Random RAM peaks lead to critical errors

[KiraZh9]

🔍 Before submitting the issue

• I have searched among the existing issues
• I am using a Python virtual environment

🐞 Description of the bug

Hello! I came across a weird unpredictable bug that I wanted to share.
Thank you for your time and support!

I'm running a transient simulation over t timesteps using a for loop in Python. At each iteration, the Fluent model is executed for a single timestep (time_step_count = 1).

During each for loop timestep, I:

1. Update the convection boundary conditions for two walls;
2. Update the volumetric heat sources for 12 cell zones based on heat generation rates;
3. Run Fluent for one timestep;
4. Extract result values using predefined Fluent report definitions;
5. Compute the heat generated using the temperature from the previous timestep and use it to update the sources for the next loop.

There are two causes or errors at random timesteps:

1. Updating boundary condition: session.setup.boundary_conditions.wall[wall_name] = {"thermal_bc" : "Convection", "h" : h, "tinf" : T}
   -> leads to immediate error and stops the for loop:
   Error: wta(1st) to string->symbol
   Error Object: #[free (3 cells)]
2. Updating the cell energy source: session.setup.cell_zone_conditions.solid[solid_name] = {"source_terms" : {"energy" : HGR}}
   -> leads to immediate error, but Fluent doesn't crash at this moment:
   Error: wta(1st) to string->symbol
   Error Object: #[free (1 cells)]
   An error occurred: wta(1st) to string->symbol
   Error Object: #[free (1 cells)]
   ->the loop continues until the launch of the next timestep session.solution.run_calculation.dual_time_iterate(time_step_count = 1, max_iter_per_step = 20) and Fluent crashes:
   Error: Attempt to mark a free block
   Error Object: #[free (1 cells)]
   Error encountered in critical code section

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 0 PID 10864 RUNNING AT MOUNTAIN24
= EXIT STATUS: -1 (ffffffff)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 1 PID 8740 RUNNING AT MOUNTAIN24
= EXIT STATUS: -1 (ffffffff)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 2 PID 7724 RUNNING AT MOUNTAIN24
= EXIT STATUS: -1 (ffffffff)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 3 PID 10068 RUNNING AT MOUNTAIN24
= EXIT STATUS: -1 (ffffffff)

===================================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= RANK 4 PID 15224 RUNNING AT MOUNTAIN24
= EXIT STATUS: -1 (ffffffff)

This critical error has occurred in over 20 runs. It's not deterministic—the timestep where it fails varies, even when running the same model. (The random seed is fixed, so it is not the issue.)

I’ve observed that system RAM usage spikes sharply just before the crash, as seen in the monitoring screenshots below:

To mitigate this, I initially added time.sleep(0.5) after each boundary and source update, which seemed to help temporarily. However, the crash returned during longer simulations. Increasing the delay to time.sleep(1) didn’t significantly improve stability.

📝 Steps to reproduce

1. Set up a transient simulation in Fluent:

• Use a model with at least two walls with convection boundary conditions and 12 cell zones with heat generation rates.
• Ensure that the model has predefined Fluent report definitions for extracting results.

2. Prepare Python script:

• Implement a loop to run the simulation for multiple timesteps.
• In each iteration, perform the following steps:
• Update the convection boundary conditions for the walls. Example code:
  session.setup.boundary_conditions.wall[wall_name] = {"thermal_bc": "Convection", "h": h, "tinf": T}
• Update the heat generation rates for the 12 cell zones. Example code:
  setup_cell_zone_HGR(session, cell_name, cell_HGR)
  Run Fluent for a single timestep (time_step_count = 1).
• Extract results using predefined Fluent report definitions.
• Compute the heat generated from the previous timestep's temperature, and update the sources for the next iteration.

3. Execute the script for multiple timesteps.

4. Monitor system RAM usage during execution (especially around the time of the crash).

💻 Which operating system are you using?

Windows

📀 Which ANSYS version are you using?

2023 R2

🐍 Which Python version are you using?

3.12

📦 Installed packages

Relevant packages:
ansys-fluent-core           0.29.0
ansys-api-fluent            0.3.35
numpy                       1.26.4
pandas                      1.5.3

[seanpearsonuk]

Hello @KiraZh9

• Have you identified which process(es) is/are consuming the RAM? Apologies if that information is already in your report.
• Would you agree that this is most likely a Fluent issue if there is a software issue? Have you contacted support?

[KiraZh9]

Hello @seanpearsonuk !

• Will come back on the first question tomorrow
• Yes, we are talking to ANSYS support as well.
  I will be honest, it's difficult to tell but it seems like a connection issue between PyFluent and Fluent because it randomly has an issue with updating either the boundary condition or the cell energy source. It always brings up this error:
  Error: wta(1st) to string->symbol
  Error Object: #[free (3 cells)]
  I double checked the values and they are alright. Moreover, it doesn't happen with the same value, every time it's different.
  I have never had an issue with updating the boundary conditions or energy sources both in GUI or in TUI, so that is why I wrote this ticket here.
  I would love to hear your thought on the issue!

[seanpearsonuk]

Hello again @KiraZh9

1. Please kindly provide details about setup_cell_zone_HGR( ) as I am unfamiliar with it. Thanks.
2. Regarding the varying outcomes between PyFluent and TUI, please note that you can record your TUI workflow to a Python journal and subsequently run the journal in either PyFluent or the equivalent Python console in Fluent. Please could you try that and let us know if the journal run produces erroneous behaviour not observed when running via the TUI.

[KiraZh9]

@seanpearsonuk

Sorry, my bad, copied the function by accident, Here is the command:
session.setup.cell_zone_conditions.solid[solid_name] = {"source_terms" : {"energy" : HGR}}

Also, today I spent some more time analyzing the errors and I updated the ticket with more info as well as correct command for cell energy source update