Sei sulla pagina 1di 9

Finite element analysis of thick cylinder with inner pressure1.

Eric Hennes, November 1999


Mechanics Department
University of Amsterdam

adressed to: MSC-Nastran, Gouda, MARC-section

Introduction & Summary................................................................................................................................ 1


Geometry, boundary conditions and material constants. ................................................................................ 2
The elastic linear case..................................................................................................................................... 3
The plastic case. ............................................................................................................................................. 3
Post processing aspects................................................................................................................................... 3
Discussion of the results................................................................................................................................. 4
Appendix I: graphical results linear elastic models at 400 MPa..................................................................... 5
Appendix II. Results for plastic quad4 and quad8 models at 400 MPa. ......................................................... 7
Appendix III. Radial stress within inner two elements: different extrapolation options ................................ 9

Introduction & Summary.


In this paper we show and discuss the results of calculation of the stresses in a infinitely
long pressurized thick-walled cylinder (say copper), using MARC 7.3 with different
element types, both in the elastic and plastic case. This small study was initiated by
inexpected stress discontinous in a more elaborate axisymmetric model, called ‘explovat’,
by Maarten Oudendijk. In order to trace, understand and eliminate these erroneous results
a very simple model is chosen exhibiting the same effects. It turns out that erroneous
results occur at the mesh boundary only when using quad4 linear elements, and increase
drastically in case of (high) plastic deformation. It is also found that this behavior is not
an artefact of the interpolation procedure applied in the postprocessing program. The
quad8 elements show nice continuous stresses that are well predicted in the elastic case.
We do not understand why the finite element analysis results in the discontinuities in the
quad4 (in particular plastic) cases: is it inherent to the analysis method or does the
analysis program need improvement/debugging?

1
Related files/documents: explovatplots.doc, thick_cylinder.xls, thick_elasq4.mud, thick_elasq8.mud,
thick_plasq4.mud, thick_plasq8.mud.

1
Geometry, boundary conditions and material
constants.
We choose, more or less arbitrarily
Inner radius ri = 10 mm
Outer radius ro = 20 mm
Young modulus E = 100000 N/mm2
Poisson’s ratio ν = 0.3
Inner pressure p = 0, 20,40, .. 400 N/mm2

and in the plastic case (see also Figure 2):


Yield stress σY = 100 N/mm2
Hardening rule (σV.M. − σY) /εV.M = 5000 N/mm2

Because of the axial symmetry we only model a small segment


z = 0, ..1 of the infinitely long tube, and impose these
boundary conditions:

∆z z=0 = ∆z z=1 = 0 ,∀ r, ∀θ

This condition assures that the tube is not allowed to change its
length, i.e. the axial strain is 0 ( ‘plain strain’). As a result the
problem is reduced to a 1-dimensional problem. Figure 1
shows the mesh.

Figure 1. FE mesh of cylinder

1200

1000
stress (Mpa)

800

600

400

200

0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
strain

Figure 2. Stress-strain relation.

2
The elastic linear case.
The case of linear material behavior is calculated with inner pressure p = 400 N/mm2.
No other non-linearities are taken into account, i.e. no geometrical effects. We have not
corrected the nodal external forces for the nodal displacement (i.e. no follower force), and
the FE-calculation is purely linear. Two types of quadrilateral elements have been used:
4-node linear (type 10) and 8-node quadratic (type 28). The finite element model results
have been obtained using the Marc/Mentat files ‘thick_elasq4.mud’ and
‘thick_elasq8.mud’. The numerical results are copied into the sheet ‘elastic 400 MPa’ of
the Excel file ‘thick_cylinder.xls’, in which also the analytical calculations are done.
In Appendix I the stresses and displacement as a function of r are shown, both for the
quad4 and quad8 models, as well as their differences with an analytical model.
These analytical results are obtained from (see standard textbooks):

r02 / r 2 − 1 r02 / r 2 + 1 2νp


σr = −p , σθ = p , σ z = ν (σ θ + σ r ) =
r / ri − 1
0
2 2
r / ri − 1
0
2 2
r / ri 2 − 1
0
2

These formulas predict how the radial stress decreases from p to 0, as r goes from ri to ro.
The circumferential stress is expected to change in our case from 666,6 to 266,6 N/m2
The axial stress is a constant, in this case 2*0.3*400/3 = 80 N/m2

The plastic case.


We have simulated the plastic case with inner pressures increasing in20 Mpa steps up to
400 MPa, using again quad4 and quad8 elements (FE-analysis files: thick_plasq4.mud
and thick_plasq8.mud). In Appendix II we present graphical results for the highest-
pressure case (numerical values in sheet ‘plastic’ of Excel file thick_cylinder.xls), where
the strains are of order 0.1.

Post processing aspects.


The above-mentioned results have been obtained using the default post processing
settings for scalars. Using these specific settings (called ‘post_extrap_linear in Mentat),
the nodal stress values are calculated from the integration point values according to these
steps (following Mentat documentation):

For each element: calculate the average of the integration point values and assign it to the
element’s centroid.
For each integration point: extrapolate linearly from the centroid through the integration
point to the corresponding nodal point.
For each node: average the extrapolation results (step 2) of all elements that share this
node.

Other optional settings include: ‘post_extrap_translate’ in which the integration point


values are copied onto the nodes without any extrapolation, and ‘post_extrap_average’, in
which the centroid value (see step 1) is copied onto the element’s nodes. In all these cases
step 3 is also applied. Finally, it is of course also possible to extract the individual values
of all integration points.

3
In order to find out the effect of these settings we have listed in Appendix III, as an
example, the radial stress of the elastic and plastic quad4 cases for different settings
(numerical values in sheet ‘postp. options’ of thick_cylinder.xls). We also show the
integration point values of the two inner elements for both quad4 and quad8 (elastic and
plastic) cases.

Discussion of the results.


Inspection of the linear results (Appendix I) leads to these three observations:
The quad4 results are very inaccurate at the borders of the mesh. This holds in particular
for the axial stress. For instance, at the inner border its error is of the order of the stress
itself.
The quad8 model does not show the above-mentioned border problems, and is also more
accurate than the quad 4 model. For instance, the total axial force is 74.77 kN (quad4:
73.48), whereas the analytical model predicts 75.40 kN.
The quad8 results have the tendency to oscillate around the correct solution, whereas the
quad4 results show a systematic error.

The same comments apply to the plastic models (Appendix II). There, in the quad4 case
the stresses at the inner mesh border are even more disastrous. These discontinuities need
to be explanated, understood, and eliminated if possible.

From the results shown in Appendix III we see that:


The above-mentioned discontinuities in the quad-4 results correspond to ‘strange’
integration point values: they tend to oscillate with an amplitude increasing with the rate
of plasticity. Because of the nodal averaging (of adjacent elements) this oscillation does
not show up at the inner nodes in an ordinary path plot. So in all path plots we only see
the underlying oscillations at the mesh border.
The ‘post_extrap_linear’ option reinforces this strange border behavior.
The ‘post_extrap_average’ option relaxes this strange behavior quite efficient.
The quad4 elastic results suggest that the upper and lower integration points have been
interchanged. If we exchange these values the resulting plot (labeled as ‘corrected’) is
very acceptable. Unfortunately this correction is not useful in the plastic case.

Final question: is this behavior to be expected in some way (and why), or do we meet
here a MARC bug?

4
Appendix I: Graphical results linear elastic models at 400 MPa.

radial stress circumferential and Von Mises stress

0
FE_quad8 1000 FE_quad8
-50
FE_quad4 900 FE_quad4
-100 800 quad4 Von Mises

-150 700 quad8 Von Mises

-200 600

500
-250
400
-300
300
-350 200
10 15 20
-400
10 15 20 r (mm)
r (mm)
Figure
6. Radial stress at p=400 MPa (elastic model) Figure 3. circumferential and Von Mises stress at
p=400 MPa (elastic model).

axial stress
ra d ia l d is p la c e m e n t

100
0 .0 8
FE_quad8 FE_quad8
95 FE_quad4
FE_quad4
0 .0 7
90

85 0 .0 6

80
0 .0 5
75

0 .0 4
70
10 12 14 16 18 20
10 12 14 16 18 20 r (m m )
r (mm)

Figure 4. Axial stress at p=400 MPa (elastic model) Figure 5. radial displacement (p=400 MPa ,elastic
model)

5
Differences of FE-stresses with analytical results.

r a d i a l s t r e s s d if f e r e n c e s
c ir c u m fe r e n tia l s tr e s s d iffe r e n c e s
2 q u a d 8 - a n a ly tic a l

q u a d 4 - a n a ly tic a l 2
1 .5 q u a d 8 - a n a y tic a l
1 .5
1 q u a d 4 - a n a ly tic a l
1
0 .5
0 .5
0
0
- 0 .5
-0 .5
-1
-1
- 1 .5
-1 .5
-2
-2
10 12 14 16 18 20
10 12 14 16 18 20
r (m m )
r (m m )

Figure 8. Radial stress: differences with analytical result at Figure 7. Circumferential stress: differences with analytical
p=400 MPa (elastic model). result at p=400 MPa (elastic model).

axial stress differences

2 FE_quad8 -anal
FE_quad4-anal
1.5

0.5

-0.5

-1

-1.5

-2
10 12 14 16 18 20

Figure 9. Axial stress: differences with analytical result at


p=400 MPa (elastic model).

6
Appendix II. Results for plastic quad4 and quad8 models at 400 MPa.

Plastic case quad4 stresses for P=400 MPa


1200
axial
1000 radial
800 circumferential
600 von Mises

400
200
0
-200
-400
0 2 4 r - ri (mm) 6 8 10

Figure 10.. Plastic case quad4 stresses for P=400 MPa

Plastic case quad8 stresses, p=400 MPa


1200
axial
1000 radial
800 circumferential
600 Von Mises
400
200
0
-200
-400
0 2 4 r - ri (mm) 6 8 10

Figure 11. Plastic case quad4 stresses for P=400 MPa.

7
Some details:

Plastic case: radial stress at p=400 Plastic case: circumferential stress at p=400

0 700
FE quad4 model
-50 650
600 elastic linear model
-100
550 FE quad8 model
-150
500
-200 450
-250 400
-300 FE quad4
350
elastic model
-350 300
FE quad8
-400 250
0 2 4 6 8 10 0 2 4 6 8 10

r - ri (mm) r -ri (mm)

Figure 13. Plastic case: circumferential stress at p=400.


Figure 14. Plastic case: radial stress at p=400.

Plastic case: axial stress at p=400


Plastic case: radial displacement

200 1.4
FE quad4 model quad4
180 1.3
elastic linear model quad8
1.2
160 FE quad8 model 1.1
140 1
120 0.9
0.8
100
0.7
80 0.6
60 0 2 4 6 8 10
0 2 4 6 8 10 r - ri (mm)
r - ri (mm)
Figure 15. Plastic case: radial displacement at p=400.
Figure 12. Plastic case: axial stress at p=400.

Plastic case: strains radial quad4


0.15 circumferential quad4
radial quad8
0.1 cicumferential quad8
0.05

-0.05

-0.1

-0.15
0 2 4 6 8 10
r - ri (mm)

Figure 16. Plastic case: strains at p=400.

8
Appendix III. Radial stress within inner two elements: different
extrapolation options
Elastic quad4 case
-300
-310
-320
-330
-340
-350
-360
post_extrap_linear
-370 analytical
-380 post_extrap_translate
post_extrap_average
-390 integration point values
-400 corrected integration points

10 10.5 11
r (mm)
Figure 17. Elastic quad4: stresses with different extrapolation options

plastic quad4 case


200
post_extrap_linear
100 elastic analytical model
post_extrap_translate
0 post_extrap_average
radial stress (MPa)

integration point values


-100
-200
-300
-400
-500
-600
-700
10 10.5 11
r (mm)
Fi
gure 19 Plastic quad4: stresses with different extrapolation options

Radial stress on integration points


quad4 and quad8 elastic & plastic cases

-300
-310
-320
-330
-340
-350
elastic analytical model
-360 q4 elastic
-370 q4 elastic 'corrected'
-380 q8 elastic
q8 plastic
-390
q4 plastic
-400
10 10.5 11
r (mm)

Figure 18 Radial stresses on integration points.


9

Potrebbero piacerti anche