Introducing imperfection 1

[futura22]

Currently i am running a load-displacement analysis for buckling of cylindrical shells. The first step is introducing imperfection on a perfect cylinder and requires combinations of eigenmodes from linear eigenvalue analysis in order to find the critical loads.

My problem is I do not know how to combine several eigenmodes and write them in input file.I only know how to specify one eigenmode at a time.Below is the expression from the input file.

*IMPERFECTION, FILE=bucklecylshell_s4r5_n1, STEP=1
1,0.0025

p/s: 1 is the eigenmode and 0.0025 is the degree of imperfections

thanx

 

[bassmanjax]

It's in the keyword manual - you can specify as many modes as you need, i.e;

*IMPERFECTION, FILE=bucklecylshell_s4r5_n1, STEP=1
1,0.0025
2,0.001
3,0.002
..

Regards



Martin Stokes CEng MIMechE

 

[futura22]

great! thanx a lot...

 

[futura22]

i have tried specifying different combinations of eigenmodes from using a unique mode to combination of all eigenmodes but still i got a value of LPF(Load propotionality factor) higher that 1.

That means the critical load afer introducing the imperfection is higher than perfect cylinder which is theoretically wrong.

Hope somebody would help detect any errors in my commands below. Much appreciated it...

*HEADING
RIKS ANALYSIS OF CYLINDER BUCKLING
(100% thickness,*IMPERFECTION)
*NODE
10, 100.,0.,0.
14, 92.39,0.,38.27
410, 100.,400.,0.
414, 92.39,400.,38.27
*NGEN,LINE=C,NSET=LOADB
10,14,1, ,0.,0.,0.
*NGEN,LINE=C,NSET=BND3
410,414,1, ,0.,400.,0.
*IMPERFECTION, FILE=bucklecylshell_s4r5_n4, STEP=1
1,0.25
2,0.25
3,0.25
4,0.25
5,0.25
6,0.25
7,0.25
8,0.25
9,0.25
10,0.25
11,0.25
12,0.25
13,0.25
14,0.25
15,0.25
16,0.25
17,0.25
18,0.25
19,0.25
20,0.25
*NGEN,NSET=BND1
10,410,10
*NGEN,NSET=REST
11,411,10
12,412,10
13,413,10
*NGEN,NSET=BND2
14,414,10
*NSET,NSET=ALL
BND1,BND2,BND3,REST
*NSET,NSET=LDB1
10,14
*NSET,NSET=LDB2
11,12,13
*NSET,NSET=LDBFIL
LDB1,LDB2
*TRANSFORM,TYPE=C,NSET=ALL
0.,0.,0., 0.,1.,0.
*BOUNDARY
BND1,YSYMM
BND2,YSYMM
BND3,ZSYMM
LOADB,1,2
LOADB,4
LOADB,6
*ELEMENT,TYPE=S4R5
1, 10,20,21,11
*ELGEN,ELSET=EALL
1, 4,1,2, 40,10,8
*SHELL SECTION,ELSET=EALL,MATERIAL=MAT
.25,
*MATERIAL,NAME=MAT
*ELASTIC
30.E6,.3
*RESTART,WRITE
**
*STEP,NLGEOM,INC=60
*STATIC,RIKS
.1,1.,,,,14,3,.6
*CLOAD
LDB1,3, 49969.31158
LDB2,3, 99938.62316
*EL PRINT, FREQUENCY=99
S,E,
*EL FILE,FREQUENCY=99
S,E,
*NODE PRINT,NSET=LDBFIL,FREQUENCY=2,GLOBAL=NO
U,
CF,
RF,
*NODE FILE,FREQUENCY=1,NSET=LDBFIL,GLOBAL=NO
U,CF,RF
*MONITOR,NODE=14,DOF=3
*END STEP

 

[bassmanjax]

My first question would be whether you need to use all 20 eigenmodes to perturb the base structure? If I use a *IMPERFECTION, I rarely use more than the first 2 logical eigenmodes.

My general workflow would be to;

1) Perform *BUCKLE analysis to get eigenmodes.
2) Run the *STATIC or *STATIC, RIKS analysis with no *IMPERFECTION. Extract and plot the load-deflection response.
3) Run 2) again, but with varying amounts of imperfection, say 10%, 20%, 30% ..etc, and plot the load-deflection responses for each run. I would start by just using the primary (first) buckling mode for the imperfection.

I usually find that after a certain value of imperfection, the buckling load does not change by more than a few percent. I then set this as the imperfection value.

Regards

Martin Stokes CEng MIMechE

 

[futura22]

Ok. How to find the bukling load?

Do you meant plotting load proportionality factor(LPF)-deflection graph and then multiply the highest value of LPF with the applied load? If it so, the LPF should always be smaller than 1 for it to be theoretically correct, is it?

 

[bassmanjax]

Do you meant plotting load proportionality factor(LPF)-deflection graph and then multiply the highest value of LPF with the applied load? If it so, the LPF should always be smaller than 1 for it to be theoretically correct, is it?

Depends on how you've set the model up. I tend to apply a load that is easy to factor, so if I'm doing a buckling pressure analysis, I will apply a pressure load of 1MPa in the Riks step, so that, buckle pressure = LPF * 1 MPa. In most cases, the LPF values will always be above 1 because I only apply a unit load of pressure.

Also, it depends on how you've predicted the buckling load. Have you done it with a hand calculation? Is it the load from the eigenvalue analysis? In either case, the FEA may well over-predict and you'll get an LPF of >1.

Regards

Martin Stokes CEng MIMechE

 

[futura22]

Basically for eigenvalue analysis, I specified concentrated force of 1500 on 2 nodes and 3000 on other 2 nodes. That means the total forces applied are 9000. This is what been done in the ABAQUS Benchmark documentation input file and I tried to imitate them in my model.

I then multiply the total force with the minimum eigenvalue to get the lowest buckling load

So I use the same load for the riks analysis. Or should I use higher load instead?

thanx

 

[bassmanjax]

There isn't anything 'wrong' with the loads that you have applied. I don't believe that you need to use 20 modes in the *IMPERFECTION to trigger the buckling, that's all. I looked at the ABAQUS benchmark, and they only use the first mode to trigger the buckling.

The load that you apply in a Riks step is a bit academic anyway, as you have to multiply the load by the LPF to get the total load.

Regards

Martin Stokes CEng MIMechE

 

[futura22]

okay, one more thing...from what i understand the LPF is the ratio of load/critical load. Does multiplying the LPF with the critical load obtained from eigenvalue analysis will give the actual critical load?

 

[bassmanjax]

LPF = Load Proportionality Factor.

No, LPF is not a ratio of anything - the LPF is part of the output from the Riks analysis, as the load magnitude is treated as an unknown. The LPF is printed to the .sta file and is used to determine the actual load on the structure;

Actual Load = LPF * Applied load

Section 6.2.4 "Unstable collapse and postbuckling analysis" in the v6.5 docs covers the principle behind the Riks analysis quite well.

Regards

Martin Stokes CEng MIMechE