Problem in Dynamic Analysis

[Sayan]

Hii,
I successfully run the file with 32 elements after attaching the beam.grd file, from which I also learnt about creating other types of meshing in grd format...Thanks a lot for that help..
I faced another peculiar problem recently for the same problem... I have validated my model and discretization of meshing to get the optimum number of elements by matching with natural frequencies obtained from the numerical result,,, After validation, I tried to solve it for getting displacement at the midpoint of the beam using geometric non-linear analysis...(Problem was, cantilever beam (0.58x0.02x0.002m, density 2778kg/m3, Youngs Modulus=70GPa,Poisson's ratio=0.3) subjected to udl of 1.5N/m amplitude harmonic force (hence in pressure format,F2 used=1.5/0.02sin(41.88*t)=75sin(41.88*t) of excitation frequency 41.88rad/s with zero initial condition....Though the program executed smoothly for 20 timestep-intervals with timestep size=0.01, but when I tried to match the result with original value there was a huge difference...at 0.2s the actual value of transverse displacement is close to 0.012 but I am getting 0.003072...I tried with different time-intervals but not getting even close to it...As natural frequencies are matching hence it can be assumed that input of units is not the problem...then what are the possible chances of error can you suggest to me.....

[kevinarden]

I did my own calculations, then performed a static solution with the 75 N/m2 and an eigenvalue analysis. I used your mesh and my own mesh and obtained good results for both.

Solution.png (13.72 KiB) Viewed 3248 times

Using a transient solution with the load applied

Boundary Condition 2
Target Boundaries(1) = 3
Name = "BoundaryCondition 2"
Force 2 = Variable time
Real MATC "75*sin(41.88*tx(0))"
End
I did not obtain a reasonable solution

load_equation.png (43.92 KiB) Viewed 3248 times

I have not found a reason why as yet.

tcase.sif

(1.96 KiB) Downloaded 209 times

[kevinarden]

Problem is the non-linear transient is not converging. Changing nonlinear itterrations to 1 was OK for the static and eigenvalue solution, but in the transient convergence is not achieved. To demonstrate I switched back to the linear solver and the results are resonable.

result.png

(55 KiB) Not downloaded yet

tcase.sif

(1.96 KiB) Downloaded 226 times

[kevinarden]

Seems in the non-linear you have to take smaller time steps to find convergance. Decrease the step size to .001 intervals to 200 then only output every 10th one. Results look reasonable.

tcase.sif

(2.29 KiB) Downloaded 212 times

[raback]

Hi

Did you try with nonlinear system relaxation factor, say in range 0.5-0.7? Maybe that helps in the convergence of the nonlinear system?

-Peter

[Sayan]

Hiii,
I have checked all your .sif files.....Thank you so much for your response and valuable time..it's really a pleasure to work with you all ..
I am trying to execute with 0.001-time step as per your latest .sif file and also relaxation factor 0.5...The only thing I am changing is, for this case to validate displacement response I don't want to apply damping as the result available to me is for without damping...Let's see what happened..

[Sayan]

Hiii,,
I just finished the execution of the displacement vs time plot considering all our previous discussions and your suggestions,,,I am very happy that the results are validated perfectly..Now I will try to move to more complex work considering damping also...
There is no word left to thank you all for your tiresome effort..Hope in the future also I will learn many things from you all..
---Sayan

[kevinarden]

Peter,
The relaxation factor of 0.5 di allow cnvergenge at the original time step size of 0.01 and 200 itterations. The results looked the same as reducing the time step ro 0.001.
Kevin

[raback]

Great Kevin, thanx for tracking this down! -Peter

[Sayan]

Hiii,
Whether "Segmentation fault-invalid memory reference" occurs in the case of a large number of timesteps intervals? As I am trying to reach 25s with a time step size 0.001,hence it needs 25000 intervals..and elmer is showing this error message at some point...If I increase the time step size from 0.001 to even 0.005, even with 0.5 non-linear relaxation factor I am not getting good results,,,,is there any alternative solution to bypass the following error?

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0xffffffff
#1 0xffffffff
#2 0xffffffff
#3 0xffffffff
#4 0xffffffff
#5 0xffffffff
#6 0xffffffff
#7 0xffffffff
#8 0xffffffff
#9 0xffffffff
#10 0xffffffff
#11 0xffffffff
#12 0xffffffff
#13 0xffffffff
#14 0xffffffff
#15 0xffffffff
#16 0xffffffff
#17 0xffffffff
#18 0xffffffff
#19 0xffffffff
#20 0xffffffff
#21 0xffffffff
#22 0xffffffff
#23 0xffffffff
#24 0xffffffff
#25 0xffffffff
#26 0xffffffff
#27 0xffffffff