Hello,
I am experiencing an issue and I think it is a bug if I understand the documentation correctly.
I have been trying to calibrate my fiber model by having a fiber be bent by a inflow of velocity. Then by measuring the deflection at the end of the fiber, make adjustments to my fiber young's modulus.
The issue:
I have made a simple 1-way coupling fix for liggghts assuming the force of drag acting on a sphere is F_drag = (b+c(v))v. Using this fix, I obtain deflection results very close (< 1% error) to what is predicted by beam theory.
However, when I try replacing my airdrag fix with a drag model provided by CFDEMcoupling, I get drastically different results. (orders of magnitude off with a percent error of ~35000%).
This has led me looking at the code base with little head way. I then decided to try the noDrag model. This seemed to do exactly what it is suppose to do and yielded no deflection. I then turned my air drag fix back on while still using the noDrag model. This resulted in a deformation percent error ~ 13000%.
Should this difference exist between using cfd-dem coupling with the noDrag model vs dem only simulations?
I have attached my test tree.
The cfd-dem simulation takes about 640 core-seconds to run on my machine (with 8 cores == 80 seconds).
| Attachment | Size |
|---|---|
 cfd_dem_tests.tar_.gz | 142.62 KB |
mschramm | Sat, 07/04/2020 - 07:45
Found Issue
Found the issue.
During the coupling process, the variable bond_hist is reset for the atom_vec_bond_gran class.
This reset causes the bond forces and bond torques to be reset, causing the increase in the deflections.