Simulation unexpected behavior

[Mohammad Wasfi]

Hello, 

I am trying to perform a simulation using the GPU module. I am dropping a rotating cylindrical screw on a granular media bed in this simulation. In addition, I am trying to perform this simulation using two different gravity values (9.81 m/s^2 and 0.11  m/s^2  ) and two different geometries (30-degree helix angle and 40-degree helix angle). First, I performed this simulation using the Earth's gravity (9.81m/s^2). I have used both geometries and the simulation was normal where the screw dropped on the bed and rotated (about its axis) which caused the screw to move on the bed. However, when I started using a gravity value of (0.11 m/s^2), one of the screws acted normally where it dropped on the bed and rotated normally similar to the earth simulation. However, when I used the second screw geometry, the screw would drop on the bed and then rotate and sly away from the bed. Note that the two different geometries had some very minimal differences especially when it comes to mass and moments of inertia. I was wondering if someone could give me some inside about this weird phenomenon. I have attached some pictures of the flying screw simulation and the regular simulations where the screw acted as expected. I have also attached my simulation file for your reference. 

Thank you so much in advance, 

[Ruochun Zhang]

Hi Mohammad,

Low gravity can make the simulation more challenging due to reduced penetration. If something in your simulation such as wheel angular velocity does not scale with gravity, then it can lead to much smaller time step size requirements (same speed + less overall penetration = small step size to sufficiently capture contact events). So yes, extraterrestrial test scenario can be more difficult, there is no reason to assume the same simulation that works on Earth will work on the moon. Again, it can be even more difficult if that specific mesh has more pointy small features. The usual treatments apply, massaging step sizes, enlarging particle sizes or CoR, smoothening meshes, you name it.

Thank you,

Ruochun