You are on page 1of 36

The Fourth International Conference on

Structural Engineering, Mechanics and


Computation: University of Cape Town

A “simple beam” revisited


Title

Roland Prukl
Pr.Eng., Dipl.-Ing., FSAICE, AIStructE
Dedicated to
Gali
Henry
Dedication

Katie
Contents
•Introduction
•FEMSA ‘92
•SEMC 2001
•SEMC 2010 - A “simple beam” revisited
•Conclusions – SEMC 2010
Contents

•SEMC 2013 - Axisymmetric models


•Conclusions – SEMC 2013
•Appeal
•Acknowledgements
Around 1980 Kamel & McCabe stated:
“The finite element method is an
approximate method based on
discretization of both geometry, loading
Introduction

and boundary conditions, as well as the


use of elements derived using various
assumptions of which the user is often
not quite aware.”
At about the same time Bathe
prophesied:
“Within a period of ten years the powerful
tool of finite element analysis will be
Introduction

available on every analysis engineer’s


desk, but not enough people will be
trained sufficiently by then to correctly
and safely apply this method.”
At the FEMSA 1992 Symposium in Cape
Town, the author presented:

The CST (Constant Strain Triangle) –


An insidious survivor from the infancy of FEA
FEMSA ‘92

The results from various finite element


programs, using eight different meshes of
plane stress elements, as well as from one
boundary element mesh, were compared.
A concrete beam, 8m long, 2m
high and 0,2m thick is loaded with
a vertical u.d.l. of 100 kN/m at the
top. The material properties are:
Young’s modulus 3,0e7 kPa and
Poisson’s ratio v = 0,2.
FEMSA ‘92

The horizontal stresses at the


centre of the span should be
–6000 kPa (compression) at the
top and +6000 kPa (tension) at
the bottom.
Four 4-noded quads Eight 3-noded triangles
FEMSA ‘92

Maximum stress Maximum stress


error = 13% error = 100%
At the SEMC 2001 Symposium in Cape Town, the
author presented:

Finite Element Analysis (FEA) tests on a simple beam


Important Information for users of FEA software.
SEMC 2001

For the same problem as used at FEMSA’92, the


results from various finite element programs, using 63
different shell and 23 solid meshes, were compared.
.
SEMC 2001

Using plane stress elements in program Strand7 for this


mesh, the maximum stress error at the bottom of the beam
at midspan can even be 116 %, i.e. instead of +6000 kPa
tension, we get -981 kPa compression !!!
Since 1986, the writer has presented a one-
week course “The Application of the Finite
Element Method in Practice” about every two
months.
In almost every course some of the participants,
SEMC 2010

when modelling a simply supported deep beam


with shell elements, make the same mistake:
They add unnecessary restrained freedoms
(translation DZ as well as the rotations RX, RY
and RZ) to all node points of the structure.
In the program Strand7, when using 3-dimensional
shell elements without any unnecessary restraints
at the nodes, the maximum error in a model with
SEMC 2010

four 4-noded elements is 14 %.


With DZ, RX, RY & RZ restrained at all nodes, the
error increases to 38 %. This error increase is, in
fact, caused only by the RZ restraint.
In the program “Prokon Frame” version W2.72, the
maximum stress error increases from 5 % to 24 %.
In the program Adina, the maximum stress error for
both cases (with and without the additional
restrained freedoms), is 13 %.
SEMC 2010
If one uses 8- or 9-noded elements in Strand7,
the maximum stress error for both models will
SEMC 2010

be 3 %.
Horizontal stresses at midspan
bottom of beam
Analysis No. 9
Program % Error
Strand7 -14
Prokon -5
SEMC 2010

Adina -13
Ansys -13
Cosmos -8
Very coarse mesh !
Conclusions SEMC 2010

• Finite Elements have to be handled with


great care.
Conclusions SEMC

• Different programs might give different


results and even with the same program
the results from different element types
might differ considerably from each other.

• All structures have to be supported


2010

carefully and unnecessary supports must


be avoided.
Axisymmetric models
Axisymmetric models

A circular slab with radius of 8 m and a


thickness of 2 m is simply supported
around the outer edge and loaded
with a UDL of 500 kPa.

The material properties are:


Young’s modulus = 3,0e7 kPa and
Poisson’s ratio v = 0,2.
Centre of slab
Axisymmetric models

Cross section of circular slab:


radius = 8 m, thickness = 2 m

The horizontal stresses at midspan should


be - 9600 kPa (compression) at the top
and + 9600 kPa (tension) at the bottom.
1) 4-noded shells, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 0%


2) 8-noded shells, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 2%


3) 8-noded solids, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 5%


4) 20-noded solids, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 2%


5) 8-noded solids, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 1%


6) 20-noded solids, using program Strand7
Axisymmetric models

Maximum stress error at midspan = 1%


7) 4-noded axisymmetric solids,
with no horizontal supports at the centre,
using program Strand7
Axisymmetric models
Centre of slab

Maximum stress error at midspan = 121%


8) 4-noded axisymmetric solids,
with horizontal supports at the centre,
using program Strand7.
Axisymmetric models
Centre of slab

Maximum stress error at midspan = 93%


9) 4-noded axisymmetric solids,
without horizontal supports at the centre,
using program Strand7.
Axisymmetric models
Centre of slab

Maximum stress error = 53%


10) 4-noded axisymmetric solids,
with horizontal supports at the centre,
using program Strand7.
Axisymmetric models
Centre of slab

Maximum stress error = 44%


11) 8-noded axisymmetric solids,
without horizontal supports at the centre,
using program Strand7.
Axisymmetric models
Centre of slab

Maximum stress error = 2%


12) 8-noded axisymmetric solids,
with horizontal supports at the centre,
using program Strand7.
Axisymmetric models
Centre of slab

Maximum stress error = 2%


Program Analysis Stresses at Bottom of Centre
Name No. Direction % Error
RR -2
7
TT -2
Fesdec
Axisymmetric models

RR 2
8
TT 2
RR 38
7
TT 121
Strand7
RR 93
8
TT 46
RR 8
7
TT 8
Lusas
RR 6
8
TT 6
Conclusions SEMC 2013

• Do not use Quad4 elements for axisymmetric


analyses in Strand7 because the program
subdivides each element into four triangles.
Conclusions SEMC

• Quad8 elements for axisymmetric analyses in


Strand7 are not being subdivided into triangular
elements.
• Mr. Stewart Morrison carried out 18 test runs for
this circular plate.
15 of these test runs produced very good
2013

results. The 3 others had errors up to 42 %.


Appeal

I appeal to all Finite Element users.

Please carry out all or some of my


tests on the software of your choice
and send them to me.
Appeal

I shall then forward you all my detailed


information on the subject.
Final note
The purpose of this presentation is to
warn users of finite element software
about possible pitfalls.
The figures produced should not be
considered as indicative of these
programs for general use.
The author wishes to emphasize that
Final note

the aberrations highlighted are specific


to the problems analysed and should
not be regarded as his preference for
one or the other package compared.
Acknowledgements
The author wishes to thank the following persons
who assisted in the preparation of this report:
• Mr. Edwin Clarke
Acknowledgements

for conducting the Strand7 & Prokon analyses


• Mr. Steve van Wyk
for conducting the Adina and Ansys analyses
• Mr. Jurgen van Wyk
for conducting the Cosmos analyses
• Mr. Stewart Morrison
for conducting the Lusas analyses
• Mr. Adrian Peirson
for proofreading
Thank you for your attention !
End

You might also like