Sie sind auf Seite 1von 9

computers & Slruclwes Vol. 52, No. 3. pp. 387-395.

1994
Copyright C 1994 Elwier Science Ltd
Pergamon 00457949(94)EO162-U Printed in Great Britain. All rights reserved
0045-7949/9457.00
+ 0.00

FINITE ELEMENT ANALYSIS ON HORIZONTAL VESSELS


WITH SADDLE SUPPORTS
L. YANG, C. WEINBERGER and Y. T. SHAH
College of Engineering, Drexel University, 32nd and Chestnut Streets, Philadelphia, PA 19104, U.S.A.

(Received 21 June 1993)

Abstract-Finite element analysis is performed to predict the stress distributions within a water-filled
horizontal vessel with hemispherical ends and two or three saddle supports. Computations are performed
for various values of design parameters. The wall thickness was varied from 0.5 to 70 mm; the vessel length
from 10 to 30 m; and the vessel radius from 0.5 to 2.0 m. The effects of the pressure and a third saddle
support are also investigated. The maximum tensile and shear stresses are compared with the predictions
of Zick’s analysis. Design correlations are obtained, based on the numerical results.

NOTATION The results gave a detailed distribution of displace-


A area ment and local stresses in the saddle supports area.
b width of saddle The results showed that the maximum tensile stress is
E Young’s modulus located near the horn of the saddle. But the effects of
g gravitational force design parameters were not investigated although the
H distance from saddle to end
effects of saddle locations were studied.
H, distance from saddle to seam
4 interpolation functions Once the required volume is set, the optimal shape
L length of vessel of the vessel can be described by the length-to-diam-
P transverse. loading per unit area eter ratio and either the diameter or the length.
Q total load per saddle Optimal shape can be defined here as that configur-
number of nodes of the element
i radius of vessel ation which will produce a design with minimum wall
r radial coordinate thickness or vessel material, consistent with maxi-
t thickness of shell mum tensile or shear stress for the material of
u displacement construction. In this work, the objective is to compute
w transverse displacement
coordinates the stress distribution within the vessel walls as a
X,Y,Z
function of pressure and the geometric variables of
Greek symbols vessel shape, wall thickness and saddle location. A
It1 By rotations of the normal to the undeformed middle finite element technique is used to determine the
surface stresses within the vessel walls. These computed
correction factor
e”
results are compared with experimental results ob-
circumferential coordinate
8’ nodal point rotations at point i tained previously [5]. The results are also used to
o normal stress evaluate the empirically-based design equations de-
bb principal stress veloped by Zick [l]. Based on the numerical results
r shear stress improved design correlations are obtained.
V Poisson’s ratio

1. INTRODUCTION
2. FINITE ELEMENT ANALYSIS
Horizontal pressure vessels are widely used in process
and power industries. For the most part these vessels Finite element analysis is used to solve numerically
are designed using empirical correlations [I]. Many the governing equations for stress within the wall
researchers have done analytical and numerical inves- material. The solutions provide a complete stress
tigations on the stress distributions within the press- distribution in the vessel. The vessel is considered as
ure vessels. Duthie and Tooth [2] presented an a shell structure due to the thin wall thickness. A
analytical shell solution using a Fourier expansion sketch of the vessel is shown in Fig. 1. The wall
technique while Stoneking and Sketch [3] performed thickness is assumed constant everywhere. The ma-
a two-dimensional finite element analysis. Widera et terial is isotropic and the theory of elasticity is used
al. [4] performed a three-dimensional finite element in the approach to the solution. The principle of
analysis on the design of horizontal pressure vessel. virtual displacements for the shell element in the .x,v

387
388 L. YANGeI ul.

fkf
R

c
z
e r

Fig. I. Sketch of a horizontal vessel with hemispherical ends.

plane is then obtained after assuming that the total /IVare the rotations of the normal to the undeformed
potential must be stationary [6]. Thus middle surface in the x-z and y-z planes, respect-
ively. For the finite element analysis we use

I 6iCrC,K dA
JA
+ 1 6yrC,y dA -
JA
1 6wp dA = 0,
JA w = f h,Wi, B,= f hi&, jJ = i: /r,e:, (3)
(1) r=l ,=I ,=1

where where hi are the interpolation functions and q is the


number of nodes of the element. 6: and 0.; are nodal
-aa, point rotations at point i.
--aw
ax Because the geometry of a pressure vessel with two

K= -2 ap
ay
a 8~._[
__”
ay
,y=

ax
ay B.”
aw
-gj+Px 1 supports is symmetric both to its vertical plane
parallel to the vessel axis and to its middle plane
perpendicular to this axis, only a quarter of the vessel
was considered as the computational domain. The
nondisplacement boundary conditions are employed
for the saddle support locations and symmetric con-
ditions are given along symmetry lines.
The first boundary condition, (BC I), expresses a
fixed location for the shell where the saddle meets the
shell:atH<z<H+b, -54”<Q<54”,r=R
Etk
c, = ___
2(1 +u) [ 1.
1 0
0 1
u, = llg = u, = 0. (4)

The stresses are obtained by the following equations The second boundary condition, (BC II) is applied to
the symmetry plane perpendicular to the z-axis and
expresses the absence of material shearing strain in
the r-z and 0-z planes and absence of displacement
in the z-direction: at z = L/2; r = R
The coordinate system is shown in Fig. 2. In above
equations, p is the transverse loading to the wall per au,laz = 0, at4,/az = 0, U, = 0. (5)
unit area, t is the shell thickness, E is Young’s
modulus, w is the transverse displacement, and /IXand The third boundary condition, (BC III) is applied to
the intersection of the cylindrical part of the shell and
the symmetry plane passing through the z-axis and
expresses the absence of shearing strain in the r-0 and
z-8 planes and absence of displacement in the 8-
direction: at 0 = O”, 180”; r = R

au,/aO = 0, au,jae = 0, ug = 0. (6)

The transverse loading on the inside surface of the


vessel is simply the normal stress or pressure arising
from the hydrostatic head of fluid of density p
Fig. 2. Deformation assumptions in analysis of plate includ-
ing shear deformations. p =pgR(l -cosQ) at r = R. (7)
Finite element analysis of a horizontal vessel 389

Fig. 3. The node and mesh layout.

The ABAQUS [7] program was employed to solve stresses as a function grid sizes. The results show that
the above set of equations. The mesh layout on the when the grid size on the cylindrical surface is 200
cylindrical surface is shown in Fig. 3(a). The S8R5 (lo,20 for r, 0), the maximum stresses level out. This
type of element which has eight nodes in one element means that the calculated results will not depend on
was used. The projection of node and element layout the grid size when the grid size is above 200 and the
of the hemispheres is shown in Fig. 3(b). Triangular grid density is above 2/m. The grid size of 200 then
elements were used in the hemisphere region and the was selected for most of future computations. For the
element pattern might be different depending on the range of parameters considered, the above grid was
number of nodal points in circumferential direction. found to be adequate.
The S4R type of element which has four nodes in one
3.2. Comparison of numerical predictions with exper-
element was used for hemispheres. The material was
imental data
assumed elastic and the Poisson’s ratio, u, was se-
lected as 0.3 for carbon steel. The computations were In order to verify the validity of numerical ap-
performed on an IBM 9121 computer. The typical proach, the calculated results were compared with the
CPU time used for one complete run was about experimental data under some typical conditions.
300 sec. Zick and Carlson [5] measured the stresses in a long
horizontal storage tank containing liquid subjected to
certain strains produced by two saddle supports by
3. RESULTS AND DISCUSSIONS
means of nondestructive tests using bonded resistance
wire strain gages. The dimensions were chosen as
3.1. Convergence test
follows: R =2.9m, L = 19.8m, b =0.3m, A = 0,
In order to determine proper grid size for conver- 2.7 m and t = 2.5 cm.
gence, a grid dependency study was performed. The Values of tensile stress were measured at the lo-
calculations were performed for the following geome- cations (point x, 0 = 85”) shown in Fig. 5. When
try: R=1.5m, L =20m, b=0.2m, A =0.75m, H, = 2.9 m, the saddles were 2.9 m from end seams,
t = 2 cm: which were used in the example by Zick [l]. the tensile stress at point x was 1.57 x 10’ Pa by
The contact angle of saddle with the vessel was experiment. When H, = 0 m, the saddles were at end
assumed to be 108”. The vessel was full of liquid seams, the tensile stress at point x was 0.92 x 10’ Pa.
water and the pressure was assumed to be ambient. The comparison of FEM results with these exper-
Figure 4 shows the maximum tensile and shear imental observations is shown in Table 1.
The predicted tensile stress at the specific point are
slightly higher than the experimental data. However,
they agree within 9%, which indicates that the pre-
sent FEM approach can predict the stress distri-
bution in the liquid vessel quite well.

3.3. Typical stress distributions


The geometric conditions given in Sec. 3.1 are
considered as the base case, namely, R = 1.5 m,
L =20m, b =0.2m, A =0.75m, t =2cm, and the
vessel full of liquid water. The contact angle of saddle
Fig. 4. Location of strain gage mounted on the vessel. with the vessel was 108”. Figure 6 shows the defor-
390 L. YANGet al.

D=3m
L=20m

Full of liquid
water

0 50 100 150 200 250

Grid Size
Fig. 5. Illustration of sufficiency of grid size = 200.

mation of the pressure vessel under the loading of and 9 contain both values for the computed stresses
water. The magnitude of deformation is exaggerated and curve fit equations. Again, the basis for these
for convenience. According to the figure, the maxi- calculations is a tank 3 m in diameter and 20 m long,
mum displacement occurs at the center of the vessel, filled with liquid water. The wall thickness was varied
although the maximum stresses occur near the saddle from 0.5 to 6 cm. These two figures indicate that both
supports. Figure 7 shows the projected contours of maximum tensile and shear stresses are inversely
the von Mises equivalent stress on the vertical plane. proportional to the wall thickness.
The equivalent von Mises stress is defined as fol-
lows [8]: 3.5. Effects of vessel length and radius on maximum
stresses
6, = (0; + ui)“‘, (8)
Figures 10 and 11 illustrate how vessel length
where e. and (T,,are principal stresses. The maximum affects the maximum tensile and shear stresses. For a
value of the von Mises equivalent stress is found near fixed wall thickness of 2 cm and the tank diameter of
the horn of the saddle, the same location as predicted 3 m, Fig. 10 shows that the maximum tensile stress
by Widera et al. [4]. Another peak value of von Mises urnaxcalculated by the FEM increases almost linearly
equivalent stress is found at the top of the middle with length of the vessel. In Fig. 11 the maximum
plane of the vessel. shear stress 7,,, is shown to vary as a second order
function of vessel length. For typical sets of values of
3.4. Effects of wall thickness on maximum stresses
vessel radius and wall thickness, Figs 12 and I3
In engineering design we are mainly concerned illustrate the behavior of the calculated maximum
with the maximum tensile and shear stresses within tensile and shear stresses as a function of vessel
the structure. Selections of structure dimensions and radius. The FEM results show that the maximum
material are based on these stresses. The system
should be designed such that the maximum stresses
are below the yield stress. Figures 8 and 9 illustrate
some typical results showing the decrease in maxi-
mum tensile stress, cr,,, and maximum shear stress,
T,,,,,~, with increasing wall thickness, t, based on the
FEM calculations. These maximum stresses can oc-
cur at any point within the tank walls but they
typically occur near the saddle locations. Figures 8

Table I, Comparison of FEM predictions with experimen-


tal data
Experiment FEM
6, (IO’Pa) 6, (IO’Pa) Difference
A = 2.9 m 1.57 1.69 1% Fig. 6. Deformation of the vessel under loading for the base
A =Om 0.92 1.01 9% case (t = 2 cm, L = 20 m, R = 3 m).
Finite element analysis of a horizontal vessel 391

KanRpi?L
%lax= (13)
21 .

According to eqns (12) and (13), o,,, is proportional


to the square of L, inversely proportional to t, and
independent of R. rmaxis proportional to L, R, and
Fig. 7. Contour of the von Mises equivalent stress for the
basecase(t=2cm,L=20m,R=3m). inversely proportional to t. By using the same tank
dimensions as in the FEM analysis, one obtains stress
predictions which can be compared directly with
tensile stress Q,,, linearly increases with the vessel those of the FEM analysis. Subject to the assump-
radius R, while the maximum shear stress rmax de- tions of elastic and isotropic behavior, the FEM
pends on R’. analysis can be assumed accurate. Therefore, these
3.6. Comparison with the prediction of Zick analysis figures enable us to quantify the validity of the Zick
design correlations. Figure 8 shows that the Zick
Figures 8-13 also allow direct comparison of the correlation is accurate and Zick’s assertion that the
finite element results with the design equation predic- tensile stress crmaXat - ’ [eqn (15)] is quite good. How-
tions of Zick. Zick [1] assumed that the vessel shell ever, Fig. 10 shows that the Zick assertion u,,,aL2
acts as a beam over the two supports to analyze the is not good, since the FEM analysis indicates that a
stresses induced in the shell by the supports. By using linear proportional relationship is more accurate, at
Zick’s method of analysis the expressions for maxi- least for a tank length below 20 m. When L is between
mum tensile and shear stresses can be expressed as 20 and 28 m, t is between 5 and 70mm, and R is
between 1.1 and 1.4 m, the Zick predictions for tensile
KI QL
fJmax
= 4nRZt (9) stress, u,,, , are within 25% of the computed values
(Figs 8, 10, and 12). However, as shown in Fig. 12,
urna. increases linearly with radius R, a dependence
(10) not shown by the Zick analysis. Over the same range
of geometric values outlined above, the Zick predic-
where Q is the total load per saddle, L, R, and t are tions for maximum shear stress, rmax, are all much
the length, radius, and thickness of the vessel, respect- lower than the computed results as shown in Figs 9,
ively. The empirical coefficients K, and K4 depend on 11, and 13. For large R and L, the predicted r,,, is
design conditions of the tank [l]. When R is relatively about 30% of the correct value and this percentage
small compared with L. the load per saddle Q could increases to about 80% as R decreases to 0.5 m and
be expressed as L decreases to 10 m. Although one might expect that
this may lead to significant problems in design, the
Q = OSzR’pgL. (11) Zick analysis normally calls for design to withstand
maximum tensile stresses, and the resulting wall
Equations (9) and (10) could be rewritten as follows: thickness is sufficient to handle the maximum shear
stress computed by finite element analysis.
K,pgL2 From Figs 8, 10, and 12 we can assume that the
Qmax= yg-- (12)
maximum tensile stress is a linear function of RL/t,

-1.0 -0.5 0.0 0.5 1.0 1.5 2.0

Mt [cm1)
Fig. 8. The maximum tensile stress as a function of wall thickness for L = 20 m. R = 3 m.
392 L. YANG et al.

0 Zick

0 AM

-1 0 1

InOkm1 )
Fig. 9. The maximum shear stress as a function of wall thickness for L = 20 m, R = 3 m.

0 10 2; 3;
L(m)
Fig. 10. The maximum tensile stress as a function of vessel length for t = 2 cm, R = 3 m.

z “NIX= 0.133L
3

&f
*0
= 2

9
P

-
C
0 10 20

L (ni 1
Fig. 11. The maximum shear stress as a function of vessel length for I = 2 cm, R = 3 m.
Finite element analysis of a horizontal vessel 393
4

o max = 0.34 + 1.4R

3-
c
2
0

b
z

0.5 1.0 1.5 2 I)

R(m)

Fig. 12. The maximum tensile stress as a function of vessel radius for t = 2 cm, L = 20 m.
4

3
h

a”
0 tick
IcO

d 2
@EM I
1

PE

1
t’
tc

0 -I-
0.5 1.0 1.5

R(m)

Fig. 13. The maximum shear stress as a function of vessel radius for t = 2 cm, L = 20 m.

0 IO 20 .

LR/t
Fig. 14. Maximum tensile stress as a function of LR/r.
394 L. YANG et a!.

element method) to predicted stress (Zick’s analysis)


by the symbol 9

(16)

The correlation factor can be derived as follows:

0, = 1230t/L2 + 16.7RIL (17)


I
P
@, = 36.3tlRL + 0.646. (18)

0: . Therefore, eqns (12) and (13) given by Zick can be


0 10 20 30 40 modified and this can be done by adding a correction
LR/t factor @ into the equations below.
For tensile stress
Fig. 15. Maximum shear stress as a function of LR/t.

(19)
then a scatter plot was obtained for a group of design
parameters (see Fig. 14) and in the parameter range
For shear stress
considered the maximum tensile stress could be ex-
pressed as

o,,, = (11.6 + 0.157RL/t)10S [Pa]. (14)


3.8. Egects of pressure on maximum stresses
For the maximum shear stress we have (see Fig. 15)
Figure 16 shows the effects of the pressure inside
the horizontal vessel on the maximum stresses. The
t,, - (43.6 + 0.775RL/t)l@ [Pa]. (15) dimensions used in the base case were considered.
According to the Zick analysis, the maximum longi-
The relative errors are 3% and 10% for u,, and t,, , tudinal tensile stress in the vessel is a linear function
respectively. The detailed results are given by Yang et of pressure and the maximum shear stress is indepen-
al. (91. dent of pressure. However, the FEM results show
that the maximum tensile stress is almost constant
3.7. Modification of Zick’s analysis when pressure is small and it increases with pressure
One of the advantages of the Zick design method more and more rapidly. When pressure is higher
is its simplicity and ease of application. In order to than 1 bar the a,,,, predicted from FEM increases
retain this ease of use and still incorporate the results at a faster rate than the one predicted from the
of the present finite element analysis, it is appropriate Zick analysis. The computed results also show that
to apply certain correction factors to the Zick predic- TmaXpredicted from FEM is a weak function of
tive equations. Defining the ratio of true stress (finite pressure.

- Zick ( tmsik )

p ( bar >
Fig. 16. Pressure effects on the maximum stresses.
Finite element analysis of a horizontal vessel 395

vessel with hemispherical ends and two and three


saddle supports. The maximum equivalent stress is
found near the horn of the saddle. For the three
saddle support vessel, the maximum tensile stress
occurs near the central saddle. The maximum values
of both tensile and shear stresses in case of three
Fig. 17. Contour of the von Mises equivalent stress for the saddle supports are considerably higher than those
case with the third saddle support (I = 2 cm, L = 20 m,
for two saddle supports. The three saddle support
R = 1Sm).
design for a horizontal pressure vessel is therefore not
recommended. The parametric study shows that the
3.9. Effects of the third saddle support maximum tensile stress is inversely proportional to
If the third saddle support is added into the system the wall thickness t, and directly proportional to
the stress distribution will be much different. The length L and radius R of the vessel. The maximum
Zick analysis does not include the effects of the shear stress is inversely proportional to the wall
multiple saddle supports. In this paper the additional thickness. It also increases with increasing vessel
saddle support is considered at the midsection of the length and radius.
vessel. The boundary condition for the third support The computed results are compared with the pre-
is expressed as: at L/2-b/2 6 z <L/2, -54” Q dictions of Zick’s analysis. The Zick analysis does not
0<54’,r=R predict maximum shear stress accurately, although it
does predict maximum tensile stress within about
u, = u,, = u2 = 0 (21) 25% for wall thickness in the range of 5-70 mm,
length of 2&28 m, and radius of 1.2-l .5 m. For
while other boundary conditions are unchanged. shorter length and larger radius, the Zick analysis
Figure 17 shows the projected contours of the von significantly underpredicts the maximum tensile
Mises equivalent stress on the vertical plane. The stress.
maximum value of the von Mises equivalent stress is
found near the horn of the third saddle. However, the
stresses on the other saddle are not reduced. Another
peak value of von Mises equivalent stress is found Acknowledgemenf-The authors gratefully acknowledge the
also at the top of the middle plane of the vessel. The support from the Cochrane Co., King of Prussia, PA.

maximum value of tensile stress is 8.37 x 10’ Pa,


which is 240% higher than the maximum tensile
stress for the two saddle support case. The maximum REFERENCES
value of shear stress is 2.13 x 10’ Pa, which is 95%
1. L. P. Zick, Stresses in large horizontal cylindrical
higher than the two saddle support case. Due to these
pressure vessels on two supports. The Welding Journals
results, the three saddle support design for a horizon- Research Supplement, September (1951).
tal pressure vessel is not recommended. 2. G. Duthie and A. S. Tooth, The analysis of a horizontal
The finite element method is well developed and a cylindrical vessel supported by saddle welded to the
very popular technique for the computer solution of vessel. The International Conference on Pressure Vessel
Technology, Part I, Design and Analysis. Tokyo, Japan,
complex problems in solid mechanics. With FEM
p. 25 (1977).
analysis we found out that the horizontal pressure 3. J. E. Stoneking and H. Sketch. The analysis of a large
vessel with two saddle supports at the ends does not horizontal saddle supported pressure vessel. ASME
act as a beam for most of the cases considered. Under Technical Paper, 77-PVP-18, New York, NY (1977).
the beam assumption, the Zick analysis can predict 4. G. E. 0. Widera, Z. F. Sang and R. Natarajan. On the
design of a horizontal pressure vessel. J. Pressure Vessel
the maximum tensile and shear stresses as a function Technol., Trans. ASME 110, 393 (1988).
of wall thickness quite well. But there are large 5. L. P. Zick and C. E. Carlson. Strain gage technique
differences between the results from two methods employed in studying propane tank stresses under ser-
when the length and diameter of vessel are varied. vice conditions. Steel. 86-88. April 12 (I 948).
6. K. J. Bathe, Finite Element Procedures in Engineering
Even so. the predictions from two methods are very
Analysis. Prentice-Hall, Englewood Cliffs, NJ (1982).
close under certain design conditions. When studying 7. ABAQUS User’s Manual, v 4.9, Hibbitt, Karlson &
the effects of design parameters, only one variable is Petersen Inc., Pawtucket, RI (1990).
changed while others are maintained constant. 8. F. P. Beer and E. R. Johnston. Design of beams and
shafts for strength. In Mechanics of Materials.
McGraw-Hill, New York (1981).
4. CONCLUSIONS 9. L. Yang, C. Weinberger and Y. T. Shah, Finite element
analysis on horizontal pressure vessels. Research report
The finite element method has been used to predict submitted to Cochrane Co., King of Prussia, PA
the stress distributions within a horizontal pressure (1993).

Das könnte Ihnen auch gefallen