Numerical Instability Issue in FDS Tunnel Fire Simulation (RESTART + Python Scripts)

[MrAhriman]

Dear FDS Community,

I am seeking your expertise regarding a numerical instability issue I encountered while using FDS for tunnel fire simulations.

Problem Description:

I utilized the RESTART function along with Python scripts to modify input files, aiming to adjust the HVAC fan airflow rate every 1 second. Specifically, I modified the value in the following RAMP definition:

&RAMP ID='VF Ramp'
      T=33.0
      ,
      F=-20  // Modified this value

However, this operation caused severe numerical instability in the simulation. The simulation frequently crashes with numerical instability errors—sometimes running slightly longer, but often failing within a short timeframe (e.g., terminating at 33 seconds, as shown in the attached error logs).

Steps Taken:

• I followed recommendations from the FDS User Manual for tunnel simulations.
• I reviewed the input files and error logs but have not yet identified the root cause.

Attachments:

• Simulation results and error logs from the failed case (33 seconds runtime, FDS-6.9.1-0-g889da6a-release).

If additional details or clarifications are needed, please let me know, and I will respond promptly.

Thank you in advance for your time and support!

jzw_0_5_pressit.csv
jzw_0_5-input.txt
jzw_0_5.out.txt

[mcgratta]

I do not know if this strategy is going to work. There are only a few parameters that can be changed after a RESTART. It sounds to me that you should not try to modify the RAMP. Is there an example in the User's Guide that you are following?

[MrAhriman]

The idea with external control is you would have a second program running like a python script or complied code that monitors FDS output files, does its calculation, and writes the update file. Not that you are doing this calculation by hand. Unless you have an extremely complicated calculation, even for relatively short DT_EXTERNAL values the FDS calculation is likely to be much longer than the external control calculation.

Thank you very much for your explanation, in fact your approach may be correct, but to follow your approach, my existing code structure needs to be adjusted a lot. This is already difficult for me to achieve as I approach graduation. So what I'm looking forward to is whether a problem with my FDS setup caused this error and how I can fix it. According to the existing logic, changing the ramp after restart is no problem from the perspective of the layman, ennnn. My shallowness led me to the wrong path.

What is the root cause of this problem, and what parameters can be modified to reduce the frequency of this problem may be my preferred answer. It may be difficult for your predecessors who specialize in fluid mechanics to accept this

Thank you again for your explanation

[MrAhriman]

You could also use the DT_SPEC, DT_TMP, and DT_UVW outputs instead of restart. These can be used to dump the gas concentrations, temperature, and flow field at the time of interest. You can then start a new simulation where you initialize the flow field using the fields output from the prior simulation. Your ramps would then start at T=0 in the new simulation.

I didn't quite understand what you meant.
In fact, my job is to collect the temperature value at the previous moment and then calculate the air volume for the next moment based on that.
Could you please elaborate on whether your method can achieve this operation?

Take a look at the section on changing the initial conditions in the user guide, specifically the section on nonuniform initial fields.

I had another thought on this which I think would work if you really want to stick to the RESTART approach. You could also set the HVAC flow rate through an EXTERNAL_FILE on the RAMP. Then, when you go to RESTART change the value within the EXTERNAL_FILE. @drjfloyd would know better if that is likely to cause it to crash immediately. I am not sure what information is stored regarding the EXTERNAL_FILE in the RESTART files.

Thank you very much for your explanation, in fact your approach may be correct, but to follow your approach, my existing code structure needs to be adjusted a lot. This is already difficult for me to achieve as I approach graduation. So what I'm looking forward to is whether a problem with my FDS setup caused this error and how I can fix it. According to the existing logic, changing the ramp after restart is no problem from the perspective of the layman, ennnn. My shallowness led me to the wrong path.

What is the root cause of this problem, and what parameters can be modified to reduce the frequency of this problem may be my preferred answer. It may be difficult for your predecessors who specialize in fluid mechanics to accept this

Thank you again for your explanation

[drjfloyd]

The intended purpose of RESTART is to allow for recovery from failures of your computing hardware during a long run. Power outage, windows blue screen of death, etc. It was not intended as a method for changing your inputs; however, because of the way the restart process is structured, in some but not all cases you can get away with modifying the input file before restarting. If this isn't working for you, then you have one of the cases where it isn't going to work and you will need to find an alternate approach. Two have been suggested to you - external control and use of FDS's built in control functions.

[MrAhriman]

The intended purpose of RESTART is to allow for recovery from failures of your computing hardware during a long run. Power outage, windows blue screen of death, etc. It was not intended as a method for changing your inputs; however, because of the way the restart process is structured, in some but not all cases you can get away with modifying the input file before restarting. If this isn't working for you, then you have one of the cases where it isn't going to work and you will need to find an alternate approach. Two have been suggested to you - external control and use of FDS's built in control functions.

I will try external control, and I might need your help. Thank you once again.

[MrAhriman]

重启的目的是允许在长时间运行期间从计算硬件的故障中恢复。停电，窗口蓝屏死机等。它不是用来改变输入的方法；然而，由于重启过程的结构，在某些（但不是所有）情况下，你可以在重启前修改输入文件。如果这种方法对你不起作用，那么你就遇到了这种方法不起作用的情况之一，你需要找到另一种方法。有两个建议给你——外部控制和使用FDS内置的控制功能。

External Control
The realization of the function is not like the traditional ramp; I need to specify the time and value.

&RAMP ID='VF Ramp',T=1,F=0/
&RAMP ID='VF Ramp',T=33,F=-20/

May I ask how I can transfer

&RAMP ID='VF Ramp',T=33,F=-20/

How about implementing it with External Control?

The examples in the help documentation are &SURF ID='HOT',TMP_FRONT=1000,RAMP_T='T RAMP',COLOR='RED'/
&RAMP ID='T RAMP', EXTERNAL_FILE=T, INITIAL_VALUE=0.081633 /

RAMP,"T RAMP" , 1
Like this, there is no time element inside.

When I:

A line of code: &RAMP ID='VF Ramp', EXTERNAL_FILE=T, INITIAL_VALUE=0/

forrtl: severe (157): Program Exception - access violation

The simulation cannot be carried out.
jzw_0_5.fds.txt
external_test_input.csv

[MrAhriman]

#13768 (comment)

maybe it is a reason,i will try this method,i don't know