Sie sind auf Seite 1von 5

# TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 2, PP. 49-53, FEB., 2013.

## Application of Finite Strip Method in Vehicle Design

Part 2 of 2: Post Buckling Analysis of thin Walled Structures
Umesh Gandhi 1, , Stephane Roussel 2, Katsuya Furusu 3, and Toshiaki Nakagawa 4

## Abstract: Thin walled parts of high strength steel, under

compressive loads are likely to buckle locally, and then
depending on geometry and material properties the section
may continue to carry additional load. For the post buckling
conditions the deformations are large but finite. Therefore we
need to consider geometrical non linearity in the calculations.
In this paper we are extending the linear finite strip element
formulation to include geometrical non linearity. Method to
derive secant and tangent stiffness matrix for nonlinear finite
strip element is developed and then the element formulation is
verified for in-plane and center load on a plate using Newton
Raphson solver. The new nonlinear finite strip element can be
useful in estimating maximum load capacity (including post
buckling) of thin walled structures from 2D data.
1

1.

u
x 1 w 2
order nonlinear

v
terms for large

2 x
displacement
2
y

w
u v

y x 2 y
P

P

w w ob L
2

w
2 x y o 0

2x
0
w

y 2

0
2 w

INTRODUCTION

## Thin walled parts of high strength steel, under

compressive loads are likely to buckle locally, and then
depending on geometry and material properties may
continue to carry additional load [1]. For the post buckling
conditions the deformations are not infinitesimally small
and therefore we need to consider geometrical non linearity
in the calculations. i.e. F=K(a)a; the stiffness K is not a
constant, but it is a function of displacement vector a. F is
force.
In the part 1 i.e. the companion paper by same authors
[2]
; we have presented the advantage of using finite strip
element in evaluating the stability of thin walled section.
The purpose of this paper is to develop geometrically
nonlinear finite strip element. We believe that such element
will help in extending CUFSM capability to estimate
maximum load capacity of thin walled structures.

2.

## displacement. These added terms make the expression

nonlinear.

STRAIN DISPLACEMENT
RELATIONSHIP

## The strain displacement relation for large out of plane

displacement for can be expressed as [3].
The u, v, and w as shown in the Fig. 1, are displacements
in x, y, and z direction in the Cartesian coordinates. The
second order terms are added to account for large
1, 2
Umesh Gandhi () and StehaneRoussel, Toyota Technical Center,
Ann Arbor,MI,USA(umesh.gandhi@tema.toyota.com and
stephane.roussel@tema.toyota.com)
3, 4
K. Furusu and T. Nakagawa, Toyota Central R&D Labs, Inc., Nagoya,
Japan (furusu@mosk.tytlabs.co.jp and tn@ket.tytlabs.co.jp)

w1

v1

(1)

w2

v2

u1

u2

b
1

Z
Y

w1

v1

w2
v2

u1

u2

## To develop the finite strip element let us choose shape

function from Cheung [3], note that same shape functions are
used for linear buckling in CUFSM [1, 4]. m is an integer.

x x u my
u 1 1 sin

b
b u 2 a'

(2)

x x v my
v 1 1 cos

b b v2 a'

(3)

## Alternatively w can also be expressed as,

w N1

N2

N3

N 4 w1 1

w2 2 T

(4)

where,
3x 2 2 x 3 my
N1 1 2 3 sin

b
b a'

RECEIVED: 19, SEP., 2012; REVISED: 15, JAN., 2013; ACCEPTED: 17, JAN., 2013; PUBLISHED: 26, JAN., 2013.

ISSN: 2345-234X

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 2, PP. 49-53, FEB., 2013.

50

2 x x 2 my
N 2 x1
2 sin

b
b a'

3x
2 x my
N 3 2 3 sin

b a'
b
2

Bob

x 2 x my
N 4 x 2 sin

b a'
b

## Strain in the finite strip can be expressed by substituting

u, v and w from (1) and (2),
B p 0 0 Bb
o
l u

1 v1 u2 v2
0 Bob 0 0
Linear
Non-linear

w1 1 w2 2 T

(5)

Let,
a u1 v1 u 2

v2

w1 1

[5-6]

Bop

Dp

E1 x E 2 0
D p x E 2
E2
0
0
0
G
D x D1
0

Db D1 D y
0
0
0 D xy

(6)
Bop

E2

and

Bob

## are same as in CUFSM [1, 3-4]. They are expressed as

(7) and (8). The nonlinear matrix is expressed as (9).

MATRIX

## Further using principal of virtual work we can express

nonlinear stiffness in terms of shape function and nodal
vector a as follows,

(11)

where,
E1

w2 2 T

3.

and

Blb

(10)

## and Db are constitutive relation based on material

properties as follows,

## Note that the displacement u, v, and w at any location in

the finite strip can be expressed as a function of x,y and
nodal parameters u1,v1,u2,v2,w1,1,w2, 2, using (2).

Non linear

Linear

B pT D B p

0
BopT D p Blb
0

K a o p o
bT

p
bT
b
0
BobT Db Bob Bl D p Bo Bl Db Bl

Ex
1 x y

Ey
1 x y

Dx

Ext 3
12 1 x y

D xy

Gt 3
12

D1
Dy

x E yt 3

12 1 x y

E yt

12 1 x y

Blb

## Note that terms in

matrix contains displacements
parameters, (w1, 1, w2 and 2) which makes the stiffness
matrix nonlinear. The nonlinear terms relates the in-plane
loads in u and v direction to the out of plane nodal
parameters w1, 1, w2, and 2.

4.

expressed as,
f K ( a )a
(11)

(7)

(8)

(9)

## TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST)

GANDHI et al.: APPLICATION OF FINITE STRIP METHOD IN VEHICLE DESIGN PART 2 OF 2: POST BUCKLING ANALYSIS OF THIN WALLED STRUCTURES.

## Where f is applied force, a is displacement vector (6)

and K(a) is nonlinear stiffness matrix.
Since K(a) is function of displacement a the system
cannot be solved using linear matrix inverse operation. We
need to use iterative approach. One of the most common
iterative approaches is Newton Raphson approach. Key
steps for Newton Raphson approach are presented below,

K ob BobT Db Bob dv
This linear stiffness part is same as in CUFSM
geometric stiffness matrix can be calculated as,

Z
Y
Tx
Txy

Kg t

Slope = Ks
Secant matrix

N1, x
Gm
N1, y

a2

Deformation

T
m

T x T xy
Gm dxdy
T
xy T y
N 2, x N 3, x N 4, x
N 2, y N 3, y N 4, y

## This is also same as in CUFSM

matrix can be calculated as,

## Note, for any point on the F-S curve, Ks is nonlinear

stiffness matrix, which is same as K(a) and Kt is tangent at
that point, which is derivative of K(a) with respect to
displacement vector a.
The typical steps used for computation based on Newton
Raphson approach are as follows,
1) Find initial value using linear stiffness matrix

0
Kn
K 21

[1, 3]

(16)

Kn Nonlinear stiffness

K12
K 22

(17)

## K12 BopT D p Blb dv

K 22

bT
b
l Db Bl dv

(18)

T
K 21 K12

a n K t 1 (at a = 0) f

## 2) Using an, calculate Kt and Ks

0 N b N b

2
x
1 ,
,

w
x
x

Blb 0
b
b

y N1 N 2

,
,
w w

y
(19)

y x
The element stiffness matrices Ks and Kt are calculated
using symbolic MATLAB.

## 3) Calculate imbalance force between internal and

external force, imbalance force should be =0 at
equilibrium,

r an K s a n a n f 0
4) Calculate next displacement vector from the
tangent matrix

a n1 a n K t 1 a n r a n

6.

## 5) Stop when displacement is not changing and

imbalance force is nearly zero.

5.

Ty

Slope = Kt
Tangent matrix

a1

, Kg

Txy

Tx

a0

[1, 3, 7-8]

Ty

r(a0)=Ks(a0)a0- f

51

VERIFICATION

## We implemented the new finite strip element as UEL in

ABAQUS and evaluated to verify the results for nonlinear
conditions. The comparison of center load and in-plane load
in a simple plate is presented in Figures 3-6.

EVALUATION OF KT TANGENT
MATRIX [3, 6]

## Derivation of Kt is available in [6]. Here we are

presenting only the expression to calculate Kt,
Kt Ko K n K g
(14)
Ko linear stiffness matrix can be calculated as,
K p
0
Ko o
b
0 K o
(15)
where,

Symmetry
u=0
v=0
=0

u= 0
v=0
=?
w=0

v (y)

uu (x)

## Fig 3.Steel plate 500x500mm 2mm thick, all edges fixed

K op BopT D p Bop du
TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST)

TRANSACTION ON CONTROL AND MECHANICAL SYSTEMS, VOL. 2, NO. 2, PP. 49-53, FEB., 2013.

52

12000
9000

## 2D data. For general loading and geometry conditions

additional development is required which is presented
in the next section.

## Finite strip 4el

Finite strip 2 el
ABAQUS 256 el, S8

8.

6000

3000

0
0

Deflection (mm)

## Fig 4. Centre load on a steel plate

Symmetry

u=0
v=imposed
=0
w=?
v (y)

uu (x)

w=
=

vertical displacement
rotation about y axis

Fig 5.Steel plate 500x500mm 4mm thick, all edges simply supported

## Mid point deflection (mm)

Finite strip 4 el

## The heading of the Acknowledgment section and the

References section must not be numbered.

ABAQUS 256 el S8

12

REFERENCES

## Step1: 10N load @ center

Step2: In-plane shortening

4
0
0

0.2

0.4

0.6

0.8

## We observe that the finite strip element matches well

with the ABAQUS results. Here we are showing ABAQUS
analysis with 256 elements; ABAQUS results with 64
elements were also similar.

7.

## Element orientation: The element stiffness matrix

derived is in global direction. In future, orientation can
be addressed in UEL with simple mapping operation.
Element size: Only one size of element is possible
defined as width and length in UEL. In future this can
be easily addressed in UEL.
In plane force: In plane for can be calculated by
integrating stress in the longitudinal direction over the
area of the cross section. A sub routine to address this
is required.
In ABAQUS with the current UEL we can use only
Newton Raphson solver. We cannot use RIKS method.
Modification in UEL, for the RHS calculations is
necessary to include the RIKS solver.

ACKNOWLEDGMENT

Finite strip 2 el

16

CONCLUSIONS

## The main purpose for this paper was to demonstrate

feasibility of finite strip element for nonlinear
geometry which is achieved. A new nonlinear finite
strip element is developed and verified for out of plane
and in-plane load through comparison with ABAQUS.
Method to use UEL in ABAQUS is developed. Use of
UEL, in developing finite element is quite helpful as
we can use ABAQUS infrastructure to assemble the
elements, apply boundary conditions and use robust
nonlinear solver.
In future such nonlinear finite strip element can be
implemented in CUFSM to extent its capability to
estimate maximum load for thin walled structures from

## [1] Schafer, B.W., dny, S. Buckling analysis of

cold-formed steel members using CUFSM:
conventional and constrained finite strip methods.
Eighteenth International Specialty Conference on
Cold-Formed Steel Structures, Orlando, FL. October
2006.
[2] Umesh Gandhi, Stephane Roussel, Katsuya Furusu,
and Toshiaki Takagawa, "Application of Finite Strip
Method in Vehicle Design Part 1 of 2: Linear Buckling
Analysis of Thin Walled Structures," TSEST
Transaction on Control and Mechanical Systems, Vol.
2, No. 1, Pp. 9-12, Jan., 2013.
[3] Cheung Y. K.; Tham Y. K. Finite strip method, 1998.
[4] CUFSM : Software for linear buckling analysis using
finite
strip
element
:
http://www.ce.jhu.edu/bschafer/cufsm/
[5] Eccher G, Rasmussen K, Zandonini R. Geometric
nonlinear isoparametric spline finite strip analysis of
perforated thin-walled structures. Thin walled
structures, 2008, V 46, P1224-1235.
[6] Zienkiewicz, O. C. The finite element method, Third
edition, Chapter 19. 1979.
[7] Dawe DJ, Wang S. Post buckling analysis of thin
rectangular laminated plates by spline FSM.
Bicentenary Conference on Thin-Walled Structures, 2
4 December 1996, University of Strathclyde, UK. 1996.
[8] ABAQUS user manual for UEL.

## TRANSACTION SERIES ON ENGINEERING SCIENCES AND TECHNOLOGIES (TSEST)

GANDHI et al.: APPLICATION OF FINITE STRIP METHOD IN VEHICLE DESIGN PART 2 OF 2: POST BUCKLING ANALYSIS OF THIN WALLED STRUCTURES.

53

w

v1

v2

u1

u2

b
1

Z
Y

v1
u1

8
9
10
11

a
a
b
E1, E2
f
G
K(a) =
Ks
Kg
Kn
Ko
Kt or KT

12

13

N1, N2,
N3, N4
x, y
t
1, 2
u
u1, u2
v
v1, v2
w
w1, w2
x
Ym

1
2
3
4
5
6
7

14
15
16
17
18
19
20
21
22
23
24
25

w
v2

u2

## Displacement vector for the finite strip (5)

Length of finite strip
Width of finite strip
Elastic modulus in X and Y direction
Applied force
Shear modulus in X-Y plane
Total stiffness matrix or secant matrix
Tangent stiffness matrix (13)
Linear stiffness matrix (14)
Geometric stiffness matrix (15)
Nonlinear stiffness matrix (16)
Integer used in the sinusoidal shape function.
m=1 used for all the calculations
y/a
Shape function for the finite strip element (3)
Poisons ratio in X and Y direction
Thickness of the finite strip element
Orientation about y axis at node 1 and node 2
Displacement function in x direction (2)
X displacement at node 1 and node 2
Displacement function in y direction (2)
Y displacement at node 1 and node 2
Displacement function in z direction (2)
Z displacement at node 1 and node 2
x/b
sin(m)

## Available online at:

http://tsest.org/index.php/TCMS/article/view/79