Learning Forum

[Brian Tang]

I am running CFX 2022R2 on Ubuntu 20.04LTS. I am running some transient (URANS) simulations involving a mixture of hydrogen and air, and have encountered a problem where the solution seems to proceed normally for some time, but then eventually stops with an error at the point it attempts to write results out. What is particularly strange though, is that it does not fail immediately - here is an example of the error in the output log for one simulation, where it fails on writing out at timestep 282. This simulation was running fine until this point, and has been writing out transient results files (which are fine) for every even timestep until this point. There is no indication that anything is wrong in the log before this point.

+--------------------------------------------------------------------+
 | Writing transient file 282_full.trn                                |
 |   Name   : Transient Results 1                                     |
 |   Type   : Standard                                                |
 |   Option : Timestep Interval                                       |
 +--------------------------------------------------------------------+
 +--------------------------------------------------------------------+
 | Notice: The maximum Mach number is  3.502E+00.                     |
 +--------------------------------------------------------------------+
Partition:    10
  1 SU_HYBARR, ,VISCTRB_FL1,ZN1                 /VERTICES,LATEST
  2 cal_VISC_TRB,LIN-LIN,DENSITY_FL1,BELG5/IP,LATEST
  3 cal_VISC_TRB, ,TKE_FL1,BELG5/IP,LATEST
  4 SU_HYBARR, ,MASCON_MT1_FL1,ZN1                 /VERTICES,LATEST
  5 SU_HYBARR, ,MASCON_MT4_FL1,ZN1                 /VERTICES,LATEST
 
 +--------------------------------------------------------------------+
 | ERROR #001100279 has occurred in subroutine ErrAction.             |
 | Message:                                                           |
 | Stopped in routine TRCPRT                                          |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 +--------------------------------------------------------------------+

 +--------------------------------------------------------------------+
 |                An error has occurred in cfx5solve:                 |
 |                                                                    |
 | The ANSYS CFX solver exited with return code 1.   No results file  |
 | has been created.                                                  |
 +--------------------------------------------------------------------+

End of solution stage.

In an attempt to narrow down the problem, I have taken this simulation and resumed it (by which I mean using the option Execution Control>Initial Values Control>Continue History From) from timestep 280 (the last successful .trn file).If Output Control>Transient Results is not removed, then the simulation crashes at the same point as above with the same error output. If Output Control>Transient Results is disabled, the simulation continues to run successfully past this point. While running, if the Save/Create Backup button is pressed in the CFX Solver manager, it once again crashes with similar output - for example:

 +--------------------------------------------------------------------+
 | Writing backup file 306_full.bak                                   |
 |   Name   : Requested from solver manager or command line           |
 |   Type   : Standard                                                |
 |   Option : End of current iteration                                |
 +--------------------------------------------------------------------+
 +--------------------------------------------------------------------+
 | Notice: The maximum Mach number is  6.738E+00.                     |
 +--------------------------------------------------------------------+
Partition:    10
  1 SU_HYBARR, ,MASCON_MT1_FL1,ZN1                 /VERTICES,LATEST
  2 SU_HYBARR, ,MASCON_MT4_FL1,ZN1                 /VERTICES,LATEST
Partition:    11
  1 SU_HYBARR, ,MASCON_MT1_FL1,ZN1                 /VERTICES,LATEST
  2 SU_HYBARR, ,MASCON_MT4_FL1,ZN1                 /VERTICES,LATEST
 
 +--------------------------------------------------------------------+
 | ERROR #001100279 has occurred in subroutine ErrAction.             |
 | Message:                                                           |
 | Stopped in routine TRCPRT                                          |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 |                                                                    |
 +--------------------------------------------------------------------+

 +--------------------------------------------------------------------+
 |                An error has occurred in cfx5solve:                 |
 |                                                                    |
 | The ANSYS CFX solver exited with return code 1.   No results file  |
 | has been created.                                                  |
 +--------------------------------------------------------------------+

End of solution stage.

It therefore looks like the issue lies with writing the results out rather than with the solution itself; please could you advise? I do not have any ideas how to proceed with debugging further at this stage.

[rfblumen]

Although it might not be related, your OS is not on the list of supported operating systems (see 1.1. System Prerequisites in the Ansys Installation Guide),

One possible reason for the error could be related to CEL expressions.  During the run, CEL expressions are evaluated using integration point information.  When writing backups or the final res file, expressions are evaluated using nodal information which can be slightly different.  It could be that a CEL expression is generating an error (i.e. divide-by-zero, invalid exponentiation) related to this.  Running the model in serial can help verify if this is the case and identify the expression causing the issue.

[DrAmine]

Are you Equilibrium Phase Change Model? I am aware about an issue resulting in similar outcome due to EPC which has been corrected in the actual release (23R1) mainly driven by the mass fraction being written. 

Which models are included?

[Brian Tang]

The forum appears to have lost one of my replies that I submitted earlier, but briefly:

1) Ubuntu 20.04 is officially supported for Ansys 2022R2, per the very installation guide that you referred to. This is specifically why I am running this distro and version of Linux.

2) I attempted to run the simulation in serial mode as suggested - it crashes at the same point with the same output as when in parallel mode. It does, however, write out a .res.err crash recovery file; but I am not sure what information this provides. Opening it in a text editor shows a short text header and footer, with a large amount of (non-text) data in the middle. Examining the header and footers did not immediately enlighten me with any information as to the source of the crash - could you perhaps provide some guidance as to what it is I should be looking for (and indeed, where to look)?

3) I'm not running a phase change model, but this does have a non-reacting, variable composition ideal gas mixture (just hydrogen and air), so a problem caused by mass fraction might make sense. Is there any easy way for me to be able to test if this is the cause? And if so, are there any workarounds for it that I can try other than moving to 2023R1?

[rfblumen]

Hi Brian,

My apologies regarding the OS - I didn't see the last one (Ubuntu 20.04) in the list in the Installation Guide.

The .res.err file crash recovery file should be readable in CFD-Post.  This may give some clue if you look at the list of variables and compare to a normal result file and/or perform other post-processing to see if the variable values/ranges are as expected.  You may also be able to use the file as an initial values file and restart your run.

[DrAmine]

I remember the fixes are within 23R1. The workaround is to avoid writing the mass fraction quantities: you check that to assess if we are talking about the same root of the issues I have in mind.

[Brian Tang]

I installed 2023R1, and experience exactly the same behaviour with the new version, so it appears not to be the same issue.

Having examined the .res.err solution in CFD post, I haven't been able to identify any problems with the solution. The flow field looks, as far as I can tell, almost identical to the last successfully written out .trn file, and I have not been able to find any variables (certainly the main ones - pressure, velocity, temperature, density, turb k.e. etc.) with obvious extreme values that I would associate with a numerical issue - it all looks nicely bounded. Indeed, the flow field looks pretty much as I would expect the flow field at this point to look.