After 2 years of bashing the mouse in Motion Analysis, or whatever they call it now, I have concluded that the precursors of failure are, in sequence of commital:-

1) Inelegantly modelled solids. There is no single rule for elegantly modelling solids.

2) Inelegant mates. There is no single rule for elegantly specifying mates.

3) Sub optimal motion property settings. There is no single rule for determining optimal settings.

You know you have done something elegantly when the person looking at your work says "Why didn't I think of that!"

The best move I made to advance my project was to spend my SW Premium subscription instead, on an industrial colour laser printer which boosted my productivity and presentation by several orders of magnitude more than an upgrade to SW would have. Specifically it enabled effortless (compared to the old ink jet) screen capture of consequtive incremental runs to paper for red pen annotation, and comparison.

Load bearing 3D Contact is interference/penetration detection plus response. It is a damped elastic system superimposed on rigid undeflecting bodies. The damped elasticity shows in, and may overwhelm, the results for the rigid bodies.

Setup requires making 17 selections over 3 dialogs. Although default values are offered, they may be useless for your job. They affect either:-

* Progress of calculation; i.e. stopping or slowing progress. Nothing speeds progress.

* Validity (forget accuracy) of results; i.e. are they in or out of the park?

* Discrimination between results and noise.

These selections are pure fudge; they have no relation to any real material property (get that out of your head).

There seems to be a limitation on incremental penetration per integration step, beyond which the algorithm cannot respond and RAM is left in an indeterminate state and the document with settings and results cannot be saved.

As contact pressure increases, it is necessary and safe to push some parameters to their limits, and to calibrate the remaining parameters on a similar but simpler model, where the expected results are agreed on beforehand by all parties.

Examples of simple models include box sliding on ramp, ball rolling in channel.

My lifes work is to simulate a choo choo train rolling around a loop track. I have tried it, and the results are awful. So I chose a similar but simpler model in an effort to determine optimal settings for my job.

I retained my model track, and substituted a ball for the train of wheelsets. I chose a ball diameter which gave a slope at the point of contact with the rail equal to the slope of the wheel tread. I chose a ball density to make the mass of the ball equal to the mass supported by one wheelset. I split the surface of the ball so that I could visually detect rolling motion, and applied friction between the ball and the rails so that the ball would roll rather than slide. I positioned the ball on the rails in the straight and gave it an initial velocity toward a curve.

I have enough information to easily calculate the tipping moment and the restoring moment in the curve. Subtracting one from the other and dividing by the distance between the rails yields the contact force on the inside rail. It equals the SW result for that force. I exported the results to a spreadsheet and created a new column being the sum of the inner and outer rail contact forces. By fortunate coincidence it was constant and equal to the weight of the ball.

The images show the ball rolling along the track, and the spreadsheet plot of results.

I understand that when transiting from a straight to an arc, centrifugal force applies instantly from the tangent point. The plot shows a damped elastic application, and a burst of noise. My media player has controls for stepping frame by frame through the movie.