Since you have different increment numbers, I presume that you are changing the timesteps as well. I've seen cracking problems give different answer when the material is very brittle, as any small changes/errors/roundoff/domain decomposition can make the crack appear at a slightly different location and propagate differently, and will result in completely different solution. However I am not sure it is the case here.

Maybe you should try to run one analysis with restart points and then just do a restart without changing anything, see if you get a similar problem. If not, then it is very likely that there is nothing wrong with the restart feature, rather the changes you made to your second analysis.

Best regards,

David

David Sziroczak thank you for your response but I did not change any variable. The restart job is continuation of previous one. Same as what you suggested on your second paragraph.

The restart file (right side) takes the values from from the last increment (332) of previous file (left side). The issue is if I just run the original input file (left side) one more increment, I see completely two different behaviors.

I didn't mentioned before but I am running a displacement controlled three point bending simulation on cortical bone. I just checked the displacement values and they are the same on both but the rest of the output data is different.

Ok, I see. So right at the time of the restart, do you get the same results on your restart model as where the original stopped? By the way your left side (previous file) is Feb 01, the right side (restart) is Jan 26, so are you sure that it is the correct file you are comparing? Sounds a bit like time travel to me :) Also, what's happening with the step time, you are increasing the increments but the time doesn't change? Is that the desired behaviour?

One more thought, if you are simulating bone, do you use and predefined fields that affect variables/behaviour? If you do, maybe they don't apply the same field values for some reason?

David Sziroczak I understand your skepticism. I wish it was a time travel. The dates are different because I run the simulation with an extra step after I found this conflict.

The image is from 2 different odb files and I showed the behavior on same time step. Since the model and the boundary conditions are the exactly the same,same step time corresponds to same displacement, i.e., 0.1197 step time is 0.01197 mm of displacement on x direction and I expect the results to be similar if not the same.

Step time on the restart job changes but this changes are very small. This is not a desired behavior. This is just a different behavior happens when I use a restart job.

I did not quite understand what you meant by predefined fields. I have different sections but these were not redefined on the restart file.

Another simpler explanation to this issue would be; I have an input file (input file 1) and if I run it 332 increments and use a restart input file (input file 2) I am having two different behaviors compared to input file 1 that run more than 332 increments. In short, the results from input file 1 on 333rd increment is completely different compared to results from input file 2 on 333rd increment. I attached a new image showing the differences on same increment.

My restart input file

"*Heading

** Job name: 12206_SecondTest Model name: 12206_2_2

** Generated by: Abaqus/CAE 6.13-3

*Preprint, echo=NO, model=NO, history=NO, contact=NO

*Restart,read,Step=1,inc=332"

Ok thanks I see what's going on, nor more jokes! I am not aware of any such limitation with XFEM, so it is not impossible that it is a bug. Saying that, your version of Abaqus is quite old. Do you have access to a more recent version perhaps? I know there has been quite a lot of improvement to the XFEM capability. If you do it might be worth checking if those behave correctly. If not and you ruled out any other errors it might be an idea to get in touch with support.