Segmentation fault in PIC simulations

MFiX
#1

I am trying to solve a problem where only energy equations for gas and solid phase are solved. The model employed for the simulations is PIC. I could not run the simulations because of the segmentation fault (invalid memory reference).

Program received signal SIGSEGV: Segmentation fault - invalid memory reference.

Backtrace for this error:
#0 0x7fddfca382da in ???
#1 0x7fddfca37503 in ???
#2 0x7fddfc0b4f1f in ???
#3 0x560c687b7c99 in calc_avg_sol_vel_
at mfix-19.1.2/model/des/pic/calc_avg_sol_vel.f90:91
#4 0x560c687aeae4 in __make_arrays_des_mod_MOD_make_arrays_des
at mfix-19.1.2/model/des/make_arrays_des.f:241
#5 0x560c685b0a74 in _main_MOD_initialize
at mfix-19.1.2/model/main.f:375
#6 0x560c68569308 in run_mfix
at mfix-19.1.2/model/mfix.f:130

For your reference, I attached the input file with this thread.

Your comments on the error are appreciated.
Thank you.
input.mfx (9.1 KB)

#2

I’m not sure how you generated this input file. There are various settings that are not consistent with the PIC model and wouldn’t have been permitted by the GUI. For example, the current PIC model does not support GARG interpolation.

#3

I agree that only LINEAR_HAT interpolation works in PIC. Also, there is no such thing as 2D in PIC simulation. It requires a 3D framework.

#4

First, I generated the input file via GUI. This file results in a memory error. So, I looked at the documentation (https://mfix.netl.doe.gov/doc/mfix/19.1.2/reference/discrete_element.html?highlight=linear_hat) and made some changes to the input file. This modified input file is supplied to the mfixsolver which is generated from tarball.

Thank you very much. I will change the interpolation scheme to linear_hat.

#5

Thank you very much for your suggestions.
I modified the input file to carry out 3D simulations. I am able to get rid of the memory reference error.
In addition, I changed the interpolation scheme to LINEAR_HAT. For your reference, I attached the input_3d.mfx (9.3 KB)
modified input file.

However, I probably misunderstood the following statement in the documentation:
“The first represents an entire misunderstanding of MFIX-PIC which is necessarily a 3-D model” (https://mfix.netl.doe.gov/doc/mfix/19.1.2/reference/pic_tips.html?highlight=pic)

Thank you.

#6

Thank you for your suggestion about our documentation. Perhaps we can make that statement more clear.

Also, be careful in your simulation. I notice two more things. (1) cgs units (which can be problematic) and (2) in terms of PIC, a very low statistical weight = 1.0. I understand why you are choosing this weight because your particles are large with respect to your mesh, but do be mindful. PIC is not DEM and will not behave the same way. There are no Newtonian mechanics, only statistical manipulations of solids stress. PIC works best in simulations with very high particle counts and large statistical weights. Good luck!

#7

I appreciate your comments on PIC simulations.

I implemented a radiation model and would like to test it on the PIC model. So, the input file is generated to make a test case in which the energy equation is solved along with a thermal radiation model. The test cases for TFM and DEM models are in the CGS system. Therefore, the CGS system is used in the input file as well. However, thank you very much for the suggestion. We will use SI units in the upcoming test cases.

I used low statical weight as the particle count in the cell is low. I will definitely use large statical weights for high solid void fractions and relatively small particles. Thank you.