Jump in Wss contour

[Ezat Shokrani]

Dear SimVascular developers,

I am going to simulate the blood flow in an idealized case. I am using a unsteady inflow waveform as an inlet and zero pressure as an outlet. I have simulated this case for three cardiac cycle. I am using a boundary layer mesh to compute the Wss more accurately on the walls. I converted the results into vtu and vtk format for the last cycle. It was surprising when when I see the contour of Wss on the wall using the Paraview software which I have attached below. I can see a separate region in Wss (blue region) which seems like a jump. I would be wondering if you could guide me why this happened.

Best regards,
Ezat Shokrani

[David Parker]

Hi Ezat,

I've not seen this kind of thing before. I'm wondering if the mesh is poorly defined in that region.

If you upload your simulation files (include the histor.dat file) someplace I can download them I'll have a look.

Cheers,
Dave

[Hadi Wiputra]

Hi Ezat,

I got a feeling that you are missing one of the geombc.dat.x (where x is the number of computation node) when running svPost . I accidentally did that and I got something similar.

Regards,
Hadi

[Ezat Shokrani]

Dear Dr. David Parker and Hadi Wiputra,


Thanks for your responses. Unfortunately, because there are a lot of files, it is difficult to me to upload all the files. In fact, I have other similar case studies like this and get good results from them from the first to the end of solution time and therefore, I am sure that my set up is right.

In fact, for this case, the mesh quality is good as the solution (WSS, velocity, pressure) from the initial time step until before an specific time was good, as shown in here

111.png (175.31 KiB) Viewed 410 times

However, at one specific time step the WSS gets some wrong amount (but not pressure or velocity), shown in here

222.png (112.93 KiB) Viewed 410 times

then, after a few time steps (For example 10 time steps), you can see the solution (WSS) is logical but not at an specific region of the fluid domain and therefore there is a jump.

333.png (148.62 KiB) Viewed 410 times

I think there is a bug here. I would be wondering if you could guide me on this issue.


Best regards,
Ezat Shokrani

[David Parker]

Hi Ezat,

Just upload the files that are created after you select the SV Create Data Files for Simulation button. I will then run a simulation to see what's going on.

Cheers,
Dave

[Ezat Shokrani]

Dear Dr. David Parker,

Thanks for your response. I have uploaded in Google Drive and sent the files to your email with this subject: Ezat Shokrani-Blood flow simulation files. The simulation time is too long. Therefore, I have uploaded some restart files there. You should run it using the Restart file 2800. After a few time step of running the case (for example 50 time steps), you can convert the results and see velocity, pressure etc. All solutions except for WSS are OK. I would be wondering if you could guide me on the issue that I have encountered.

Best regards,
Ezat Shokrani

[David Parker]

Hi Ezat,

I downloaded your simulation files and continued the simulation from 2800 to 2890. The vWSS looks fine.

I am thinking that the results for the region of the mesh where the vWSS is blue were not converted successfully. Did you convert the files using SV? If so then check to see if the results were any errors when converting the results (select the Show Details button on the popup that appears after the conversion finishes).

If you are running on an HPC cluster it might be that a node you where running went down.

Cheers,
Dave

[Ezat Shokrani]

Dear Dr. David Parker,


Thanks for your repose and running the simulation files.
I am using the SV to convert the files. I have attached the file related to converting the results. I can see some errors here when the results are converted.

I think you are right that one of CPU went down, because when I recreated the simulations files (without any changing in the set up like you) and run it again using the restart file 2800, I got good WSS contour as well. I do not know how I can stop happening this specially when the simulation time is large. I am using a virtual machine which has a GUI (Ubuntu platform). I would be wondering if you could help me


Best regards
Ezat Skorani

[David Parker]

Hi Ezat,

Yes, the `ERROR parsing header: Unexpected end of line.` lines means that a restart file was not written successfully.

You are running an Ubuntu virtual machine on what host OS? So this is not an HPC cluster? In that case it is unlikely that a CPU (i.e. node) is going down, doesn't happen very often. It is more likely that you are exceeding some resource limit like job time or disk space, something that would stop the job.

Your mesh is pretty fine, probably don't need so many elements to resolve the flow. You could try reducing the number of elements in your simulation and see if it completes.

Cheers,
Dave

[Julian Suk]

Hi Ezat and David,

I am experiencing the same problem right now. I run (the September 2021 release of) svSolver in a (twice nested) for-loop in bash for different shapes and boundary conditions using OpenMPI. For different shapes, consistently around the third or fourth loop iteration (for boundary conditions), some restart file fails to write (without error message) which causes

Code: Select all

ERROR parsing header: Unexpected end of line.

in svPost.

I have confirmed that when trying to re-simulate with the very same loop iteration files manually, all goes well.

I have live-monitored the computer's CPU and RAM when this failed write happens and neither of them is close to its limit. Could it be that there is some resource specific to bash, OpenMPI or svSolver that is the bottleneck here?

Please let me know if someone found a fix. Best regards,

Julian