum
damage model. Int. J. Mech. Sci. 49, 200–209.
Varma, N.S.P., Narasimhan, R., 2008. A numerical study of the effect of loading con-
ditions on tubular hydroforming. J. Mater. Process. Technol. 196, 174–183.
Woo, D.M., Hawkes, P.J., 1968. Determination of stress–strain characteristics of tubu-
lar materials. J. Inst. Met. 96, 357–359.
Xing, H.L., Makinouchi, A., 2001. Numerical analysis and design for tubular hydro-
forming. Int. J. Mech. Sci. 43, 1009–1026.
 4
International Journal of Mechanical Sciences 104 (2015) 112–125

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

Analysis and experiment on wrinkling suppression for hydroforming

of curved surface shell
Yi-Zhe Chen a, Wei Liu a,b,n, Yong-Chao Xu a,b, Shi-Jian Yuan a,b
a
School of Materials Science and Engineering, Harbin Institute of Technology, Harbin 150001, China
b
National Key Laboratory for Precision Hot Processing of Metals, Harbin Institute of Technology, Harbin 150001, China

art ic l e i nf o a b s t r a c t

Article history: Wrinkling on the unsupported area of some deep drawn shells with curved surface and thin thickness is
Received 21 May 2015 extremely hard to control. Sheet hydroforming process is of concern to eliminate this defect, but the
Received in revised form analytical prediction of a proper liquid pressure path is far from perfection. In this article, an analytical
8 October 2015
model on the wrinkling suppression for the hydroforming of curved surface shell was proposed and
Accepted 12 October 2015
achieved by the combination of stress model and geometrical model. An experimental setup and
Available online 19 October 2015
simulation model were designed to verify the accuracy of analytical model. The mechanisms for the
Keywords: wrinkling-suppression effect of the liquid pressure were analyzed by stress analysis. What's more, the
Analytical model effects of different parameters for the required liquid pressure were discussed using multiple linear
Wrinkling suppression
regression analysis. Results show the proper liquid pressure can be used to form wrinkling-free shell.
Hydroforming
Both the experimental results and the stress distribution agree well with the analytical prediction. The
Curved surface shell
most influential parameters for the liquid pressure are the axis length ratio and the surface quality of the
punch. It can be seen that the sheet hydroforming process is very effective for the wrinkling suppression
of deep drawn shells with large area of unsupported area, meantime the analytical model can be used in
engineering applications and reducing simulation time.
& 2015 Elsevier Ltd. All rights reserved.

1. Introduction prediction and prevention of wrinkling during the deep drawing

Curved surface shells formed by conventional deep drawing
(CDD) are used in a wide variety of manufacturing applications
such as aerospace, aviation and automobile industries [1,2]. Nor-
mally, the flange area of a blank is subjected to a tensile stress in
radial direction, and a compressive stress in circumferential dire-
ction during the drawing process [3,4]. Thus, the rupture caused
by very high tensile stress and wrinkling caused by high com-
pressive stress are recognized as the major obstacles in the deep
drawing operation [5]. Many approaches were proposed to meet
the control of rupture, while things changed recently. The pre-
valent deformation instability, wrinkling, appears to be particu-
larly pronounced due to great usage of thin and high strength
sheet metals [6,7]. As a result of this undesirable instability, the
geometrical precision and mechanical properties of formed parts
maybe affected. What's more, it can even result in the damage of
drawing molds and the disruption of production [8]. Therefore the
n
Corresponding author at: School of Materials Science and Engineering, Harbin
Institute of Technology, Harbin 150001, China. Tel.: þ 86 451 86413115;
fax: þ86 451 86415776.
E-mail address: liuw@hit.edu.cn (W. Liu).
are these years main concerns [9–11].
It is easy to understand that there are two kinds of wrinkling,
flange wrinkling and unsupported wrinkling (or side-wall wrink-
ling) in the sheet metal forming due to the compressive hoop
stress. One can control the flange wrinkling conceivably with a
pressure imposed by a blank holder. As a kind of basic sheet metal
forming method, CDD can be used to fabricate curved surfa-
ced shell without wrinkling defect successfully with the ratio of
thickness to diameter larger than 3%. Many researchers perfor-
med some analytical models, experiments and simulation works
[3,12,13]. It is convincible to say that the control of wrinkle in the
flange area is not questionable to handle [14]. While things get
tricky in the unsupported area, because the blank is free of normal
constraint, the compressive stress needed to initiate the wrinkling
is smaller than that in the flange area [15]. As shown in Fig. 1a,
when curved surface shells were deep drawn by CDD, a cir-
cumferential compressive stress on the unsupported area is inev-
itable, which will result in serious wrinkling defect especially
when the ratio of thickness to diameter is very small. Cao and
Wang et al. [9,16,17] presented many results on the prediction of
unsupported wrinkles by the combination of theoretical and
simulation analysis, which agreed well with the experimental
results. Shaffaat et al. [15] made some prediction of wrinkling in

http://dx.doi.org/10.1016/j.ijmecsci.2015.10.002
0020-7403/& 2015 Elsevier Ltd. All rights reserved.
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 113

Nomenclature t blank thickness

φ ratio of t and (2R0)
Pr required liquid pressure f interfacial shear stress
Pw wrinkling-suppression liquid pressure F blank holder force
P rc rupture liquid pressure n strain hardening exponent
a radius of long axis in vertical section K strength coefficient
b radius of short axisin vertical section R anisotropy coefficient
λ ratio of a and b R function of R
ð
x; yÞ  coordinate on the punch surface μ1 friction coefficient of the punch
x ;y coordinate of point A μ2 friction coefficient of the blank holder and the die

h h
a; yh coordinate of point B σr radial stress
x0 the derivative of x σθ circumferential stress
h punch displacement σe equivalent stress
S1 area of Zone I εr radial strain
S2 area of Zone II εθ circumferential strain
ρ radius of Zone II εe equivalent strain
O σ rc rupture tensile stress

h  center of Zone II
Ohx ; Ohy coordinate of Oh σ r ðxh Þ deformation tensile stress at point A
Lh gap between point A and B σ θ ðx h Þ circumferential stress at point A
αh degree of the tangent line at point A σ f ðxh Þ friction radial stress at point A
R0 initial radius of the blank P r =K normalized required liquid pressure
Rh current radius of the blank beta standardized coefficient
R0 initial radius of Zone I and Zone II sig: Partial regression coefficient
R0

Fig. 1. Comparison of CDD and SHP (a) and the zones of the blank in the hydroforming process (b).

conical deep drawn parts using a new deflection function based on
Cao's work, and it turned out the prediction shared same trend
with the experimental results. However, their achievements were
restricted in the CDD process. What's more, it should be pointed
out that not only an accurate prediction but also a great control on
the wrinkling defect is of remarkable concern.
Sheet hydroforming process (SHP) has a growth in popularity
recently because of the profitable advantages of high accuracy and
quality, low cost and increased formability [18–22]. In addition, as
a result of a controllable liquid pressure, this process is therefore
helpful on the controlling of wrinkles, especially for the curved
surface shells with large unsupported area. Many studies have
shown that SHP does prevent the wrinkling instability. Yossifon
et al. [23–25] first proposed many considerable investigations on
the permissible liquid pressure path of cylindrical parts. Lo et al.
[26] studied the wrinkling phenomenon of hemispherical drawn
parts, and found that the working-zone was narrower than that in
cylindrical parts. Abedrabbo et al. [27] researched the wrinkling
behavior during the hydroforming of aluminum alloy hemisph-
erical part and they focused on the effect of different yield function
on the liquid pressure. For deep drawn shells with large unsup-
ported area, the “backward-bulging effect”, which means that the
blank can be bugled backward because of the effect of liquid
pressure, is the key to control the development of wrinkles in SHP,
as shown in Fig. 1a. [28,29].
Hence, there are two main problems need to be solved. Firstly,
the unsupported wrinkling is still extremely troublesome to
compress when the radio of thickness to blank diameter (φ) is too
small, and the geometry of the formed shell is very complex. For
instance, the φ of the propellant storage tank is always lower than
0.5%. Secondly, for curved surface shells with large unsupported
area, the theoretical and experimental works on the wrinkling
suppression by sheet hydroforming are still insufficient. Since the
varying liquid pressure will cause different sheet deformation
behavior, this complex hydroforming process is too complicated to
analyze. Thus it seems impossible to obtain the analytical results
about the appropriate liquid pressure before simulations and
experiments to save costs and time. Therefore, the prediction and
control of this challenging problem are of greater theoretical and
industrial interests.
In this paper, an analytical method was developed to obtain
proper prediction of liquid pressure for the hydroforming of a
curved surface shell by the combination of geometrical and stress
model. The accuracy of this model was verified by experimental
results and simulation works. The mechanisms for the wrinkling-
suppression effect of the liquid pressure were processed by stress
analysis. As final step, to find the influence of various parameters
114 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125

on the liquid pressure, a multiple linear regression analysis was

conducted using statistical method to find the effects of experi-
mental parameters on the liquid pressure.

2. Analytical model

As a result of the controllable liquid pressure in the sheet

hydroforming of curved surface parts with large unsupported area,
a favorable “backward-bulging effect” can be produced, as men-
tioned before in Fig. 1a. It is plausible, therefore, to make an
assertion that the wrinkles in unsupported area can be cleared up
with a “constant dynamic backward-bulging effect” for the formed
parts using sheet hydroforming process. In practical engineering,
ellipsoidal components are widely used as the dome of propellant
storage tanks, pressure vessels, artistic buildings, and end closures
in many industries. Thus, an geometrical model of an ellipsoidal
shell fabricated with aluminum alloy sheet is considered as a case
to analyze in the following sections.
As shown in Fig. 1b, the blank in the ellipsoidal shell defor-
mation process is divided into three regions including Zone I, Zone
II and Zone III. Zone I is the region tightly contacted with punch by
liquid pressure. Zone II is the unsupported region in CDD process,
but it is bulged backward by liquid pressure in the hydroforming
process. Zone III is the flange area below the blank holder. It is
assumed that there is no relative motion between the punch and
the coordinate system, which means that the deep drawing pro-
cess in the analytical model is changed into a relative movement of
blank holder and the die. Meanwhile, it is assumed that the point
A, where the blank touches the punch, shares the same y coordi-
nate value with the flange area all the time during the drawing
process, as shown in the figure. Furthermore, the section arc of
Zone II is assumed as a part of a circle, thus the curvature of this arc
ρ is equal to the radius of this circle [30–32].
It is easy to understand that this assertion extremely simplify
the sheet deformation behavior, meantime, it makes the calcula-
tion of σ r and σ θ possible. The area calculation for different
deformation region plays an important role in obtaining the stress
distribution results [9,15,26,27]. During the sheet hydroforming Fig. 2. Flow chart of the calculation procedure for analyzing the required liquid
process, the uncertain cross-section shape of the blank under pressure.
various liquid pressure makes the theoretical analysis harder. But
things get better in the present model. Firstly the section profile is
assumed as a more regular one in order to simplify the calculation (3) The normal stress caused by the liquid pressure is much
of deformation area. Then, basing on the area changing of the smaller compared with in-plane stress, so that the normal
blank, the strain and the stress distribution of deformation area stress can be ignored in this process [26].
can be calculated according to the stress model. Finally, an (4) Compared with the size of final shape and blank, the gap
expected proper liquid pressure can be generated according to the between the punch and the die is too small therefore it is
neglected.
wrinkling-suppression liquid pressure and rupture liquid pressure.
(5) The blank anisotropy can be depicted by the anisotropy coef-
Thus this process looks more like a reverse-thinking process. Now
ficient R value [24,26].
the calculation process therefore is shown in Fig. 2. The outer
radius Rh at different punch stroke can be obtained through the
difference of original area and the area of Zone I and Zone II. 2.1. Geometrical model
According to the equilibrium equation of flange area for an axi-
symmetric shell system, Rh is a key parameter to calculate σ r and Based on the assumptions about the “backward bugled area”
σ θ . Then the required liquid pressure P r at point B can be obtained. (Zone II), the geometrical model is proposed as below. In the deep
Of course this process will be shown more specifically in the next drawing process, the coordinate ðx; yÞ is satisfied the ellipsoidal
section. All the calculations were processed using the Matlab geometric equation as follows.
numerical computing language. Some assumptions were intro- x2 y2
duced to render this analysis. þ ¼ 1; a 4 b 40 ð1Þ
a2 b2

(1) The thickness and the volume of the blank are constant Thus when the punch moves down with a distance of h, the y
through the whole process [15,17,26,27]. coordinate of point A is yh , it is easy to see that:
(2) The bending and unbending effects are neglected during the
calculation process [23].  b þ h ¼ yh ð2Þ
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 115

Then the x coordinate and the x0 which indicates the derivative

of x, can be expressed by y as follows, respectively.
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a 0 ay
b  y2 ; x0 ¼ f ðyÞ ¼  qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
x ¼ f ðyÞ ¼ ð3Þ
b 2
b b  y2

As mentioned before, S1 and S2 are the area of revolution sur-

faces: Zone Iand Zone II, respectively. It is easy to find the upper
and lower limits of S1 are  b and yh , the expression of S1 is shown
as follows.
Z yh qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
02
S 1 ¼ 2π f ðyÞ 1 þ f ðyÞdy
b
Fig. 4. Stress state for an axisymmetric element in the Zone III.
And the form of S1 after integral is:
8 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9 yh
>
>
> π ay b4 þ a2  b2 y2 þ >
>
>


model and analytical strategy can be used further for any other
>
< 2 >
= geometry shell of sheet hydroforming process as long as the area
b r ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
S1 ¼   ð4Þ
>
>
> p ab π2
ffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
ln y a  b þ b þ a  b y >
2 4

2 2

2 >
>

of the formed parts can be calculated.
>
: a2  b2 >
;
b 2.2. Stress model
According to previous assertion, the section curve in Zone II is a
part of the circle, thus the coordinate ðx; yÞ satisfies: By virtue of the assumption that the thickness remains con-

2 stant, the current flange area radius, Rh , can be obtained as follows.
ðx  Ohx Þ2 þ y  Ohy ¼ ρ2 ð5Þ

 π R20 ¼ S1 þS2 þ π R2h  π a2 ð9Þ
where ρ is the radius of this circle, and Oh Ohx ; Ohy is the
center coordinate. Meanwhile, the degree of the tangent line at meantime
   
point A is defined as αh , as shown in Fig. 3. Therefore the calcu- π R20  R02 ¼ π R2h  a2 ð10Þ
lations of these parameters are shown as follows.

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi then
2
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi a3 yh b  b  y2h rffiffiffiffiffiffiffiffiffiffiffiffiffiffi
a a 2 S1 þS2
Ohx ¼ þ b  yh ; Ohy ¼ yh þ
2 ð6Þ R0 ¼
2 2b 3
2b xh π
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
b xh a  Ohx π R20 þ π a2  S1 S2
sin αh ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi; ρ¼ ; L ¼ a  xh ð7Þ Rh ¼
4 2 sin αh h π
a yh þ b xh
4 2

Of course the upper and lower limits of this zone are Ohy þ ρ Fig. 4 illustrates the stress state of an element in the flange
and yh . Here we can see S2 is: area, as well as the Zone III, the interfacial shear stress, f is
Z Ohy þ ρ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi expressed as follows:
02
S 2 ¼ 4π f ðyÞ 1 þ f ðyÞdy F μ2
yh f¼   ð11Þ
π R2h  a2
Combining Eqs. (5)–(7), the form of S2 after integral is:
8
pffiffiffiffiffiffiffiffiffiffiffiffi ffi
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 9 Ohy þ ρ where F and μ2 mean the blank binder force and Coulomb friction
> a  2a þ 2b
a 2
b  yh 2 >
>
> 4 π a
þ a
b
2
y 2 U >
> coefficient. Combined with the Tresca criterion and the Material
>
> p
2
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b x 2 2b h >
>
>
>
h
>
> behavior model named Hollomon Law
>
>
4 2
a y þb x 4 2 >
>
< 2
h
 pffiffiffiffiffiffiffiffiffi3
h
=
S2 ¼ 3 2 2

ð8Þ σ r  σ θ 6 σ max  σ min ¼ σ e ð12Þ
>
> 4
a yh b  b y
h
5 >
>
>
> y  y þ
p ffiffiffiffiffiffiffiffiffiffiffi >
>
σ e ¼ K ðεe Þn
h
>
>
2b3 xh
b  y2h >
> ð13Þ
> >
>
2
a
 2b
a
> arcsin
> þ4 π 2
y >
: a  O sin αh ;
sin αh
hx
yh
thus the equilibrium equation in the flange can be written:
d t

Here we can see that the deep drawing displacement affects t σ r þ σ r  σ θ þ 2f ¼ 0 ð14Þ
the S1 seriously. Meanwhile with the drawing process going on, dr r
the gap between the punch and the die drops rapidly, which d t
reduces the S2 obviously. It should be noticed that this geometrical t σ r þ K ðεe Þn þ 2f ¼ 0 ð15Þ
dr r
Because of the assumption that no liquid flows into flange area
as a result of the sealing effect, the boundary conditions are the
prescribed pressure at the far edge σ r ¼ 0 when r ¼ Rh , therefore
the radial stress σ r ðr Þ on any point with current radius r is shown:
Z Rh
2f 1
σ r ðr Þ ¼ ðRh r Þ þ UK ðεe Þn dr ð16Þ
t r r
Including the normal anisotropy of the blank in the formula-
tion, the equivalent strain rate is donated as below by employing
the incremental theory of Hill [33]:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1 þR 2R
dεe ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffi dε2r þ dεr dεθ þ dε2θ ð17Þ
Fig. 3. Geometrical relationship of parameters in Zone II. 1 þ 2R 1þR
116 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125

the relationship between equivalent strain and radial strain can be

seen as below:
pffiffiffi pffiffiffi 2ð1 þRÞ R0
εe ¼ Rεr ; R  ; εr ðaÞ ¼ ln ð18Þ
1 þ 2R a
Therefore, the radial stress at the point with radius of r can be
expressed:
Z Rh
2f 1 n
σ r ðrÞ ¼ ðRh  rÞ þ UK ðεr Þn R2 dr ð19Þ
t r r
Here the radial stress at the point of die corner σ r ðaÞ can be
obtained, as shown below:
Z Rh
2f 1 n
σ r ðaÞ ¼ ðRh  aÞ þ U K ½εr ðaÞn R2 dr ð20Þ
t a r
Then using the relationship of σ r  σ θ ¼ σ e , the circumferential
stress at this point σ θ ðaÞ can be shown as follows:
n
σ θ ðaÞ ¼ σ r ðaÞ  K ½εr ðaÞn R2 ð21Þ
Fig. 5. An example of rupture occurs at the point A.
Thus according to the stress distribution of point B, the
wrinkling-suppression liquid pressure P w can be calculated as Meantime, the friction shear stress satisfies the equation:
below: P rc Uxh U μ1
σ f ðxh Þ ¼ ð25Þ
2σ r ðaÞ U t U sin αh t
Pw ¼
Lh It is well known that the rupture defect is commonly defined at
Here the point where diffused necking begins associated with the fall-
n o ing of the load. This condition yields [34]:
RR n
pffiffiffiffiffiffiffiffiffiffiffiffiffiffi
2
2 2ft ðRh  aÞ þ a h 1r UK ½εr ðaÞn R2 dr U t U pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
b xh ffi
a4 y2h þ b4 x2h dσ e 1 þ 2R
Pw ¼ ð22Þ ¼ σe ð26Þ
a  xh dεe 1þR

It is clear that with the drawing process going on, the radial Combined with the Material behavior model named Hollomon
stress keep increasing, meantime the sin αh becomes larger as Law and disregarding any effect of the hoop stress on tensile
well. However, the gap between the punch and the die Lh goes instability, thus the rupture tensile stress can be shown as follows:


down sharply. Hence, it can be predicted that the wrinkling- 1 þR n þ 1 n
suppression liquid pressure P w will increase very quickly, and σ rc ¼ K U pffiffiffiffiffiffiffiffiffiffiffiffiffiffi Un ð27Þ
1 þ 2R
this can be verified in the latter section.
Combine equation (23)–(27), finally the rupture liquid pressure
2.3. Rupture pressure analysis P rc is expressed as below:
 n þ 1 h RR n i
t UK U p1ffiffiffiffiffiffiffiffiffi
þR ffi
1 þ 2R
U nn  t U 2ft ðRh  aÞ þ xhh 1r U K ðϵr Þn R 2 dr
With the deep drawing process going on, the gap between the P rc ¼ ð28Þ
punch and the die gets smaller soon, and this needs a pretty huge xh U μ1
liquid pressure to maintain the “constant dynamic backward- Combine the wrinkling-suppression liquid pressure P w and the
bulging effect”. There is no doubt that the excessive pressure is rupture liquid pressure P rc , the required liquid pressure P r there-
inappropriate and it will lead to rupture. Thus an upper limit of the fore can be obtained using the former part of P w and the latter part
liquid pressure P rc is needed. of P rc . Thus the circumferential stress at the point A can be
Contrary to the CDD process, the point A is more possible to expressed as below:
have a crack defect, as shown in Fig. 5. It is easy to understand this 8 R n n
9
>
< Pw U xth U μ1 þ 2ft ðRh  aÞ þ xRh 1r U K ðεr Þn R2 dr  K ðεr Þn R2 ; P rc 4 P w >
=
difference. As a result of the effect of liquid pressure, the blank
σ θ ðxh Þ ¼ Prc U x U μ 2f h
R Rh 1 n n
is tightly pressed on the punch surface, and therefore the cir- >
: h 1
þ t ðRh  aÞ þ xh r U K ðεr Þ R dr  K ðεr Þ R ; P rc o P w >
n 2 n 2
;
t
cumferential strain increment in Zone I is zero as the drawing
processes. Consequently, a higher tensile stress can be presumably ð29Þ
operative before rupture occurs. Thus the radial stress at this point It is clear that both P w and P rc are affected by many parameters.
is concerned to reach the rupture radial stress. Thus it is not easy to confirm which parameter is more influential
The tensile stress at this point is the largest one in the deep directly from the expression. This point will be discussed in the
drawing process, because not only the tensile stress σ r ðxh Þ caused fifth section.
by symmetric deformation but also the friction radial stress σ f ðxh Þ
in the Zone I caused by liquid pressure can be found here. There-
fore, the tensile stress is composed of deformation radial stress 3. Experimental procedure
and friction shear stress.
σ rc ¼ σ r ðxh Þ þ σ f ðxh Þ ð23Þ 3.1. Mechanical properties of material

where σ rc presents the rupture tensile stress. Here the σ r ðxh Þ 2219 aluminum alloy was prepared in this research to verify
can be expressed as below: the wrinkling suppression effect of the sheet hydroforming pro-
Z cess. The thickness and diameter of this circular blank is 1.484 mm
Rh
2f 1 n
and 330 mm, which means the ratio of the thickness and the
σ r ðxh Þ ¼ ðR  aÞ þ U K ðεr Þn R2 dr ð24Þ
t h xh r diameter in this research, is lower than 0.45%. The blank was in
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 117

Table 1 solution state and the mechanical properties are presented

Mechanical properties of AA2219 in solution state in Table 1.
Mechanical parameters value

Yield stress, σ s (MPa) 125 3.2. Experimental setup and simulation model
Tensile stress, σ b (MPa) 291
Fracture elongation, δ (%) 25 A double action 2000 kN hydroforming press in Harbin Insti-
Strength coefficient, K (MPa) 558
tute of Technology was used to carry out experimental studies.
Strain hardening exponent, n 0.30
Anisotropy coefficient, R0 ; R45 ; R90 ; R 0.47,0.87,0.42,0.66 This machine is coupled with a high pressure intensifier, as shown
in Fig. 6a, to supply the designed liquid pressure in the die. The

Fig. 6. The double action press and high pressure intensifier (a), the experimental molds (b), the experimental control system (c), the molds size (d) and the simulation
model (e).
118 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125
sheet hydroforming mold is composed of four parts: a punch, a
blank holder, a die and a liquid inlet as shown in Fig. 6b. As shown
in Fig. 6c, the experimental control system consists of three sec-
tions: experimental setup, hydraulic drive system and computer
control system. Here we can see apparently that the displacement
and the force of the punch can be obtained through two sensors.
Meantime, the liquid pressure, monitored by the pressure sensor
A, is supported by the intensifier. The precision of the pressure
sensor is 0.1 MPa, and the deep drawing velocity is too small to
cause any pressure fluctuation larger than 0.5 MPa. Moreover, all
of the previous numerical signals obtained by the sensors can be
integrated to the computer system. The punch displacement, the
blank holder force, and the liquid pressure can be controlled
simultaneously to ensure the accuracy of experimental process.
The size of the mold can be seen in Fig. 6d as well. Each experi-
ment was repeated three times to obtain a good repeatability.
In this study, some essential data can be provided by a
numerical simulation for the analysis procedure. The sheet hydr-
oforming process simulation was completed by the finite element
code LS-DYNA3D. As shown in Fig. 6e. The quarter 3-D model was
built in the analysis to take advantage of axisymmetric system and
to reduce the calculation time. The punch, the blank holder, and
the die were modeled as rigid bodies. The blank consisting 20,168
elements with size of 1 mm was modeled as deformable using the
Barlat's anisotropic material. The frictional effect was taken into
account by the Coulomb law. The friction coefficient between the
punch and blank is 0.12 because of the directly rigid contact. As a
result of the use of sealing between the blank holder and the
blank, the die and the blank, the coefficient was decreased to 0.05.
According to the experimental design, the travel distance of the
punch was 80 mm.
3.3. Multiple linear regression analysis
To study the effect of different geometrical and mechanical
parameters on the required liquid pressure, a multiple linear
regression analysis was conducted in this paper. In statistics, linear
regression is an approach for modeling the relationship between a
scalar dependent variable and one or more explanatory variables.
For more than one explanatory variable, the process is called
multiple linear regression. In linear regression, data are modeled
using linear predictor functions, and unknown model parameters
are estimated from the data. Linear regression was the first type of
regression analysis which was studied rigorously, and to be used
extensively in practical applications. In this study, with various
parameters in the analytical model, many required liquid pressure
can be obtained. Then the results were statistically analyzed by the
Statistical Product and Service Solutions (SPSS) statistics software
which can be applied for statistical analysis to find the effects of
independent variables on a dependent variable. Therefore, the
influence of these parameters can be shown easily to make a
better reset about the experimental and simulation design.
4. Verification and discussion
Deal with the mechanical properties of the AA2219 and the
geometrical size of final shape, the wrinkling-suppression liquid
pressure P w and the rupture pressure P rc can be calculated
immediately using the analytical model built previously, as shown
in Fig. 7. It is clear that with the drawing process going on, the
wrinkling-suppression pressure, which is shown as square in the
picture, increases obviously and the incremental diversification
becomes larger. Meantime, the rupture pressure shown as circle
went down since the drawing began. It is easy to understand why
the pressure of P rc is huge at first. The reason is that the Zone I is
Fig. 7. The theoretical results of liquid pressure in SHP and the fitted liquid
pressure path.
extremely small at the first 20% of the punch stroke, which leads to
an inconvincible and exceptionally high pressure value. However,
things change after the former period. As mentioned before, the
tensile stress at point A is consist of friction shear stress σ f ðxh Þ and
symmetric tensile stress σ r ðxh Þ, σ f ðxh Þ becomes larger with the
increasing of S1. However, σ r ðxh Þ gets smaller during the process,
leading to an integrated influence on the rupture pressure. Thus
the rupture pressure seems to be almost constant after the punch
stroke reaches 60%. Basing on the theoretical analysis, the fitted
liquid pressure path therefore can be established in this AA2219
ellipsoid deep drawing part, as shown in Fig. 7. The liquid pressure
path is more possible to be fitted using the Sigmoidal Fit, which is
shown as follows. It is easy to see that the liquid pressure can be
fitted very well.
a
y¼ ð30Þ
1 þ bexpð  kxÞ
Furthermore, this analytical method can be used for some
materials that the stress–strain curve can be depicted with
Material behavior model named Hollomon Law, including alumi-
num alloys, carbon steels and interstitial free steels, etc. Moreover,
even the material hardening behavior does not follow the Hollo-
mon Law, we can still use the combination of other yield criterion
and material flow rule to get the final result.
4.1. 1. Experimental results
The wrinkling suppression effect can be seen in Fig. 8. As
shown in Fig. 8a, there are 6 different liquid pressure paths in
these experiments, and the comparison of formed parts with dif-
ferent liquid pressure is presented. It should be pointed out that
for industrial application, a similar and simpler path (path 6) was
more suitable to prove the wrinkling-suppression effect of the
liquid pressure, as well as saving time and costs. What's more, it
was easy to design the comparative tests (path1–5) when using
the path 6 shown in Fig. 8a now instead of the fitted curve in Fig. 7.
All the path curves have the same characteristic. Therefore, the
wrinkling-suppression effect can be seen more clear and accurate
with simpler design. Note that the path 6 is nearly as same as the
fitted liquid pressure path in Fig. 7. It is clear that the wrinkling
can be weakened with the increasing of the liquid pressure. The
wrinkling numbers under various pressures of 5 MPa, 10 MPa,
15 MPa and 20 MPa are 18, 24, 38 and 42, respectively. Moreover,
the defect almost disappear when the Pr ¼25 MPa, and there is no
wrinkling at all when the P r is equal to 30 MPa in the end. Here we
can see a good agreement between the experimental phenomenon
and the prediction using the proposed analytical model. Still note
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 119

Fig. 8. The experimental path (a) and the formed parts under various liquid pressure (b).

that the optimal path is just the case that the pressure is enough to
prevent wrinkling, which means the part can be formed success-
fully in this way. The lower limit of liquid pressure is too hard to
calculate because of the extremely complex plastic deformation in
the sheet hydroforming.
4.2. Stress analysis
To verify the accuracy of the analytical model, the circumf-
erential stress distribution, which is a key parameter for the
wrinkling suppression, of the formed parts between the path 6 in
the simulation and the calculation results are shown in Fig. 9.
Fig. 9a displays the comparison of circumferential stress of point A
between analytical model and simulation results. It is easy to find
that the results are in excellent agreement with each other. The
stress drops a little at first, and then it increases when the punch
stroke reaches 0.4. Consequently, there is a peak value for cir-
cumferential stress when the punch stroke is 0.6. Note that the
sharp change of analytical stress is a result of the limitation of P rc .
It can be explained by the statement below. At the point where
120 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125
Fig. 9. Comparison of the analytical calculation and simulation results of circumferential stress at point A (a) and point B (b).
Fig. 10. Von Mises stress distribution along the circumferential direction at (a) 20%, 40% of punch stroke and (b) 60%, 80% of punch stroke.
punch stroke reached 0.6, there was almost no “backward bugling
effect” because of the rapidly decreased gap between the punch
and the die. What's more, the liquid pressure also nearly reached
the peak at this point. This implies the peak value of hoop stress.
After this, the drawing process became similar with a traditional
one, resulting in the continuous drop of circumferential stress.
Furthermore, a same comparison is shown in Fig. 9b of the var-
iation of circumferential stress at point B. The simulation results
have a similar change with the calculation values. The largest error
between these two lines, which is nearly 50 MPa, is located at the
middle punch stroke point. It appears that the compressive cir-
cumferential stress at the point B keeps increasing continuously,
which is well accepted in the deep drawing technology. Therefore,
this method can be used directly in the deformation analysis to
predict the experimental process, and saving the calculation time
in numerical simulation as well. Compared with the finite element
method (FEM), the present analytical model has some advantages.
On one hand, FEM can be looked as a kind of “Trial and error
method” based on computer essentially, which is characterized by
repeated, varied attempts that are continued until success. While
the presented analytical model is a positive strategy to find app-
ropriate liquid pressure during the whole drawing process basing
on the stress model and geometrical model of the formed parts.
On the other hand, the appropriate liquid pressure path can be
generated very efficiently once the analytical model was proposed.
However the simulation work will cost a lot of time for calculating,
especially for some large size components. What's more, the
accuracy of FEM is influenced by the built model, but the influ-
ential factors on theoretical model is less than that in FEM.
To find the reason of wrinkling suppression effect caused by the
hydroforming process more specifically, the stress distribution in
the unsupported area both by CDD and SHP with liquid pressure of
30 MPa (path 6) were conducted. The simulation model of CDD
was built according to the hydroforming one except the loading of
liquid pressure.
As shown in Figs. 10–12, each figure is consist of four different
drawing periods including 20%, 40%, 60% and 80% of the punch
stroke (PS). Several conclusions are shown as below: (a) Both the
Von Mises stresses in two kinds of drawing processes rise with the
increasing of punch stroke. Compared with the conventional deep
drawing, the Von Mises stress is increased a lot as a result of the
presence of liquid pressure through this process. What's more, the
Von Mises stress in the hydroforming is almost 100 MPa higher
than that in conventional. It is speculated that this is because the
more serious deformation in SHP, which will lead to a larger work
hardening effect, as shown in Fig. 10. (b) The radial stress in SHP is
nearly 100 MPa larger than that in conventional drawing at the
20% of the punch stroke, whereas there is almost no difference
between the radial stress of the hydroforming and the conven-
tional ones at the 40%, 60% and 80% of the punch stroke, respec-
tively, as shown in Fig. 11. The increasing of radial stress in SHP at
the beginning is mainly caused by the “backward-bugling effect”
deformation, then with the deep drawing process going on, the
unsupported area becomes smaller that results in the little
variation of radial stress between the hydroforming and the con-
ventional one. (c) There is no significant change of circumfer-
ential stress by SHP during all the drawing process. While at the
60% of the punch stroke, the hoop compress stress in the
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125
Fig. 11. Radial stress distribution along the circumferential direction at (a) 20%, 40% of punch stroke and (b) 60%, 80% of punch stroke.
Fig. 12. Circumferential stress distribution along the circumferential direction at (a) 20%, 40% of punch stroke and (b) 60%, 80% of punch stroke.
conventional deep drawing becomes larger and the stress even
reaches  250 MPa at 80% of the punch stroke. It is clear that there
already was serious wrinkling occurred at this period. Here we can
say that the most effective reason for the wrinkling suppression
during the hydroforming is the extraordinary decreasing of com-
pressive circumferential stress and the significantly improvement
of Von Mises stress. These results eliminate the chance of wrink-
ling and enhance the work hardening effect of the blank.
5. Effect of parameters on liquid pressure
Since the accuracy of analytical model has been proved by the
experiment and simulation work, it is reasonable to make a better
comprehension about the effect of different geometric and material
parameters on the sheet hydroforming process. Therefore, the requ-
ired liquid pressure was calculated by the composed method using
Matlab with various input values. Here there are three kinds of
input values we checked, the first part is geometric parameters
including the axis length ratio λ, the ratio between thickness and
diameter φ, and the second part is the process parameters including
the friction coefficient μ1 , μ2 , and the blank holder force F. The third
part is the mechanical properties consisting n-value, R-value.
5.1. Effect of parameters on the normalized liquid pressure
Fig. 13a is the calculation results of normalized liquid pressure
P r =K under different axis length ratios λ. It is clear that λ strongly
121
affects the normalized pressure in the former half of the punch
stroke. P r is larger when the λ is smaller. It can be explained that
the outer radius of the blank rises with the improvement of λ, thus
the required tensile deformation is relatively becoming lower,
which can result in a decrease of the liquid pressure. In Fig. 14b,
the ratio of thickness and diameter φ is seen to have an obvious
effect on the normalized liquid pressure: not only the wrinkling-
suppression liquid pressure P w but also the rupture liquid pressure
P rc improve significantly during this process. This is easy to
understand that the required tensile stress which makes a same
level deformation will go up if the thickness is thicker, and it looks
like φ affects the rupture liquid pressure P rc more than the
wrinkling-suppression liquid pressure P w .
Fig. 14a illustrates the dependence of normalized liquid pres-
sure on the n-value. When the K stays constant but n varies a lot, a
higher n means a lower true stress before the true stress gets to K.
That is why an increase in n strongly reduces the wrinkling-
suppression liquid pressure P w . However, the rupture liquid
pressure increases with n, which means that the material with a
higher n can withstand greater pressure before rupture occurs.
From Fig. 15b, it can be seen that R-value has limited influence on
the normalized liquid pressure. This can be explained from the
following points. Before the liquid pressure was normalized, the
difference of liquid pressure between different R-values is about
0.5 MPa, but the effect is minished a lot after normalizing. What's
more, under the current stress model and geometrical model, the
R-value plays more important role in the homogeneous deforma-
tion process, but not in the calculation of wrinkling-suppression liquid
122 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125
Fig. 13. Comparison of normalized liquid pressure with different parameters (a) axis length ratio and (b) ratio of thickness and diameter.
Fig. 14. Comparison of normalized liquid pressure with different parameters (a) strain hardening exponent n and (b) anisotropy coefficient R.
pressure and rupture liquid pressure. Thus the effect of R-value is
indeed not as obvious as other parameters [12].
Fig. 15 corresponds to the required liquid pressure with various
friction coefficient μ1 , μ2 and blank holder force F, respectively. It
is observed that μ1 just has an impact on the rupture liquid
pressure P rc but no effect on the wrinkling-suppression liquid
pressure P w . While the variation of μ2 changes not only P rc but also
P w as shown in Fig. 15b. The increasing of μ2 rise the wrinkling-
suppression liquid pressure, whereas it reduces the rupture pres-
sure. It is easy to understand that a bigger μ2 indicates a larger
friction force in the flange area which will result in a large
wrinkling-suppression pressure. But the tensile radial stress at the
point A increases with the μ2 , and this leads to a drop of rupture
liquid pressure. A same but lower influence of blank holder force
can be found in Fig. 15c, it is clear that the blank holder force F
influences the P rc and P w simultaneously, but its effect is limited.
5.2. Multiple linear regression analysis
As shown in Table 2, the normalized liquid pressure at 20%,
40%, 60% and 80% of the punch stroke, defined as P r0:2 =K, P r0:4 =K,
P r0:6 =K and P r0:8 =K, respectively, with various parameters can be
seen clearly. Thus there are 4 calculation models in this table. It is
understandable to see that some variables may have a kind of
linear relationship with another one, or the presence of a para-
meter would make the other existing variable non-significant,
which is known as the multicollinearity. To overcome any poss-
ible multicollinearity problem among the parameters, the Step-
wise analysis method was used when doing the multiple linear
regression analysis in SPSS. Using the Stepwise method, the non-
significant parameter, which was significant before another vari-
able included into the model, will be removed to increase the
accuracy of regression model. Meantime, the effect of parameters
to the target parameter will be evaluated independently in this
way. Then all the parameters included in the analysis are calcu-
lated into the model step by step until no variables in the equation
can be removed. It is clear from Table 2 that all of these models are
meaningful from the F Test results. Meantime, all the R Squares of
the models are larger than 0.895, which means that almost all of
the variation in the normalized liquid pressure can be explained
by the change of parameters. The Standardized coefficient Beta
and Partial Regression coefficient Sig., which represent the influ-
ence level of each variable to the dependent variable and whether
there is obvious correlation between variables of the dependent
variable, respectively, were calculated and shown in Table 3.
In order to compare the results more directly, the calculation
results are shown in Fig. 16. Note that we use the absolute value of
all the results to get a better comparison effect. From Fig. 16 we
can see that in the former stage of the drawing including 20%, 40%
and 60% of the punch stroke, the axis length ratio λ plays the most
influential role for the liquid pressure changing, then followed by
the friction coefficient μ2 , strain hardening exponent n, the ratio
between thickness and diameter φ, the blank holder force F. Note
that the Beta values of λ and n are below zero, meaning that these
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 123

Fig. 15. Comparison of normalized liquid pressure with different parameters (a) friction coefficient μ1 , (b) friction coefficient μ2 and (c) blank holder force.

Table 2
Normalized liquid pressure with different parameters and punch stroke.

n φ μ1 μ2 λ R F P r0:2 =K P r0:4 =K P r0:6 =K P r0:8 =K

0.2 0.0025 0.12 0.05 1.5 1.0 500,000 0.00253 0.007543 0.02239 0.03765
0.3 0.0025 0.12 0.05 1.5 1.0 500,000 0.00202 0.006041 0.01863 0.03945
0.4 0.0025 0.12 0.05 1.5 1.0 500,000 0.00163 0.004885 0.0156 0.04074
0.3 0.0020 0.12 0.05 1.5 1.0 500,000 0.00169 0.005039 0.01554 0.05932
0.3 0.0030 0.12 0.05 1.5 1.0 500,000 0. 00,236 0. 007043 0. 02,173 0. 02904
0.3 0.0025 0.08 0.05 1.5 1.0 500,000 0. 00202 0.006041 0.01863 0.03128
0.3 0.0025 0.16 0.05 1.5 1.0 500,000 0.00202 0.006041 0.01863 0.04470
0.3 0.0025 0.12 0.05 1.4 1.0 500,000 0.00288 0.008539 0.02653 0.03944
0.3 0.0025 0.12 0.05 1.6 1.0 500,000 0.00137 0.004142 0.01252 0.03939
0.3 0.0025 0.12 0.05 1.5 0.5 500,000 0.00205 0.00613 0.01891 0.03931
0.3 0.0025 0.12 0.05 1.5 1.5 500,000 0.00200 0.005993 0.01843 0.03952
0.3 0.0025 0.12 0.05 1.5 1.0 250,000 0.00185 0.00553 0.01705 0.03931
0.3 0.0025 0.12 0.05 1.5 1.0 750,000 0. 00219 0. 00656 0.02021 0.03952
0.3 0.0025 0.12 0.10 1.5 1.0 500,000 0.00237 0.00707 0.02178 0.03805
0.3 0.0025 0.12 0.15 1.5 1.0 500,000 0.00271 0.00810 0.02493 0.03652
ANOVA and Model summary F Test 195.524 206.674 227.344 51.131
R2 0.991 0.991 0.992 0.895

two variables are negatively correlated with the liquid pressure.
Meantime, the anisotropy coefficient R-value has a negatively
limited effect on the liquid pressure as well. The relationship
between the wrinkling-suppression liquid pressure and another
friction coefficient μ1 is also not obvious. Then, it is observed that
μ1 and φ have an great effect on the normalized liquid pressure of
P r0:8 =K. This is because the surface quality of the punch really
matters the radial tensile stress the most. To design an appropriate
experiment and simulation for the ellipsoidal parts hydroforming
process, the axis length ratio and the friction coefficient of the
punch are of concern most for the wrinkling-suppression liquid
pressure and rupture liquid pressure, respectively.
6. Conclusion
In this article, an analytical model on wrinkling suppression
of a curved surface shell formed by sheet hydroforming was
124 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125

Table 3 Acknowledgments
Standardized coefficient Beta and Partial Regression coefficient Sig. in the four
models.
The authors would like to thank the financial support from
P r0:2 /K P r0:4 /K P r0:6 /K P r0:8 /K the National Natural Science Foundation of China (Nos. 50905041
and 51375114) and the support from the Program for Changjiang
Beta Sig. Beta Sig. Beta Sig. Beta Sig.
Scholars and Innovative Research Team in University (No. IRT
λ  0.700 0.000  0.692 0.000  0.724 0.000  0.001 0.989 1229). Special thanks to Mr. Zhi-Biao Wang for his help on the
μ2 0.454 0.000 0.460 0.000 0.465 0.000  0.146 0.124 statistic analysis.
n  0.417 0.000  0.418 0.000  0.351 0.000 0.088 0.368
φ 0.311 0.000 0.315 0.000 0.320 0.000 0.383 0.001
F 0.158 0.001 0.161 0.001 0.163 0.000 0.006 0.952
μ1 0.000 1.000 0.000 1.000 0.000 1.000  0.856 0.000
R  0.023 0.499  0.022 0.519  0.025 0.434 0.006 0.952 References

[1] Cwiekala T, Brosius A, Tekkaya AE. Accurate deep drawing simulation by

Fig. 16. Standardized coefficient Beta in the four models with different parameters.
proposed and achieved. The accuracy was verified by the experi-
mental results and simulation works. The mechanisms for the
wrinkling-suppression effect of the liquid pressure were analyzed.
The influence of various parameters on the normalized liquid
pressure was also given. Conclusions can be drawn as follows:
(1) The theoretical analysis method of the unsupported area for
hydroforming of curved surface shell was proposed by the
combination of geometrical model and stress model.
(2) The wrinkling-suppression liquid pressure and rupture liquid
pressure of the ellipsoidal shell were calculated analytically. It
can be used for the prediction of liquid pressure for well-
formed parts with no wrinkling and rupture.
(3) The experiment and simulation works of AA2219 hydroform-
ing parts proved that the proper liquid pressure can be used
for manufacturing wrinkling-free part. Both of the formed
parts and stress distribution agree well with the analytical
prediction. Furthermore, the most effective reason for the
wrinkling suppression in this case is the extraordinary decre-
asing of compressive circumferential stress and the signifi-
cantly improvement of Von Mises stress.
(4) The most influential parameter for the wrinkling-suppression
liquid pressure is the axis length ratio, however, the surface
quality of the punch plays the most important role in calcu-
lating the rupture pressure.
The summarized conclusions reveal that the sheet hydro-
forming process is very effective for the wrinkling-suppression of
deep drawn shells with large area of unsupported area, meanwhile
the analytical model can be used in engineering applications and
reducing simulation time.
combining analytical approaches. Int J Mech Sci 2011;53(5):374–86.
[2] Halkaci HS, Turkoz M, Dilmec M. Enhancing formability in hydromechanical
deep drawing process adding a shallow drawbead to the blank holder. J Mater
Process Technol 2014;214(8):1638–46.
[3] Yu TX, Johnson W. The buckling of annular plates in relation to the deep-
drawing process. Int J Mech Sci 1982;24(3):175–88.
[4] Taghipour E, Assempour A. The effects of proportional loading, plane stress,
and constant thickness assumptions on hydro-mechanical deep drawing
process. Int J Mech Sci 2011;53(5):329–37.
[5] Lee MG, Kim C, Pavlina EJ, Brrlat F. Advances in sheet forming—materials
modeling, numerical simulation, and press technologies. J Manu Sci Eng
2011;133 061001-12.
[6] Jha AK, Murty S, Sreekumar K, Sinha PP. High strain rate deformation and
cracking of AA 2219 aluminium alloy welded propellant tank. Eng Fail Anal
2009;16(7):2209–16.
[7] Wang XF, Yang QS, Law S. Wrinkled membrane element based on the
wrinkling potential. Int J Solids Struct 2014;51(21):3532–48.
[8] Zhang SH, Lang LH, Kang DC, Danckert J, Nielsen KB. Hydromechanical deep-
drawing of aluminum parabolic workpieces—experiments and numerical
simulation. Int J Mach Tool Manuf 2000;40(10):1479–92.
[9] Wang X, Cao J. On the prediction of side-wall wrinkling in sheet metal forming
processes. Int J Mech Sci 2000;42(12):2369–94.
[10] Kawka M, Olejnik L, Rosochowski A, Sunaga H, Makinouchi A. Simulation of
wrinkling in sheet metal forming. J Mater Process Technol 2001;109(3):283–9.
[11] Correia JP, Ferron G, Moreira LP. Analytical and numerical investigation of
wrinkling for deep-drawn anisotropic metal sheets. Int J Mech Sci
2003;45:1167–80.
[12] Wang X, Cao J. An analytical prediction of flange wrinkling in sheet metal
forming. J Manuf Process 2000;2(2):100–7.
[13] Agrawal A, Reddy NV, Dixit PM. Determination of optimum process para-
meters for wrinkle free products in deep drawing process. J Mater Process
Technol 2007;191(1–3):51–4.
[14] Senior BW. Flange wrinkling in deep-drawing operations. J Mech Phys Solid
1956;4:235–46.
[15] Shafaat MA, Abbasi M, Ketabchi M. Investigation into wall wrinkling in deep
drawing process of conical cups. J Mater Process Technol 2011;211(11):1783–
95.
[16] Cao J, Boyce MC. Wrinkling behavior of rectangular plates under lateral con-
straint. Int J Solids Struct 1997;34(2):153–76.
[17] Cao J. Prediction of plastic wrinkling using the energy method. J Appl Mech
1999;66(3):646–52.
[18] Nakamura K, Nakagawa T. Sheet metal forming with hydraulic counter pres-
sure in Japan. Ann CIRP 1987;36(1):191–4.
[19] Zhang SH, Danckert J. Development of hydro-mechanical deep drawing. J
Mater Process Technol 1998;83(1):14–25.
[20] Yuan SJ, Han C, Wang XS. Hydroforming of automotive structural components
with rectangular-sections. Int J Mach Tool Manuf 2006;46:1201–6.
[21] Chen YZ, Liu W, Yuan SJ. Strength and Formability Improvement of Al-Cu-Mn
Aluminum Alloy Complex Parts by Thermomechanical Treatment with Sheet
Hydroforming. JOM 2015;67(5):938–47.
[22] Lang L, Danckert J, Nielsen KB. Study on hydromechanical deep drawing with
uniform pressure onto the blank. Int J Mach Tool Manuf 2004;44(5):495–502.
[23] Yossifon S, Tirosh J, Kochavi E. On suppression of plastic buckling in hydro-
forming processes. Int J Mech Sci 1984;26(6):389–402.
[24] Yossifon S, Tirosh J. Buckling prevention by lateral fluid pressure in deep-
drawing. Int J Mech Sci 1985;27(3):177–85.
[25] Yossifon S, Tirosh J. On the permissible fluid-pressure path in hydroforming
deep drawing processes—analysis of failures and experiments. J Manuf Sci Eng
1988;110(2):146–52.
[26] Lo S, Hsu TC. Wilson WRD. An analysis of the hemispherical-punch hydro-
forming processes. J Mater Process Technol 1993;37(1):225–39.
[27] Abedrabbo N, Zampaloni MA, Pourboghrat F. Wrinkling control in aluminum
sheet hydroforming. Int J Mech Sci 2005;47(3):333–58.
[28] Liu W, Xu Y, Yuan S. Effect of pre-bulging on wrinkling of curved surface part
by hydromechanical deep drawing. Procedia Eng 2014;81:914–20.
[29] Xu Y, Li F, Liu X, Yuan S. Effects of pre-bulging on 2024 aluminum alloy
complex-shaped components. JOM 2011;63(5):39–41.
 Y.-Z. Chen et al. / International Journal of Mechanical Sciences 104 (2015) 112–125 125

[30] Teng BG, Yuan SJ, Wang ZR. Experiment and numerical simulation of hydro- [33] Hill R. The mathematical theory of plasticity. Oxford: Clarendon Press; 1950.
forming toroidal shells with different initial structure. Int J Press Vessels Pip [34] Yossifon S, Tirosh J. Rupture instability in hydroforming deep-drawing pro-
2001;78(1):31–4. cess. Int J Mech Sci 1985;27(9):559–70.
[31] Yuan S, Teng B, Wang ZR. A new hydroforming process for large elbow pipes
[J]. J Mater Process Technol 2001;117(1):28–31.
[32] Teng BG, Yuan SJ, Wang ZR. Effect of the initial structure on the hydro-forming
of toroidal shells. J Mater Process Technol 2002;123(1):18–21.
 6
International Journal of Mechanical Sciences 107 (2016) 180–187

Contents lists available at ScienceDirect

International Journal of Mechanical Sciences

journal homepage: www.elsevier.com/locate/ijmecsci

Analysis of shape variation during hydro-forming of ellipsoidal shells

with double generating lines
S.J. Yuan a,n, W.W. Zhang a,b
a
National Key Laboratory of Precision Hot Processing of Metals, Harbin Institute of Technology, Harbin 150001, China
b
Institute of Electronic Engineering, China Academy of Engineering Physics, Mianyang 621900, China

art ic l e i nf o a b s t r a c t

Article history: In order to obtain an ideal ellipsoidal shell by the controlling of inputting pressured water volume, the
Received 16 September 2015 curvature and volume variation have been analyzed for hydro-forming of ellipsoidal shell with double
Received in revised form generating lines. The geometric criteria of double generating lines changing into single generating line
26 December 2015
was derived, and the mathematical model between shell volume and axis length ratio was developed. It
Accepted 5 January 2016
is shown that the volume variation is only associated with the initial and final axial length ratios, and can
Available online 11 January 2016
be theoretically predicted. An experiment was conducted on hydro-forming of the ellipsoidal shells with
Keywords: the axis length ratio 1.5, 1.7 and 2.2, respectively, and the effect of internal pressure was discussed on the
Ellipsoidal shell deformation of the shell shape, curvature and volume variation. It is experimentally shown that the short
Spherical shell
axis is linearly increased, the long axis is slightly changed and the shell volume is exponentially
Hydro-forming
increased. It is proved that the theoretical prediction of the volume variation has a good agreement with
Hydro-bulging
Volume variation the experimental results, which indicates that the needed ideal ellipsoidal shell can be successfully
formed by controlled of inputting pressured water volume.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction wrinkling stress. In order to avoid the latitudinal compressive

Ellipsoidal shells have been usually used as the water tanks and
bottom of fuel tanks in the rockets due to the advantages of the
small wind area, lower centroid and space saving. Since the die-
less hydro-forming technology of closed shells was introduced [1],
it was widely used to manufacture ellipsoidal shell with various
sizes and materials. The experiments were early dealt with for the
hydro-forming of ellipsoidal shells with axis length ratios 1.88,
1.67 and 1.25, and the influence of the axis length ratios, lateral
petal numbers and materials were deeply discussed on the plastic
deformation and wrinkling behavior [2]. It is experimentally
proved that the wrinkling occurs in the equatorial plane for the
ellipsoidal shells with axis length ratio 1.88 and 1.67 respectively,
but it is well formed for the ellipsoidal shell with axis length
ratio 1.25.
More comprehensive theoretical analysis was introduced on
the hydro-forming of p ellipsoidal
ffiffiffi shells [3,4]. It is pointed out that
the axis lengthpratio
ffiffiffi 2 is a critical value. When the axis length
ratio exceeds 2,there is a latitudinal compressive stress in the
equatorial plane, which will lead to wrinkling occurrence if the
value of latitudinal compressive stress is beyond the critical
n
Corresponding author.
E-mail address: syuan@hit.edu.cn (S.J. Yuan).
stress, an ellipsoidal shell with double generating lines was
designed as the pre-form shell [5]. In this design, the ellipsoidal
shell segment which has strong possibility of wrinkling occurrence
can be replaced by another
pffiffiffi ellipsoidal shell segment with axis
length ratio less than 2. Experiments were subsequently carried
out for hydro-forming of ellipsoidal shells with double generating
lines, and the variation of stress state and shell shape were ana-
lyzed. It is shown that a sound ellipsoidal is obtained, and the
wrinkling is successfully avoided. Similarly, this idea can be also
used to resolve the problem that the polar plate is not deformed
during hydro-form of prolate ellipsoidal shell [6].
Stress analysis on the ellipsoidal shell with double generating
lines subjected to internal pressure was theoretically carried out,
and the initial yield pressure was derived [7]. Stress locus of
typical points was experimentally illustrated, and the expanding of
plastic deformation was analyzed. It is shown that the plastic
deformation first occurs on the pole and then expands towards to
equatorial line, and the sheet materials are deformed prior to the
weld seams along latitudinal direction. The initial yield pressure of
the shell obtained by experiment is well agreed with the theore-
tical prediction.
In order to reverse an optimum pre-form structure according to
needed ellipsoidal shell, a mathematical response model between
axis lengths and structural parameters was developed on the basis
of the Box–Behnken design method [8]. For the needed ellipsoidal

http://dx.doi.org/10.1016/j.ijmecsci.2016.01.007
0020-7403/& 2016 Elsevier Ltd. All rights reserved.
 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187 181

Nomenclature α0 Dividing globe angle

a1 Long axis of first ellipsoidal shell
r First principal curvature radius a2 Long axis of second ellipsoidal shell
R Second principal curvature radius b1 Short axis of first ellipsoidal shell
ϕ Rotating angle b2 Short axis of second ellipsoidal shell
a Long axis λ1 Axis length ratio of first ellipsoidal shell
b Short axis λ2 Axis length ratio of second ellipsoidal shell
λ Axis length ratio λF Axis length ratio of final ellipsoidal shell
p Internal pressure ϕ1 Rotating angle of first ellipsoidal shell
t Thickness ϕ2 Rotating angle of second ellipsoidal shell
σφ Longitudinal stress V Final volume
σθ Latitudinal stress V0 Initial volume
α Theoretical dividing globe angle ðx0 ; y0 Þ Coordinate on dividing globe angle

shell with final axis length ratio λ ¼ 1:5, the optimum initial axis where the latitudinal stress varies from tensile state to compres-
length ratio of pre-form shell is λ ¼ 2:0  2:2. By comparison with sive state. In the previous research [5], the relationship between
the experimental results, it is indicated that the predicted axis theoretical globe dividing angle α and axis length ratio λ has been
lengths calculated by reverse design are in good agreement with given as follows:
the experimental data, and the deviation is no more than 1%. pffiffiffiffiffiffiffiffiffiffiffiffiffi
However the forming pressure during optimization deeply depend λ2 2
α ¼ arc tan ð3Þ
on the materials and axis length ratio, the optimum pre-form shell 2
λ
is not generally applied to any situation.
Fig. 2 shows the variation of theoretical globe dividing angle α
The materials of ellipsoidal shells widely used in engineering
with the increase of initial axis length ratio λ. There is a peak value
manufacture are always stainless steel, low carbon steel and low
of theoretical globe dividing angle α.
alloy steel, and the maximum diameter is up to 3 m. For the
Derivative operation on Eq. (3), it is expressed as follows:
ellipsoidal shells with various materials and initial axis length
ratios, the forming pressure and axis lengths variation are still λð4  λ2 Þ
different. In order to obtain the needed ellipsoidal shell, it is α0 ¼ f 0 ðλÞ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffi ð4Þ
ðλ þ λ 2Þ λ  2
4 2 2
essential to evaluate the forming pressure, constantly record the
axis lengths and roundness during hydro-forming. Hence, it is a For α0 ¼ 0, it is shown that when the initial axis length ratio is
time consuming process. λ ¼ 2:0, the maximum value of theoretical globe dividing globe
To obtain the needed ideal ellipsoidal shell with single gen- angle is αmax ¼ 19:51.
erating line and realize shape controlling by inputting pressured Based on the mechanical analysis mentioned above, it is the-
water volume, a mathematical model between volume flow and oretically proved that there is a strong possibility of wrinkling
axis length ratio was developed by theoretical analysis. An occurrence between dividing line and equatorial line. Fig. 3 shows
experiment was carried out and the effect of internal pressure was the geometric structure of the ellipsoidal shell with double gen-
discussed on shell shape, axis lengths, curvature and volume erating lines, which is composed of two ellipsoidal shells with
variation. Finally, the sound ellipsoidal shell is obtained. different axis lengthpratios.
ffiffiffi The axis length ratio of the first ellip-
soidal shell
pffiffiffi is λ 1 4 2 , and that of the second ellipsoidal shell is
λ2 o 2. The initial axis length ratio of the ellipsoidal shell with
2. Theoretical analysis double generating lines is defined as λ ¼ a2 =b1 , where a2 is the
length of the long axis and b1 is the length of the short axis. In
2.1. Design of ellipsoidal shell with double generating lines order to obtain double tensile stress state for any position of
ellipsoidal shell, the dividing globe angle is α0 4 αmax .
Fig. 1 shows the geometry of an ellipsoidal shell subjected to
the internal pressure, where r is the first principal curvature
radius, R is the second principal curvature radius, ϕ is rotating
angle between revolution axis y and second principal curvature
radius R, a is radius of the long axis and b is radius of the short
axis. The axis length ratio is defined as λ ¼ a=b. The general geo-
metric relationship is expressed in Eq. (1).
1
r ¼ λak R ¼ λak
3
k ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1Þ
ðλ  1Þ sin ϕ þ 1
2 2

When the ellipsoidal shell is subjected to the internal pressure,

according to Laplace equation, the longitudinal stress σ φ and the
latitudinal stress σ θ can be expressed as follows [3]:
 
p pR R
σ φ ¼ R; σ θ ¼ 2 ð2Þ
2t 2t r
where, p is the internal pressure; t is the thickness. pffiffiffi
For the ellipsoidal shell with initial axis length ratio λ 4 2,
there is a theoretical globe dividing angle α on the ellipsoidal shell, Fig. 1. Ellipsoidal shell revolving along the y-axis.
182 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187

For the ellipsoidal shell shown in Fig. 3a, the coordinate ðx0 ; y0 Þ a1 and a2 can be derived as follows:
on dividing globe angle is satisfied with the geometric equation as vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u
a1 u 1 þ λ1 tan2 α0
2
¼t
follows:
ð6Þ
1 þ λ tan2 α0
a2 2
2
x20 λ1 y20 x2 λ y2
2 2
þ 2 ¼ 1 and 02 þ 2 2 0 ¼ 1 ð5Þ Based on the Eq. (6), it can be seen that the long axes for the
a21 a1 a2 a2
two shell segments are not independent. When the initial axis
length ratio and dividing globe angle are constant, the Eq. (6) is
The coordinate ðx0 ; y0 Þ on dividing globe angle is satisfied
also constant.
tanα0 ¼ y0 =x0 . Based on Eqs. (4) and (5), the relationship between

2.2. Geometric criteria of double generating lines changing into

single generating line

Because the needed shell shape is an ideal ellipsoid with single

generating line, it is very essential to analyze the curvature var-
iation and give the critical internal pressure.
Fig. 4 shows the variation of curvature radius on the dividing
globe angle. With the increasing of internal pressure, the curvature
radius on the both shell segments will be close to each other, and
the double generating lines are gradually changing into single
generating lines.
According to Eq. (1), the curvature radius for both of the shell
segments on dividing globe angle can be expressed as Eq. (7):
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

R1 ¼ λ1 a1 = λ21  1 sin2 ϕ1 þ 1
r
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi

R2 ¼ λ2 a2 = λ22  1 sin2 ϕ2 þ 1 ð7Þ

Fig. 2. Variation of theoretical dividing globe angle with the increasing of axis On the dividing globe angle ðx0 ; y0 Þ, sin ϕ1 ¼ x0 =R1 and
length ratio. sin ϕ2 ¼ x0 =R2 , hence the relationship between curvature radius

Fig. 3. Design of the ellipsoidal shell with double generating lines: (a) 2D schematic view [5]; (b) 3D schematic view.
 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187 183

Fig. 4. Curvature radius variation on dividing globe angle during hydro-forming of ellipsoidal shell with double generating lines: (a) Early stage; (b) Middle stage;
(c) Last stage.

and axis lengths can be written as follows.

a41 a21 x20

R21 ¼ 2
 2
þ x20
b1 b1
a42 a22 x20
R22 ¼ 2
 2
þ x20 ð8Þ
b2 b2

During the hydro-forming of ellipsoidal shell with double

generating lines, the coordinate ðx0 ; y0 Þ long axis a2 and short axis
b1 are easy to recorded with the increasing of internal pressure.
Hence the curvature radius can be transformed as follows
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u 4 2
ux y þ x2 ðb2  y2 Þ2
R1 ¼ t 0 0 2 0 1 2 0
ðb1  y20 Þ
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ða22  x20 Þ2 þ x20 y20 Fig. 5. Schematic view of angles on the dividing line of the ellipsoidal shell.
R2 ¼ ð9Þ
y20
can be written as follows.
Next step, the geometric criteria of double generating lines
changing into single generating line can be written as follows: 4π a32
V¼ ð13Þ
R1 ¼ R2 . 3λF
where, λF is the final axis length ratio.
2.3. Mathematical model of volume variation According to Eqs. (12) and (13), the volume variation can be
expressed as follows.
In Fig. 5, the coordinate ðx0 ; y0 Þ on dividing globe angle is
satisfied with tan α0 ¼ y0 =x0 , sin ϕ1 ¼ x0 =λ1 a1 k1 and sin ϕ2 ¼ V V0 1
V0 ¼ ¼  rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 3=2 rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi  1
x0 =λ2 a2 k2 , hence the relationship between rotating angle ϕ and V0 λF λ41 1 þ λ1 tan2 α0
2
λF λ42
λ 1  þ λ
dividing globe angle α0 can be derived as follows: cot α0 þ λ1 1 þ λ2 tan α0 cot α0 þ λ2
2 4 2 2 2 4
1 2

ð14Þ
λ41 λ42
cos 2 ϕ1 ¼ and cos 2 ϕ2 ¼ ð10Þ Based on the theoretical analysis, it can be seen that the
cot2 α0 þ λ1 cot2 α0 þ λ2
4 4
volume variation depends on the initial axis length ratio and final
Suppose that the pre-form shell is an ideal ellipsoidal shell with axis length ratio, and does not depend on the materials and dia-
double generating lines, the volume before hydro-forming can be meter to thickness ratio. Hence the shell shape can be controlled
derived by triple integral and expressed as follows by inputting pressured water volume.

4π a31 4π a32
V0 ¼ ð1  cosϕÞ þ cosϕ2 ð11Þ
3λ 1 3λ2
3. Research procedure
According to Eq. (10) and Eq. (11), the volume of ellipsoidal
shell before hydro-forming can be re-written as follows. 3.1. Structure of pre-form shell
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u u
4π a31 u λ 4
4π a3u
λ42 Table 1 shows the main structural parameters of the ellipsoidal
V0 ¼ ð1  t 1
Þþ 2t
ð12Þ
3λ 1 cot2 α0 þ λ1
4 3λ2 cot2 α0 þ λ4 shells with double generating lines. Fig. 6 shows the pre-form
2
shell, which is composed of 12 pieces of lateral petal and 2 pieces
When the hydro-forming process is finished, an ideal ellipsoi- of polar plate. The major diameter of the ellipsoidal shells are all
dal shell with single generating line is obtained, and the volume 440 mm, and the diameter of polar plate is 100 mm.Three
184 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187

Table 1 4. Experimental results and discussion

Structural parameters of the ellipsoidal shells with double generating lines.
4.1. Discussion on shell shape and axis lengths
Shell a1 =mm b1 =mm λ1 a2 =mm b2 =mm λ2 λ α0 =1

Shell-1 230 143 1.6 220 157 1.4 1.5 24 Fig. 8 shows the shape variation during hydro-forming of
Shell-2 240 132 1.8 220 157 1.4 1.7 24 ellipsoidal shell with double generating lines. There is no wrink-
Shell-3 300 100 3.0 220 157 1.4 2.2 24 ling occurrence during deformation process for all the three shells.
At the early stage of deformation (p r 1:5MPa), plastic deformation
occurs on the polar plate and the short axis is elongated, and it
quickly reaches to the initial designed dimension. Meanwhile, the
dividing globe angle is obviously spread out, and the curvature
radius of the first shell segment is significantly varied. At the
middle stage of deformation ð1:5MPa o p r 3:0MPaÞ, the second
shell segment is to deform, and the curvature radius is slightly
varied. The lateral petals are deformed towards to a smooth arc,
and the double generating lines are gradually changing into single
generating lines with increasing of internal pressure. At the last
stage of deformation ð3:0MPa o p r 4:5MPaÞ, plastic deformation
is expanded to the equatorial plane, the cross-sectional shape of
the equatorial plane is changed from polygon to a perfect circle
gradually.
Fig. 9 shows the variation of axis lengths during hydro-forming
of ellipsoidal shells with double generating lines. For the ellip-
soidal shell with the axis length ratio λ ¼ 2:2, at the early stage of
deformation ðp r1:5MPaÞ, when the internal pressure is 0.5 MPa,
the short axis reaches to the design dimension. At the middle stage
of deformation ð1:5MPa o p r 3:0MPaÞ, the long axis is slightly
shortened and short axis is significantly elongated. At the last
stage of deformation ð3:0MPa op r 4:5MPaÞ, the whole shell is an
ideal ellipsoid with single generating line. At this time, the needed
ellipsoidal shell can be obtained by adjusting the forming pressure.
When the internal pressure is 3.3 MPa, the axis length ratio is 1.45.
When the internal pressure is 3.6 MPa, the axis length ratio is 1.41.
When the forming pressure is 4.5 MPa, and the long axis and short
Fig. 6. Structural parameters of pre-form shell.
axis are 214.0 mm and 168.1 mm respectively. Accordingly the axis
lengths variation are  2.7% and 68.1%, respectively.
It can be seen that the variation of axis lengths for the ellip-
ellipsoidal shells with initial axis length ratio 1.5, 1.7 and 2.2 were soidal shell with the axis length ratio λ ¼ 1:7 and λ ¼ 1:5 respec-
designed to comprehensively understand the curvature and tively are similar to the ellipsoidal shell with the axis length ratio
volume variation. λ ¼ 2:2. For the ellipsoidal shell with the axis length ratio λ ¼ 1:7,
Stainless steel SUS304 with the thickness 1.1 mm was used in when the forming The forming pressure is 4.5 MPa, and the long
this paper, and the mechanical properties at room temperature axis and short axis are 215.8 mm and 174.0 mm respectively.
were obtained by uniaxial tensile tests. The yield strength is Accordingly the axis lengths variation are  1.9% and 31.8%,
σ s ¼ 303MPa, the tensile strength is σ b ¼ 800MPa, the elongation respectively. For the ellipsoidal shell with the axis length ratio
is δ ¼ 54% and the strain harden component is n ¼ 0:4 λ ¼ 1:5, when the forming The forming pressure is 4.5 MPa, and
the long axis and short axis are 217.1 mm and 181.0 mm respec-
tively. Accordingly the axis lengths variation are  1.3% and 26.5%,
3.2. Experimental setup
respectively.
Based on the shell shape analysis mentioned above, it is well
The experimental setup during hydro-forming of ellipsoidal
known that: 1) During hydro-forming of ellipsoidal shell, the short
shell is shown in Fig. 7a, and it contains the control system,
axis is significantly linearly increased with the internal pressure,
hydraulic drive system and oil–water pressure transducer. The oil–
and the long axis is slightly shortened; 2) The bigger the initial
water pressure transducer is used to transfer oil pressure into the axis length ratio is, the larger the elongation of short axis is; 3)
water pressure, so that the ellipsoidal shell can be formed by water Deformation first occurs on the first shell segment, and then is
medium. The pressure of the ellipsoidal shell inside could be spread to the second shell segment.
controlled by a servo system according to the loading path from It can be concluded from the experimental results that the
the computer control system. It should be paid attention that the wrinkling can be avoided by designing the ellipsoidal shell with
volume variation of ellipsoidal shell is calculated by the displace- double generating lines as pre-form shell.
ment of piston at water pressure.
In order to accurately record the shell shape during hydro- 4.2. Discussion on curvature radius on dividing globe angle
forming of ellipsoidal shell, a 3D video measurement was carried
out as shown in Fig. 7b. The industrial CCD camera is used as Owing to the needed shell is an ideal ellipsoid with single
image sensor, and the image is decoded by computer. According to generating line, it is essential to analyze the critical internal
the triangle measuring principle, the 3D coordinates can be cal- pressure that double generating lines are changing into single
culated for any step. Hence the variation of shell shape can be generating line. Fig. 10 shows the curvature variation on the
recorded. dividing globe angle during hydro-forming, it can be seen that the
 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187 185

Fig. 7. Experimental setup: (a) Pressure and volume control; (b) Shell shape measurement.

Fig. 8. Deformation of shell shape during die-less hydro-forming of ellipsoidal shell with double generating lines: (a) Initial state; (b) p ¼ 3.0 MPa; (c) p ¼ 3.6 MPa;
(d) p ¼ 4.5 MPa.

curvature radius on the first shell segment is monotonously
decreased, while the curvature radius on the second shell segment
is first slightly increased and then significantly decreased.
According to the geometric criteria of double generating lines
changing into single generating line, for the ellipsoidal shell with
the axis length ratio λ ¼ 2:2, the critical internal pressure is
3.3 MPa. For the ellipsoidal shell with the axis length ratio
λ ¼ 1:7and λ ¼ 1:5, the critical internal pressures are both 3.0 MPa.
4.3. Discussion on shell volume
Based on the axis lengths and curvature analysis mentioned
above, the relationship between volume variation and axis length
ratio will be discussed and compared with mathematical model in
detail.
Fig. 11 shows the volume variation with the increasing of
internal pressure during hydro-forming of ellipsoidal shell with
186 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187

Fig. 9. Axis lengths variation during hydro-forming of ellipsoidal shell with double
generating lines.

the axis length ratio λ ¼ 1:5, λ ¼ 1:7 and λ ¼ 2:2. The volume var-
iation is exponentially increased. Especially at the last stage of
deformation, there is a significant increase of volume even if there
is a slight increase of internal pressure. It also can be seen that, the
bigger the initial axis length ratio is, the larger the volume varia-
tion is. When the forming pressure is 4.5 MPa, the shell volume
variation are 16.6%, 26.1% and 41.0%, respectively.
According to the axis lengths, curvature and volume analysis,
the relationship between shell volume variation and axis length
ratios can be obtained. Fig. 12 shows the volume variation with the
final axis length ratio under the condition of different initial axis
length ratios. Obviously, with the increasing of internal pressure,
the final axis length ratio becomes smaller, accordingly the volume
variation becomes more significant.
More importantly, the volume variation obtained by experi-
ment has a good agreement with the theoretical model prediction.
It is experimentally proved that the volume variation depends on
the initial axis length ratio and final axis length ratio. If the initial
structural parameters are fixed, the volume variation is only varied
with the final axis length ratio. One needed ideal ellipsoidal shell
is associated with one fixed volume variation, namely the needed
shell shape can be controlled by inputting pressured water.

5. Conclusion

Theoretical analysis was carried out on ellipsoidal shell with

double generating lines, and the geometric criteria of double
generating lines changing into single generating line was derived
and the mathematical model between shell volume and axis
length ratio was developed. Experimental researches on the
ellipsoidal shells with axis length ratio λ ¼ 1:5, λ ¼ 1:7and λ ¼ 2:2
were conducted, and the effect of internal pressure was discussed
on the axis lengths, curvature and volume variation. The main
conclusions are drawn as below: Fig. 10. Curvature radius variation on dividing angle during hydro-forming of
ellipsoidal shell with double generating lines: (a) λ ¼ 2:2; (b) λ ¼ 1:7; (c) λ ¼ 1:5.
1) During hydro-forming of ellipsoidal shells, the short axes are
significantly linearly increased with the internal pressure, and
the long axes are slightly shortened. The bigger the initial axis 2) On the dividing globe angle, the curvature radius on the first
length ratio is, the larger the elongation of short axis is. When shell segment is monotonously decreased, while the curvature
the final forming pressure is 4.5 MPa, the short axes are radius on the second shell segment is first slightly increased and
elongated 26.5%, 31.8% and 68.1%, and the long axes are then significantly decreased. According to the geometric criteria
shortened 1.3%, 1.9% and 2.7% for the ellipsoidal shells with axis of double generating lines changing into single generating line,
length ratio λ ¼ 1:5, λ ¼ 1:7 and λ ¼ 2:2, respectively. the critical internal pressure is about 3.0 MPa.
 S.J. Yuan, W.W. Zhang / International Journal of Mechanical Sciences 107 (2016) 180–187
Fig. 11. Shell volume variation during hydro-forming of ellipsoidal shells with
double generating lines.
Fig. 12. Shell volume variation with axis length ratios.
3) During hydro-forming of ellipsoidal shells, the volume variation
is exponentially increased. The bigger the initial axis length
ratio is, the larger the volume variation is. When the final
forming pressure is 4.5 MPa, the shell volume variation are
16.6%, 26.1% and 41.0% , respectively.
4) It is proved that the volume variation depends on the initial axis
length ratio of the preform shell and the final axis length ratio.
The needed ellipsoidal shell shape can be controlled by input-
ting pressured water volume.
187
Acknowledgments
This paper was financially supported by program for Chang-
jiang Scholars and Innovative Research Team in University (No.
IRT1229). The authors would like to take this opportunity to
express their sincere appreciation.
Reference
[1] Wang. ZR, Wang. T, Kang. DC, et al. The technology of the hydro-bulging of
whole spherical vessels and experimental analysis. J Mech Work Technol
1989;18(1):85–94.
[2] Yuan SJ, Zeng YS, Wang ZR. The integral hydro-bulging forming of ellipsoidal
shells. In: Altan T, Lange K, (editors), Proceedings of the 5th international
conference on technology of plasticity, vol. 2. Columbus, Ohio, USA; 1996. p.
943–6.
[3] Zeng. YS, Yuan. SJ, Wang FZ, et al. Research on the integral hydro-bulge forming
of ellipsoidal shells. J Mater Process Technol 1997;72(1):28–31.
[4] Zeng. YS, Wang. ZR, Yuan. SJ. The numerical simulation of the integral hydro-
bulging of ellipsoidal shells. J Mater Process Technol 1997;72(3):358–62.
[5] Yuan. SJ, Zhang WW, Teng. BG. Research on hydro-forming of combined ellip-
soidal shells with two axis length ratios. J Mater Process Technol 2015;219
(5):124–32.
[6] Zhang WW, Yuan. SJ. Research on hydro-forming of combined prolate ellip-
soidal shell with double generating lines. Int J Adv Manuf Technol 2015. http:
//dx.doi.org/10.1007/ s00170-015-7351-4.
[7] Zhang WW, Yuan. SJ. Stress and deformation analysis on hydro-forming process
of ellipsoidal shells with double generating lines. J Mech Eng 2015;51(18):83–
90 in chinese.
[8] Zhang WW, Yuan. SJ. Pre-form design for hydro-forming process of combined
ellipsoidal shells by response surface methodology. Int J Adv Manuf Technol
2015;81(9–12):1977–86.
 7

Journal of Materials Processing Technology 174 (2006) 42–55

Applications of hydroforming processes to automobile parts

Soo-Ik Oh a , Byung-Hee Jeon b , Hyun-Yong Kim c , Jae-Bong Yang a,∗
a School of Mechanical and Aerospace Engineering, Seoul National University, Seoul 151-744, Republic of Korea
b Mechanical Engineering and Design, Induk Institute of Technology, Seoul, Republic of Korea
c Mechanical Engineering, Kangwon National University, Chunchon, Republic of Korea

Received 26 April 2001; received in revised form 7 March 2002; accepted 12 December 2004

Abstract

Hydroforming process recently draws attention of automotive industries due to its advantages like the increased strength, weight reduction,
improved quality and reduced tooling cost. This paper summarizes some of our experiences of tube and sheet hydroforming process design
by simulation technique through actual tryout. Parts included in this paper are tie bar, subframe and engine mount bracket. The simulations
for the entire hydroforming processes were performed and their results were used to predict and fix the forming failures like the wrinkling
and the fracture. Based on this experience, we also suggested necessary improvements in the design procedure.
© 2006 Published by Elsevier B.V.

Keywords: Hydroforming; Tie bar; Subframe; Engine mount bracket

1. Introduction
Hydroforming process recently draws attention of auto-
motive industries due to its advantages over conventional
methods. It can produce wide range of products including
subframes, camshafts, radiator frames, front and rear axles,
crankshafts, seat frames and exhaust manifolds with cheaper
production cost by reducing overall number of processes
[1–6]. The advantages of hydroformed parts include the
increased strength to weight, weight reduction through more
efficient section design, reduction of number of parts and
subassemblies, improved quality of formed components in
terms of dimensional stability and repeatability, and reduced
tooling developing costs.
Hydroforming process is mainly subdivided into the tube
and the sheet hydroforming. Tube hydroforming is mainly
used to produce tube-like products with varying cross-
sectional shape along the length direction [7–12]. Internal
hydraulic pressure is applied to expand the tube to fill the
cavity of the die, which is designed to yield the shape of the
final product. Tube hydroforming starts from a tube that has
∗ Corresponding author. Tel.: +82 2 880 1909; fax: +82 2 880 1910.
E-mail address: jbyang@snu.ac.kr (J.-B. Yang).
been precut into proper length. The tube requires pre-bending
as a pre-forming process. The tube must be bent to the approx-
imate centerline of the finished part prior to hydroforming to
enable the tube to be placed in the die cavity. During the
bending process, the tube undergoes considerable deforma-
tion including thinning. The tube is then placed into the die
and the die closes. Hydraulic liquid fills the tube with two
side cylinders close around the ends of the tube. Simultane-
ously, the liquid is pressurized and cylinders are pushed in
from the side. Material of the tube yields and flows into die
cavity and the part is formed.
There are several methods that can be used in tube bend-
ing [13,14]. Some of these methods are compression bending,
press bending, three-roll bending and rotary draw bending.
The most commonly used bending method for hydroforming
is the rotary draw bending. In rotary draw bending, tube is
locked to the bend die by the clamp. As the bend die rotates,
the pressure die advances with the tube. In this process, a
mandrel may be used to prevent excessive collapse and wrin-
kling in the bend region. Tube hydroforming process requires
precise control of various forming conditions such as die clos-
ing, internal pressure, end sealing and axial feeding [15–19].
Bursting takes place when pressure is applied too rapidly
while enough material is not fed into the die, and too much

0924-0136/$ – see front matter © 2006 Published by Elsevier B.V.

doi:10.1016/j.jmatprotec.2004.12.013
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55 43

feeding of material tends to cause buckling. So, the success-

ful tube hydroforming depends on the reasonable combina-
tion of the internal pressure and axial feeding at the tube
ends.
The sheet hydroforming is a process for producing hollow
components from flat sheet metal blanks [20]. The basic prin-
ciple of the process is to introduce a liquid medium between
two metal sheets and to form these into a hollow component
under the fluid pressure building up. The shape of the finished
component geometry is defined by forming dies. Transport
of the volume flow or build up of internal pressure is made
externally in a high pressure unit and is introduced between
the metal sheets. To build up internal pressure for the sheet
hydroforming, the sheets must be sealed in their flange areas
to hold the internal pressure. This can be done by connecting Fig. 2. Initial mesh of the bending tool.
the sheets by means of welded seam on their circumference.
In this paper, we summarized our experience of hydro-
forming process design starting from simulations to actual
tryouts and validations. The parts included are tie bar, engine
mount bracket and subframes. Parts are made either by tube
or sheet hydroforming.

2. Applications

2.1. Tie bar

Tie bar is an automotive part, which is placed at the front
of passenger room and supports the instrument panel and the
steering wheel. Fig. 1 shows the shape of the tie bar. This
figure also shows several different cross-sections along the
length of the part. Initial tube outer diameter is 48 mm and
thickness is 2.0 mm.
2.1.1. Pre-bending process
Pre-bending simulation of the tube is performed using
a conventional rotary draw bending machine and a bend
die, respectively. During pre-bending process, the tube is
Fig. 3. Bent tube shape: (a) without a wiper die and (b) with a wiper die.
subjected to six bends, with 76 mm centerline bend radius.
Bending angles are 35◦ , 58◦ and 52◦ , respectively.
The material of the tube is STKM11A and its proper-
ties are obtained from tension test. The following are the
material properties: density is 7.8 t/m3 , Young’s modulus is
206.0 GPa, yield stress is 215.1 MPa and stress–strain rela-
tionship is σ̄ = 468.3(ε̄ + 0.022)0.206 MPa.
The pre-bending model using a rotary draw bending
machine is composed of a tubular blank and a tool set. The

Fig. 1. Shape and sections of tie bar.

44 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
Fig. 4. Predicted bent tube geometry.
blank is just a piece of tube. The tool set for pre-bending
model simulates the real tool on the rotary draw bending
machine. The set is composed of a bend die, a pressure die,
a clamp die and a wiper die. Fig. 2 shows the tool set and the
initial tubular blank.
The simulation of the pre-bending process is conducted
with PAM-STAMP. In bending process, the wrinkling and
the buckling may appear depending on the geometry factors
such as bend radius of bend die and diameter and thickness
of the tube. Without a wiper die among the tool set, wrin-
kling occurs in compressive region of the tube. Fig. 3a shows
the bent tube shape with a wrinkling, after the first bend-
ing operation is completed. A wiper die is used to prevent
the wrinkling. Fig. 3b shows the bent tube shape without
a wrinkling, when the bending operation is completed. It
Fig. 5. Predicted thickness and effective plastic strain distribution of the bent tube: (a) thickness distribution and (b) effective plastic strain distribution.
Fig. 6. Bend die and initial tubular blank.
appears that the clearance between the tube and the wiper
die is important. When the clearance is larger than 0.6 mm,
the wrinkling appears again. Springback in bending pro-
cess is a factor that affects dimensional accuracy of the tie
bar. So, the amount of springback must be estimated and
added in bending process. Six bending and springback sim-
ulations are conducted to adjust the precise positions and
angles. Fig. 4 is the view of the final part after all the bends
have been formed. In tensile region of the tube, buckling
occurs, but it is recovered in hydroforming process. Fig. 5a
is the thickness distribution of the tube after all the six
bends have been formed. In tensile region with 58◦ bend-
ing angle, thickness is 1.7 mm minimum and in compressive
region, 2.4 mm maximum. The effective plastic strain distri-
bution of the tube is given in Fig. 5b. The maximum value
is 0.32.
The pre-bending model using a bend die is composed of
a tubular blank and an upper and lower die. Fig. 6 shows
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55 45

Fig. 7. Predicted thickness and effective plastic strain distribution of the bent tube: (a) thickness distribution and (b) effective plastic strain distribution.

Fig. 8. Section shapes of bent tube.

the bend die and the initial tubular blank. To get the preform
shape, three bending operations are required. Fig. 7a is the
thickness distribution of the tube after all the six bends have
been formed. In tensile region with 58◦ bending angle, thick-
ness is 1.83 mm minimum and in compressive region, 2.3 mm
maximum. The minimum thickness is increased by material
flow into tensile region. The effective plastic strain distribu-
tion of the tube is given in Fig. 7b. The maximum value is 0.29.
Fig. 8 is the shape of sections B, C and D of the bend-
ing simulation using the rotary draw bending machine and
the bend die. Fig. 9 is the thickness distribution of section
C. In case of bending with the mandrel, the section shape
remains close to the circle, but the thickness reduction is the
largest. When bend radius is small, the deformation of section
is increased and the thickness is reduced. In case of the bend-
ing with the bend die, the shape of section is similar to that
of draw bending, but the thickness reduction is the smallest. Fig. 9. Thickness distribution of section C.
46 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55

Fig. 10. Tube hydroforming simulation model and result: (a) F.E. model for tube hydroforming process and (b) predicted final geometry of the tie bar.

2.1.2. Hydroforming process the section C. In experiment, the tube was bent with draw
The hydroforming simulation model is composed of the bending machine. In hydroforming simulation, the thickness
pre-bent tube, which is the result of pre-bending simulation, reduction of the tube, which was bent with the draw bending
and the split dies in order to form the designed part geome- machine, is about 0.4 mm in tensile region (b–c–d), but in
try. Fig. 10a shows the pre-bent tubular blank and the upper case of the bending with the bend die, it is about 0.2 mm.
and lower dies. The parting line of the die is designed to In the process design of tie bar hydroforming, it was found
avoid undercut. Through the 2D section simulation of hydro- that:
forming process, the appropriate internal pressure value and
• The pre-bending operation affects the thickness distribu-
the die closing velocity are determined. During die closing,
tion considerably.
internal pressure is applied in order to avoid collapse of the
• It is recommended to use die bending rather than draw
tube. Fig. 10b is the view of the final part after hydroform-
bending for quality improvement.
ing process has been completed. Fig. 11 shows the picture of
the hydroformed tie bar. Fig. 12 shows the tube wall thick-
ness distribution of the final product, which is formed with 2.2. Engine mount bracket
maximum internal pressure 60 MPa, axial feeding 30 mm
and the pre-bent tube produced by the bend die. The min- Engine mount bracket is an automotive part, which is
imum thickness of the hydroformed part is 1.45 mm and the placed between the engine and the auto body, and prevents
maximum is 3.4 mm. By the axial feeding, the maximum
thickness appeared at the both sides of the tube. Fig. 13 shows
the thickness distributions of simulations and experiment of

Fig. 11. Picture of the hydroformed tie bar. Fig. 12. Predicted wall thickness distribution.
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55 47

Fig. 13. Thickness distribution in section C.

Fig. 14. Engine mount bracket: (a) components of engine mount bracket and (b) assembly shape.

the transform of the engine vibration. Fig. 14 shows the com-
ponents and the assembly shape of the conventional stamped
engine mount bracket. It was attempted to use both tube as
well as sheet hydroforming processes in making the engine
mount bracket.
2.2.1. Tube hydroforming
In this application, the engine mount bracket is pro-
duced by tube hydroforming process. Outer diameter of
the initial tube is 41 mm and thickness is 2.6 mm. Pre-
bending simulation of the tube using a conventional rotary

Fig. 15. Thickness distribution of pre-bent tube: (a) without mandrel and (b) with mandrel.
48 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
Fig. 16. F.E. model for tube hydroforming.
draw bending machine is performed. Bend radius is 50 mm
and bending angle is 90◦ . The material of the tube is
SAPH38P and its properties are obtained from tension
test.
Fig. 15 shows the thickness distribution of the bent tube
with mandrel and without mandrel. In case of the bending
without the mandrel, minimum thickness is 2.21 mm and
maximum thickness is 3.22 mm. In case of the bending with
the mandrel, minimum thickness is 1.79 mm and maximum
thickness is 3.31 mm.
Fig. 16 shows the hydroforming simulation model which
is composed of the upper, lower dies and pre-bent tube. Dur-
ing die closing, the internal pressure is not applied. After
die closing, the internal pressure is increased rapidly to
240 MPa. Fig. 17 shows the tube wall thickness distribu-
tion of the final product, which is formed with the pre-
bent tube produced with mandrel. The minimum thickness
of the hydroformed part is 1.3 mm and the maximum is
3.12 mm.
Fig. 17. Thickness distribution of hydroformed tube.
Fig. 18. Production process of engine mount bracket.
2.2.2. Sheet hydroforming
In this application, the engine mount bracket is produced
by sheet hydroforming process. The material of the sheet
blank is SAPH38P and blank thickness is 2.6 mm. Fig. 18
shows the production process of the engine mount bracket
using the sheet hydroforming process. Fig. 19 shows the sheet
hydroforming simulation model, which is composed of the
upper, lower dies and welded blank.
Fig. 20 shows the section shape in processing the sheet
hydroforming analysis. In hydroforming simulation, the
internal pressure is maximum 85 MPa, the flow rate into
the welded blank is 0.1 m3 /s and friction coefficient is 0.05.
Fig. 21 shows the side view and the top view of the deformed
shape of hydroformed part. This figure shows good agree-
ment between the simulation and experiment results.
Fig. 22a shows the thickness distribution of hydroformed
part. The minimum thickness of the hydroformed part is
2.09 mm and the maximum is 2.88 mm. In the ‘a’, ‘b’ and ‘c’
Fig. 19. F.E. model for sheet hydroforming.
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
Fig. 20. Section shapes of sheet hydroforming analysis.
Fig. 21. Comparison between the simulation and the experiment result: (a)
prototype, (b) analysis, (c) prototype and (d) analysis.
regions, the thickness is smallest and the fracture is expected.
In the prototype engine mount bracket, the fracture occurred
in the ‘a’ and ‘b’ regions. Fig. 22b shows the thickness dis-
tribution of section A–A in simulation and experiment. This
figure shows good agreement between the simulation and
experiment results.
In the process design of engine mount bracket hydroform-
ing, it was found that:
Fig. 22. Thickness distribution of hydroformed engine mount bracket: (a) thickness distribution and (b) thickness distribution of section A–A .
49
• It is recommended to bend the tube without mandrel for
quality improvement.
• Thickness reduction is smaller in sheet hydroforming
rather than tube hydroforming.
2.3. Sheet hydroformed subframe
Subframe is an automotive part, which is placed at the
bottom of engine room and supports the engine and absorbs
the crash energy at the car collision. Fig. 23 shows the shape
of the conventional stamped subframe, which is composed
of mainly four parts, No. 1, No. 2, left and right mem-
ber. This figure also shows the target part, which would be
produced by the sheet hydroforming process. Target parts
are the No. 1 member and the upper parts of the left and
right member. For the conventional manufacture of the tar-
get part, two tool sets composed of three tools each are
required, while for the sheet hydroforming process a sin-
gle forming tool consisting of top and bottom die is suffi-
cient. In the stamped subframe, the blank thickness of the
upper part is 2.6 mm and the blank thickness of the lower
part is 2.3 mm. In hydroformed subframe, the blank thick-
ness is the same of the stamped subframe. In the left and
right member of stamped subframe, there is a reinforce-
ment part, which increases the crashworthiness characteristic
of subframe. In hydroformed subframe, the reinforcement
part is eliminated through the crash simulation results with
LS-DYNA3D.
The material of the sheet metal is SAPH38P and its
properties are obtained from tension test. The following
are the material properties: density is 7.8 t/m3 , Young’s
modulus is 206.0 GPa, yield stress is 276.91 MPa and
stress–strain relationship is σ̄ = 629.6(ε̄ + 0.002)0.184 MPa.
Elastic–plastic work hardening and anisotropy model is
used.
Fig. 24 shows the section shapes of the subframe, which
have the different drawing depth of the upper and lower half
of the welded blanks. This figure also shows the results of 2D
section hydroforming simulation. These shows the deformed
blank shapes with the wrinkling. In order to prevent the wrin-
50 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55

Fig. 23. Subframe produced by conventional stamped and welding method: (a) stamped subframe and (b) target part which is redesigned to hydroformed part.

Fig. 25. Parting line moving of the left and right member.

geometry. Fig. 26 shows the geometry of welded blank and

Fig. 24. Section shapes and 2D analysis of sheet hydroforming.
kling, drawing depth of the upper and lower die should be
equaled. Fig. 25 shows the side view of the subframe. The
parting line is moved to the middle of the side plane in order
to prevent the wrinkling.
The sheet hydroforming simulation model is composed
of welded blank and a split die for forming the designed part
a set of upper and lower die.
In sheet hydroforming simulation, the welded blank is
placed into the dies and the upper die moves down until the
two dies are closed. Then, hydraulic liquid fills the inside of
the welded blank. The following are the simulation condi-
tions. The flow rate into the welded blank is 0.5 m/s, die clos-
ing velocity is 5.0 m/s and friction coefficient is 0.05. Fig. 27
shows the forming pressure and blank holding force curve.
Initially, the pressure is increased slowly to 10 MPa until
35 ms, and next the pressure is increased rapidly to 80 MPa
for the calibration. The blank holding force is increased to
21,000 kN.

Fig. 26. F.E. model for welded sheet hydroforming simulation.

 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
Fig. 27. Forming pressure and blank holding force curve.
Fig. 28 shows the deformed shape and thickness distri-
bution of hydroformed subframe. This figure also shows the
picture of the hydroformed prototype. In the results of sim-
ulation, minimum thickness of the hydroformed subframe
is 1.31 mm and maximum thickness is 3.61 mm. This figure
also shows the wrinkling in the ‘a’, ‘b’, ‘c’ and ‘d’ regions.
In the prototype subframe, wrinkling is also appeared.
Fig. 29 shows the deformed shape in the section D–D . It
shows why the wrinkling appears. When the die is closed, the
welded blank is bent according to the geometry of the parting
line. As the internal pressure is applied, the blank at the ‘a’,
‘b’, ‘c’ and ‘d’ regions is under compressive force. In order
to prevent the wrinkling in these regions, the blank under
compression is needed to be stretched. The increased blank
holding force restrains the metal flow into the die cavity, and
it helps the blank overcoming the compressive force.
In order to prevent the wrinkling, the blank holding force
is increased 20,000 kN when the internal pressure is applied.
Fig. 28. Simulation results and the picture of prototype subframe: (a) deformed shape, (b) thickness distribution and (c) prototype subframe.
51
Fig. 30 shows the internal pressure and blank holding force
curve. The blank holding force is increased and maintained
during 15 ms. This figure shows the deformed shape of the
hydroformed subframe. The wrinkling in the ‘a’ and ‘b’
regions is disappeared, but the wrinkling in the ‘c’ and ‘d’
regions still remains. For the successful hydroforming with-
out the wrinkling, the bent angle of the welded blank seen
in the section D–D is to be small through the moving of the
parting line or the modification of the shape in the ‘c’ and ‘d’
regions.
In the process design of subframe sheet hydroforming, it
was found that:
• The parting line is selected to equal the drawing depth of
upper and lower die.
• The control of blank holding force helps prevent the wrin-
kling.
2.4. Tube hydroformed subframe
The application discussed here concerns the tube hydro-
forming process of the subframe. Fig. 31 shows the picture
of the hydroformed subframe. The material of the tube is
STKM11A. Initial outer diameter of the tube is 89.0 mm and
thickness is 2.8 mm.
2.4.1. Pre-bending process
Pre-bending of the tube is performed by the draw bend-
ing machine with mandrel. During the pre-bending process,
the tube is subjected to two bends, with 215.0 mm centerline
bend radius. Bending angles are 88◦ and 84◦ , respectively.
The simulation of the pre-bending process is conducted with
PAM-STAMP. Fig. 32a shows the thickness distribution of
the tube after two bends have been formed. The minimum
thickness is 2.43 mm in tensile region with 88◦ bending
angle and the maximum thickness is 3.28 mm in compressive
region. Fig. 32b shows the thickness distribution of section
52 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55

Fig. 29. Deformed shape in section D–D : (a) deformed shape and (b) section D–D .

A–A . This figure shows the good agreement in simulation
and experiment results. Fig. 33 shows that the wrinkling
appears in the compressive region of the bent tube.
Through the bending simulation, we know that the wrin-
kling appears when the clearance between the tube and the
mandrel is 2.0 mm and disappears when the clearance is
0.5 mm. Consequently, in order to prevent the wrinkling
appearance, the clearance between the tube and the mandrel
is to be small.
2.4.2. Hydroforming process
The simulation of hydroforming process is performed. The
F.E. model of the upper die, lower die and pre-bent tube is
represented in Fig. 34. In simulation, the forming parameters
like the internal pressure and axial feeding are the same of the
real hydroforming process. The internal pressure maintains
12.0 MPa until the upper and lower dies are closed. After die
closing, the pressure is increased rapidly to 100 MPa for the
calibration. The amount of axial feeding is 15 mm at both

Fig. 30. Simulation results of subframe sheet hydroforming and control curve: (a) forming pressure and blank holding force curve and (b) deformed shape.
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55 53

Fig. 31. Tube hydroformed subframe. Fig. 34. F.E. model for subframe tube hydroforming.

ends of the tube. Fig. 35a shows the tube wall thickness dis-
tribution of the final product. The minimum thickness of the
hydroformed part is 2.2 mm in tensile region with 88◦ bend-
ing angle, and the maximum is 3.4 mm. Fig. 35b shows the
thickness distribution of the section A–A . This figure shows
the maximum 0.3 mm thickness difference in simulation and
experiment results.
Fig. 35c shows the shape of section A–A . In this figure,
the ‘d’ region is the tensile region of the bent tube and the
‘b’ region is the compressive region. The bent tube is first

Fig. 32. Thickness distribution of pre-bent tube (with mandrel): (a) thickness distribution of pre-bent tube and (b) thickness distribution in section A–A .

Fig. 33. Wrinkling appearance in the compressive region of the bent tube: (a) photo of pre-bent tube and (b) thickness distribution of pre-bent tube.
54 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
Fig. 35. Thickness distribution of hydroformed subframe: (a) thickness distribution, (b) thickness distribution in section A–A and (c) shape of section A–A .
contact with the die in the ‘b’ region, so the bent tube in the
‘d’ region is expanded by the internal pressure. In the ‘d’
region, the thickness is smallest and the fracture is expected.
In the prototype subframe, the fracture occurred in the ‘d’
region, and it was removed by the moving the fracture part
of bent tube near to the die.
In the process design of subframe hydroforming, it was
found that:
• The pre-bending operation affects the thickness distribu-
tion considerably.
• The location of bent tube and weld seam relative to die
appears to be important.
3. Conclusion
In this paper, the experiment and simulation results of
the hydroformed automotive parts, tie bar, subframe and
engine mount bracket were presented. The simulations have
been performed for the entire hydroforming process and their
results were used to predict and fix the forming failures.
Tie bar has been produced by the tube hydroforming pro-
cess. In order to test the effects of forming parameters, com-
puter simulations of pre-bending and hydroforming operation
were performed.
Engine mount bracket has been produced by the tube and
sheet hydroforming process. Compared with the stamped
part, the production cost of hydroformed part was reduced
about 30%, and the number of production process was
reduced from 11 to 4.
The front parts of the conventional stamped subframe have
been redesigned to the sheet hydroformed part. For the hydro-
formed subframe, a single forming tool consisting of top
and bottom die has been produced, and the reinforcement
part, which increased the crashworthiness characteristic of
the stamped subframe, has been eliminated through the crash
simulation.
Subframe has been produced by the tube hydroforming
process. Pre-bending of the tube has been performed by the
draw bending machine with mandrel. In hydroforming pro-
cess, the fracture occurred in the tensile region of the bent
tube. The fracture was removed by the moving the fracture
part of bent tube near to the die.
Through the hydroforming experiments and simulations,
the reasonable combination of the internal pressure and axial
feeding was found to be necessary for the successful tube
hydroforming. So, the sensitivity analysis with respect to the
internal pressure and axial feeding was developed for the
optimal process design of hydroforming. It was applied to
the optimization of tube hydroforming process, and the tube
thickness variation was minimized [21,22].
The advantages of hydroformed parts include the
increased strength, weight reduction, improved quality and
reduced tooling costs. In hydroforming part design, the shape
design to meet the strength requirement and to reduce the
 S.-I. Oh et al. / Journal of Materials Processing Technology 174 (2006) 42–55
weight is important. So, the sensitivity analysis with respect
to section shape of hydroforming part was developed. It was
applied to the shape design of tube hydroforming part, and
optimal shape with the increased strength and reduced weight
was achieved [23].
References
[1] S. Nakamura, H. Sugiura, H. Onoe, K. Ikemoto, Hydromechanical
drawing of automotive parts, J. Mater. Process. Technol. 46 (1994)
491–503.
[2] M. Mason, Tube hydroforming—advancements using sequenced
forming pressures, innovations in hydroforming technology, Tube
Pipe Assoc. Int. (1996).
[3] L. Wu, Y. Yu, Computer simulations of forming automotive structural
parts by hydroforming process, Numisheet’96 (1996) 324–329.
[4] F. Dohmann, Ch. Hartl, Tube hydroforming—research and practical
application, J. Mater. Process. Technol. 71 (1997) 174–186.
[5] M. Ahmetoglu, T. Altan, Tube hydroforming—state-of-the-art and
future trends, in: SAE, International Congress and Exposition,
Detroit, MI, 1999.
[6] J.B. Yang, B.H. Jeon, S.I. Oh, Simulation of tube hydroforming for
an automotive part, Numisheet’99 (1999) 491–496.
[7] S. Fuchizawa, Deformation of metal tubes under hydrostatic bulge
forming with closed die, Adv. Technol. Plast. (1990) 1543–
1548.
[8] S. Fuchizawa, M. Narazaki, Bulge test for determining stress–strain
characteristics of thin tubes, Adv. Technol. Plast. (1993) 488–
493.
[9] T. Hiroi, H. Nishimura, Bulging limit of thin walled aluminum tubes
with a surface defect, Adv. Technol. Plast. (1993) 527–532.
[10] F. Dohmann, A. Bohm, K. Dudziak, The shaping of hollow shaft-
shaped workpieces by liquid bulge forming, Adv. Technol. Plast.
(1993) 447–452.
55
[11] H.J. Kim, B.H. Jeon, H.Y. Kim, J.J. Kim, Finite element analysis
of the liquid bulge forming processes, Adv. Technol. Plast. (1993)
545–550.
[12] F. Dohmann, Ch. Hartl, Liquid bulge forming as a flexible production
method, J. Mater. Process. Technol. 45 (1994) 377–382.
[13] L. Oldacre, Introduction to tube bending for hydroforming applica-
tions, in: TPA’s 2nd Annual Automotive Tube Conference, 1997, pp.
37–68.
[14] T. Granelli, Examining the variables in production tube bending, in:
TPA’s 2nd Annual Automotive Tube Conference, 1997, pp. 69–84.
[15] D. Schmoeckel, C. Hessler, B. Engel, Pressure control in hydraulic
tube forming, Ann. CIRP 41 (1992) 311–314.
[16] S.C. Rama, A FEM model to predict pressure loading cycle for
hydroforming processes, in: SAE, International Congress and Expo-
sition, Detroit, MI, 1999.
[17] W. Rimkus, H. Bauer, M.J.A. Mihsein, Design of load-curves for
hydroforming applications, J. Mater. Process. Technol. 108 (2000)
97–105.
[18] M. Ahmed, M.S.J. Hashmi, Finite-element analysis of bulge forming
applying pressure and in-plane compressive load, J. Mater. Process.
Technol. 77 (1998) 95–102.
[19] M. Ahmed, M.S.J. Hashmi, Estimation of machine parameters for
hydraulic bulge forming of tublar components, J. Mater. Process.
Technol. 64 (1997) 9–23.
[20] D. Schmoeckel, C. Hessler, R. Huber, Metal forming of tubes and
sheets with liquid and other flexible media, Ann. CIRP 48 (1999)
1–17.
[21] J.B. Yang, B.H. Jeon, S.I. Oh, Design sensitivity analysis and opti-
mization of hydroforming process, J. Mater. Process. Technol. 113
(2001) 666–672.
[22] J.B. Yang, B.H. Jeon, S.I. Oh, A study on optimal process design
of hydroforming process using sensitivity analysis, J. Korean Soc.
Technol. Plast. 9 (2000) 723–730.
[23] J.B. Yang, B.H. Jeon, S.I. Oh, Optimal shape design of hydroformed
part considering crashworthiness, Trans. KSME, submitted for pub-
lication.
 8

ARCHIVES OF CIVIL AND MECHANICAL ENGINEERING

Vol. VIII 2008 No. 3

Automotive component development

by means of hydroforming
A. KOCADA, H. SADOWSKA
Warsaw University of Technology, ul. Narbutta 85, 02-524 Warszawa, Poland

Hydroforming processes have become popular in recent years, due to the increasing demands for
lightweight parts in various fields, such as bicycle, automotive, aircraft and aerospace industries. This
technology is relatively new as compared with rolling, forging or stamping, therefore there is not much
knowledge available for the product or process designers. Comparing to conventional manufacturing via
stamping and welding, tube (THF) and sheet (SHF) hydroforming offers several advantages, such as de-
crease in workpiece cost, tool cost and product weight, improvement of structural stability and increase of
the strength and stiffness of the formed parts, more uniform thickness distribution, fewer secondary op-
erations, etc. The paper presents extensive possibilities of component development in automotive industry
by means of hydroforming processes. There are also presented some examples on computer modelling of
these processes and limiting phenomena.

Keywords: hydroforming, THF, SHF, FEM

1. Introduction
Hydroforming uses fluid pressure in place of the punch as comparing with a con-
ventional tool set to form the component into the desired shape of the die. Generally,
hydroforming processes would be classified as tube or sheet hydroforming depending
on the initial shape of workpiece. In the tube hydroforming process (THP), the initial
workpiece is placed into a die cavity, which corresponds to the final shape of the com-
ponent, Figure 1. Next, the dies are closed under the force and the tube is internally
pressurized by a liquid medium to effect the expansion of the component (internal
pressure, pi) and axially compressed by sealing punches to force material into the die
cavity (axial force, 2). Hence the component is formed under the simultaneously con-
trolled action of pi and axial force. The process should be controlled to avoid failures
such as buckling, wrinkling and bursting.
Appropriate fundamentals to determine process controls were developed by ex-
perimental approaches as well as by means of FE simulations, e.g. [1–6]. Water–oil-
emulsions are typically used media to apply internal pressure, which is usually in-
creased to 250 MPa, and in certain cases up to 600 MPa. The necessary amount is in-
fluenced significantly by the wall thickness of the component, the material strength
and hardening as well as by the component shape [7–10].
The hydroforming is very useful for producing whole components that would oth-
erwise be made from multiple stampings joined together. For example, a typical chas-
56 A. KOCADA, H. SADOWSKA

sis component that would normally be made by pressing up to six channel sections and
joining by spot welding can be hydroformed as a single part. Other examples of hy-
droformed parts from automotive industry are shown on Figure 2.

Fig. 1. The hydroforming principles: a – tool setup, b – initial tube, c – final product (T-joint)

In the majority of cases the complexity of the components requires that additional
preforming operations are considered together with the hydroforming process itself.
These preforming operations can involve bending and mechanical forming of the ini-
tial component to ensure that it is capable of insertion into the hydroforming die or to
obtain an optimized material distribution [13].

Fig. 2. Industrial examples of tube and shell hydroforming: (a) – engine cradle [0], (b), (c) – Pontiac Sol-
stice and parts made by hydroforming [12], (d) – exhaust system of Kawasaki [14]
 Automotive component development by means of hydroforming 57

In the last few years, the demand for weight reduction in modern vehicle construc-
tion has led to an increase in the application of hydroforming processes for the manu-
facture of automotive lightweight components made from steel or aluminium. It re-
sults from improvements in stiffness and crash behaviour due to the reduction of
welding seams, and with reduced assembly costs. Currently, use of aluminium alloy
which is light weight material is increasing. Extruded aluminium profiles have been
used for automobile frame parts [10] to get higher stiffness and light weight, Figure 3.
Many progresses have been made in forming of square sectioned profile by stretch
bending [12, 15] and hydroforming [16–17].

Fig. 3. Part of the subframe (a), photographs of forming of the extruded profile:
bending (b), pressing (c), hydroforming (d) [21]

The use of lightweight materials such as aluminium and magnesium can reduce the
weight of passenger vehicles up to 40–75% by replacing ferrous auto body structures
and body panels [18]. It was reported that a 10% weight reduction in an average auto-
motive body could improve the fuel efficiency by 6–8% [19]. However, the cost for
the lightweight materials is estimated to be higher than that of the mild steel structures
because of the raw material price and the production costs with the existing manufac-
turing technologies. For aluminium alloys, the cost increases up to 30–100% are ex-
pected while it is forecast to be around 50–150% for the magnesium alloys [18]. On
the other hand, since 80% of the total energy consumption throughout the life cycle of
an automobile occurs during the utilization (driving) period, the use of lightweight
parts is still seen as a prominent, long-term and cost effective response to the fuel effi-
ciency and the emission reduction demands [20].
Considerable mass savings are possible through eliminating the flanges required
for welding and using thinner steel. In spite of that, stiffness is still maintained and the
discontinuous spot-welded joints are eliminated [21, 23–24]. Figure 4 presents main
part of exhaust system for motorcycle Kawasaki Zx 10 R produced by conventional
methods and by hydroforming. Shape of hydroformed parts is more complicated but
58 A. KOCADA, H. SADOWSKA

very smooth and harmonic. What more, cost of hydroforming of these complicated
parts has been lower than cost of producing and joining many parts by conventional
processes.

Fig. 4. Comparison of exhaust systems produced by conventional methods (a)

and by hydroforming (b) [14]

Hydroforming has already been widely used in the US – more than a million en-
gine cradles a year are produced by hydroforming processes. Some 2.8 million com-
ponents a year for one of Chrysler’s model are produced by hydroforming, too. In
Europe the technology is being used in sub-frames for models such as Ford’s Mondeo
and General Motors’s Vectra. Hydroforming is well advanced in Germany. However,
as hydroforming – particularly high-pressure hydroforming – is at the frontier of mod-
ern steel technology, many designers and engineers still need convincing of its capa-
bilities [24].

2. Sheet hydroforming
Today, predominantly tubular material is considered for the mass production of
hydroformed parts. Hydroforming of sheet material (SHF) is up to now mainly used
for small batch production due to a comparatively high cycle time. Furthermore, sheet
hydroforming requires higher clamping forces than tube hydroforming, causing more
cost-intensive presses. However, advances in process and press technology, e.g. [3–6],
increasingly contribute to a wider industrial application of sheet hydroforming, in par-
ticular for the flexible manufacture of small batch sizes. One of the first examples for
an industrial application is the sheet hydroforming of roofs for luxury class cars, Fig-
ure 5.
Figure 6 shows the scheme of a sheet hydroforming. When the punch pushes the
sheet metal into the die cavity, within which oil or other liquids are contained, pres-
sure pi that can press the sheet metal tightly onto the punch will be generated. At the
same time, the liquid in the die cavity will flow out between the upper surface of the
die and the sheet metal, what results in reduction of frictional forces. By this process,
the limit drawing ratio value of sheet metal can be increased. The liquid can be used as
a punch, a draw die or an assisting way to improve sheet formability. Actually, almost
all of the materials used in conventional stamping can be used in sheet hydroforming.
 Automotive component development by means of hydroforming 59

Depending on the different means, the liquid pressure in the die cavity is from around
30 to 150 MPa, but the usage of 200 MPa has also been reported [13, 25]. Roof for
luxury class car (Figure 5), deep and partially conical cup (Figure 7) or cups with
complicated geometries of bottoms (Figure 6) are examples of parts made by sheet hy-
droforming.

Fig. 5. Roof for luxury class cars [23]

Fig. 6. Scheme of sheet hydroforming (a) and examples of products (b) [13]

It is well known that formability of the lightweight materials usually increases at

elevated temperature levels [27–28].Warm forming technology with selective heating
enables manufacturing of lightweight parts by utilizing the increased formability at
elevated temperature [28]. However, it is quite difficult to determine optimal tempera-
ture distributions in tooling [29].
Warm hydroforming technology for lightweight materials is currently being devel-
oped to achieve reduced number of manufacturing steps and part consolidation. It makes
use of the improved formability at elevated temperature while using the tooling and
hydraulic medium as means of transporting heat as well as mechanical/hydraulic force
at relatively low levels, Figure 8.
60 A. KOCADA, H. SADOWSKA

Fig. 7. Scheme of deep drawing by hydroforming (a) and example of product (b) [26]

Fig. 8. An example of loading profiles: (a) – controllable variables (blank holder force,
hydraulic pressure and punch speed), (b) – temporal control of loading profiles [30]

Fig. 9. Scheme of the hydroforming of metal pairs (a) [26]

and industrial examples of hydroformed metal pairs (b) [31–32]
 Automotive component development by means of hydroforming 61

A special case of sheet hydroforming is the hydroforming of welded-closing sheet

metal pairs. Figure 9a shows the scheme of the process whereas Figure 9b shows in-
dustrial examples. The hydroforming technology of sheet metal pairs was developed
some years ago [5, 26] but the number of applications in automotive industry increased
especially in the last years. In the scheme (Figure 9a), the periphery of the sheet metal
can be welded using laser welding. Then a liquid medium can be filled between the
blanks, and pressurization can be effected by a hydraulic system. It is very difficult to
realize radial feeding using this method, as it is essentially a pure bulging deformation.
To some extent, this technology is similar to tube hydroforming.

3. Tube hydroforming
Grey et al. in 1939 [33] formed a T-joint using a seamless copper tube and applied
for and achieved a US patent in the 1940s, which gave an indication of the coming pe-
riod of tube hydroforming. Until now, the forming of T-shape used for a joint is still
a problem in hydroforming. Figure 1 shows one of the schemes of tube hydroforming.
When using seamless tube or welded tube, the blank can be formed into the shape
of the die cavity by internal pressure and when the side punches move in. Tube hydro-
forming has many advantages such as part consolidation, weight reduction, improved
part strength and stiffness, highly accurate dimensions and low springback, lower tool-
ing cost and fewer integrated processes, etc. which all promote rapid spreading of this
technology in the automotive, household and aerospace industries [2, 34]. Welded
tube blanks can be subjected to an internal pressure as high as 300 MPa, but excep-
tionally a pressure of up to approximately 700 MPa can be used [34–36].
Common semi finished products used for initial workpieces are longitudinally
welded tubes made of conventional steel like unalloyed and stainless steel, and in
some cases of aluminium alloys [37–38]. Today, steel suppliers offer new steel grades,
e.g. high strength steels [39–40], competing with aluminium as new materials for
lightweight constructions combined with hydroforming.
One of the leading automotive companies using hydroforming is BMW [41]. In the
newest models of BMW (e.g. BMW M3) hydroformed exhaust components are used
(Figure 10). New technology of tube hydroforming has become so profitable from
many points of view, that other car producers have introduced new components into
cars, as for example Audi (Figure 11).
There exists a considerable interest to reduce vehicle weight through the adoption
of lightweight materials, such as aluminium alloys, while maintaining energy absorp-
tion and component integrity under crash conditions. The interaction between tube hy-
droforming and behaviour during crash events was studied using lightweight automo-
tive structural members [43]. There was used a high-pressure hydroforming process in
which tubes with various corner radii in the tube cross-section were produced, Figure
12. Next the tubes were subjected to axial upsetting.
62 A. KOCADA, H. SADOWSKA

Fig. 10. Exhaust system of BMW M3 (a) and its elements formed by hydroforming (b)
as well as BMW rear axles with hydroformed components (c) [42]

Fig. 11. Audi TT Coupe with hydroformed rear axle components [42]
 Automotive component development by means of hydroforming 63

Fig. 12. Cross-section of non-hydroformed and hydroformed tubes [43]

The tubular hydroforming process can be used to produce a bumper stay, which se-
cures the bumper beam to the vehicle body (see Figure 13b). When the automobile is
hit from the front or behind, the bumper beam collapses and the impact force is trans-
mitted to the left and right front frames, respectively, through the bumper beam and
bumper stays. The impact energy is absorbed by plastic deformation of the bumper
beam and bumper stays. The conventional bumper structure is assembled from several
parts (Figure 13a), so several manufacturing processing steps are needed, and the
structure is somewhat complex. Most research work on bumper stays has focused on
using reinforcing members that have complicated shapes [44–45]. Hydroformed
bumper stay (Figure 13c,d) is rather simple in the shape but its ability to absorb energy
through plastic deformation is relatively high.
In most of the tube hydroforming processes, the decrease in wall thickness is pre-
vented by compressing the tube in the axial direction simultaneously with the action of
the internal pressure (see example in Figure 1). If the internal pressure is too small, the
axial compression causes the wrinkling of the tube wall. Hence the paths of internal
pressure and axial compression in the tube hydroforming are keys to prevent the oc-
currence of these defects. The finite element simulation has been employed to deter-
mine the pressure paths [16, 47–50]. A pulsating hydroforming process of tubes has
been developed for the forming of hollow products with a complex shape [51]. An im-
provement of the formability by means of the pulsating hydroforming have been in-
vestigated [52–53] and simulated by the finite element method [52].
The deformation behavior of the tube in the hydroforming is greatly influenced by
the friction at the interface between the tube and die [54]. Although the friction during
forming is reduced by ultrasonic vibration of dies [55], the friction in the pulsating hy-
droforming has been hardly decreased by the oscillation of internal pressure [56]. In
the pulsating hydroforming, the local thinning is prevented by the uniform expansion,
and thus the formability is improved. Figure 14 shows the difference between shape of
free bulged tube with constant and pulsating pressure. The tube is uniformly expanded
by repeating the appearance and disappearance of the small wrinkling, and thus the lo-
cal thinning is prevented. The pulsating hydroforming is attractive for the improve-
ment of the formability of tubes.
64 A. KOCADA, H. SADOWSKA

Fig. 13. A conventionally produced bumper stay (a), hydroformed bumper stay (b), bumper stay hydro-
forming process (c) with detailed shape of tube wall (d) [46]

Fig. 14. Difference between constant and pulsating hydroforming: tube before forming (a), deforming
shape of free bulging region for high constant pressure (b) and for pulsating pressure (c) [53]

Bellows are quite important group of products made by hydroforming of tubes,

Figure 15. They are commonly used in piping systems to absorb expansion and me-
chanical movement and they have widespread applications in industrial and chemical
plants, power systems, heat exchangers, and automotive vehicle parts [57]. Special
stands for expanding the tube wall by internal pressure and axial compression of the
tube in order to produce bellows were developed in Metal Forming Department at
Warsaw University of Technology in 1960s. There were produced bellows with one
and multi-layer walls [58–59].
 Automotive component development by means of hydroforming 65

Fig. 15. The hydroformed bellows [57, 58]

Fig. 16. An example of hole made by hydropiercing [60]

Cost-effectively produced, correctly positioned holes are essential, functional re-

quirements in nearly all structural automotive components [60]. Holes can be cut into
a part by drilling/milling, laser or plasma cutting, flow drilling, and punching in the
die (hydropiercing), or post-piercing. Hydropiercing can produce various hole styles
just in the primary process. Figure 16 shows an example hydropierced hole where
a slug was pushed out from the cavity by internal pressure. Holes that cannot be hy-
dropierced must be done in a secondary cutting or piercing operation at additional cost
and reduced positional accuracy and repeatability.

4. Product development cycle

The development of hydroformed parts for series production necessitates efficient
methods to meet the requirements of short development times, and high part quality
with an optimized process chain. An important factor in achieving these short devel-
opment lead times is the use of CAD, CAM and CAE techniques, including process
simulations exploiting the potential of FEA [70]. This is particularly important as the
spectrum of the hydroformed components increases continuously in quantity and com-
plexity.
66 A. KOCADA, H. SADOWSKA

Finite element analysis (FEA) has been commonly used in numerical simulation of
hydroforming. As exemplary for joints, mainly two models with shell [61–65] and
solid elements [66–69] have been reported in most of the recent papers. It is rather dif-
ficult to say which model would be good enough for a specific application. Hence,
a comparison of calculated results obtained by means of different numerical models
could be very helpful in finding some indications for using those models.
Typical steps and phases of a product development cycle for a hydroformed com-
ponent are illustrated in Figure 17. The basis of this cycle is the CAD model of the
component. The first analysis of the component, e.g. the determination of cross-sec-
tional circumferences, corner radii, bent mean axis [71] and the application of one-step
simulations, gives a rough estimation as to whether the part can be manufactured by
hydroforming or not. It gives also a first indication of forming loads such as internal
pressure, axial forces and the force to close the die during the forming process.

Fig. 17. Characteristic product development cycle for hydroformed components [70]

If the feasibility of the part is proven, and additional forming steps such as bending
and pre-forming are determined, a more accurate simulation can be performed by fi-
nite element simulation methods. The forming die shape of the hydroforming die
model is usually generated directly from the CAD-data, complete with the tool parting
line and the end feed sections of the workpiece.
Even during the experimental prototyping phase, FEA is used to optimize the proc-
ess design. Necessary modifications of the dies can also be investigated by “virtual
manufacturing”. This enables the saving of time and costs for the whole product de-
velopment cycle. Successful prototyping showing the part’s feasibility and process re-
liability is followed by the design, manufacture and try-out of the series production
tooling. To ensure adequate tooling life time and a constant part quality, the FEA is
used in this stage to determine the stresses and elastic deformation acting on the tool
 Automotive component development by means of hydroforming 67

elements due to the forming loads. In the context of the above described component
and process design, the FEA is used:
x to check the production feasibility of the component,
x to analyze and optimize the final component quality and expected process reli-
ability,
x to determine an indication of the required process forces for the die and machine
design, e.g. [62].
The forming possibilities of the hydroforming process are crucial for product de-
velopment. The tube model passes through the individual forming steps, so that the
strains and stresses from each preceding step are considered [9]. For the tube model,
four-node shell elements with complete integration over the shell thickness work satis-
factorily. In practically all cases, the component curved surfaces are the most critical
areas in the hydroforming process. Very good simulation results can be achieved by
the detailed reproduction of the bending process. Final dimensions can be calculated
with an accuracy variation of less than 2% by such coherent simulation.
There was also shown [9] the comparison of calculated component properties with
the properties of the manufactured component for an aluminium rear axle component,
manufactured by pre-forming and hydroforming. The forming limit curve (FLC) of the
aluminium material was used to estimate the process feasibility with the FEA during
the component and process design [73]. This was done on the assumption that the ratio
of maximum to minimum strain is almost constant during the forming. Besides other
boundary conditions, the choice of the friction law and friction coefficients is impor-
tant for the accuracy of the simulation results. For most if the simulations, Coulomb’s
law of friction was used. A modified friction test, working with pressurized tubes, en-
ables the determination of suitable values for the friction coefficient [72].
The optimisation of the hydroformed engine cradle geometry with the aid of FEA
in the course of an experimental prototyping was presented in [74]. In this example the
first execution of the pre-forming operation resulted in wrinkles on a bent area of the
part. It was not possible to flatten those wrinkles with the internal pressure by the sub-
sequent hydroforming operation. To remove this potential failure, different variants of
pre-forming die shapes were investigated by FEA and the most suitable one was
translated into the new die cavity. A saving of time and costs of about 70% was
achieved with the aid of FEA in comparison to an experimental try-out.
Wide spreading of the HF has been limited by a kind of secrecy on knowledge
bases and the lack of specific material specifications for incoming shapes and tubes.
Hence, new component applications cause many problems that usually must be solved
individually. Excellent example can be hydroformed X and T-joint. Hydroforming
process feasibility of X-joint has been discussed in [75–76]. Experimental results of
X-joint hydroforming have been compared with the results of extensive FEA simula-
tions in order to find a method of failure prediction.
68 A. KOCADA, H. SADOWSKA

5. Concluding remarks
Sheet and tube hydroforming processes have been placed at the frontier of modern
metal forming technologies. The future of the processes remains quite prospective due
to the development of computer modeling but there have been some limitations on
prediction of forming limits. As the industry has developed, it has become standard
practice that hydroforming requires an intense focus on process simulation to predict
as many difficulties as possible beforehand.
As comparing with conventional metal forming technologies, hydroforming is still
quite new. Following the development of the relevant technologies, such as equipment
design and manufacturing, automatic control systems or ultra pressure units, many dif-
ferent methods have been invented and hydroforming can be applied successfully in
large volume production. The future of the hydroforming still remains quite exciting
and prospective though a huge achievement has been obtained at this time mainly for
tube hydroforming in the automotive industry. Once a break-through component is
found in sheet hydroforming in some application fields, possible development will be
even much faster.

References

[1] Chaupczak J.: Application of hydroforming for making T-joints and X-joints (in Polish),
Scientific Papers of Technical University of Kielce, Mechanics, 41, 1988.
[2] Ahmetoglu M., Sutter K., Li X.J., Altan T.: Tube hydroforming: current research, Journal
of Materials and Technology, 98, 2000, pp. 224–231.
[3] Kleiner M. et al.: New 100.000 kN press for sheet metal hydroforming, Proceedings of the
International Conference on Hydroforming, Stuttgart, Germany, 6th/7th November, 2001,
pp. 351–362.
[4] Geiger M.: Neue Entwicklungen beim IHU von Blechen und deren Einsatzpotenziale für
Leichtbauteile, Proceedings of the Third Chemnitz Car Body Colloquium, Chemnitz,
Germany, 25th/26th September, 2002, pp. 225–241.
[5] Novotny S., Hein P.: Hydroforming of sheet metal pairs from aluminum alloys, Proceed-
ings of the SheMet’99, 9th September, 1999, pp. 591–598.
[6] Novotny S., Geiger M.: Process design for hydroforming of light weight metal sheets at
elevated temperatures, Journal of Material Process and Technology, 138, 2003, pp. 594–
599.
[7] Altan T.: Formability and Design Issues in Tube Hydroforming, Proceedings of the In-
ternational Conference on Hydroforming, Stuttgart, Germany, 12th/13th October, 1999,
pp. 135–151.
[8] Dohmann F., Hartl Ch.: Hydroforming – a method to manufacture light-weight-parts,
Journal of Material Process and Technology, 60, 1996, pp. 669–676.
[9] Hartl Ch., Bauer H., Haas A.: Optimierungserfolge durch Simulation in der Innen-
hochdruck-umformung, EFB-Kolloquium Wirtschaftliche Blechumformung durch Simu-
lation, Fellbach b. Stuttgart, Germany, March, 1999, pp. 9–10.
 Automotive component development by means of hydroforming 69

[10] Vollesten F. Sprenger A., Krauss J.: Extrusion, channel, and profile bending: a review,
Journal of Materials and Technology, 87, 1990, pp. 1–27.
[11] Kelder M., Edgar K.: Hydroforming dual-phase steel, The Tube & Pipe Journal 7th
December, 2004, pp. 53–57.
[12] Maki T., Walter C.: Liquid curves – Sheet hydroforming helps the sporty Solstice stand
out, Stamping Journal, 8th May, 2007, pp. 32–37.
[13] Zhang S.H., Danckert J: Development of hydro-mechanical deep drawing, Journal of
Materials Processing Technology, 83, 1998, pp.14–25.
[14] http://www.carpimoto.com/content/prod/l_EN/s_1_Home/pr_16884_Akrapovic_Complete
_Racing_Hydroforming_Exhaust_System_Kawasaki_ZX_10R_04_05_full_titanium_Oval_
muffler_with_titanium_sleeve.htm.
[15] Paulsen F., Welo T., Sovik O.P.: A design method for rectangular hollow section in bend-
ing, Journal of Materials and Technology, 113, 2001, pp. 699–705.
[16] Yang J.B., Jeon B.H., Oh S.I.: Design sensitivity analysis and optimization of the hydro-
forming process, Journal of Materials and Technology, 113, 2001, pp. 666–672.
[17] Asnafi N., Nilsson T., Lassl G.: Tubular hydroforming of automotive side members with
extruded aluminum profiles, Journal of Materials Processes Technology, 142, 2003, pp.
93–142.
[18] Carpenter J.A.: The Freedom CAR Challenge and Steel, American Iron and Steel Insti-
tute, Great Designs in Steel Seminar, Livonia, MI, 2004, pp. 96–111.
[19] Schultz R.A.: Aluminum for light vehicles – an objective look at the next ten to twenty
years, in: 14th Int’l Aluminum Conference, Montreal, Canada (Ducker Research), 15th
September, 1999, pp. 243–250.
[20] Mildenberger U., Khare A.: Planning for an environment-friendly car, Technovation, 20,
2000, pp. 205–214.
[21] Choi Y., Yeo H.T., Park J.H., Oh G.H., Park S.W.: A study on press forming of automo-
tive sub-frame parts using extruded aluminum profile, Journal of Materials Processing
Technology, 187–188, 2007, pp. 85–88.
[22] Godwin J.: Materials World, Hydroforming Techniques, Vol. 6, No. 8, 1998, pp. 483–84.
[23] Hartl Ch.: Research and advances in fundamentals and industrial applications of hydro-
forming, Journal of Materials Processing Technology, 167, 2005, pp. 383–392.
[24] USS – materials available in http://ussautomotive.com/auto/steelvsal/alintensive.htm.
[25] Vollertsen F., Prange T., Sander M.: Hydroforming: needs, developments and perspec-
tives, Proceedings of the Sixth International Conference on Technology Plasticity, Ad-
vanced Technology of Plasticity, Vol. 6, Berlin, Germany, 1999, pp. 1197–1210.
[26] Kleiner M., Gartzke A., Kolleck R., Ramer J., Weidner T.: Finite element simulation for
high pressure sheet metal forming (HBU process) and tool construction, Adv. Technol.
Plast. 2, 1996, pp. 975–983.
[27] Li D., Ghosh A.: Tensile deformation behavior of aluminum alloys at warm forming tem-
peratures, Mater. Sci. Eng. A, 352, 1–2, 2003, pp. 279–286.
[28] Naka T., Torikai G., R. Hino, F. Yoshida: The effects of temperature and forming speed
on the FLD for type5083 Al–Mg alloy sheet, Journal of Materials and Technology, 113,
2001, pp. 648–653.
[29] Kim H.S., Koc M., Ni J.: Determination of app.ropriate temperature distribution for
warm forming of aluminum alloys, Trans. NAMRC 32, 2004, pp. 573–580.
70 A. KOCADA, H. SADOWSKA

[30] Zhongqi Yu, Zhongqin Lin, Yixi Zhao: Evaluation of fracture limit in automotive alumi-
num alloy sheet forming, Materials and Design, 28, 2007, pp. 203–207.
[31] Vahl M., Hein P., Bobbert S.: Hydroforming of sheet metal pairs for the production of
hollow bodies, 2000, available on http://www.edpsciences.org/articles/metal/pdf/(2000)/
10/ p1255.pdf.
[32] Vollrath K.: Hydroforming gains ground in German, 2002, available on http://www.the
fabricator.com/Hydroforming/Hydroforming_Article.cfm?ID=106.
[33] Grey J.E., Devereaux A.P., Parker W.N.: Apparatus for making wrought metal T’s, US
Patent 2, 203, 1939, pp. 868.
[34] Constantine B., Roth R., Clark J.P.: Substituting tube-hydroformed parts for automotive
stamping: an economic model, J. Mat., 53, 8, 2001, pp. 33–38.
[35] Zimniak Z., Piela A.: Finite element analysis of a tailored blanks stamping process, Jour-
nal of Materials Processing Technology, 106, 2001, pp. 254–260.
[36] Kuvin B.F.: Hydroformer evolves into complete source for prototype and short-run sheet,
Metalforming, 12, 2002, pp. 26–29.
[ 37] Luecke H.-U., Hartl Ch., Abbey T.: Hydroforming, Proceedings of the International
Conference on Sheet Metal, Erlangen Nuremberg, Germany, 27th/28th, September, 1999,
pp. 607–614.
[38] Leitermann W. et al.: Innovative Forming Technologies for Space Frames, Proceedings
of the International Conference on Advanced Technology of Plasticity, Nuremberg, Ger-
many, 19–24 September, 1999, pp. 1183–1188.
[39] Flehmig T., Schwarz S.: Hydroforming complex hollow sections, Steel Grips 1, 6, 2003,
pp. 408–412.
[40] Kroef et al.: Untersuchungen von Stahl und Bestimmung von Umformcharakteristiken fur
hydraulische Umformprozesse, Proceedings of the Third Chemnitz Car Body Colloquium,
Chemnitz, Germany, September 25/26, 2002, pp. 69–80.
[41] http://nsm.eng.ohio-state.edu/THFModule1/html/app.lications.html.
[42] http://www.salzgitterhydroforming.de/westsachsen/crimmitschau/salzgitter/de/produkte /
abgaskomponenten/bsp_kruemmer/.
[43] Williams B.W., Oliveira D.A., Simha C.H.M., Worswick M.J., Mayer R.: Crashworthi-
ness of straight section hydroformed aluminum tubes, International Journal of Impact En-
gineering, 2007, pp. 1451–1464.
[44] Oyama M., Masuta N.: Automotive bumper stay structure, US Patent 5441319, 1995.
[45] Kariatsumari K., Kobe: Car body energy absorber and bumper stay, US Patent 6481690
B2, 2002.
[46] Sohn S.M., Kim B.J., Park K.S., Moon Y.H.: Evaluation of the crash energy absorption
of hydroformed bumper stays, Journal of Materials Processing Technology, 187–188,
2007, pp. 283–286.
[47] Rimkus W., Bauer H., Mihsein M.J.A.: Design of load-curves for hydroforming
app.lications, Journal of Materials Processing Technology, 108, 1, 2000, pp. 97–105.
[48] Fann K.-J., Hsiao P.-Y.: Optimization of loading conditions for tube hydroforming, Jour-
nal of Materials Processing Technology, 140, 1–3, 2003, pp. 520–524.
[49] Aue-U-Lan Y., Ngaile G., Altan T.: Optimizing tube hydroforming using process simula-
tion and experimental verification, Journal of Materials Processing Technology, 146, 1,
2004, pp. 137–143.
 Automotive component development by means of hydroforming 71

[50] Imaninejad M., Subhash G., Loukus A.: Loading path optimization of tube hydroforming
process, International Journal of Machine Tools and Manufacture, 45, 12–13, 2005, pp.
1504–1514.
[51] Yogo T., Ito M., Mizuno T.: Digital master system of hammering hydroforming, (in Japa-
nese), Journal of the Japan Society for Technology of Plasticity, 45, 527, 2004, pp. 1022–
1025.
[52] Mori K., Patwari A.U., Maki S.: Improvement of formability by oscillation of internal
pressure in pulsating hydroforming of tube, Annals of the CIRP 53, 1, 2004, pp. 215–218.
[53] Mori K., Maeno T., Maki S.: Mechanism of improvement of formability in pulsating
hydroforming of tubes, International Journal of Machine Tools & Manufacture, 47, 2007,
pp. 978–984.
[54] Manabe K., Amino M.: Effects of process parameters and material properties on defor-
mation process in tube hydroforming, Journal of Materials Processing Technology, 123,
2, 2002, pp. 285–291.
[55] Murakawa M., Jin M.: The utility of radially and ultrasonically vibrated dies in the wire
drawing process, Journal of Materials Processing Technology, 113, 1–3, 2001, pp. 81–86.
[56] Mori K., Maeno T., Bakhshi-Jooybari M., Maki S.: Measurement of friction force in free
bulging pulsating hydroforming of tubes, in: P.F. Bariani et al. (Ed.), Advanced Technol-
ogy of Plasticity, Edizioni Progetto Padova, Padova, 2005, pp. 192–200.
[57] Kang H., Lee M. Y., Shon S. M., Moon Y. H.: Forming various shapes of tubular bellows
using a single-step hydroforming process, Journal of Materials Processing Technology,
194, 2007, pp. 1–6.
[58] Erbel S., Kuczyski K.: Metalforming of bellows part 1 (in Polish), Mechanik, Vol. 4,
1962, pp. 180–183.
[ 59] Erbel S., Kuczyski K.: Metalforming of bellows part 2 (in Polish), Mechanik Vol. 5,
1962, pp. 281–285.
[ 60] Morphy G.: Tube Hydroforming Design Flexibility—Part VII: Holes, Exclusive to The
Fabricator.com, February, 2005, available on http://www.thefabricator.com/ Hydroform-
ing/ Hydroforming_Article.cfm?ID=1047.
[61] Hama T., Ohkubo T., Kurisu K., Fujimoto H., Takuda H.: Formability of tube hydroform-
ing under various loading paths, Journal of Materials Processing Technology, 177, 2006,
pp. 676–679.
[62] Jirathearanat S., Hartl Ch., Altan T.: Hydroforming Y-shapes – product and process de-
sign using FEA simulation and experiments, Journal of Materials and Technology, 146,
2004, pp. 124–129.
[63] Kim T.J., Yang D.Y., Han S.S.: Numerical modelling of the multi-stage sheet pair hydro-
forming process, Journal of Materials Processing Technology, 151, 2004, pp. 48–53.
[64] Plancak M., Vollertsen F., Woitschig J.: Analysis, finite element simulation and
experimental investigation of friction in tube hydroforming, Journal of Materials Proc-
essing Technology, 170, 2005, pp. 220–228.
[ 65] Ray P., Mac Donald B.J.: Determination of the optimal load path for tube hydroforming
processes using a fuzzy load control algorithm and finite element analysis, Finite Ele-
ments in Analysis and Design, 41, 2004, pp. 173–192.
[66] Aydemir, A., Vree, J.H.P., Brekelmans W.A.M., Geers M.G.D., Sillekens W.H., Werkho-
ven R.J.: An adaptive simulation approach designed for tube hydroforming processes,
Journal of Materials Processing Technology, 159, 2005, pp. 303–310.
72 A. KOCADA, H. SADOWSKA

[67] Mac Donald B.J., Hashimi M.S.J.: Finite element simulation of bulge forming of a cross-
joint from tubular blank, Journal of Materials Processing Technology, 103, 2000, pp.
333–342.
[ 68] Mac Donald B.J., Hashimi M.S.J.: Tree-dimensional finite element simulation of bulge
forming using a solid bulging medium, Finite Element Analysis and Design, 37, 2001, pp.
107–116.
[69] Zadeh H.K., Mashhadi M.M.: Finite element simulation and experimental in tube hydro-
forming of unequal T shapes, Journal of Materials Processing Technology, 177, 2006, pp.
303–310.
[7 0] Dohmann F., Hartl Ch.: Hydroforming-applications of coherent FE Simulations to the
development of product and processes, Journal of Materials and Technology, 150, 2004,
pp. 18–24.
[71] Hartl Ch.: Theoretical fundamentals of hydroforming, Proceedings of the International
Conference on Hydroforming, Stuttgart, Germany, 12th/13th, October, 1999, pp. 28–37.
[72 ] Vollertsen F.: Challenges and chances of hydroforming of aluminum alloys, in: O. Hahn,
X.-S. Yi (Eds.), Proceedings of the Chinese–German Ultralight Symposium, DVS-
Berichte Band, Vol. 218, 2001, pp. 71–91.
[73] Bauer H., Hartl Ch., Haas A.: Hydroforming-optimized process design with the aid of the
FEA, Proceedings of the Advanced Technology Symposium on Tube Hydroforming, Iron
and Steel Society, Detroit, MI, USA, 20–22 June, 1999, pp. 98–105.
[74] Böhm A., Hartl Ch., Abbey T.: Process and tool technology for hydroforming: case study
and technical and economic considerations, Proceedings of the International Conference
on Tube Hydroforming, Columbus, USA, 13–14, June, 2000, pp. 123–128.
[75 ] Kocada, A., Sadowska, H.: An approach to process limitations in hydroforming of X-
joints as based on formability evaluation, Journal of Materials Processing Tech, Volume:
177, Issue: 1–3, 3th July, 2006, pp. 663–667.
[76] Sadowska H., Kocada A.: Comparison of numerical models for hydroforming of X-
shapes, Computer Methods in Materials Science, Vol. 7, No. 2, 2007, pp. 294–298.

Wykorzystanie ksztatowania hydromechanicznego do rozwoju czci samochodowych

Ksztatowanie hydromechaniczne zaczto coraz czciej stosowa w ostatnich latach

w zwi zku z rosn cymi potrzebami obni
enia wagi ró
nych wyrobów w przemyle lotniczym,
kosmicznym i motoryzacyjnym. Jest to stosunkowo moda technologia w porównaniu z wal-
cowaniem, kuciem, czy toczeniem i st d stosunkowo mao jest dostpnych danych niezbd-
nych do projektowania wyrobów i procesów. W porównaniu z klasycznymi procesami tocze-
nia i spawania, ksztatowanie hydromechaniczne rur (THF) i blach (SHF) przynosi wiele ko-
rzyci: zmniejszenie kosztu wyrobu, kosztu narzdzi, wagi wyrobu, popraw stabilnoci kon-
strukcji, zwikszenie wytrzymaoci i sztywnoci wyrobu, bardziej równomierny rozkad gru-
boci, zmniejszenie iloci operacji dodatkowych, itd. W referacie zaprezentowano rozlege
mo
liwoci rozwoju produkcji wyrobów poprzez zastosowanie procesów ksztatowania hy-
dromechanicznego, ze szczególnym uwzgldnieniem wyrobów dla przemysu motoryzacyj-
nego. Podano tak
e uwagi na temat komputerowego modelowania tych procesów i zjawisk
ograniczaj cych.
 9

Available online at www.sciencedirect.com

ScienceDirect
Procedia Engineering 81 (2014) 1035 – 1042

11th International Conference on Technology of Plasticity, ICTP 2014, 19-24 October 2014,
Nagoya Congress Center, Nagoya, Japan

Blank hydroforming using granular material as medium

- investigations on leakage
Martin Grüner*, Tobias Gnibl, Marion Merklein
Department Mechanical Engineering, Institute of Manufacturing Technology,
Friedrich-Alexander Universität Erlangen-Nürnberg, Egerlandstraße 13, 91058 Erlangen, Germany

Abstract

For high and ultra-high strength steel grades a forming at elevated temperatures is advantageous due to higher forming limits
and reduced flow stresses resulting in low process forces. For highly complex parts out of these materials hydroforming is an
appropriate production method allowing undercuts and homogeneous stretching of the material. In case of forming temperatures
above 350 °C typical liquid forming media no longer can be used and gases are in the focus of interest. The big disadvantage of
blank hydroforming at elevated temperatures results from the increase of friction coefficient by temperature and the need of
sealing the cavity by high contact pressures in the flange area of the part. To overcome this challenge a third class of media,
granular material like sand or small ceramic spheres can be used as forming media. For investigations of the process limits,
especially the sealing limit of granular material, an experimental tool was built up.
© 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
© 2014 The Authors. Published by Elsevier Ltd.
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of Nagoya University and Toyohashi University of Technology
Selection and peer-review under responsibility of the Department of Materials Science and Engineering, Nagoya University
Keywords: Hydroforming; Granular material; Leakage

1. Introduction

The International Energy Agency (IEA, 2013) shows in their statistics a steady increase in CO 2 emissions from
fossil fuel combustion. The Kyoto protocol as the Doha amendment describes defined pledges of CO 2 emission
with about 42%, transport with about 22% and industry with about 21%. For the year 2050 the SRU (2005)

* Corresponding author. Tel.: +49-9131-85-28317; fax: +49-9131-85-28866.

E-mail address: martin.gruener@fau.de

1877-7058 © 2014 Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license
(http://creativecommons.org/licenses/by-nc-nd/3.0/).
Selection and peer-review under responsibility of the Department of Materials Science and Engineering, Nagoya University
doi:10.1016/j.proeng.2014.10.137
1036 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042

predicts fuel consumptions of 1.6 l/100 km for diesel-engined vehicles and 2 l/100 km for petrol-engined vehicles,
respectively. The classic approaches to reduce fuel consumption in automotive industry are increasing efficiency of
engines, improving aerodynamics and reducing the weight. Weight reduction can be reached on the one hand by
using materials with low density or on the other hand by using a low amount of material with high strength.
Typical for the second approach is the InCar (2009) project by ThyssenKrupp or the ULSAB (2002) project of the
American Iron and Steel Industry using high strength and ultra-high strength steel grades. For processing these
steel grades elevated temperatures are often advantageous due to reduction of flow stress and increase in
formability. These elevated temperatures are usually no problem for deep drawing operations but for hydroforming
operations if it is necessary to form highly complex parts. Liquids as a forming medium are often limited to about
350 °C and gases suffer the problem of compressibility and leakage so they are often used in tube hydroforming
processes in which leakage can be controlled well.
The use of a new kind of medium for hydroforming at elevated temperatures up to 600 °C was investigated by
Grüner and Merklein (2011). Small ceramic spheres with diameters smaller than 1 mm as medium show sufficient
temperature stability and seem to show very low leakage tendencies. The typical process limit of leakage of
hydroforming processes does not limit forming operations with this medium. Similar to deep drawing operations
wrinkles occur in the flange area of the drawn parts but no leakage can be observed. This behavior leads to the need
of detailed research of leakage using small ceramic spheres as a forming medium for hydroforming processes.

2. Experimental setup for investigations on leakage

2.1. Used Materials

For all presented experiments ceramic beads typically used for shot peening operations were used. The
presented results base on the coarse sphere diameter distribution with diameters between 600 microns and 850
microns. The material can be purchased under the name “Zirblast B20” and is distributed by SEPR Keramik
GmbH & Co.KG. In as-delivered condition the medium has only a small fraction of fine powder. Under
compression load some of the spheres brake to fine powder and the volume decreases as the density increases. This
behavior is well known and can be described analytically as in numerical simulation as shown by Grüner and
Merklein (2009).

2.2. Experimental tool and experiments

Taking into account that leakage can only occur if the dimension of a gap is higher than the diameter of the
smallest ceramic sphere an experimental tool was constructed which allows investigating two different scenarios of
leakage. The tool is designed for use in a universal testing machine allowing high accuracy in positioning and
having the possibility of measuring force and displacement of the upper tool. In case one a defined gap exists,
similar to a displaced blank holder, and the pressure within the medium increases. For these investigations the
setup shown in Fig. 1(a) is used. Using spacer blocks it is possible to define the gap between the upper tool
cylinder and the lower tool cylinder and to keep it constant throughout the complete experiments. Different heights
of the spacer block allow testing different gap heights.
At the beginning of an experiment upper tool cylinder and lower tool cylinder are not connected. The lower tool
is in its top position and the position of the upper tool cylinder allows filling the lower cylinder with ceramic
spheres. After filling with medium the upper cylinder is positioned on the spacer blocks and the upper tool is
connected to the lower tool by screws. By moving the upper tool with constant velocity of 10 mm/min against the
lower tool the punch pressurizes the ceramic spheres. The experiments are stopped if a pressure of 100 N/mm² is
reached inside the ceramic spheres or in case of leakage if the punch would get into contact with the upper tool.
During the experiments with different gap heights from 0 mm up to 6 mm force-displacement curves are recorded.
In case two a constant force is set for the universal testing machine and an initial closed gap is slowly opened.
For these experiments it is necessary to modify the tool as shown in Fig. 1(b). The spacer blocks are removed and
the screw nut is placed on the screw thread of the punch. A connector allows moving the lower tool cylinder
 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042 1037

upwards and downward by rotating the alignment pins. An additional incremental measuring sensor placed on the
plate of the upper tool cylinder allows recording the gap height during the experiments.

Fig. 1. Experimental tool: configuration for defined constant gaps (a); for defined constant forces (b).

At the beginning of the experiments the screw nut is in its top position and the lower tool cylinder is filled with
ceramic spheres. Afterwards a constant force is applied pressing the upper tool cylinder against the lower tool
cylinder. Now the screw nut is turned and the lower tool cylinder moves downwards. In this situation two different
load paths can occur. In case of path one the applied force is transmitted from the top tool to the lower tool to the
screw nut to the punch. This means that the gap between upper and lower tool cylinder does not increase its height
and keeps closed. In case of path two the force is transmitted from the upper tool to the ceramic spheres to the
punch, the lower tool cylinder is unloaded and the gap between upper and lower tool cylinder increases. If the gap
height gets too big leakage will occur and the medium no longer transmits the load resulting in a reduction of gap
height.

3. Experimental setup for forming experiments

3.1. Experimental tool and experiments

Supplementary to the experiments on leakage parts are formed out of high and ultra-high strength steel at
temperatures up to 600 °C using the described ceramic spheres as forming medium. For the forming experiments
the tool shown in Fig. 2 is used. The lower tool is fixed to the table of a hydraulic press, typically used for
hydroforming, with a ram force of 6300 kN. The punch is connected to the drawing cushion with forces up to
2500 kN. The upper tool is fixed to the ram and additionally the die is connected with the ram cushion with forces
up to 4x400 kN. The die as well as the blankholder can be heated by heating cartridges up to 500 °C. Thermo
couples within die and blankholder allow building up a closed loop control for high temperature accuracy. For
higher forming temperatures the blank is preheated in a furnace and temperature loss during transport is taken into
account. A preheating is also done for temperatures up to 500 °C to get almost isothermal conditions. To prevent
extensive heating of the table and the ram of the hydraulic press water-cooled plates are used, being separated from
the heated tool components by insulation plates.
For the forming experiments two different steel grades are used. On the one hand the complex phase steel
CP800 with a sheet thickness of 1.5 mm and on the other hand the martensite phase steel MS1200 with a sheet
thickness of 1.8 mm is used.
At the beginning of the forming experiments the reservoir within the blankholder is filled with the granular
material and the preheated blank is centered on the blankholder. In a second step the ram moves downwards and
applies the blankholder force to the flange area of the blank. For the forming operation the punch moves upwards
and pressurizes the granular medium which leads to a forming of the blank.
1038 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042

Fig. 2. Experimental tool for forming experiments.

With forming times less than two seconds the process is relatively quick for hydroforming and the temperature
of the blank during the process can be assumed as constant. After forming the blank the tool opens and the
produced components are cooled down outside the tool. The granular material is removed from the tool with a
vacuum cleaner and refurbished by removing fine powder resulting from broken spheres.

4. Results for investigations on leakage

4.1. Constant gap heights

For the experiments with constant gap height gaps of 0 mm, 2 mm, 2.5 mm, 3 mm, 4 mm and 6 mm were
investigated. With a gap height of 0 mm no leakage can occur and only the compressibility of the medium itself is
measured. The results for different gap heights are shown in Fig. 3.
It shows that for a gap height of 0 mm and 2 mm the desired pressure of 100 N/mm² can be reached for the
coarse medium B20. A gap of 2.5 mm allows a pressure of about 90 N/mm² after a punch displacement of 25 mm.
The constant rising of the pressure indicates that with additional punch travel a pressure of 100 N/mm² would be
possible although a small amount of leakage is present. A further increase of the gap height to 3 mm leads to a
significant drop in resulting pressure. For 4 and 6 mm gap height almost no pressurizing of the medium is possible
and the medium flows through the gap.
As a conclusion out of these experiments it can be stated that gaps with a height of up to 300% of the biggest
diameter of the spheres and about 400% of the smallest diameter of the spheres within the diameter distribution can
be sealed by the medium. This means that typical sealing limits for hydroforming processes depending on the
flange contact pressure do not exist if this kind of medium is used but it has to be taken into account that gaps in
the flange area of a tool always mean zero flange contact pressure so wrinkling will occur at the produced parts.
 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042 1039

Fig. 3. Stress-displacements curves for constant gap heights using coarse ceramic spheres as a medium.

4.2. Constant loads

As described in chapter 2.2. two different load cases exist for experiments with constant force and opening gaps.
Fig. 4 shows the typical medium pressure-time curves calculated from the measured values recorded by the load
cell of the universal testing machine. Up to about 180 s the load is transmitted from the upper tool cylinder directly
to the lower tool cylinder resulting in a drop of recorded medium pressure to zero due to rigid tool components and
velocity of the universal testing machine if the lower tool cylinder moves downwards. At a time of about 180 s the
load is transmitted to the medium and the gap starts to open.

Fig. 4. Medium pressure over time for constant force and opening gap.

Fig. 5(a) shows the load path in initial condition if the compression load is transmitted to the tool components.
Due to the high stiffness of the steel components compared to the ceramic spheres an opening of the gap leads to
an immediately drop of measured force to zero. In the second case presented in Fig. 5(b) the load is transferred
from the upper tool to the ceramic spheres to the punch. In this case an increasing gap does not lead to an
immediate drop in pressure. For a drop in pressure granular material has to pass the gap between upper and lower
tool cylinder.
1040 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042

Fig. 5. Load paths for experiments with constant load in initial condition (a), during experiment (b).

Fig. 6 shows the data recorded by the incremental measuring sensor during the same experiment. Up to about
180 s only small gaps up to 0.5 mm can be observed resulting from non-infinite speed of closed loop control of the
universal testing machine. At about 180 s the gap starts to open as the load is transmitted to the medium. After
400 s with a gap height of about 3.5 mm to 4 mm no further strong increase of the gap height can be observed.

Fig. 6. Gap height over time for constant medium pressure and opening gap.

Experiments with different constant medium pressures of 2.5 N/mm², 7.5 N/mm² and 10 N/mm² also show that
gaps with up to at least 3 mm can be sealed. Due to friction between the screw thread of the punch and the screw
nut experiments with constant force and opening gap can only be conducted for pressures up to 10 N/mm2. At
higher forces it is not possible to rotate the screw nut even if a lever is used on the alignment pins.
 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042 1041

5. Results from forming experiments using granular material as a forming medium

In order to check the possibility of the granular material of increasing the process window additional forming
experiments were conducted. The tool concept presented by Grüner and Merklein (2011) was used in which a
punch is pressurizing the medium and blank-holder force is applied by elastic tool deformation of the insulation
plates. As shown in Fig. 7 (a) for the complex phase steel CP800 for low blank-holder forces and a temperature of
400 °C a forming is possible although wrinkling occurs. A forming pressure of 100 N/mm² can be reached and no
leakage could be observed.

Fig. 7. Wrinkles in the flange area of the formed cup (a); fracture at the bottom of the cup (b).

For a temperature of 500 °C and higher blank-holder forces a fracture appears at the transition area from the
bottom of the cup to the radius, as it can be seen in Fig. 7 (b). In a blank hydroforming process using a fluid or a
gas as a forming medium this would lead to a suddenly drop of the forming pressure. In case of a granular material
as a forming medium the fracture cannot be detected during the process, only at the end if the part is taken out of
the tool. Bridging effects and the compression stresses leading to compaction of the medium seal the fracture and
the desired forming pressure is reached.
Especially for the ultra-high strength martensite phase steel MS1200 an increase of forming temperature from
room temperature to 600 °C shows a great improvement in forming and the benefit of granular material as a
forming medium for high temperatures. As shown in Fig. 8 (a) at room temperature for a forming pressure of
70 N/mm² no die contact takes place in the bottom area of the cup whereas for a forming temperature of 600°C,
shown in Fig. 8 (b), the forming is significantly improved.

Fig. 8. Increase in forming for martensite phase steel MS1200: Room temperature (a); 600 °C (b).

6. Summary and outlook

A tool for investigations on the sealing limit of granular material was constructed and successfully built up. The
presented results show that typical flange-contact-pressure dependent sealing limits of sheet metal hydroforming
processes should not be used for granular material used as a forming medium due to the low tendency to leakage of
this material. For the coarse medium with sphere diameter distributions form 600 microns up to 850 microns being
here under investigation even gaps with a height of 2.5 mm can be used without significant leakage. Forming
1042 Martin Grüner et al. / Procedia Engineering 81 (2014) 1035 – 1042

experiments at different temperatures showed that the granular material will seal gaps like wrinkles or fractures by
bridging effects. In this case wrinkling and fracture will give the process limits as in deep drawing processes and
not leakage.
Further experiments will be done with two additional sphere diameter distributions. The first one consists of
spheres with diameters from 63 microns up to 125 microns and the second one of spheres from 250 microns up to
450 microns. With these additional experiments it is expected to find a relation between sphere diameters and gap
height that can be sealed.

Acknowledgements

The presented results were achieved through investigations within the research project ”Warm forming of high
strength steel using granular material as a hydroforming medium“ which is sponsored by the German Research
Foundation (DFG).

References

International Energy Agency: CO2 emissions from fuel combustion – highlights (2013 Edition). OECD/IEA, 2013. Information
from http://www.iea.org
United Nations – Framework Convention on Climate Change. Information from http://unfccc.int/resource/docs/convkp/kpeng.pdf
United Nations – Framework Convention on Climate Change. Information form http://unfccc.int/files/kyoto_protocol
/application/pdf/kp_doha_amendment_english.pdf
Sachverständigenrat für Umweltfragen (SRU): Potentiale und Instrumente zur CO2-Verminderung von PKW – Auszüge aus dem
Sondergutachten Umwelt und Straßenverkehr. 2005. Information from http://www.umweltrat.de/SharedDocs/Downloads
/DE/06_Hintergrundinformationen/2005_08_SG_Strassenverkehr_Auszuege.pdf?__blob=publicationFile
InCar – Der innovative Lösungsbaukasten für die Automobilindustrie. ThyssenKrupp AG. 2009. Information
from http://incar.thyssenkrupp.com/download/Broschueren/InCar_D.pdf
ULSAB-AVC – Advanced Vehicle Concepts, Overview Report. Worldsteel Association. 2002. Information
from http://www.autosteel.org/~/media/Files/Autosteel/Programs/ULSAB-AVC/avc_overview_rpt_complete.pdf
Grüner, M.; Merklein, M.: Influences on the Molding in Hydroforming Using Granular Material as a Medium. In: Chung, K.; Han, H. N.; Huh,
H. Barlat, F.; Lee, M.-G. (Edtrs.): AIP Conference Proceedings Volume 1383 - The 8th International Conference and Workshop on
Numerical Simulation of 3D Sheet Metal Forming Processes (NUMISHEET 2011), American Institute of Physics, 2011, 645-652
SEPR ZirPro Department: Zirblast Ceramic Beads. Information from http://www.zirpro.com/uploadedFiles/SGzirpro/Documents/SGZirPro-
Zirblast-TDS-201204-E.pdf
Grüner, M.; Merklein, M.: 2009, Mechanical Behaviour of Ceramic Beads Used as Medium for Hydroforming at Elevated Temperatures. Key
Engineering Materials, 410-411, 61-68
 10
Combined tube and double sheet hydroforming for the manufacturing of complex parts

M. Merklein, M. Geiger ( I ) , M. Celeghini

Chair of Manufacturing Technology, Institute of Mechanical Engineering
University of Erlangen-Nuremberg, Germany

Abstract
Modern lightweight construction, especially in the automotive industry, requires more and more complex
components, which can be manufactured in one process step using the hydroforming technology. The
combination of the tube and double sheet hydroforming is a new forming process, where a tube and two
blanks are formed simultaneously in a die cavity, combining the advantages of both hydroforming variants.
This paper deals with the fundamental considerations and investigations related to connection between tube
and double sheet. The finite element analysis and laboratory trials were used in order to design the shape of
the die cavity and to avoid wrinkles, material tearing and the collapse of the tube section during forming. The
paper will also illustrate an analytical model for the prediction of the edge shape in the constrained bulging of
a rectangular cup together with several technical solutions, which enabled a complete forming of the
investigated part. Finally, the definition of a hydroforming material factor based on the analytical model of the
hydraulic bulging process enables the right choice of sheets with different material strength and thickness for
the hydroforming of hybrid components

Keywords:
Hydroforming, sheet metal, tube

1 INTRODUCTION
Lightweight design is a key solution for the modern
manufacturing technology, especially in the automotive
industry [ I ] . Load optimized parts are often characterized
by very complex shapes and by an efficient use of different
materials. These essential properties can be achieved with
the hydroforming technology, which uses a fluid under high
pressure to form tubes, profiles, single and double sheets
into a die cavity. This paper shows an innovative process
chain for the manufacturing of complex shaped parts by
combining the tube and double sheet hydroforming
technologies into one single process.
2 INTEGRATED PROCESS CHAIN
2.1 Description of the process steps
In the first step of the developed process chain the lower
sheet and the tube are placed into the tool by an
articulated robot. The upper sheet is then positioned onto
the tube (Figure I-a) and onto special guiding elements
(Figure I-b), who secure a horizontal position of the sheet
during the following tool closing operation.
During the tool closing operation the sheets are preformed
by the tube into the die cavity in the junction area between
tube and double sheet. Then the tube is upset with an axial
force by the docking cylinder (Figure I-b). The
hydroforming fluid flows through a channel in the docking
cylinder and forms the two blanks and the tube against the
die cavity. At the end of the forming process the clinching
punches, which were retracted in apposite holes in the
blank holder, are actively moved and join the two sheets in
six points in the flange (Figure 1-c). Due to the
hydroforming operation a joining by hydroforming between
tube and double sheet is also realized. Finally the
component can be taken out of the press and finished by
additional welding and cutting operations in the adjacent
laser station (Figure I-d).
2.2 Combination of tube and sheet hydroforming
The advantage of combining tube and double sheet
hydroforming lies in the possibility of manufacturing a
complex shaped part with three different materials in just
one process. The material strength and the wall thickness
of the two sheets and tube can be optimized for any
specific application. In addition, the component presents
three characteristic areas. The tube area has no flange
and thus reveals a high stiffness, due to the closed cross
section profile. The sheet area has larger cross sections
and presents a flange, which is required for the assembly
and joining with the adjacent components. The junction
area is essential for the connection between the tube and
the two sheets and is also the most critical area during the
forming process.

Figure 1: Integrated process chain
3 FAILURE MODES
A systematic investigation of the possible failure modes
was performed by finite element analyses using the
commercial program Optris2000 (Esi Group) as well as by
laboratory trials. The combined tube and double sheet
hydroforming presents several failure modes. Some of
these have already been observed in previous
investigations related to sheet or tube hydroforming, while
others are specific for the realized process combination
and are close related to the characteristics of the junction
area.
3.1 Failures in the sheet and tube area
The sheets may burst in the corner areas in case of high
blank holder forces or excessive drawing depth and
irreversible wrinkles can be built in case of convex binder
surfaces [2]. In this case the action of the pressure on the
bent profile leads to compression stresses in the sheet and
finally to an irreversible wrinkling (Figure 2-a).

Figure 3: Tube failure modes

Figure 2: Sheet wrinkling in curved binder areas
Such a convex binder surface is obtained by each sheet in
proximity of the junction area. In case of a circular cross
section of the tube, the bending angle of the sheet is
almost qS= 90". Figure 2-b shows the shape initially
conceived for the junction area with a rounding radius of
8 mm and a bending angle of go", which leads to wrinkles
in the side walls of the rectangular cup.
The tube may fail either by bursting in case of insufficient
axial load or by wrinkling and buckling in case of
insufficient inner pressure, as already observed in
conventional tube hydroforming [3].
3.3 Insufficient forming of the component
An insufficient sealing of the parts during the process leads
to a low maximum pressure especially in the hydro-
calibrating stage. In the preforming stage a nitrogen
accumulator with a high flow rate (9 Vmin) is used to
quickly fill the part with the fluid and form it up to a
maximum pressure of about 20 MPa. But in the hydro-
calibrating stage the fluid is pumped into the part using the
pressure intensifier, which has a much lower flow rate
(2.7 Vmin). Eventual leakages cannot be compensated
anymore by the flow rate and a low maximum pressure is
reached. This leads to an insufficient forming of the corner
areas of the two sheets and of the tube as shown in
Figure 4-a.

3.2 Tube failures in the junction area

Three additional failure modes can be identified in the
junction area during the forming process. The first failure
mode (vertical crushing) occurs during the tool closing
operation (Figure 3-a), as the tube has to form the blanks
into the upper and lower die cavity in the junction area.
Using the same thickness to for the tube and for the sheet,
large tube diameters D (e.g. D/to= 60) lead to an
insufficient stiffness of the tube cross section. The profile
collapses and the tube is crushed between the two binders.
For smaller diameters (e.g. D/to = 50), the profile stiffness
is high enough to enable a preforming of the blanks into
the die.
The second failure mode (buckling) occurs in the tube Figure 4: Sealing problem at the junction area
docking operation, as the hydraulic cylinder presses the The most critical area for the sealing of the parts is the
tube against an opposite circumferential bead in the die junction between the tube and the two sheets. Figure 4-b
(Figure 3-b). This bead is required to obtain a counterforce shows a cross section with the profiles of the dies, of the
necessary to upset the tube during forming. The bead tube and of the two sheets. Due to the necessity of
leads to a reduction of the circumferential length and thus bending the sheets around the tube, for a given die radius
to tangential compression stresses in the tube, which may Rd, the resulting blank radius Rb will result in Rb = Rd + to,
lead to a local buckling of the profile. where to is the sheet thickness. This suggests that a very
The third failure mode (horizontal crushing) occurs during small die radius is necessary in order to minimize the
the sheet hydroforming process (Figure 3-c). The blank dimension of the critical leakage gap. However, even when
draw-in generates tangential compression stresses in the using a die radius equal to zero (sharp edge), still a
flange, which can lead to a lateral crushing of the tube residual leakage gap will remain, which cannot be sealed
profile in the areas close to the binder, as also displayed in by an appropriate contact pressure between the sheets
Figure 2-b. and the tube.
3.4 Insufficient sheet draw-in in the junction area
The circumferential bead in the die cavity close to the
junction area causes a local limitation of the sheet draw-in.
This can result in an excessive stretching of the material
during the forming of the rectangular cup edge close to the
junction area and so in premature tearing of the sheets.
Thus, a safety factor has to be foreseen in designing this
critical edge of the die cavity. Therefore an analytical
model of the edge forming process was built up for the
correct design of the edge rounding radii. In order to obtain
a very simple analytical model, several approximations
concerning the mechanical properties of the material, the
tensile state and the bulge geometry must be considered
[4]. In particular the material strength cf and the thinning
are associated so that throughout deformation the product
oft = T remains constant. Furthermore will be assumed
that T, = T2 = T [4]. Under these conditions the relation
between the inner pressure, the membrane curvature and
the material properties is expressed by the equation 1,
where p, and p2 are the principal radii of curvature.
This equation must be satisfied in every area of the part
during a bulging operation. Figure 5-a shows a perspective
view from the finite element analysis of the constrained
bulging of a rectangular cup together with two cross
section planes. The plane EX2is normal to the short edge
while the plane Ey2 is normal to the long edge of the
rectangular die cavity. Figure 5-b shows a schematic top
view of the rectangular die and of the contact area
between the sheet and the die bottom. The sheet profile in
the edges is assumed to be elliptical with axes a, and b, in
the section EX2 (Figure 5-d) and axes a, and by in the
section EL2(Figure 5-c).
Figure 5: Analysis of the edge forming process
In order to find a correlation between the profiles in the two
sections and the die geometry, the equation 1 will be
applied in the points A,, B,, A, and B,. The curvature of the
membrane in the points A, and A, in the top view will be
assumed equal to the curvature of the inscribed ellipse
with axes a, and b, (Figure 5-b). With these hypotheses
the principle curvature radii can be calculated for the four
considered points in the two section planes EX2and EYT
and P2Bx co
(3)
andP,By +
Substituting equations (2) and (3) into equation (1) results
in the equations (4) and (5).
~ : bx +-=o
Section E ~ 5-- a0
b: a x b i (4)
Section Eyz.
2
Considering that the pressure is the same in each cross
section, equation 6 can be used to correlate the equations
4 and 5 together in a system which can be easily solved in
numerical way.
bx -
Pressure: - - by
-
2 2 (6)
ax ay
Assuming a given die cavity geometry (axes a, and b,) and
an axis of one elliptical profile (e.g. a, in the cross section
€,J, the three residual axes of the profiles can be obtained
by the developed system. Figure 6 shows the sheet profile
in the sections EX2and Ey2,which were determined by the
FE-analysis for a rectangular die with dimensions
300x200~60mm3 at 15.2 MPa inner pressure. Measuring
the dimension of the axis a, = 35.0 mm from the FE-
analysis and putting this value together with the values
a, = 150 mm and b, = 100 mm into the equations, the
system predict the values of the axes b, and a, and b,.
Figure 6: Comparison between FE-analysis and model
The comparison between the profiles from the FE-analysis
and the analytical model in Figure 6 suggests a sufficient
accuracy of the model for the prediction of the edge shape
in the constrained bulging process. The developed
equation system implicates a straight connection between
the dimensions of the die cavity and the elliptical profiles in
the cross sections along the short and long edge of the
part. In particular the relations confirm that the rounding
radii of the short edge are bigger than those in the long
edge and that this difference increases when the die cavity
geometry becomes more elongated. The difference in the
rounding radii in the short and in the long edge can be also
detected in the incomplete forming of the sheet shown in
Figure 4-a.
The developed analytical model was used in the tool
design phase for the definition of the rounding radius for
the cavity edge close to the junction area. Using the cavity
dimensions (ao= 110 mm and a. = 78 mm) and targeting a
rounding radius in the short edge of 25 mm (b, = 25 mm),
the model predicted an elliptical profile in the critical long
edge with b y = 20.63 mm and a y = 19.75 mm. An
overdimensioned constant rounding radius of 25 mm in the
long edges enables a sufficient safety factor and secures
good forming results even in case of a very low sheet
draw-i n.
4 DEVELOPED TECHNICAL SOLUTIONS
In order to avoid the described failure modes several
additional technical solutions were developed. Numerical
investigations showed that the sheet wrinkling could be
avoided by reducing the bending angle to 4 = 60" while
increasing the die radius to R d = 12 mm, as shown in
Figure 7.
Figure 7: Optimized shape of the junction area
But earlier investigations also showed that a sufficiently
high blank holder force in the first forming stage can help in
preventing the wrinkle building [5]. Furthermore a high
blank holder force reduces the blank draw-in and can help
in avoiding the lateral crushing of the tube. For this reason,
further numerical simulations were performed using a
higher blank holder force with a third configuration of the
junction area (Rd= 12 mm and 4 = 90"). This solution
enabled a wrinkle free forming of the part and was adopted
in the final tool design.
The tube failures were avoided by placing a steel metal
ring of 5 mm thickness in the tube before the tool closing
operation (Figure 8-a). This solution enables a local
reinforcement of the cross section with high stiffness and
prevents the collapse of the tube profile during the whole
process.
Figure 8: Local tube reinforcement and sealing solutions
In order to avoid the leakages in the junction area, two
additional sealing solutions were developed. The first
solution uses a sealing profile placed at the outer surface
of the tube, which fills the critical leakage area assuring the
tightness up to medium internal pressures (Figure 8-a).
The second solution was developed for improving the
sealing also at higher internal pressures. A sealing medium
made of solid particles (e.g. plastic fibers) is mixed to the
hydroforming medium during the forming and stops the
leakage as it occurs by filling the gap (Figure 8-b) [6].
Basic investigations with laboratory tools enabled the
identification of the most effective type and concentration
of the sealing medium, which was then adopted in the
forming trials [7].
In particular double sheet hydroforming tests were
performed using a constant blank holder force in the
preforming phase (fBH) up to the tightness limit. Then the
sheets were hydro calibrated by using a very high blank
holder force (5000 kN). The material was stretched until
the bursting failure occurred in proximity of the bottom
edge. The comparison between the two process windows
in Figure 9 shows the effect of the sealing medium on the
double sheet hydroforming process (material FeP04 -
to = 1.5 mm; cup diameter 150 mm and height 60 mm).
The process window displaces towards higher maximum
pressure (+loo% ca.) and lower values of the optimal
blank holder force (-15% ca.).
Figure 9: Displacement of the process window with sealing
medium
The described tool design, sealing solutions and process
strategies allowed the forming and joining of tube-double
sheet components using a mild steel for the sheet (FeP04
- t o = 1.0 mm) and both an IF260 (to = 1.2 mm) and a
DP600 (to = 1.0 mm) steel for the tubular blank. Figure 9
shows a part formed up to a maximum pressure of
29.2 MPa and joined by six clinching points in the flange
area and by hydroforming in the junction area.
Process parameters:
- Calibrating pressure pmax = 29.2 MPa
- Tool closing force Fmax = 4500 kN
Figure 9: Tube-double sheet component
5 HYDROFORMING OF HYBRID COMPONENTS
As already mentioned, one of the main advantages of the
tube-double sheet components is the combination of
several materials, which differ in mechanical properties
and sheet thickness. In this way hybrid components can be
manufactured in order to enhance the functionality by
minimizing the weight of the part.
Figure 10 shows some double sheet hydroforming tests
performed with a complex shaped part using different
sheet types. The results show that not only the bursting
pressure (pmx), but also the location of the failure is
dependent on the material combination used. In
Figure 10-a tearing occurs in the lower sheet in the front
area of the part while in Figure 10-b, using a different
material combination, tearing occurs in the upper half in
the corner at the backside of the component, which is
characterized by the highest drawing depth. This
phenomenon can be interpreted by analyzing the forming
behaviour of the different sheets used.
Figure 10: double sheet hybrid components
5.1 Hydroforming Material Factor HMF
For this purpose a factor quantifying the pressure
necessary for a given forming result in defined tool
geometry was developed using the analytical modeling of
the bulging test [7, 81. The pressure p o 5 required to obtain
a pole height (h) half of the inner tool radius (R) in a
hydraulic bulging process follows the hyperbolic law written
in equation (7) with the constant H M f (Figure 11).
Figure 11: Hydraulic bulge test model
This constant H M f is defined as Hydroforming Material
Factor and for the given tool geometry is only dependent
on the sheet material properties as written in equation 8. R
is the die opening radius, to is the initial sheet thickness, K
and n are the Hollomon approximation coefficients and r is
the mean anisotropy.
HMF
P0.5 =R (7)
The H M f is directly related to the inner pressure
necessary to obtain in a hydraulic bulge process a dome
height equal to the half of the die radius. The higher the
H M f value of the sheet, the higher will be the necessary
pressure in the process. Equation 7 shows that the
dimension of the H M f is expressed by [Pressure] x
[Length]. Thus using MPa for the pressure and mm for the
die diameter the H M f values are reported in MPa.mm.
The H M f can be calculated directly from standard material
parameters obtained in uniaxial tensile tests. It enables the
classification of the sheets for hydroforming processes in
an easy and reproducible way according to DIN EN 10002.
The H M f values of several materials (mild steel, high
strength steel, aluminum) with different sheet thicknesses
were first calculated from equation 8 using the mechanical
properties from uniaxial tensile tests. Table 1 shows the
obtained H M f values.
I I Mild steel I Hiah strenath steel I Aluminum I
Table 1: calculated H M f values for different materials
In order to verify the analytical model hydraulic bulge tests
were performed in laboratory trials using the same sheet
materials and measuring the dome height during the test
with a tactile sensor. The die opening radius was
R = 75 mm while the clamping force was 5000 kN. The
inner pressure values were also recorded and
synchronized with the measured dome height values. By
this way it was possible to determine experimentally the
pressure po5, necessary to get the specific dome height
h = 0.5 x R = 37.5 mm.
Equation 7 suggests that the product between this
pressure and the die radius R can also deliver an
experimental value of the H M f . Figure 11 shows the
comparison between the calculated and experimentally
determined H M f values. This comparison indicates a
sufficient quality of the developed model for the calculation
of the H M f values.
Figure 11: Comparison between calculated and
experimentally determined H M f values
The diagram illustrates also how both the mechanical
strength and the initial sheet thickness influence the H M f .
Sheet materials with same mechanical properties but
different thicknesses have different H M f values (e.g.
DP450 in 0.98, 1.4, 1.85 mm thickness). On the other side
a material with a lower strength but higher thickness can
have the same HMF value as another one with higher
strength but lower thickness. This is the case of the two
materials AA5754 - 2.54 mm and FeP04 - 1.5 mm which
have similar H M f s around 800 MPa.mm.
5.2 Influence of the sheet material on the process
parameters
In order to analyze the influence of the sheet material on
the process parameter during the hydroforming process,
several hydraulic cup drawing tests were performed with
the considered materials. The initial blank diameter was
250 mm, the cup diameter 150 mm and the cup height
60 mm. The sheets were hydroformed using a constant
blank holder force in the preforming operation ( f B H p r e ) up
to the tightness limit. Then the sheets were clamped by a
high blank holder force (5000 kN) and hydro calibrated up
to the bursting pressure pmaFFigure 12 shows the process
window curves obtained by plotting the bursting pressure
against the preforming blank holder force for the
investigated materials.
Figure 12: Displacement of the process window with
different sheet materials
The diagram indicates that the choice of the sheet material
has an eminent influence on the dimension and location of
the process window. With increasing H M f values the
windows displace towards higher blank holder forces and
higher bursting pressures. In addition the dimension of the
window increases, so that a sort of scaling of the process
window with the H M f can be observed. The curves of
sheet materials with similar H M f values (e.g. AA5754 -
2.54 mm and FeP04 - 1.5 mm) are located very close to
each other. For every sheet material an optimal preforming
blank holder force ( f B H O p f ) can be detected, which allows a
maximum of the bursting pressure (pmaxOpJ.
In order to show more directly the dependency between
the process parameters and the H M f , for each curve the
maximum bursting pressure pmaxOpfand the optimal
preforming blank holder force are plotted versus the H M f
values of the different material in Figure 13.
The diagram demonstrates the linear dependency between
the process parameters and the H M f . With increasing
HMf-values, higher pressure and blank holder forces are
needed for forming of the component. The calculation of
the H M f allows prediction of the optimal blank holder force
within preforming operation and prediction of the maximum
bursting pressure. Furthermore an optimal material
combination in double sheet hydroforming can be obtained
by choosing two sheets with similar HMF values.
Figure 13: Correlation between HMF and process
parameters
6 SUMMARY
The described investigations showed the possibility of
manufacturing complex components by combining tube
and double sheet hydroforming into one single process.
For this purpose several technical solution were needed to
avoid failures in the sheet, in the tube and in the junction
area. In particular the design of the die cavity in the
junction area is necessary together with special
reinforcement and sealing solutions for the successful
forming of the part. An analytical model for the forming of
the edges in a rectangular die was adopted for the design
of the die rounding radius in the junction area. In case of
different thickness and strength of the two sheets a correct
choice of the two materials can be obtained by using the
presented Hydroforming Material Factor H M f .
7 ACKNOWLEDGEMENT
The presented work is supported by the German Research
Foundation within the Collaborative Research Center 396
"Robust, shortened process sequences for sheet light-
weight parts" (project C3).
8 REFERENCES
Kleiner, M., Geiger, M., Klaus, A,, Manufacturing of
Lightweight Components by Metal Forming, Annals of
the ClRP 52/2, 2003, 521-542.
S. Bobbert: Process limits for the hydroforming of
sheet metal pairs, 6th Int. Conf. on Advanced
Technology of Plasticity, 19./24.09.1999, Nuremberg
- Germany, 1261-1266.
Schmoeckel, D., Geiger, M., Hielscher, C., Huber, R.:
Metal forming of tubes and sheets with liquid and
other flexible media, Annals of the ClRP 48/2,
497-514.
Marciniak, Z. Duncan, J. L.: The mechanics of sheet
metal forming. London - United-Kingdom, Edward
Arnold, 1992.
Dehghani, M.M., Jinag, W., Rasty, J.: An
investigation of hydroforming of sheet metals with
varying blank holding loads. Computational Methods
in Materials Processing Technology, MD-Vol 39 /
PED-Vol 61, ASME, 1992, 87-96.
M. Celeghini, Verfahren zur wirkmedienbasierten
Umformung von Blech-Werkstucken und Wirkmedium
zur Verwendung bei diesem Verfahren. Patents
DE19959769C2 and EP1270106Al.
M. Celeghini, Wirkmedienbasierte Blechumformung:
Grundlagenuntersuchungen zum Einfluss von
Werkstoff und Bauteilgeometrie. PhD Dissertation,
University Erlangen-Nuremberg, Bamberg, (2004).
Geiger, M.; Celeghini, M.: Double Sheet
Hydroforming of Complex Hollow Parts. Proc. of the
7th ICTP, Yokohama, 2002, 991-996.
 11
Materials and Design 32 (2011) 2650–2662

Contents lists available at ScienceDirect

Materials and Design

journal homepage: www.elsevier.com/locate/matdes

Comparative investigations on numerical modeling for warm hydroforming

of AA5754-O aluminum sheet alloy
Hasan Gedikli a,b, Ömer Necati Cora a, Muammer Koç a,c,⇑
a
NSF I/UCR Center for Precision Forming, Virginia Commonwealth University, Richmond, VA 23284, USA
b
Karadeniz Technical University, Department of Mechanical Engineering, Trabzon 61080, Turkey
c _
Department of Industrial Engineering, Istanbul _
Sßehir University, Istanbul, Turkey

a r t i c l e i n f o a b s t r a c t

Article history: This study aimed to determine the proper combinations of numerical modeling conditions (e.g. solver,
Received 20 August 2010 element type, material model) for warm hydroforming of AA5754-O aluminum alloy sheets. Assessment
Accepted 13 January 2011 of finite element analyses (FEA) is based on comparison of numerical results and experimental measure-
Available online 18 January 2011
ments obtained from closed-die forming, hydraulic bulge and tensile tests at different temperature (25–
300 °C) and strain rate (0.0013–0.013 1/sec) levels. Thinning (% t) and cavity filling ratios (CFR) on the
Keywords: formed parts were taken as comparison parameters. Several numerical analyses employing different ele-
A. Non-ferrous metal and alloys
ment types, solution methods and material models were performed using the commercially available FEA
B. Film and sheet
C. Forming
package LS-Dyna to determine the best combination of modeling options to simulate the actual warm
hydroforming operation as accurately as possible. Analyses showed that relatively better predictions
were obtained using isotropic material model, shell elements and implicit solution technique when com-
pared with experimental results.
Ó 2011 Elsevier Ltd. All rights reserved.

1. Introduction or using new forming technologies such as hydroforming [10]. It

Lightweight materials, in particular, aluminum alloys have been
widely used in automative and aircraft industry as the high
strength-to-weight ratio of aluminum results in significant weight
and fuel savings [1–4]. In addition to weight reduction, utilization
of aluminum offers some other advantages such as better corrosion
resistance, higher recyclability potential and increased energy
absorption during a crash situation. 5XXX alloys, in particular, have
the highest formability, and are used in automotive inner panels
[5]. Automotive industry has a special interest in AA5754 because
of its high ductility, lightweight, strength and weldability
properties [6]. However, because of their susceptibility to micro-
structural damage, aluminum alloy sheets generally exhibits a
lower level of formability compared to typical sheet steels [7].
Furthermore, utilization of aluminum alloys in the automotive
industry has been far behind of steel because of cost and formabil-
ity issues at room temperature [8]. On the other hand, this alloy
presents some problems such as surface roughening during
deformation, yield point phenomena, and the Portevin–Le
Chatelier (PLC) effect [9]. Therefore, innovations are imposed to
achieve higher formability of aluminum including the attempts
of increased forming temperature [10–15], heat treating [16,17]
⇑ Corresponding author at: NSF I/UCR Center for Precision Forming, Virginia
Commonwealth University, Richmond, VA, 23284, USA. Tel.: +1 804 827 7029.
E-mail address: mkoc@vcu.edu (M. Koç).
was reported that limiting drawing ratio of AA5754 aluminum al-
loy cup can be increased from 1.9 (at room temperature) to 2.7
when the forming die is heated to 250 °C [5].
Quite a few studies investigating the proposed solutions for im-
proved formability of AA 5754 are available in literature. Multiaxial
as well as uniaxial tests were performed with AA5754-O by
Iadicola et al. to better predict the AA5754-O deformation behavior
[18]. Similarly, Mahabunphachai et al. conducted a series of tensile
(at strain rate of 0.0083/s) and hydraulic bulge tests (at strain rates
of 0.0013/s and 0.013/s) for AA5754-O sheet blanks at different
temperatures ranging from 23 °C to 260 °C [19]. As a general obser-
vation, significantly improved formability beyond 200 °C, and at
low strain rates were reported [17–19].
Numerical analysis, especially the finite element method (FEM),
has been extensively used in automotive design and forming pro-
cesses to accurately predict deformation mechanics. It is vitally
important for understanding, and forecasting the complex defor-
mation behaviors that take place during sheet forming processes.
For example, Ahmed and Hashmi modeled the hydraulic bulging
process with combined pressure and in-plane compressive loads
on the sheet-plate by finite element method. They used elastic, lin-
early plastic (bi-linear) isotropic material models with 2D 4-node
quadrilateral elements that allow large-deformation and large-
strain analysis [20]. Wowk investigated strain rate sensitivity of
AA5754 sheets experimentally employing very wide range of strain
rates (0.001/s–1500/s), and then numerically implemented these

0261-3069/$ - see front matter Ó 2011 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2011.01.025
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662 2651

data into rate-sensitive Voce material model available in Ls-Dyna Table 1

[6]. Likewise, Smerd experimentally studied the deformation Test types and conditions for determination of mechanical properties of AA5754-O
[19].
behavior of aluminum alloy sheets at high strain rates (between
600/s and 1500/s), and temperature levels of 25–300 °C using a Material Temperature Material Constitutive equations Lankford’s
tensile split Hopkinson bar apparatus [21]. He used the Johnson– testing (°C) model coefficients

Cook and Zerilli–Armstrong constitutive material models for Tensile Room Power law r ¼ 482:74  e0:3215
numerical simulations of different aluminum alloys. The Zerilli– test 150 Power law r ¼ 346:33  e0:2355

r 0 = 0.656
(TT) 260 Power law r ¼ 174:76  e0:1209 
r 45 = 0.644
Armstrong model was noted to be suitable for a wide range of 
Bulge 150 Strain rate r ¼ 435:59  e0:3127  e_ 0:0422 r90 = 0.752
FCC materials, but it underestimates the strength of BCC materials,
test power law
and is only valid for high strain rates (104/s–106/s) and relatively (BT) 260 Strain rate r ¼ 474:71  e0:1844  e_ 0:1766
low temperatures [6]. power law
Reliability of finite element analyses for forming processes is
highly affected by variety of factors including accuracy of geometric
and material models, reasonability of assumptions and simplifica-
tions, element type and size, solution algorithm. Anisotropic
material models, which include temperature variation effects, have Table 2
been widely used for more realistic modeling of warm forming Barlat YLD2000 material model anisotropy coefficients for AA5754 [15].
processes of lightweight alloys. Barlat et al. developed a series of
Anisotropic coefficients for YLD2000-2d Temperature
constitutive equations for anisotropy, and implemented those in
150 °C 260 °C
past 20 years such as 3-parameter Barlat Yld89 [22], Yld91 [23],
Yld96 [24], YLD2000-2d [25], Yld2004 models [26]. After him, a1 0.9471 0.9814
several researchers implemented these models in their numerical a2 1.0863 1.0423
a3 0.9085 0.9321
modeling studies. Abedrabbo et al. used Barlat’s Yld96 temperature- a4 0.9955 0.9813
dependant anisotropic material model for thermo-mechanical a5 1.0009 0.9933
coupled FEA of AA3003-H111 aluminum alloy sheet forming a6 0.9826 0.9759
[13,14]. Similarly, in another work by the same authors, Barlat’s a7 1.2016 2.4780
a8 1.1459 1.0181
Yld2000-2d was used to simulate forming of AA5754-O and

B SECTION A-A

A
A SECTION B-B

(a)

(b)
Fig. 1. (a) Shape and dimensions of non-axisymmetric die used in closed-die warm hydro-forming experiments (dimensions in mm), and (b) sample part with thickness and
CFR measurement profiles and locations.
2652 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662

AA5182-O aluminum alloys [27–29]. Korkolis used Yld2000-2d Bar-
lat model in FEA analysis of forming of aluminum tubes [30]. How-
ever, thus far, effects of material models, element types and
solution methods on the numerical analysis for the warm hydro-
forming of aluminum alloy AA5754 have not been investigated in
detail.
In this study, first, a series of closed-die warm hydroforming
experiments were performed to determine the effect of process
parameters, such as temperature, pressure, and pressure rate on
the forming limits of the AA 5754-O alloy. Then, a set of finite ele-
ment analyses were performed using different combinations of ele-
ment types (solid vs. shell elements), solution procedures (explicit vs.
implicit), and material models (isotropic strain rate power law, aniso-
tropic 3-parameter Barlat, and anisotropic Yld2000-2d). Comparison
of FEA predictions with experimental measurements (such as cav-
ity filling ratio-CFR and thinning-% t) were used to verify the accu-
racy of the numerical models and to reveal the best combination of
modeling variables. Next section introduces the experimental
conditions while the third section presents the FEA results of
AA5754 warm hydroforming process. Results and discussions are
followed by conclusions in section four and five, respectively.
2. Closed-die warm hydroforming experiments
This study was intended to investigate the limits of warm
hydroforming process and its proper simulation methods that will
further help determination of the optimal process conditions and
understanding of the forming behavior for different materials of
interest under this forming process. To this goal, a set of closed-
die hydroforming experiments were conducted in order to study
the effects of pressure and temperature on the forming limits of
the AA5754-O. The pressure and temperature levels were selected
as 20–30 MPa and 150 °C, and 260 °C, respectively. The pressure
rate was fixed to be 0.22 MPa/s for all tests. Each condition was
tested at least three times to address the repeatability. The die/part
geometry of warm hydroforming experiments is shown in Fig. 1.
After the tests performed, a 3D optical photogrammetric sys-
tem, ARAMIS was employed to capture the final profiles/shapes
of the hydroformed parts to obtain the coordinate information on
the part surface. Cavity filling ratio (CFR) was calculated using
the coordinate information distribution along different contours
on the part (Profile A–A, short side and Profile B–B, long side as
shown in Fig. 1a). These profiles were selected along and across
the rolling direction so that effect of anisotropy can be revealed.
Along the same contours, thickness (t) measurements were per-
formed on 10 different locations. At each location, at least six thick-
ness measurements were performed (Fig. 1b).

3. Numerical analyses

Numerical analyses were performed using the commercial ex-

Fig. 2. (a) Finite element model of closed-die warm hydroforming, and (b) FE mesh
of blank at the final stage of process.
plicit and implicit finite element code Ls-Dyna. Necessary mechan-
ical properties for modeling of AA5754-O sheet metals were
obtained from authors’ previous study as well as from literature
as given in Table 1, and Table 2 based on the well agreement of
the flow curves obtained for current study and the others available
in literature [12,19].
To take advantage the symmetry of the part, a quarter-model of
the closed-die warm hydroforming setup, as shown in Fig. 2a, was
considered in the FEA model to reduce the simulation time and in-
crease the model accuracy using relatively higher number of ele-
ments. The sheet blank with an initial thickness of 1 mm was
modeled as an elasto-plastic material while the upper and lower
dies were defined as rigid bodies using 3D shell elements. Two
different element types, namely 3D shell and 3D solid elements,

Table 3
Details of numerical simulations performed.

Sim. Gr. No Test conditions Material models used Element type Solution procedure Flow curve data
Anisotropic Isotropic
3-Parameter Barlat Barlat’s YLD2000 Strain rate power law Shell Solid Explicit Implicit Tensile test Bulge test
1 (a) 150 °C & 20 MPa U U U U
2 U U U U
3 U U U U
4 (b) 150 °C & 30 MPa U U U U
5 U U U U
6 U U U U
7 (c) 260 °C & 20 MPa U U U U
8 U U U U
9 U U U U
10 (d) 260 °C & 30 MPa U U U U
11 U U U U
12 U U U U
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662 2653

Fig. 3. Hydroformed specimens at temperature 150 °C and pressure of 20 MPa: (a) experimental, and (b) FEA.

Fig. 4. Hydroformed specimens at temperature 150 °C and pressure of 30 MPa: (a) experimental, and (b) FEA.

Fig. 5. Hydroformed specimens at temperature 260 °C and pressure of 20 MPa: (a) experimental, and (b) FEA.

were used in simulations. In the simulations where shell elements
employed, shell element type 16 with four nodes, and full integra-
tion feature was utilized. On the other hand, simulations per-
formed using 3D solid elements utilized type 2 elements with
eight nodes which had full integration and selective-reduced capa-
bility. Sheet blank was modeled with 225 shell element at the
beginning of the simulation whereas, with adaptive meshing fea-
ture, the number of elements went up to as high as 5000 elements
as the forming progressed which can be seen in Fig. 2b. For the
simulations in which 3D hexagonal solid elements were used, con-
stant number of 3600 elements were used since adaptive reme-
shing feature was not available with this element type. Adaptive
2654 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662

Fig. 6. Hydroformed specimens at temperature 260 °C and pressure of 30 MPa: (a) experimental, and (b) FEA result.

Fig. 7. Effect of pressure on thickness change for elements S7444 and S7728 as indicated in the figure (for TT-Shell-Implicit-30 MPa condition).

Fig. 8. Effect of temperature on thickness distribution along Profile A (for TT-Shell-
Implicit-30 MPa condition).
mesh parameters were based on multiple criteria and the time
interval between adaptive refinements was set to 0.01 s. Adaptive
error tolerance for total angle change in degrees relative to the sur-
rounding element for each element to be refined was selected as
ºC)
Fig. 9. Effect of temperature on cavity filling ratio along Profile A (numerical
solution was obtained for TT-Shell-Implicit-30 MPa condition).
2°. The other adaptive remeshing criterion selected was the abso-
lute minimum shell thickness, and pre-defined as 0.9 mm.
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662 2655

(a)

(b)
Fig. 10. Effects of solver, material and element type on cavity filling for profile A at 150 °C and 30 MPa forming conditions: (a) die shape based filling comparisons, and (b)
cavity filling ratio comparisons.

Mechanical properties for AA 5754 sheet blank such as Young’s
modulus, Poisson’s ratio and density were entered into model as
68 GPa, 0.33 and 2.66 g/cm3, respectively.
The contact conditions in both upper die-blank and lower die-
blank pairs were imposed as ‘‘forming one way surface to surface’’
contact. Coulomb friction model with a coefficient of friction of
0.1 was employed for all contacting surfaces after testing the effect
different of friction coefficients varying between 0.01 and 0.5. Sym-
metrical boundary conditions were imposed to perpendicular
edges at the periphery of the model. To simulate the blank holder
force, the sheet blank was subjected to clamping force of 1000 kN
in between lower and upper dies. The sheet blank was also ex-
posed to maximum hydroforming pressure of 30 MPa with a
0.22 MPa/s pressure rate.
Since the element type, material model and solution method af-
fect several response factors such as accuracy, solution time, con-
vergence of the numerical approximation; different combination
modeling variables need to be tested to have a well-established
model [31–34]. Numerical simulations were conducted using 3D
shell and 3D solid elements; explicit and implicit solution tech-
niques; isotropic strain rate power law material model, and aniso-
tropic material models such as three-parameter Barlat (3P-Barlat),
2656 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662
(a)
(b)
Fig. 11. Effects of solver, material and element type on cavity filling for profile B at 150 °C and 30 MPa forming conditions: (a) die shape based filling comparisons, and (b)
cavity filling ratio comparisons.
and Barlat Yld2000-2d (YLD2000) at different four test conditions:
(a) 150 °C/20 MPa, (b) 150 °C/30 MPa, (c) 260 °C/20 MPa, and (d)
260 °C/30 MPa. A total number of 48 simulations, as presented in
Table 3, were performed using an Intel Core i7 2.8 GHz CPU with
8 GB of RAM. Prior to these simulations, a non-isothermal FEA
was carried out to find out the temperature variation on the sheet
blank and found that sheet blanks can be modeled as equally
heated (isothermal condition), and hence, structural-only type of
finite element analyses were conducted.
4. Results and discussion
The experimental and numerical results of formed parts from
the closed-die hydroforming tests are presented in Figs. 3–6 for
150 °C/20 MPa, 150 °C/30 MPa, 260 °C/20 MPa, 260 °C/30 MPa con-
ditions, respectively.
4.1. Effects of forming pressure and temperature on sheet
hydroformability
Effects of pressure and temperature on the formability charac-
teristics of AA 5754-O sheet blanks were investigated experimen-
tally and numerically using two different responses: (a)
distribution of thinning in the A-A (Profile A) and B–B (Profile B)
sections (as described in Fig. 1), and (b) cavity filling ratio (CFR).
In addition, strain distribution and the highest strain levels were
determined from the finite element analyses as can be seen from
Figs. 3–6. It was noted that both pressure and temperature have
significant effect on the strain levels experienced. Maximum strain
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662 2657

Die Profile A
Experimental
BT-Shell-Imp-No Springback
BT-Shell-Imp-Springback

Fig. 12. Cavity filling for Profile A after springback analysis (for 150 °C and 30 MPa conditions).

Die Profile B
Experimental
BT-Shell-Imp-No Springback
BT-Shell-Imp-Springback

Fig. 13. Cavity filling for Profile B after springback analysis (for 150 °C and 30 MPa conditions).

35
Experimental 8
BT-Solid-Explicit 6
30 BT-Solid-Implicit
TT-Solid-Explicit
TT-Solid-Implicit
10
25
4
Thinning (%)

20 3
2
1
15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile B - Distance from Center (mm)

Fig. 14. Thinning distribution obtained using solid elements for the profile B at 260 °C and 30 MPa.

level was acquired as 0.29 for 150 °C/20 MPa case while this value 30 MPa forming conditions, correspondingly. Thinning variations
went up to 0.33 for 150 °C/30 MPa case. Similarly, strain levels of with respect to pressure were presented for two elements that
0.37 and 0.40 were obtained for 260 °C/20 MPa and 260 °C/ attained the maximum thinning values on the Profile A (S7444)
2658 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662

35
Experimental
BT-Shell-Explicit
TT-Shell-Explicit
30 YLD2000-Shell-Explicit
3P-Barlat-Shell-Explicit

25

Thinning (%)
20

15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile A - Distance from Center (mm)

(a)
35
Experimental
BT-Shell-Explicit
30 TT-Shell-Explicit
YLD2000-Shell-Explicit
3P-Barlat-Shell-Explicit
25
Thinning (%)

20

15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile B - Distance from Center (mm)
(b)
Fig. 15. Thinning distribution obtained with shell elements using explicit solver for: (a) Profile A at 260 °C–30 MPa, (b) Profile B at 260 °C–30 MPa conditions.

and Profile B (S7728) contours as presented in Fig. 7. Numerical
simulations were conducted using implicit solver and isotropic
material model based on tensile test data (Profile A, TT150 and Pro-
file B, TT150) and bulge test data (Profile A, BT260 and Profile B,
BT260). Results showed that thickness values of the elements
S7444 and S7728 rapidly decreased with the increasing pressure
up to 20 MPa, and then it was stabilized. In both temperature lev-
els, the thickness along the Profile A declined relatively faster than
that of along Profile B up to 5 MPa. After 5 MPa level, the thickness
along Profile B decreased faster than that for along Profile A.
Fig. 8 shows the thickness distribution of the formed parts at
different temperature levels (for TT-Shell-Implicit-30 MPa condi-
tion). As expected and observed from the experimental results,
the CFR of AA 5754-O sheet alloy is increased with increasing tem-
perature as shown in Fig. 9.
4.2. Effects of solver type, element type and material data source in
FEA predictions
To analyze the effect of solver, element, and material data
source (tensile or bulge test) on the prediction accuracy of simula-
tions, both cavity filling ratios (CFR) and part thickness values from
experimental measurements and FE analyses were compared.
Experimental cavity filling ratio (CFR) (cavity filling area/die
profile area) for the Profile A and Profile B were measured as 90%
and 93%, respectively for the warm hydroforming condition of
30 MPa/150 °C. Figs. 10 and 11 illustrate these comparisons. Max-
imum differences between numerical and experimental results
were obtained as around 9% on the Profile A, and 6% on the Profile
B. Comparisons were based on the 2D (filled area) measurements
and grouped for explicit and implicit analyses. Figs. 10 and 11
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662 2659

35
Experimental
BT-Shell-Implicit
30 TT-Shell-Implicit
YLD2000-Shell-Implicit
3P-Barlat-Shell-Implicit
25

Thinning (%) 20

15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile A - Distance from Center (mm)
(a)
35
Experimental
BT-Shell-Implicit
30 TT-Shell-Implicit
YLD2000-Shell-Implicit
3P-Barlat-Shell-Implicit
25
Thinning (%)

20

15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile B - Distance from Center (mm)
(b)
Fig. 16. Thinning distribution obtained with shell elements using implicit solver for: (a) Profile A at 150 °C/30 MPa, (b) Profile B at 150 °C/30 MPa, (c) Profile A at 260 °C/
30 MPa, (d) Profile B at 260 °C/30 MPa conditions.

revealed that, in all simulations, the explicit analyses yielded high-
er cavity filling ratios compared to the implicit analyses. This can
be explained by the fact that explicit analysis results deviate from
the exact solutions as the certain critical time step size is surpassed
[31].
It was also noticed from Figs. 10 and 11 that the numerical anal-
yses yielded higher cavity filling ratios when compared to the
experimental values, in general. This was assumed to be resulted
from negligence of springback in the analyses that commonly
experienced in sheet metal forming. It was reported that elevated
springback related issues are experienced in aluminum sheet use
compared to steels [35,36], and these issues are diminished or less
experienced with increasing forming temperature, advanced form-
ing techniques techniques such as hydroforming and superplastic
forming [37].
Therefore, a set of simulations in which springback effect in-
cluded were performed for some of the simulation cases (with flow
curves obtained from bulge tests using shell elements and implicit
solution techniques) as presented in Figs. 12 and 13. It was ob-
served that simulations taking springback into account yielded
slightly lower cavity filling ratios and were considered more reli-
able than non-springback type analyses. It was also concluded that
shell elements led to higher filling ratios than solid elements owing
to their higher number of elements and stiffness matrix
differences.
In order to analyze the effects of element type and solution pro-
cedures on simulation results and accuracy, thinning of the actual
hydroformed parts were compared with the simulation results.
Numerical and experimental thinning comparisons are shown in
Figs. 14–16 at different test conditions for Profile A or B.
2660 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662

35
Experimental
BT-Shell-Implicit
TT-Shell-Implicit
30 YLD2000-Shell-Implicit
3P-Barlat-Shell-Implicit

25

20
Thinning (%)

15

10

0
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile A - Distance from Center (mm)
(c)
35
Experimental
BT-Shell-Implicit
30 TT-Shell-Implicit
YLD2000-Shell-Implicit
3P-Barlat-Shell-Implicit
25

20
Thinning (%)

15

10

-5
-4 1 6 11 16 21 26 31 36 41 46 51 56
Profile B - Distance from Center (mm)
(d)
Fig. 16 (continued)

Fig. 14 shows the simulation results obtained with using solid
elements and with both implicit and explicit solution procedures
for 260 °C and 30 MPa conditions along profile B direction.
Although there is no significant difference between the simulation
results wherein solid elements used, FE analyses that utilized: (a)
flow curves obtained from bulge tests and (b) explicit procedure
provided better approximation for thinning predictions, in general.
However, in terms of predicting the thinning in the critical regions
where the highest thinning occurred, the FE model that was based
on flow curves from tensile tests with solid elements and implicit
solution procedure (TT-Solid-Implicit) yielded closer results to
experimental part measurements as can be seen from Fig. 14. It
can be concluded that finite element analyses resulted in reason-
ably acceptable predictions for thinning within the deviation range
of 5–15%.
Figs. 15 and 16 illustrate thinning for hydroformed parts ob-
tained from simulations that performed with shell elements, and
using explicit (Exp) and implicit (Imp) solvers at different forming
conditions. FE analyses facilitated TT-Shell-Exp, BT-Shell-Exp,
3P_Barlat-Shell-Exp and YLD2000-Shell-Exp combinations pro-
vided conforming approximations to experimental results for pro-
file A, up to the contour length of 20 mm. In the maximum
thinning location, the analysis employed YLD2000-Shell-Exp fea-
tures resulted in higher accuracy and it was followed by TT-
Shell-Exp, 3P_Barlat-Shell-Exp and BT-Shell-Exp, respectively. For
profile B, on the other hand, analyses with different numerical
preferences yielded very close results to each other up to the
contour length of 35 mm. All numerical results except YLD2000-
Shell-Exp were in close proximity, however; beyond 35 mm
TT-Shell-Exp and 3P_Barlat-Shell-Exp were closer to experimental
values. TT-Shell-Exp was determined to yield best results. The total
computational time was 15 h, approximately.
In the analyses in which implicit solver were exploited, the FE
models with TT-Shell-Imp, BT-Shell-Imp, YLD2000-Shell-Imp, and
 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662
35
30
25
20
Thinning (%)
15
10
0
-4 1 6 11
Profile A - Distance from Center (mm)
Fig. 17. Thinning distribution obtained with shell and solid elements at 150 °C and 30 MPa for profile A.
3P_Barlat-Shell-Imp arrangements resulted in very close approxi-
mations (5% difference) to experimental data in Profile A, up to
a contour distance of 25 mm (i.e., measurement location #6). After
this critical region where maximum thinning is usually observed
at, the difference between experimental and numerical analyses
went up to 10% levels, which is still in acceptable range. Similar
observations were made for Profile B. As a general conclusion,
the locations where maximum thinning occurred were estimated
moderately better with the recently developed material models
(3P_Barlat, YLD 2000). These material models were proven to yield
accurate results, especially where anisotropy is present [12,29]. As
overall evaluation, FE models that utilized TT-Shell-Imp combina-
tions showed more consistent predictions. The corresponding com-
putational time was recorded as short as 13 min.
Thinning comparison given in Figs. 14–16 showed that, the dif-
ferent numerical solution attempts were able to approximate the
critical thinning regions with varying degree of deviations from
experimental values, limited up to 15%, at most. For flat sections
of the hydroformed parts, the numerical results were highly accu-
rate (less than 1% difference between experimental measurements
and numerical predictions in some cases) while in curved parts of
sheet blanks, deviation were usually in 5–15% range, which is still
in admissable margin of prediction error. Uncertainties and
approximations in temperature distribution, friction and material
modeling may have contributed to these deviations. It should also
be noted that there are also several other factors that cause dis-
crepancies such as manual measurement errors, inability to obtain
the numerical results from exact corresponding location where
manual measurements were obtained (element size related is-
sues). Based on the previous knowledge available in the literature,
anisotropy of AA 5754-O can be assumed insignificant and it is not
responsible from the deviation of FEA results from experimental
observations [38].
Solid elements, in general, resulted in better predictions, partic-
ularly for the flat region of the parts (i.e., up to 30 mm distance
from center point, Fig. 15) especially with the strain rate power
law material model. Nevertheless, for the curved regions of the
formed part (i.e. after 30 mm contour distance in Fig. 17), the re-
sults obtained with shell elements were closer to measured thick-
ness values than the results obtained with solid elements. In
2661
Experimental
TT-Solid-Implicit
TT-Shell-Implicit
TT-Shell-Explicit
16 21 26 31 36 41 46 51 56
contrast, finite element analyses that use solid elements are re-
garded as more realistic compared to analyses in which shell ele-
ments utilized [39]. This can be explained with the adaptive
remeshing feature that was used in conjunction with the shell ele-
ments. Different from the shell elements, the analyses in which so-
lid elements (3D, hexagonal) were employed performed without
adaptive remeshing feature. Another factor for the difference be-
tween the analyses is the fact that the response to bending and
straining are obtained with different stiffness matrices in solid
and shell elements formulations.
5. Conclusions
Forming characteristics of AA5754-O aluminum alloy under
warm hydroforming process conditions were numerically investi-
gated and were compared with the experimental findings obtained
at different elevated temperature and pressure values. Thinning
distribution and the cavity filling ratio (CFR) values of the formed
part were taken as measures to compare and assess the effect of
the temperature and pressure as well as the accuracy of numerical
models.
As expected, increasing temperature and pressure values re-
sulted in an increased cavity filling ratio and thinning. For the
closed-die hydroforming problem discussed in this study, no sig-
nificant advantage was observed by using the anisotropic material
models (3-parameter Barlat, and YLD2000) over isotropic material
models (strain rate power law). Nevertheless, the locations on the
part that underwent excessive thinning were predicted slightly
better with the recently developed material models that take
anisotropy into consideration. It is believed that these models
would result in better predictions for the materials exhibit higher
degree of anisotropy. In terms of the effect of material test type
(bulge vs. tensile test), the numerical models that were built on
the flow stress curves obtained from tensile tests resulted in closer
prediction to the experimental results. In comparison to solid ele-
ments, shell elements were found to be more appropriate to pre-
dict the thinning in the part. From overall observations, the best
combination for the FEA parameters is found to be ‘‘TT-Shell-
Imp’’ with isotropic material model based on tensile tests, shell
2662 H. Gedikli et al. / Materials and Design 32 (2011) 2650–2662
elements and implicit solver, which offered better predictions and
a significant savings in simulation time.
Acknowledgments
The authors are thankful to National Science Foundation (NSF)
for the partial support on this project through NSF ENG/CMMI
Grants 0703912; and NSF IIP IUCRC Grant 0638588.
References
[1] Verbrugge M, Lee T, Krajewski PE, Sachdev A, Bjelkengren C, Roth R, et al. Mass
decompounding and vehicle lightweighting. Mater Sci Forum 2009;618–
619:411–8.
[2] Mordike BL, Ebert T. Magnesium properties-applications-potential. J Mater Sci
Technol 2001;302:37–45.
[3] Greene DL, DiCicco J. Engineering-economic analyses of automotive fuel
economy potential in the United States. ORNL/TM-2000/26, Oak Ridge
National Laboratory, Oak Ridge, TN; 2000.
[4] Montalbo T, Lee TM, Roth R, Kirchain R. Modeling costs and fuel economy
benefits of lightweighting vehicle closure Panels. 2008-01-0370, SAE
international; 2008.
[5] Bolt PJ, Lamboo NAPM, Rozier PJCM. Feasibility of warm drawing of aluminum
products. J Mater Proc Technol 2001;115:118–21.
[6] Wowk DL. Effect of prestrain on the strain rate sensitivity of AA5754 sheet.
Ph.D dissertation, Queen’s University, Canada; 2008.
[7] Hosford W, Caddell R. Metal Forming. PTR Prentice Hall; 1993.
[8] Ayres RA. Alloying aluminum with magnesium for ductility at warm
temperatures (25–250 °C). Metal Trans A 1979;10:849–54.
[9] Robinson JM, Shaw MP. Microstructural and mechanical influences on dynamic
strain aging phenomena. Int Mater Rev 1994;39:113–22.
[10] Li D, Ghosh AK. Tensile deformation behavior of aluminum alloys at warm
forming temperatures. Mater Sci Eng 2003;352:279–86.
[11] Li D, Ghosh AK. Biaxial warm forming behavior of aluminum sheet alloys. J
Mater Proc Technol 2004;145:281–93.
[12] Abedrabbo N, Pourboghrat F, Carsley JE. Warm forming of aluminum alloys
using a coupled thermo-mechanical anisotropic material model. NUMISHEET
2005. In: Proceedings of the 6th international conference and workshop on
numerical simulation of 3D sheet metal forming processes, Detroit, MI, USA;
August 15–19 2005.
[13] Abedrabbo N, Pourboghrat F, Carsley JE. Forming of aluminum alloys and
elevated temperatures – part 1: material characterization. Int J Plasticity
2006;22:314–41.
[14] Abedrabbo N, Pourboghrat F, Carsley JE. Forming of aluminum alloys and
elevated temperatures – part 2: numerical modeling and experimental
verification. Int J Plasticity 2006;22:342–73.
[15] Abedrabbo N, Pourboghrat F, Carsley JE. Forming of AA5182-O and AA5754-O
at elevated temperatures using coupled thermo-mechanical finite element
models. Int J Plasticity 2007;23:841–75.
[16] Krajewski PE, Carsley JE. Heat treatment effects on bending in AA6111.
Aluminum 2003. In: Das SK, editor. TMS; 2003, p. 25–32.
[17] Lee TM, Hartfield-Wunsch SE, Xu S. Demonstration of the preform anneal
process to form a one-piece aluminum door inner panel. 2006-01-0987. SAE
international; 2006.
[18] Iadicola MA, Foecke T, Banovic SW. Experimental observations of evolving
yield loci in biaxially strained AA5754-O. Int J Plasticity
2008;24(11):2084–101.
[19] Mahabunphachai S, Koç M, Carsley JE. Numerical and experimental
investigations on deformation behavior of aluminum 5754 sheet alloy under
warm hydroforming conditions. In: Proceedings of NUMIFORM 2010, Pohang,
S Korea; 2010 June 13–16.
[20] Ahmed M, Hashmi MSJ. Finite-element analysis of bulge forming applying
pressure and in-plane compressive load. J Mater Process Technol
1998;77:95–102.
[21] Smerd RO. Constitutive behavior of aluminum alloy sheet at high strain rates.
M.S. thesis, Waterloo University, Canada; 2005.
[22] Barlat F, Lian J. Plastic behavior and stretchability of sheet metals. part I: a
yield function for orthotropic sheets under plane stress conditions. Int J
Plasticity 1989;5:51–66.
[23] Barlat F, Lege DJ, Brem JC. A six-component yield function for anisotropic
metals. Int J Plasticity 1991;7:693–712.
[24] Barlat F, Maeda Y, Chung K, Yanagawa M, Brem JC, Hayashida Y, et al. Yield
function development for aluminum alloy sheets. J Mech Phys Solids
1997;45(11/12):1727–63.
[25] Barlat F, Brem JC, Yoon JW, Chung K, Dick RE, Lege DJ, et al. Plane stress yield
function for aluminum alloy sheets – part 1: theory. Int J Plasticity
2003;19:1297–319.
[26] Barlat F, Aretz H, Yoon JW, Karabin ME, Brem JC, Dick RE. Linear transformation
based anisotropic yield functions. Int J Plastic 2005;21:1009–39.
[27] Abedrabbo N. Experimental and numerical investigations of stamp
hydroforming and ironing of wrinkling in sheet metal forming. M.S. Thesis,
Michigan State University, USA; 2002.
[28] Abedrabbo N, Zampaloni MA, Pourboghrat F. Wrinkling control in aluminum
sheet hydroforming. Int J Mech Sci 2005;47:333–58.
[29] Zampaloni M, Abedrabbo N, Pourboghrat F. Experimental and numerical study
of stamp hydroforming of sheet metals. Int J Mech Sci 2003;45:1815–48.
[30] Korkolis YP, Kyriakides S. Path-dependent failure of inflated aluminum tubes.
Int J Plasticity 2009;25:2059–80.
[31] Ls Dyna, Keyword user’s manual, Livermore Software Technology Corporation,
Livermore, CA, USA; 2007.
[32] Beams JW. Mechanical properties of thin films of gold and silver. In:
Neugebauer CA, Newkirk JB, Vermilyea DA, editors. Structure and properties
of thin films. New York, NY: John Wiley and Sons; 1959. p. 183.
[33] Kim J, Kangl YH, Choi HH, Hwang1 SM, Kangl BS. Comparison of implicit and
explicit finite-element methods for the hydroforming process of an
automobile lower arm. Int J Adv Manuf Technol 2002;20:407–13.
[34] Harewood FJ, McHugh PE. Comparison of the implicit and explicit finite
element methods using crystal plasticity. Compos Mater Sci 2007;39:481–94.
[35] Mahabunpachai S, Koç M. Investigations on forming of aluminum5052 and
6061 sheet alloys at warm temperatures. J Mater Des 2010;31:2422–34.
[36] Lee W, Kim D, Kim J, Chung K, Hong SH. Analysis of forming process of
automotive aluminum alloys considering formability and springback. Key Eng
Mater 2007;345–346:857–60.
[37] Zhang SH, Wang ZR, Xu Y, Wang ZT, Zhou LX. Recent developments in sheet
hydroforming technology. J Mater Process Technol 2004;151:237–41.
[38] Van den Boohaard AH. Thermally enhanced forming of aluminum sheet. Ph.D.
Dissertation, University of Twente, The Netherlands; 2002.
[39] Menezes LF, Teodosiu C. Three dimensional numerical simulation of deep
drawing process using solid finite elements. J Mater Process Technol
2000;97:100–6.
 12
Journal of Materials Processing Technology 211 (2011) 150–166

Contents lists available at ScienceDirect

Journal of Materials Processing Technology

journal homepage: www.elsevier.com/locate/jmatprotec

Database for real-time loading path prediction for tube hydroforming using
multidimensional cubic spline interpolation
Angshuman Ghosh, Karan Deshmukh, Gracious Ngaile ∗
Department of Mechanical & Aerospace Engineering, North Carolina State University, Box 7910, Raleigh, NC 27695, USA

a r t i c l e i n f o a b s t r a c t

Article history: Tube hydroforming (THF) is a metal-forming process that uses a pressurized fluid in place of a hard tool to
Received 8 December 2009 plastically deform a given tube into a desired shape. In addition to the internal pressure, the tube material
Received in revised form 4 September 2010 is fed axially toward the die cavity. This process has various applications in the automotive, aerospace,
Accepted 17 September 2010
and bicycle industries. Accurate coordination of the fluid pressure and axial feed, collectively referred to
as a loading path, is critical to THF. Workable loading paths are currently determined by trial and error,
which can be time consuming.
Keywords:
This study discusses an innovative technique for developing an interactive, real-time database that
Tube hydroforming
Loading path
would be able to predict loading paths for many THF components and hence reduce the computational
Interpolation time required. By classifying most of the commercial THF parts into families, parameters such as material
properties, part geometry, and tribological factors were simulated by category and stored in the database.
Multidimensional cubic spline interpolation was implemented to enable an end user to request from the
database a loading path for a wide range of conditions. Test results from the database for different THF
families were shown to approximate the simulated results. In addition to reducing the computation time,
the use of interpolation techniques eliminates the need for carrying out multiple simulations for similar
THF parts.
© 2010 Elsevier B.V. All rights reserved.

1. Introduction
Hydroforming has found significant industrial utility in the
present decade, although research on tube hydroforming started in
the 1940s (Koc and Altan, 2001). Automotive parts that are typically
produced using this process include exhaust manifolds, chassis,
engine cradles, and radiator frames. Hydroforming allows weight
reduction without compromising strength, which is of great sig-
nificance in the automotive and aerospace industries. For example,
conventional fabrication of a chassis involves welding several parts
together, whereas hydroforming can manufacture a chassis in a
single step, thus reducing weight and secondary operations.
Tube hydroforming (THF) is a metal-forming process in which
a tube is plastically expanded into a die cavity by the simultane-
ous action of fluid pressure and axial material feed, such that the
tube takes the shape of the die cavity. Success of THF depends on
various process and material parameters. The two most important
process parameters are internal pressure (Fig. 1a) given to the tube
and axial material feed (Fig. 1b) applied to the ends of the tube.
A graphical representation of the two parameters is called a load-
∗ Corresponding author.
E-mail address: gracious ngaile@ncsu.edu (G. Ngaile).
ing path (Fig. 1c). Loading path depends on process conditions and
materials and is independent of process time. Also, loading path
has to be in a particular process window for the part to be success-
fully formed, as shown in Fig. 1c. Inaccurate loading path will lead
to part failures such as wrinkling, bursting, or buckling. The details
on the modes of failure and mathematical models for failure pre-
dictions are given by Dohmann and Hartl (1997) and Xia (2001),
respectively.
Prediction of the proper loading path for a particular set of pro-
cess conditions is one of the biggest challenges in THF. Since this
process is relatively new, much is not known. Previous attempts to
predict loading paths have mostly involved trial-and-error. Reduc-
ing the computational time required for loading-path prediction
would increase the utility of THF. The following are some of the
attempts to develop a faster and more systematic approach to
loading-path prediction:
Trial and Error FEA Simulation approach: One of the first improve-
ments over trial-and-error was the use of FEA analysis, where
iterative FE simulations are carried out until acceptable forming
results are obtained. This process is still extremely inefficient and
time consuming (Strano et al., 2004). To reduce the number of
simulations, the minimum axial feed of the materials can be estab-
lished by carrying out simulation without forced axial feeding, i.e.,

0924-0136/$ – see front matter © 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jmatprotec.2010.09.012
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166
Fig. 1. Process window for THF process and loading path.
the axial feed obtained from this simulation is due to the action
of internal pressure only. This technique is known as self-feeding
and provides the initial loading path (Strano et al., 2004).
Optimization approach: This approach aims at optimizing local
tube-wall thickness distribution by sensitivity analysis. Repeated
simulations are carried out until an optimal loading path is
obtained (Gelin and Labergere, 2002).
Adaptive approach: The adaptive method is faster than the above
methods in that it determines the process-loading curve by run-
ning a single simulation. The method monitors the failure modes
of wrinkling and bursting with the use of indicators inside the
FEA simulation itself. This is done by defining failure indica-
tors, namely, the wrinkle indicator and the bursting indicator
(Nordlund, 1998). The control strategy identifies the wrinkle or
bursting with the help of these indicators and gives feedback to
the simulation, where changes are made to prevent failure
Fuzzy Load Control method and Neural Network Analysis method:
Fuzzy Load Control (FLC) involves development of fuzzy logic rules
based on previous knowledge of the THF process window (Ray
and Mac Donald, 2004). Using these rules, failure indicators based
on threshold for wrinkling and buckling, are decided. The sim-
ulation procedure differs from the adaptive approach such that
here the failure indicators are decided based on fuzzy logic rules.
Ray and Mac Donald (2004) used FLC to predict loading paths for
T-shape THF. Lin and Kwan (2004) presented four-layer and five-
layer abductive network models to predict the process parameters
of THF for an acceptable T-shape product. They chose a material
with a constant shear friction and varied the geometrical parame-
ters of the die cavity and the internal pressure input to get a total
of 75 sets of parameters. Then they obtained the training data by
performing FEA simulation on those 75 sets using the commercial
FEA simulation software DEFORM 3D.
All the methods discussed above expend considerable time in
searching for a loading path that will result in a successful part.
Moreover, most of those methods can only be used to obtain a
loading path for a specific material and geometry, i.e., they must
be rerun if the THF geometry or material has changed. This study
proposes the development of a database that through interpolation
could instantly provide continuous loading-path data for THF parts
with different materials and geometries.
2. Objectives and approaches
The objectives of this study are to (a) develop a database contain-
ing loading-path data for forming numerous THF components with
different materials, geometries, and tribological conditions and (b)
151
establish a multidimensional interpolation scheme in conjunction
with the FE-based simulation database that will facilitate deter-
mination of real-time continuous loading paths for THF parts. The
approaches taken to achieve the above objectives are presented in
the flowchart shown in Fig. 2.
3. Classifications of THF components
A thorough review on commercially available THF components
was carried out in order to classify THF families for the database as
shown in Fig. 3. However, the scope of this study is limited to the
families that require both axial feed and internal pressure for form-
ing. Hence seven families were selected for this study namely Bulge
shape (B), Single Y shape (SY), Aligned Double Y shape (DY), Sin-
gle T shape (ST), Aligned Double T shape (DT), Non Aligned Double
T shape – opposite side (DTOS) and Non Aligned Double T shape –
same side (DTSS). Fig. 4 shows schematics of the seven THF families.
4. Pressure curve generation and strategy
Pressure and material feed profiles are the two components
that build up the loading paths for THF. This section will focus on
pressure curve generation and the next section will be devoted on
material feed curves. The strategy taken in generating the loading
paths was to obtain a generic pattern for the pressure profile. That
is, the pressure curves which will have the sample pattern for all
THF families. Having a generic pressure profile will lead to a robust
interpolation scheme as compared to using different patterns of
pressure profiles for different THF families. It should be noted that
the loading path for THF is composed of the pressure loading and
material feed profiles as shown in Fig. 1a and b. Theoretically, infi-
nite routes are possible to obtain the loading path (material feed vs.
pressure) shown in Fig. 1c. For example, three different scenarios
to obtain the same loading path shown in Fig. 1c can be examined.
In scenario A, one could fix material feed profile and vary the pres-
sure profile, whereas in scenario B, pressure profile could be fixed
and the material feed profile could be varied. In scenario C, both
pressure and material feed profiles could be varied. The common
boundary conditions for all three scenarios are (a) at the end of
the process a certain maximum forming pressure will be needed
to ensure that the desired corner radii are formed and (b) at the
end of the process a certain amount of material should have been
fed to the die cavity to ensure that the part is successfully hydro-
formed. In this study scenario B was adopted, where the pattern of
the pressure profile is fixed for all THF families.
The unit pressure curve shown in Fig. 5 was adopted for this
study. The specific profile in Fig. 5 shows various stages in the THF
152 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Database for Continuous Real Time

Loading Path Prediction in THF

Review on Commercially Classification of Classified THF

Available THF Components THF Components Components

Calculation of Maximum Instantaneous

Selection of THF
Pressure for Free Expansion Bulge
Components for
using Thin Walled Pressure Vessel
Current Study Classification
Theory and Theory of Plasticity
Section-3

Maximum Instantaneous
Scaling of the Maximum Instantaneous
Pressure Equation,
Pressure Equation to Incorporate the
Scaling Factors and Unit
THF Parameters not considered
Pressure Profile

Generation of the Selection of

Pressure Curve using Dominant Process
Pressure Curve Unit Pressure Profile Parameters in THF
Section-4

Axial Feed Data Generation Determination of Matrix

by FEA Simulation Pre-Processing for
for FEA Simulation
FEA Simulation of
Section-5 Varying the Selected
THF
Process Parameters

FEA Simulation of THF: Material

Axial Velocity Feed is determined by
Trial & Error with Prior Experience

Material Axial Material Axial Displacement

Displacement Feed Data Analysis:
Feed obtained from FEA
Data are Stored Monotonicity & Trend
Simulation

According to the Selected Cubic

Spline Multidimensional Interpolation Selection of Appropriate
Scheme, Derivation of the Necessary Multidimensional Interpolation
Equations for Current Study Scheme

Interpolation Scheme to Development of the

obtain Continuous Loading Interpolation Scheme using Data Analysis
Path Data in THF Appropriate Programming Tool & Interpolation
Section-6

With the Interpolated Material Axial

Feed Data and Calculated Pressure
Curve, Perform the THF Simulation

Validation of the Results

Concluding
Results
Remarks
Section-7, 8

Fig. 2. Flowchart showing the approaches taken in this study.

 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166 153

Fig. 3. Classification of axial feed and pressure driven THF components.

process. Yielding: This part of the curve should maintain a linearly By assuming that the material follows the power law given in Eq.
increasing profile during yielding. Expansion: During this stage the (7) and relating Eqs. (2), (5), (6) and (7), an expression could be
gradient is reduced. This enhances more material feed compared derived for instantaneous pressure (Pi ) as a function of K, tube geo-
to pressure. Calibration: During this stage, the pressure increases metrical variables, stress and strain as shown in Eq. (8), where, K
to its maximum value to fill the corners and radii. is the material strength coefficient and n is strain hardening coeffi-
This unit pressure profile would then be multiplied by the max- cient. Since Eq. (8) gives the instantaneous pressure of the system
imum pressure to obtain the actual pressure curve for a specific at the final time step, this is the maximum instantaneous pressure
THF geometry. Using a third degree polynomial shown in Eq. (1), (Pf ) for the bulged tube.
unit pressure profile is estimated for a unit process time, where T  √ n
1/2
denotes the time. It should be noted that a different unit pressure K (2/ 3)(1 + ˇ + ˇ2 ) .εa
2ti
profile could have been adopted. Pi = (8)
di − ti (1 − ˛ + ˛2 )
1/2
3 2
Pu = 1.67T − 4T + 3.33T (1)
In Eq. (8) ˇ and ˛ need to be determined. Using the volume con-
stancy given in Eq. (9), ˇ (ˇ = εb /εa ) could be expressed in terms
4.1. Determination of the maximum instantaneous pressure
of εa and εr as shown in Eq. (10). Using the relation of principal
stresses, ˛ could be obtained from Eqs. (11) and (12). By multi-
The prediction of instantaneous pressure for bulge family is pro-
plying the maximum instantaneous pressure (Eq. (8)) by the unit
posed based on the assumptions that the tube is thin and obeys
pressure (Pu ), Eq. (1), Eq. (13) can be obtained.
membrane theory and the maximum thinning before fracture is
30%. Fig. 6 shows a bulge tube under internal instantaneous pres- εa + εb + εr = 0 (9)
sure (Pi ) and material feed (f) supplied by the punches. The figure ε 
r
also shows an element of the bulge tube showing tangential stress ˇ=− +1 (10)
( a ) and axial stress ( b ). However the radial stress ( r ) is assumed εa
to be negligible compared to the other stresses. ¯
 ε2
 ¯
 ε2

1 = ε1 + ; 2 = ε2 + (11)
The instantaneous pressure (Pi ) at the instant of final forming ε̄ 2 ε̄ 2
for a thin cylindrical tube could be given by Eq. (2), where, ti is the
2 [ε2 + (1/2)ε1 ] [ˇ + (1/2)] 2ˇ + 1
minimum final instantaneous thickness of the tube at 30% thin- ˛= = = = (12)
ning (ti = 0.7t0 ) in the expansion zone and di is the maximum final 1 [ε1 + (1/2)ε2 ] [1 + (1/2)ˇ] 2+ˇ
instantaneous diameter in the expansion zone (Asnafi, 1999). Due
to plane stress condition ( r = 0), the radial strain (εr ) and tangen-  √ n
1/2
tial strain (εa ) are defined by Eqs. (3) and (4) respectively. Using the K (2/ 3)(1 + ˇ + ˇ2 ) .εa
2ti
Von Mises criteria, the effective stress and strain can be defined by P = (1.67T 3 − 4T 2 + 3.33T )
di − ti (1 − ˛ + ˛2 )
1/2
Eqs. (5) and (6) respectively. (13)
2a ti It should be noted that Eq. (13) represents initial pressure load-
Pi = (2)
d ing path for bulge tube hydroforming. Since the bulge analysis did
t  not consider friction (free expansion analysis), and tube length, the
i
εr = ln (3) pressure may not be adequate. Thus, this profile is used as initial
t0
 d −t  input to FE simulation. Through FE iterations this pressure profile
εa = ln i i
(4) was scaled up to 1.35 to successfully form the Bulge shape with
d0 − t0 maximum thinning below 30%. To determine the pressure profile
2 1/2
for the other THF families, Eq. (13) derived for the Bulge shape
ε̄ = √ εa (1 + ˇ + ˇ2 ) (5) was used as an initial pressure curve and FE iterations were car-
3
ried out to determine the scaling factors. Table 1 shows the scaling
1/2 factors for all the families selected for this study. The scaling factors
¯ = (1 − ˛ + ˛2 ) a (6)
give satisfactory results within the following range of the parame-
where ˛ =  b / a and ˇ = εb /εa ters. Friction coefficient from  = 0.001 to  = 0.25 for all families,
Length of the tube from 150 mm to 400 mm for Bulge, 150 mm to
¯ = K(ε̄)n (7) 300 mm for Single Y and Single T, 200 mm to 350 mm for Double
154 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Fig. 4. THF families used in the database.

Table 1
Scaling factors for pressure curve generation.

THF family Scaling factor THF family Scaling factor

Bulge shape (B) 1.35 Aligned Double T shape (DT) 2.45

Single Y shape (SY) 2.35 Non Aligned Double T shape – opposite side (DTOS) 2.45
Aligned Double Y shape (DY) 1.80 Non Aligned Double T shape – same side (DTSS) 2.00
Single T shape (ST) 2.45
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166 155

1 Geometrical parameters
0.9
0.8
Calibration
0.7 Initial outer diameter of the tube, did ∈ [ad , bd ], where id
Pressure

0.6
0.5 Expansion = 0, . . . , nd
0.4
0.3 3 2
Pu = 1.67T - 4T + 3.33T
0.2 Yielding Initial thickness of the tube, tit ∈ [at , bt ], where it
0.1
0 = 0, . . . , nt
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time

Fig. 5. Unit pressure profile. Initial length of the tube, lil ∈ [al , bl ], where il = 0, . . . , nl

Protrusion/bulge position along the length of the tube,

× la ∈ [ala , bla ], where ila = 0, . . . , nla

Protrusion/bulge height, hih ∈ [ah , bh ], where ih = 0, . . . , nh

Angle of the protrusion, i ∈ [a , b ], where i = 0, . . . , n

Fig. 6. Bulge shape THF.
Y and Aligned Double T, Double T unaligned same side (DTSS) and
Double T unaligned opposite side (DTOS).
5. Axial feed curve generation using FEA
The THF process depends on the material, tribological and geo-
metrical parameters. The following nine parameters from different
categories are taken as the governing variables for this study.
Material parameters
Material strength coefficient, KiK ∈ [aK , bK ],
= 0, . . . , nK
The material feed data corresponding to all these nine variables
are represented by FiK in i id it il il ih i . However the material feed data
a
itself is an array of 101 displacement values corresponding to the
incremental time interval, i.e. (FiK in i id it il il ih i ) , iT = 0, . . . , 100.
a iT
Table 2 shows the simulation matrix for the database devel-
opment. However in order to analyze the data, additional sets of
simulations were also carried out. As seen in Table 2, only one node
is used for each of the last 6 dimensions namely d, t, l, la , h and  to
reduce the number of simulations. In order to develop the database
27 sets of parameters for each family, adding up to 189 sets were
chosen. The simulation parameters are presented in the following
order with the specified unit unless otherwise mentioned. (K, n, ,
where iK d, t, l, la , h, ), where K is in MPa; d, t, l, la and h are in mm; and  is
in◦ .
5.1. Inputs, boundary conditions, and FE results

Strain hardening coefficient, nin ∈ [an , bn ], where in The FEA analysis package used in this study is Abaqus 6.7, which
= 0, . . . , nn is an explicit code. In the simulation, it is ensured that tube ends
are always in contact with punches. The assumptions made for the
simulations are rigid die and punch, deformable tube with shell
Tribological parameters type elements, simulation process time of 0.01 s and 30% maximum
wall thinning. In this study ¯ = K ε̄n material model is selected and
Friction coefficient, i ∈ [a , b ], where i = 0, . . . , n coefficient of friction () is prescribed at the tube-die/tube-punch

Fig. 7. Pressure curve variations for Bulge shape (B).

156 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Table 2
Simulation matrix.

Families K (MPa) n  d (mm) t (mm) l (mm) la (mm) h (mm)  (◦ )

 500
  0.1
  0.001 
Bulge shape (B) 1000 0.425 0.1 {30} {2} {200} {100} {11} {90}
 1500
500
  0.60
0.1
  0.001
0.25

Single Y shape (SY) 1000 0.425 0.1 {30} {2} {150} {70} {36} {60}
 1500
500
  0.60
0.1
  0.001
0.25

Aligned Double Y shape (DY) 1000 0.425 0.1 {34.925} {1.651} {200} {87.3} {45} {60}
 1500
500
  0.60
0.1
  0.001
0.25

Single T shape (ST) 1000 0.425 0.1 {30} {2} {150} {75} {30} {90}
 1500
500
  0.60
0.1
  0.001
0.25

Aligned Double T shape (DT) 1000 0.425 0.1 {30} {2} {200} {100} {30} {90}
 1500
500
  0.60
0.1
  0.001
0.25

Non Aligned Double T shape – opposite side (DTOS) 1000 0.425 0.1 {30} {2} {200} {50} {30} {90}
 1500
500
  0.60
0.1
  0.001
0.25

Non Aligned Double T shape – same side (DTSS) 1000 0.425 0.1 {34.925} {1.651} {200} {58.4} {30} {90}
1500 0.60 0.25

interface. Fluid pressure load was given uniformly to the inner sur-
face of the tube. The punch velocity profiles to form an acceptable
part were obtained by iterative simulations until an acceptable part
is formed. The typical thickness variation maps for the different
families are shown in Table 3.
The representative pressure curves, axial material feeds (dis-
placement) are discussed for different families with respect to
different variables. Fig. 7 shows the variation of the pressure curves
for different variables for the Bulge family. These profiles are based
on Eq. (13). From curves B1 and B2, it could be observed that, as

Fig. 8. (a) Material axial displacement feed variations for Aligned Double Y shape (DY) and (b) material axial displacement feed variations for Aligned Double Y shape (DY).
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166
Fig. 9. Loading path variations for Single T shape (ST).
the value of strength coefficient K increases, the internal pressure
increases. This is due to the fact that the flow stress increased with
increase in the strength coefficient, K. It is also found from curves
B1 and B3 that the internal pressure increases with decrease in the
strain hardening coefficient n. This is due to the fact that for a mate-
rial that obeys the power law, lower n value will results in higher
stress at strain levels below 1.0.
In Fig. 7, the curves B1, B4, B7 and B8 fall into the same curve,
since the variations of coefficient of friction (), initial length of the
tube (l) and the bulge position (la ) are not included in the maximum
instantaneous pressure Eq. (13). It should be noted, however, that
the effects of , l and la will be factored in the material feed curves.
Since the initial diameter of the tube (d) and bulge height (hb ) are
inversely proportional to the maximum instantaneous pressure,
the pressure decreases with increasing d and hb as shown by curves
B1–B5 and B1–B9 respectively. As can be seen in Eq. (13) the initial
tube thickness (t) is directly proportional to the maximum instan-
taneous pressure, thus the pressure increases with increasing t as
shown in the curves B1 and B6.
Fig. 8a and b shows the representative material displacement
variation for DY family with respect to the change in the parame-
ters. Curves DY1L and DY1R show that, the left axial feed is greater
than the right axial feed. The same observation is found for asym-
metric shape like SY. Since the Y protrusion is inclined towards
the right side for both the SY and DY families, it is easier to feed
the material from the left side for such geometry and so the left
feed is greater than the right feed. The DTSS shape and the DTOS
shape are always symmetric for all geometric variations, thus the
material feed for both left and right sides is the same. However, the
remaining families such as B, ST and DT could be either symmetric
or asymmetric based on the geometry. When the position of the
bulge/protrusion is at the center of the tube (la = 0.5l), the families
are symmetric.
From DY1L and DY4L curves, it could be observed that material
feed decrease with decrease in the coefficient of friction (). This
is attributed to the fact that with higher friction more material will
be needed at the die cavity to compensate for rapid thinning. From
curves DY1L and DY5L it could be observed that as the diameter
of the tube (d) increases, the axial feed also increases. In case of
a constant and uniform thinning, the axial feed does not change
with variation of initial tube thickness (t). However some varia-
tions are observed in DY1L and DY6L, since the thinning is not
constant and uniform. The axial feed increases with increasing ini-
tial tube length (l) and protrusion height (h) as seen in DY1L–DY7L
and DY1L–DY9L respectively. The position of the protrusion (la ) also
157
changes the axial feed based on the symmetry of the THF part. From
the curves DY1L and DY10L, it is found that with increasing protru-
sion angle (), the axial feed reduces in the left side and increases
in the right side. This is due to the fact that as the angle increases,
the shape tends to be more symmetric. Finally the representative
loading paths obtained from the pressure curves and the material
axial feed curves, are shown in Fig. 9 for the ST family.
6. Data analysis and interpolation
6.1. Data analysis
In order to obtain continuous data for loading path in the spec-
ified range, the database developed need to be interpolated. In
order to select an appropriate interpolation scheme the data was
first analyzed. Most interpolation schemes require that the data be
monotonic. Monotonic data is defined as that data which is either in
an increasing order or in a decreasing order or that which remains
the same. However, strictly monotonic data refers to the data being
only in increasing or decreasing trend.
Figs. 10–13 show the representative material feed data varia-
tions with respect to the different variables for monotonicity. As
seen in the figures the material feed data was found to be mono-
tonic for all sets of simulations and for all time steps. Fig. 10 shows
that the axial feed increases monotonically as material strength
coefficient, K, increases. Similar trend is observed for strain hard-
ening coefficient (Fig. 11). Fig. 12 shows that the right side material
feed increases with increase in tube length for the Bulge shape. It
Fig. 10. Material axial displacement feed variations with respect to material
strength coefficient (K); DTSS (K, 0.6, 0.1, 30, 2, 200, 50, 30, 90); values of K are
500, 750, 1000, 1250 and 1500 MPa.
158 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Table 3
Thickness variation.

B (500, 0.6, 0.1, 30, 2, 200, 100, 11, 90)

SY (1500, 0.425, 0.1, 30, 2, 150, 70, 36, 60)

DY (1500, 0.6, 0.001, 34.925, 1.651, 200, 87.3, 45, 60)

ST (500, 0.6, 0.1, 30, 2, 150, 75, 30, 90)

DT (1500, 0.6, 0.001, 30, 2, 200, 100, 30, 90)

DTOS (500, 0.6, 0.1, 30, 2, 200, 50, 30, 90)

DTSS (1500, 0.6, 0.001, 34.925, 1.651, 200, 58.4, 30, 90)

should be noted that due to non-symmetric of the part the left side parameter known as the coefficient of determination (R2 ), which
material feed will decrease with increase in the tube length. Similar varies from 0 to 1. The closer is the value of R2 to 1, the better is the
trend is observed with DY shape as the protrusion angle is varied approximation. The coefficient of determination is defined by Eq.
as shown in Fig. 13. (14).
The characteristic behavior of the data was also tested. The tests
included (a) linear, (b) logarithmic, (c) polynomial, 2nd and 3rd
order, (c) power, and (d) exponential. The accuracy of the approx- SSE
imation of a data set by a particular curve was quantified by a R2 = 1 − (14)
SST
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166 159

Pressure Curve
a 50 B-3
45
40
B-2

Pressure (MPa)
35
B-1
30 B-Ref
25
20 B-1
B-2
15
B-3
10
B-Ref
5
0
0 0.002 0.004 0.006 0.008 0.01
Time (sec)

Material Axial Feed (Displacement)

Fig. 11. Material axial displacement feed variations with respect to strain hardening b 20
B-Ref
18 B-1
coefficient (n); DTOS (1500, n, 0.1, 30, 2, 200, 50, 30, 90); values of n are 0.1, 0.3, 0.425,
16
0.52 and 0.6.

Axial Feed (mm)

14
12 B-2
10
B-1
8 B-3
B-2
6
B-3
4
B-Ref
2
0
0 0.002 0.004 0.006 0.008 0.01
Time (sec)

Fig. 14. (a) Pressure curves and (b) material axial feed vs. time.

Trend analysis was performed by choosing 5 nodes for each

Fig. 12. Material axial displacement feed variations with respect to initial tube
length (l); B (1500, 0.52, 0.1, 30, 2, l, 100, 11, 90); values of l are 150, 200, 250,
300 and 400 mm.
parameter variation. Table 4 shows the R2 values for the trend anal-
ysis of the data. For the six types of curves tested, it is observed that
the linear, logarithmic, power and the exponential curves fail to fit
the data satisfactorily for all variables as their R2 values are low. The
polynomial curves give better approximations for all the variables.

6.2. Interpolation scheme – multidimensional cubic spline

interpolation

Based on the data analysis presented above the cubic spline

interpolation scheme is chosen to approximate the material feed
function. Since the material feeds depend upon the nine vari-
ables namely material strength coefficient (K), strain hardening
coefficient (n), friction coefficient (), initial outer diameter of
tube (d), initial thickness of the tube (t), initial length of the tube
(l), the position of the protrusion/bulge along the length of the
tube (la ), protrusion/bulge height (h) and the protrusion angle ();
the problem becomes a multidimensional interpolation problem
Fig. 13. Material axial displacement feed variations with respect to protrusion angle where the cubic spline approximation is to be used. It should
(); DY (1500, 0.52, 0.1, 30, 2, 200, 100, 36, ); values of  are 60◦ , 70◦ , 75◦ , 85◦ and be noted that except the Y-families the protrusion angle is con-
90◦ .
stant and equal to 90◦ , which makes the interpolation problem for
 2   2
those families eight dimensional. However for generalization pur-
where SSE = (Yi − Ŷi ) , SST = (Yi2 ) − (( ˙Yi ) /N), Yi is the pose nine-dimensional interpolation would be discussed. The cubic
actual data, Ŷi is the approximated data and N is the number of spline interpolation scheme used in this study is based on the work
data points. of Habermann and Kindermann (2007).

Table 4
R2 values for trend analysis.

Trend R2 value of material axial displacement feed data at T = 0.01 s

K n  d t l la h 

Left Right Left Right Left Right Left Right

Linear 0.9579 0.9617 0.8525 0.7804 0.9954 0.8441 0.9764 0.9841 0.9828 0.9449 0.9732 0.8606 0.9842
Logarithmic 0.9882 0.8314 0.9565 0.9850 0.9953 0.9540 0.9945 0.9765 0.9507 0.9760 0.9616 0.8175 0.9655
2nd order polynomial 0.9943 0.9856 0.9714 0.8791 0.9968 0.9737 0.9919 0.9887 0.9945 0.9873 0.9808 0.9917 0.9994
3rd order polynomial 0.9999 0.9968 0.9865 0.8929 0.9998 0.9978 0.9990 0.9999 0.9972 0.9874 0.9985 0.9988 0.9995
Power 0.9883 0.8457 0.9802 0.9838 0.9967 0.9607 0.9900 0.9807 0.9689 0.9824 0.9680 0.8363 0.9512
Exponential 0.9485 0.9669 0.8431 0.7417 0.9916 0.8556 0.9885 0.9228 0.9910 0.9613 0.9769 0.8778 0.9740
160 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Fig. 15. Successful THF parts and a part with an under fill.

Here the nine variables are represented by the corresponding thickness of tube, l-grid for initial length of tube, la -grid for the
nine grids namely; K-grid for strength coefficient variable, n-grid position of the protrusion/bulge along the length of tube, h-grid for
for strain hardening coefficient variable, -grid for friction coef- protrusion/bulge height and -grid for protrusion angle. The first
ficient, d-grid for initial outer diameter of tube, t-grid for initial grid K (aK , bK ) = {K0 , . . . , KnK }, nK ∈ N is defined on the interval
[aK , bK ] as a set of points KiK ∈ [aK , bK ]; where iK = 0, . . ., nK , aK =
K0 K1 · · ·KnK = bK and KiK is the equidistant nodes. Here N refers
DY Pressure Curve
to the set of Natural numbers {0, 1, 2, 3, . . .}. For example if the
a 90 DY-2
database is generated for the material strength coefficient (K) val-
80 DY-3
ues of {500, 750, 1000, 1250, 1500}MPa, then K (500, 1500) ={K0 ,
70
K1 , K2 , K3 , K4 }={500, 750, 1000, 1250, 1500}. The second grid is
Pressure (MPa)

60 DY-1, DY-Ref

50 DY-EXP defined as n (an , bn ) = {n0 , . . . , nnn }, nn ∈ N on the interval [an ,

DY-1
40 DY-2
DY-3
30 DY-Ref
20 DY-EXP Table 5
10 Interpolation scheme test data.
0
(K, n, , d, t, l, la, h, )
0 0.002 0.004 0.006 0.008 0.01
Time (sec) Bulge
B-Ref (500, 0.6, 0.1, 30, 2, 200, 100, 11, 90)
Material Axial Feed (Displacement)
B-1 (550, 0.55, 0.09, 30, 2, 200, 100, 11, 90)
b 45 B-2 (600, 0.45, 0.08, 30, 2, 200, 100, 11, 90)
40 DY-1L DY-RefL B-3 (750, 0.40, 0.01, 30, 2, 200, 100, 11, 90)
35 DY-1L Double Y
Axial Feed (mm)

30 DY-EXPL
DY-RefR DY-2L DY-Ref (1500, 0.6, 0.001, 34.925, 1.651, 200, 87.3, 45, 60)
DY-2L DY-3L
25 DY-RefL DY-1 (1450, 0.55, 0.005, 34.925, 1.651, 200, 87.3, 45, 60)
20 DY-3L
DY-EXPL DY-2 (1400, 0.45, 0.01, 34.925, 1.651, 200, 87.3, 45, 60)
DY-1R
15 DY-2R
DY-3 (1275, 0.45, 0.05, 34.925, 1.651, 200, 87.3, 45, 60)
DY-EXPR
10 DY-3R DY-EXP (1426, 0.502, 0.05, 34.925, 1.651, 200, 87.3, 45, 60)
DY-RefR Double T same side
5 DY-1R DY-2R, DY-3R DY-EXPR
0 DTSS-Ref (1500, 0.6, 0.001, 34.925, 1.651, 200, 58.4, 30, 90)
0 0.002 0.004 0.006 0.008 0.01 DTSS-1 (1450, 0.55, 0.005, 34.925, 1.651, 200, 58.4, 30, 90)
Time (sec) DTSS-2 (1400, 0.45, 0.01, 34.925, 1.651, 200, 58.4, 30, 90)
DTSS-3 (1275, 0.45, 0.05, 34.925, 1.651, 200, 58.4, 30, 90)
Fig. 16. (a) Pressure curves and (b) material axial feed vs. time. DTSS-EXP (1426, 0.502, 0.05, 34.925, 1.651, 200, 58.4, 30, 90)
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166 161

DTSS Pressure Curve

a 100
90
DTSS-2

DTSS-3
80

Pressure (MPa)
70 DTSS-1, DTSS-Ref

60
DTSS-EXP DTSS-1
50 DTSS-2
40 DTSS-3
DTSS-Ref
30
DTSS-EXP
20
10
0
0 0.002 0.004 0.006 0.008 0.01
Time (sec)

Material Axial Feed (Displacement)

b 30 DTSS-EXP DTSS-Ref

Axial Feed (mm)

25 DTSS-1

20

15 DTSS-3 DTSS-1
DTSS-2
10 DTSS-2 DTSS-3
DTSS-Ref
5
DTSS-EXP
0
0 0.002 0.004 0.006 0.008 0.01
Time (sec)

Fig. 18. (a) Pressure curves and (b) material axial feed vs. time.

, d, t, l, la , h, }={1500 MPa, 0.6, 0.1, 30 mm, 2 mm, 200 mm, 100

mm, 36 mm, 60 ◦ }. This material feed data FiK in i id it il il ih i which is
a
extracted from the simulation results is stored in the database.
In order to interpolate the material feed data FiK in i id it il il ih i
a
stored in the database, a cubic spline of order three and smooth-
3,2
ness two snK ,nn ,n ,nd ,nt ,nl ,nl ,nh ,n ∈ S3,3,3,3,3,3,3,3,3 (K , n ,  ,
a
d , t , l , la , h ,  ), can be used, if it satisfies
the interpolation conditions given by Eq. (15). Here
S3,3,3,3,3,3,3,3,3 (K , n ,  , d , t , l , la , h ,  ) is the
function space of all splines of degree three and smoothness
two on all grids. It should be noted that a spline of order three
and smoothness two is a piecewise polynomial function with
third order polynomial pieces. The polynomial pieces contain
derivatives up to the second order.

sn3,2
K ,nn ,n ,nd ,nt ,nl ,nl ,nh ,n (KiK , nin , i , did , tit , lil , laila , hih , i )
a

= FiK in i id it il ila ih i (15)

Now using the B-Spline functions and the Natural Spline conditions,
the interpolating spline s could be defined as given in Eq. (16).
nK +3nn +3n +3nd +3nt +3 nl +3 nla +3nh +3n +3

s(K, n, , d, t, l, la , h, ) =
Fig. 17. Successful THF parts and parts with underfill. jK jn j jd jt jl jla jh j

CjK jn j jd jt jl jl j j ujK (K)ujn (n)uj ()ujd (d)ujt

a h 
bn ] as a set of equidistant points nin ∈ [an , bn ]; where in = 0, . . .,
× (t)ujl (l)ujl (la )ujh (h)uj () (16)
nn and an = n0 n1 · · ·nnn = bn . In a similar fashion the remaining a

grids are defined. The interpolation data is the material feed and
is defined as FiK in i id it il il ih i ∈ R; where iK = 0, . . ., nK ; in = 0, . . ., nn ,
a detailed in Appendix A. Once the coefficients in Eq. (16) are known,
i = 0, . . ., n and so on. Here R denotes the set of Real numbers.
However the material feed data itself is an array of 101 displace-
ment values corresponding to the incremental time interval, i.e.
(FiK in i id it il il ih i ) , iT = 0, 1, 2, . . ., 100. For example, the material feed
a iT
required on the left side of the tube at time 0.01 s (iT = 100) for Dou-
ble Y-shape part is 45.63 mm for the forming conditions of {K, n,
The coefficients CjK jn j jd jt jl jl jh j are calculated in a structured way as
a
the value of the spline for any sets of the nine variables (K, n, ,
d, t, l, la , h, ) in the given interval could be obtained. This is the
interpolated value of the material feed for that particular set of
variables.
To illustrate the interpolation approach used in this study, a
2-diminsional cubic interpolation scheme is discussed in detail in
162 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

Fig. 19. Successful THF parts and parts with underfill.

the Appendix. For this illustration K-grid and n-grid variables were interpolation would lead to a successfully formed part, several tests
used. were carried out. The tests also determined the maximum accept-
able spacing of the data nodes. The user inputs were tested for the
Bulge, Double Y, and Double T families, as shown in Table 5.
7. Results and discussions
In Table 5, “Ref” represents an existing data node in the database,
whereas “1”, “2”, and “3” refer to user-provided sets of parame-
The loading paths (pressure vs. time, and material axial feed
ters. “1” was the closest to “Ref”, “2” was farther away from “Ref”,
vs. time) established from the simulations for several THF fam-
and “3” bisected “Ref” and the adjacent existing data node. Hence
ilies were used to build a database. The multidimensional cubic
the material feed data for “1”, “2”, and “3” were derived from the
spline interpolation was used to determine the loading paths for
database using the multidimensional cubic spline interpolation.
any THF part that fell within the ranges of the variables used in the
Furthermore, DY-EXP and DTSS-EXP are the sets used for exper-
database. The database stored the data in a 9-dimensional matrix.
imental validation which will be discussed in the next section.
The cubic spline interpolation scheme was carried out using Matlab
Fig. 14a and b shows the pressure curves and the material
and Visual BASIC. To check whether the loading paths produced by
axial feeds for Bulge-shaped THF families B-Ref, B-1, B-2, and B-3.
Fig. 15a–c shows the thickness variation obtained from simulations.
Table 6 These figures suggest that the interpolated axial-feed data yield sat-
Dimensional comparison of simulation and experiment. isfactory results for sets B-1 and B-2. However B-3 shows a slight
Protrusion height from the center of the tube, h (mm) underfill. The results imply that the spacing of the data nodes needs
to be refined to provide an acceptable continuous loading path for
Simulation Experiment
the whole range. Therefore, it can be concluded that if the data
DY 43 42 nodes for the variables for the Bulge family were spaced at inter-
DTSS 30 30
vals of 100 MPa for K, 0.15 for n, and 0.02 for , the database would
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166
Fig. 20. (a) Hydroforming test rig and (b) tube hydroforming tooling.
give a loading path that forms the part satisfactorily for any set of
parameters within the range.
Fig. 16a and b shows the pressure curves and the material axial
feeds for Double-Y-shaped THF parts. Fig. 17a–d implies that DY-1
and DY-2 yield successful parts, whereas DY-3 and DY-EXP yield
parts with underfill. This implies that the data nodes for Y-shape
families should be spaced at intervals of 100 MPa for K, 0.15 for n,
and 0.01 for .
Fig. 18a and b shows the pressure curves and the material axial
feeds for Non-aligned Double T shape – same-side (DTSS) THF parts.
Fig. 19a–d shows that the DTSS-1 and DTSS-2 set give satisfactory
results, whereas DTSS-3 and DTSS-EXP fail to form acceptable parts.
Hence DTSS families will also require data node spacing of 100 MPa
for K, 0.15 for n, and 0.01 for .
8. Experimental validation
In order to validate the proposed methodology, THF experi-
ments for aligned Double Y shape (DY) and Non-aligned double
T-shape (DTSS) families were carried out. The loading paths used in
the experiment were obtained by multidimensional interpolation
scheme.
Fig. 21. (a) Cross-section of DY die, dimensions are in mm and (b) cross-section of DTSS die, dimensions are in mm.
163
8.1. Test set-up and experimental procedures
The experimental set-up for THF is shown in Fig. 20a and b. The
tooling set-up consists of split dies and die holders for DY and DTSS
families, two axial cylinders (50 t capacity each) to provide the axial
feed, and a 150 t hydraulic press which is used to clamp the split
dies during the operation. The maximum fluid pressure that can
be attained in this hydroforming test rig is 20,000 psi (140 MPa).
The split dies were made of A2 steel and hardened to 62 HRC.
Fig. 21a and b shows the cross-section views of the split dies show-
ing the dimensions, which actually are the part dimensions sought
to be achieved by the experiments. The tubular material used for
the tests was stainless steel SS304 with material strength coeffi-
cient K = 1426 MPa and the strain hardening exponents n = 0.502.
The dimensions of the tube samples are given in Table 5.
Prior to the experiments, the tube and the dies surfaces were
cleaned by acetone. Lubrication was provided by a Teflon sheet
which was wrapped around the tube. For both DY and DTSS THF
experiments, the friction coefficient at the interface was estimated
as 0.05 for the applied Teflon sheet. The loading paths used for these
experiments were obtained from the database using the developed
interpolation scheme. All the experiments were carried out in 20 s.
It should be noted that all the loading paths residing in the database
were obtained at a simulation time of 0.01 s. The difference in the
process time should not account for any errors because the loading
path for cold forming is independent of the process time.
8.2. Experimental results and discussions
The hydroformed parts DY and DTSS are shown in Fig. 22a and b,
respectively. The dimensions of the protruded sections were mea-
sured and compared with the dimensions obtained from the FE
simulations as shown in Table 6. The protrusion height of DTSS
part matched very well between experiment and FE simulations.
The protrusion height measured from DY hydroformed part was
found to be 1 mm less than the protrusion height obtained from the
simulation. The deviation may be attributed to non-linear friction
exhibited at the interface. As mentioned earlier, a friction coeffi-
cient of 0.05 was assumed in the simulation. A slight increase in
the interface friction has the propensity to hinder material flow
to the die cavity. For the DY part more material was supposed to
be fed from the left side. Hence, the effect of friction was more
pronounced with this geometry compared to the DTSS geome-
try.
Wall thickness distribution for the DTSS and DY parts was also
compared between experiment and FE simulations as shown in
Fig. 23a and b. The thickness measurements were taken using
Vernier caliper after cutting the specimens along the longitudi-
nal direction. The thickness distribution plots only show measured
values from the protrusion side of the parts. It can be seen that
the thickness distributions obtained from the experiments closely
match FE simulation results. The percentage difference between
164 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166
Fig. 22. Split dies, initial tube and final part for (a) DY-EXP and (b) DTSS-EXP.
Fig. 23. (a) Thickness distribution comparison for DY part, final part length
138.5 mm and (b) thickness distribution comparison for DTSS part, final part length
152.6 mm.
experiment and simulation falls in the range from 0 to +10% for the
DY case and in −6.6% to +6.2% for the DTSS case.
9. Conclusions
A database scheme for real-time loading-path prediction for THF
parts was developed. The database contained material axial feed for
a wide variety of THF components. THF parts were classified into
families based on the geometrical configurations. The THF fam-
ilies identified were Bulge, Single Y, Aligned Double Y, Single T,
Aligned Double T, Non-aligned Double T – opposite side, and Non-
aligned Double T – same side. In order to obtain the loading path for
these THF families, consideration was given to geometric param-
eters, material parameters, and tribological parameters. With the
aid of FEA, simulations were carried out to establish the loading
paths. The variations of axial feed with respect to all variables were
analyzed for monotonicity, which is one of the critical conditions
for determining appropriate interpolation. The cubic spline inter-
polation was employed to provide a real-time loading path. The
validation of the proposed interpolation scheme was carried out in
two stages. In stage one, the interpolated loading paths were vali-
dated using FEA. That is, FEA was carried out using the interpolated
loading paths and then the formed geometry was assessed to deter-
mine if the part is successfully formed. The second validation was
carried out by actual THF experiments for DY and DTSS families.
The following conclusions are drawn from this study:
• Material feed data obtained from FE simulations have shown to
vary monotonically for all parameters (geometric, material, and
tribological), irrespective of THF family. This indicates that multi-
dimensional interpolation of the data to obtain the loading paths
could be done with high precision.
• The tested data to obtain the loading paths for various THF fami-
lies reveal that the interpolated loading paths are accurate if the
node-spacing interval is appropriate. Bulge-shaped parts require
node spacing intervals of 100 MPa for K, 0.15 for n, and 0.02 for
. Y-shaped parts require 100 MPa for K, 0.15 for n, and 0.01 for
. T-shaped parts also require 100 MPa for K, 0.05 for n, and 0.01
for .
• Experiments carried out for DY and DTSS families have shown
that the loading paths determined through the cubic spline inter-
polation scheme results in the part geometries that are very close
to the ones predicted by FEA.
• Multidimensional cubic spline interpolation allows the database
to output the required loading path within 0.85 s using Microsoft
Windows XP Professional x64 Edition, Version 2003; Intel®
Xeon® 4 CPU at 2.33 GHz; and 16.0 GB of RAM.
The method presented in this study to determine loading paths
for THF parts could reduce the cost and time incurred in THF oper-
ations. Furthermore, this method eliminates the need for running
multiple FE simulations for similar parts.
Acknowledgments
The authors would like to acknowledge the National Science
Foundation, through which this work was funded under Project
No. DMI-0448885. Any opinions, findings, and conclusions or rec-
ommendations expressed in this material are those of the authors
and do not necessarily reflect the views of the National Science
Foundation
Appendix A. 2-Dimensional cubic spline interpolation
To illustrate the interpolation scheme used in this study, mate-
rial strength coefficient (K) and strain hardening coefficient (n)
are chosen as variables, hence two grids namely K-grid and n-
grid will be used. The first grid K (aK , bK ) = {K0 , . . . , KnK }, nK ∈ N
is defined on the interval [aK , bK ] as a set of points KiK ∈ [aK , bK ];
where iK = 0, . . ., nK , aK = K0 K1 · · ·KnK = bK and KiK is the equidis-
tant nodes. Similarly the second grid is defined as n (an , bn ) =
{n0 , . . . , nnn }, nn ∈ N on the interval [an , bn ] as a set of equidistant
points nin ∈ [an , bn ]; where in = 0, . . ., nn and an = n0 n1 · · ·nnn = bn .
The interpolation data is the material feed and is defined as FiK in ∈ R;
where iK = 0, . . ., nK and in = 0, . . ., nn . However the material data
itself is an array of 101 displacement values corresponding to incre-
mental time interval, i.e. (FiK in )i , iT = 0, . . ., 100.
T
Now a so-called cubic spline of order three and smoothness
two sn3,2
K ,nn
∈ S3,3 (K , n ), where S3,3 (K , n ) is the function space
of all degree three and smoothness two splines on both K and n
grids, interpolates the data FiK in , if it satisfies the (nK + 1) × (nn + 1)
interpolation conditions. Although the cubic spline which is a piece-
wise third-order polynomial could be calculated by an intuitive
approach, it has certain disadvantages based on computation time
and effort. To overcome this problem, the notion of basis func-
tions of function space S3,3 (K , n ) is introduced where B-Spline
functions are the most commonly used basis functions (De Boor,
 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166 165

2001). Though one can use different ˚ functions, there could be some
variation in the numerical answers which would not be significant.
sn3,2
K nn
(KiK , nin ) = FiK in , where iK = 0, . . . , nK and in = 0, . . . , nn
Eq. (A.4) could thus be re-written as follows.
(A.1)

nK +3nn +3

sn3,2
K ,nn
(K, n) = CjK jn ujK (K)ujn (n) (A.9)
K − K(jK −1)−3 KjK − K jK =1 jn =1
Bj3,K (K) = 3−1,K
B(j −1)
(K) + Bj3−1,K (K)
K K(jK −1) − K(jK −1)−3 K KjK − KjK −3 K

(A.2) Since ujK and ujn vanish outside the respective bounded inter-
val [KjK −4 , Kj ] ∩ [aK , bK ] and [njn −4 , njn ] ∩ [an , bn ] due to the
bounded support supp (ujK ) = [KjK −4 , KjK ] ∩ [aK , bK ] and supp
 (ujn ) = [njn −4 , njn ] ∩ [an , bn ], Eq. (A.9) is further changed to the fol-
1 if KjK −1 ≤ K ≤ KjK
where, jK = 1, . . . , nK + 3 and Bj0,K = lowing form.
K 0 otherwise

n − n(jn −1)−3 njn − n

mK mn

Bj3,n (n) = 3−1,n

B(j (n) + Bj3−1,n (n) sn3,2
K ,nn
(K, n) = CjK jn ujK (K)ujn (n) (A.10)
n n(jn −1) − n(jn −1)−3 n −1) njn − njn −3 n
jK =lK jn =ln
(A.3)
 
where lK = (K − aK )/hK + 1, mK = min(lK + 3, nK + 3), ln =
  
1 if njn −1 ≤ n ≤ njn (n − an )/hn + 1, mn = min(ln + 3, nn + 3); •
denotes the floor
where, jn = 1, . . . , nn + 3 and Bj0,n =
function.
n 0 otherwise
The required equations for 2-dimensional cubic spline inter-
Now the nK + 3 B-Spline basis functions Bj3,K and nn + 3 B-Spline polations are defined and now the procedure to obtain the
K
basis functions Bj3,n form a basis of the function space S3,3 (K , n ). interpolated value of material feed F at the point (K, n) in the defined
n
boundary is presented below in steps.
The cubic spline sn3,2
K ,nn
or simply snK nn could therefore be presented Step 1: The first step is to calculate the coefficients Cj∗ q of the
as the linear combination of the tensor products of B-Spline basis K
following splines.
functions on grid K and n with coefficients CjK jn .

nK +3nn +3

nK +3
sn3,2
K ,nn
(K, n) = CjK jn Bj3,K (K)Bj3,
n
n (n) (A.4) sq,K (K) = Cj∗ q ujK (K) ∈ S3 (K ) (A.11)
K K
jK =1 jn =1 jK =1

If Eqs. (A.1) and (A.4) are equated, they would yield the values of
coefficients CjK jn and with the known values of CjK jn , the value of where q = 0, . . ., nn and the interpolation condition is,

any sn3,2
K nn
(K, n)
within the interval [aK , bK ] and [an , bn ] could be
calculated which is the interpolated value of material feed F at that sq,K (K) = FiK q , where iK = 0, . . . , nK (A.12)
point (K, n). Here the equidistant data nodes are chosen, so that the
following equations could be written. Since S3 (K ) is in nK + 3 dimensional space, nK + 3 interpolation con-
ditions are required to calculate the unique interpolating spline
bK − aK
KiK = aK + iK hK ; interval [aK , bK ]; hK = ; iK function. However from Eq. (A.13) only nK + 1 conditions are found.
nK
In order to get the remaining conditions the second-order deriva-
= 0, . . . , nK (A.5) tives of the spline function at the two boundaries are equated with
some constants. If the spline is assumed as ‘Natural Spline’ then
these constants would be zero (Behforooz and Papamichael, 1979).
bn − an Finally Eq. (A.11) could be re-written as Eq. (A.15) in terms of known
nin = an + in hn ; interval [an , bn ]; hn = ; in = 0, . . . , nn
nn values.
(A.6)

3
Now for equidistant nodes, the B-Splines shown in Eq. (A.4) could 
sq,K (K0 ) = Cj∗ q uj (K0 ) = 0 (A.13)
be explicitly presented in terms of some functions u. K K
jK =1
K − a 
K
Bj3,K (K) = ujK (K) = ˚ − jK + 2 , where jK
K hK

nK +3

= 1, . . . , nK + 3 (A.7) sq,K (KnK ) = Cj∗ q uj (KnK ) = 0 (A.14)
K K
jK =nK +1

n − a 
n
Bj3,n (n) = u (n) = ˚
jn − jn + 2 , where jn
mK
n hn sq,K (KiK ) = Cj∗ q ujK (KiK ) (A.15)
K
= 1, . . . , nn + 3 (A.8) jK =lK

 

3
where q = 0, . . ., nn , lK = (KiK − aK )/hK + 1, mK = min(lK + 3, nK + 3),
(2 − |t|) , 1 ≤ |t| ≤ 2 iK = 0, . . ., nK .
where, ˚(t) = 4 − 6|t|2 + 3|t|3 , |t| ≤ 1 After re-arranging Eqs. (A.13), (A.14) and (A.15) and also using
0, elsewhere Eq. (A.12), the following matrix equation is written, which could be
166 A. Ghosh et al. / Journal of Materials Processing Technology 211 (2011) 150–166

solved for the Cj∗ q .

K
⎛ ⎞
⎛ u (K ) u (K ) u (K ) 0 ··· ··· 0
⎞ ⎛ ∗
C1q ⎞
1 0 2 0 3 0 0
··· ··· ⎜ ⎟ ∗
C2q
⎜ u1 (K0 ) u2 (K0 ) u3 (K0 ) u4 (K0 ) 0 ⎟⎜ ⎟ ⎜ F0q ⎟
⎜ 0 u2 (K1 ) u3 (K1 ) u4 (K1 ) u5 (K1 ) ··· 0 ⎟⎜ ⎟ ⎜ F1q ⎟
∗
C3q
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ .. .. .. .. .. .. .. ⎟⎜ ⎟ ⎜ .. ⎟
..
⎜ . . . . . . . ⎟⎜ ⎟=⎜ . ⎟. (A.16)
⎜ ⎟⎜ ⎟ ⎜ ⎟
⎜ 0 ··· 0 u(KnK −1 ) u(KnK −1 ) u(KnK −1 ) u(KnK −1 ) ⎟ ⎜ Cn∗ +1q ⎟ ⎜ FnK −1q ⎟
⎝ 0 ⎠ ⎜ ∗K ⎟ ⎝ ⎠
··· ··· 0 u(KnK ) u(KnK ) u(KnK ) ⎝ Cn +2q ⎠ FnK q
un +1 (KnK ) un +2 (KnK ) un
K
0 ··· ··· ··· (KnK ) Cn∗ 0
K K K +3
K +3q

Step 2: The second step is to solve the cubic spline interpolation

problem sjK ,n ∈ S3 (n ).
nn +3
sjK ,n (n) = CjK jn ujn (n), jK = 1, . . . , nK + 3
jn =1
and the interpolation condition is,
sjK ,n (nq ) = Cj∗ q , q = 0, . . . , nn
K
With similar reasoning and approaches, the value of CjK jn could be
obtained using Eqs. (A.17) and (A.18).
Step 3: Since the coefficients CjK jn in Eq. (A.9) are known, for any
(K, n) value in the given interval, the corresponding spline could be
calculated, which is the cubic spline interpolated value of the mate-
rial feed (F) at that particular (K, n). Eq. (A.9) could be presented in
matrix notation as expressed in Eq. (A.19), where ujK (K), ujn (n) and
CjK jn denote (nK + 3) × 1, (nn + 3) × 1 and (nK + 3) × (nn + 3) matrices
respectively.
s(K, n) = [ujK (K)]T CjK jn [ujn (n)]
where jK = 1, . . ., nK + 3 and jn = 1, . . ., nn + 3
Non-equidistant nodes: For non-equidistant nodes similar
approach used for equidistant nodes could be applied, if the nodes
are mapped by re-scaling approach using certain grid function. This
grid function maps the non-equidistant nodes by a strictly mono-
tone, continuous one-to-one function to structure an equidistant
grid.
References
(A.17)
Asnafi, N., 1999. Analytical modeling of tube hydroforming. Thin-Walled Structures
34, 295–330.
Behforooz, G.H., Papamichael, N., 1979. End conditions for cubic spline interpolation.
IMA Journal of Applied Mathematics 23, 355–366.
(A.18) De Boor, C., 2001. A Practical Guide to Splines, revised edition Springer Publication,
New York.
Dohmann, F., Hartl, Ch., 1997. Tube hydroforming – research and practical applica-
tion. Journal of Materials Processing Technology 71, 174–183.
Gelin, J.C., Labergere, C., 2002. Application of optimal design and control strategies to
the forming of thin walled metallic components. Journal of Materials Processing
Technology 125–126, 565–572.
Habermann, C., Kindermann, F., 2007. Multidimensional spline interpolation – the-
ory and applications. Computational Economics 30, 153–169.
Koc, M., Altan, T., 2001. An overall review of the tube hydroforming (THF) technology.
Journal of Materials Processing Technology 108, 384–393.
Lin, F.C., Kwan, C.T., 2004. Application of abductive network and FEM to predict
an acceptable product on T-shape tube hydroforming process. Computers &
Structures 82, 1189–1200.
(A.19) Nordlund, P., 1998. Adaptivity and wrinkle indication in sheet-metal form-
ing. Computational Methods in Applied Mechanical Engineering 161,
127–143.
Ray, P., Mac Donald, B.J., 2004. Determination of the optimum load path for tube
hydroforming processes using a fuzzy load control algorithm and finite element
analysis. Finite Elements in Analysis and Design 41, 173–192.
Strano, M., Jirathearanat, S., Shr, S., Altan, T., 2004. Virtual process development in
tube hydroforming. Journal of Materials Processing Technology 146, 130–136.
Xia, Z.C., 2001. Failure analysis of tubular hydroforming. Journal of Engineering
Materials and Technology 123, 423–429.