Desgrid error when using PSD

MFiX Bug report
#1

Error: Solver crash!
Program received signal SIGSEGV: Segmentation fault - invalid memory reference.
Backtrace for this error:
#0 desgrid_MOD_desgrid_neigh_build
at des/desgrid_mod.f:1060
#1 neighbour_mod_MOD_neighbour
at des/neighbour.f:65
#2 des_time_march_MOD_des_time_step
at des/des_time_march.f:234
#3 run_dem
at mfix.f:211
#4 run_mfix
at mfix.f:146
#5 main_MOD_run_mfix0
at main.f:79
#6 start_thread
at /build/glibc-SzIz7B/glibc-2.31/nptl/pthread_create.c:477
Segmentation fault (core dumped)

Primary job terminated normally, but 1 process returned
a non-zero exit code. Per user-direction, the job has been aborted.

Dear sir!
When I use PSD in the simulation, the error above always appears.For example,It runs successfully when I set standard deviation to 0.0005,however ,this error appears when I just change it to 0.001.this is one of the case.Could you give me some advice,please?

Screenshot from 2023-05-08 20-58-27
Screenshot from 2023-05-08 20-58-27794×417 43.2 KB

Screenshot from 2023-05-08 20-59-01
Screenshot from 2023-05-08 20-59-01828×467 305 KB

Screenshot from 2023-05-08 21-00-37
Screenshot from 2023-05-08 21-01-42

#2

Please attach your input files - go to MFiX main menu, “Submit bug report”, attach the ZIP file here. Thanks.

#3

did you change the d_p0 value to a lengthscale close to the min of the size distribution? Maybe 1mm or little over?

1 Like
#4

Dear sir,I have changed d_p0 value in my case according to your advice.However,it still didn’t work.
Besides,I noticed a case named hopper_dem_custom_psd in MFIX tutorials. its dp0 value is 0.002m which is much larger than the min(0.000292m) of the size distribution.What’s more,this case runs successfully.
dp0
psd

dp0

#5

You may need to increase the spring stiffness and/or run the neighbor search more often.

#6

Thank you sir! But I don’t know how to run the neighbor search more often.Do you mean make the neighbor_seach_n larger?
Besides,I want to know the minimal size of grid.I noticed in previous post that desgrid size should not be less than 2diarlm_factor(Grid-based neighbor search algorithm, why the desgrid size should not be less than 2*dia*rlm_factor?).However,in MFIX tutorials(hopper_dem_custom_psd),the largest diameter of particles ,the rlm and grid size are 3.33mm,1.2, 5mm respectively.Thus,according to the post, 2diarlm=23.331.2=7.99>5mm.So is it right?
maxdiameter
cell size

Apart from that,In my case, I used pure granular flow. The largest diameter of particles is 3mm. I want to know whether it is right if I set grid size to 4mm?

#7

You need to decrease neighbor_seach_n to run the neighbor search more often. The grid spacing you show above is for the fluid cells. You may need to change the des grid spacing. Please attach your files if you want someone to take a look.

#8

error_report_2023-05-19T194512.724101.zip (4.5 MB)
Dear sir!
I have changed both of the parameters you mentioned,but this error still appeared later.
So I attach it for you to check.Thank you !

#9

Can you please attach your psd.txt file?

#10

psd.txt (137 Bytes)

#11

What parameters have you tried to change?

  1. Your spring stiffness is 100 N/m which is very small. Have you tried to increase it?
  2. Your neighbor_seach_n is set to 25 which is the default value. Have you tried to decrease it?

You are also over-decomposing the parallel partition. Please try NODESI=4, NODESJ=6 and NODESK=2 (40 cores) to see if this helps.

#12

Thank you very much sir