I keep getting the following error when modeling a contact:
---- START: WARNING ----
The Integrator has detected a singular Jacobian matrix.
This problem may be caused by:
(a) A constraint equations that became almost redundant, e.g., the model
is at a singular configuration or at a bifurcation point.
(b) A massless PARTS in the model.
The Calahan Sparse Solver has detected an unexpected singularity at row 1271.
The corresponding equation is: Contact/Torque3D/39 Y Torque
A minimization algorithm will be used to find a solution on the
projected constraint manifold.
---- END: WARNING ----
There are multiple contacts in the model, many of which are identical to the one that I am getting the error at so I was wandering if there is any way to find out what is this singularity related to. I've already trying isolating it (by deactivating the other contacts) but I could not draw any solid conclusions.
The first thing is to double check the model has no massless parts. Then you could run the same simulation up to fraction of a second before the issue happens with a very short timestep and visualise/plot what is going on with that Contact. Alternatively is to run debug/eprint.
Josef, also make sure to have realistic stiffness values. Use HertzWin to estimate the parameters. If anything, start on the low side for stiffness. After setting up the material and geometry in HertzWin, enter a small force, read the resulting displacement. Create an Excel sheet and make a table with displacement (x) and Force (y). Type in the values you had from HertzWin. Now repeat this for several forces, I usually use something like 100, 200, 500, 1000, 2000, 5000, 10000, 50000 [N]
Plot this in Excel, create a trendline of POWER type and set "Display Equation on Chart". Now you can directly read the stiffness and exponent values. (You might have to increase the number of digits shown for the equation.)
Make sure to convert stiffness properly if converting between millimeter and meter . Yes, the conversion between those two are 1000. But the stiffness has the unit Force/length^exp.