This camera recording shows a wheel rolling on a rail. Other views show that it is tracking correctly.
Part way through the journey the wheel penetrates the rail and falls to the centre of the earth.
I don't know why, or what to do about it.
There's a few potential causes:
Max time step too large
Stiffness not high enough
Contact not tessellated with enough detail (if using facetted contact).
Penetration volume calc leading to spurious opposing force direction.
From the look at the start of the animation, it appears there's a significant amount of penetration to start with. I could be wrong and it may just be the perspective of the cutting plane with respect to the wheel. But if that's the case, crank up the stiffness and damping an order of magnitude for the contact
If facetted contact, try to increase accuracy (number of tessellations). Otherwise move to precise contact.
Other things you can do to help are:
a) ensure the geometry of the wheel is a single solid body (if it's thin walled then at causes problems with large penetration).
b) Simplify rail geometry (if the I section and lower portion plays no roll with the wheel. Just have a body with the top flange geometry of the rail (Follow the KISS principal at all times)
Hope this helps,
I discovered that WMP can step thru each frame, so I raised the video capture fps to 30.
The frame below shows that the wheel starts penetrating the rail before it enters the curve. Other views show that the train follows the curve even while the wheels are falling through the rails. I have previously demonstrated that penetration is a consequence of contact pressure. The gravity load on the wheel tread is much greater than the centrifugal load on the wheel flange, so flange contact is honoured when tread contact is not.
The integration settings are:-
fps = 1000
precise contact = yes
accuracy = 1E-5
Integrator = S12
max iteration = 50
initial step = 1E-4
min step 1E-9
max step = 1E-3
jacobian = every
3D contact settings are:-
stiffness = 1E12
exponent = 2
damping = 5E6
penetration = 1E-4 metres
Presentation requires that the rail look like a rail.
What should I tweak next?
Can you see what the magnitude of the contact force is at that time frame. Based on the stiffness that should give you an idea about how much penetration should occur.
Your contact parameters should be high enough to resist most forces.
The only other though is if there is a constraint or motion that is forcing the penetration. Are you just applying rotating motion to the wheels? If so is it defined with respect to the bogie or whatever is holding the wheel? Thati s a crucial point. If no reference is specified it is with respect to global ground and that locked to yaw orientation of the wheel.
I have not measured anything yet, and the only driver is initial velocity. This is the crawl, walk, run paradigm.
My model is more complex than previous versions.
I will tinker with other aspects while I mull over this.
Retrieving data ...