Error using "Spray Model" for GDI SI Engine Combustion
 

Hello everyone,

I have been trying to simulate GDI SI Engine Combustion by using Spray model. I found the tutorial case in Converge 2.4, namely, "Tumble_GDI_SAGE", which was used to set up the spray model. I have used two different versions of Converge, 2.3 and 2.4 and I am facing simulation errors at one particular Crank Angle.

So, my injection starts at around -280 CAD for an injection period of around 50 CAD. Now, I run my simulation from 1080 CAD to 1800 CAD. Now, injection happens from 1160 CAD to 1210 CAD.

In Converge 2.3, at around 1356 CAD, the memory exceeded 125 GB and I got a "malloc " error, which is because the RAM in the system is 128 GB. This was strange to me and I further used Converge 2.4 to simulate this case by converting the files to 2.4 format (based on Converge 2.4 manual).

Now, in Converge 2.4, for the same simulation, the memory required does not exceed 70 GB. However, the simulation again crashes at around 1356 CAD. For this simulation, the error is however different. It is a segmentation fault.

Updated parallel map by METIS is ready
The parallel block size (dx, dy, dz) is 1.600000e-02 1.600000e-02 1.600000e-02.
There are 477373 active cells in the domain
for rank 0, tot_num_inputs[0]=18
for rank 0, tot_num_inputs[1]=8
Number of frozen blocks== 37; Number of frozen islands==8



Number of vertices, constraint, and CPUs
289 1 36

starting METIS ..... ..

Updated parallel map by METIS is ready
The parallel block size (dx, dy, dz) is 1.600000e-02 1.600000e-02 1.600000e-02.
Freezing extra blocks.
--------------------------------------------------------------------------
mpirun noticed that process rank 9 with PID 2079 on node kim-e320 exited on signal 11 (Segmentation fault).
--------------------------------------------------------------------------

I am new to Spray modeling and hence I am trying out with the same settings as in the Converge 2.4 "Tumble_GDI_SAGE" case.

I am unsure as to how to go about resolving this error. Any leads are greatly appreciated. Thank you! :)

Balavignesh

Balavignesh,

Your memory utilization appears to be high. I have a case with 1.2M cells and the maximum memory utilization is 30 GB. Can you share plots of Total_MEM vs. Crank from the data in mem_usage.out and Total_Cells vs. Crank from the data in cell_count_ranks.out?

What value are you using for parallel_scale in inputs.in?

Best regards,

Tristan

It is a good idea to simulate GDI SI engine combustion by using spay model. I used to make extensive nozszle flow and GDI spray simitations.