Did not sample any reaction for nuclide U235

User Support
1

Hello,

I am getting the following error while running an eigenvalue calculation with OpenMC 0.13.2:

RuntimeError: Did not sample any reaction for nuclide U235 application called MPI_Abort(MPI_COMM_WORLD, -1) - process 0 [unset]: write_line error; fd=-1 buf=:cmd=abort exitcode=-1 : system msg for write_line failure : Bad file descriptor

I have been running alternate cases of the same problem with varying material temperature without this or any other problems. Any thoughts/suggestions on what could be the reason for this and possible troubleshooting options?

I also tried to take a look at the particle file which interestingly shows the particle type as 0. What does particle of type ā€˜0ā€™ signify?

I would really appreciate any help. Thanks in advance.

PS - Would have been ideal to share the problem model, but I am afraid, it wouldnā€™t be possible.

2

Hello All,

I just wanted to check if anybody has any suggestions on this.

Have a great day/week ahead!!!

3

Standard MC logic says that your particle collided with something, decided it was U-235, then failed to identify the specific reaction. By design, this error condition should not generally occur.

While I understand you may not be at liberty to provide the input, you can perhaps answer some general questions to help the community support you:

  1. What nuclear data tables are you using? Are they custom-built or did you use standard tables distributed with OpenMC?

  2. Is this a coupled neutron/photon problem or are the physics limited to neutron physics only?

  3. Does the problem reproduce if the code is compiled with full debugging and no optimizations?

  4. Does the error reproduce if you simplify the model?

1 Like
4

Hi @AlexandreTrottier

Thanks for this explanation.

I am using the official ENDF/B-VII.1 library distributed with OpenMC.

Itā€™s a neutron physics problem only.

I havenā€™t tried this option yet but try it.

I have been progressively simplifying the model and the problem does reoccur. Any suggestions on what simplifications might be useful?

Thanks once again.

5

Hello.

Without more information on the model, I can only offer general suggestions. I would simplify the fuel such that it includes only U-235, but adjust the density to compensate the impact on reactivity. Likewise I would simplify the moderator/coolant. Overall, you want to keep the fuel to moderator ratio about the same and minimize any impact on the neutron spectrum.

If you can reproduce the problem using a model which can be shared, such as a simple pincell, then this can help with diagnostics. You want to eliminate the possibility that the problem arises from your OpenMC build, as opposed to your model problem.

I would also set a limit on lost particles, to eliminate the possibility that a geometry error is leading to unexpected behavior.

Finally, is the error condition occurring at the first collision with U-235, or are you able to complete some cycles?

6

Hello,

Thanks again for sharing these suggestions.

I am using prebuilt OpenMC 0.13.2 available on conda-forge. The model should not have a problem, I feel, as I have already used it for hundreds of simulations previously, but I am still investigating if thereā€™s any bug in it.

The error condition occurs after more than 6500 cycles with multiple generations per cycle.

7

The symptoms are very odd. How much wall clock time is needed to manifest the error condition? I am concerned that running any debugging will be difficult considering the high cycle count.

Is this a depletion case? That increases the complexity of the diagnosis.

If you cannot reproduce this error except with a high cycle count, then I suggest the following procedure:

  1. Compile with debugging flags.
  2. Examine the code to identify the block where the error condition occurs, you can do this using the error message string.
  3. Launch using a debugger.
  4. Add a conditional breakpoint at the start of that block, with condition that cycle count is 6500.
  5. Run the code and hope you can examine the conditions leading to the error.

Otherwise, I am running out of suggestions.

8

Actually, can you see if the error manifests using OpenMC 0.12.2 ?

9

Thanks so much for these valuable suggestions. Iā€™ll try them and share any updates.

The error occurs after 20 mins with 24 threads which, as you mentioned, is making the diagnosis extremely painful. Itā€™s not a depletion case.

10

Sorry, just wanted to confirm if thatā€™s OpenMC 0.12.2 or 0.13.2?

11

The older version.
I would suggest you also run the problem with some energy cutoff limits to see if that helps understand the error condition.

12

It seems like this might be related to an issue that was fixed with this pull request:

Note that that fix is included in release 0.13.3, so upgrading to 0.13.3 might resolve the issue here.

13

Thanks @paulromano for this insight. Iā€™ll try the running the problem using version 0.13.3.