Discussion Forum

Hello all,

I am learning COMSOL and working my way up from less complex models. I have a decent understanding of stationary and time-dependent studies with fixed geometry, and I am attempting to do a 2D study with moving geometry. The study is of a paddle-like object moving through the air and rotating. I am using deformed geometry, as the translation of the paddle is given, and the deformation of the paddle itself is essentially zero.

My end goal is something similar to this blog post: www.comsol.kr/blogs/how-to-improve-your-paddle-stroke-with-simulation/
Though without the boat hull.

I have set up the geometry and meshed as per the COMSOL moving geometry tutorials. If I set the BC of the paddle to be flow velocity zero, I can run the simulation and watch the paddle move in the appropriate motion. However, if I set the paddle BC to "moving wall" with wall velocity set to anything other than zero, the solver fails on the first timestep with a "nonlinear solver did not converge" error. Additionally, if I let the zero-wall-velocity simulation run for a while, eventually it will throw the same error and stop. If I look at the velocity field results, there are spotty patterns of higher velocity that spontaneously form throughout the mesh. All velocities are extremely small (10^-12 m/s), but the spontaneously forming velocities are ~100 times greater, so I think they might be a symptom of whatever problem I'm facing.

I have rebuilt the problem with the same geometry, except with only rotation, and have run into exactly the same problem. Changing the mesh does not affect anything, nor does making the movement of the paddle smaller (set it to move only a fraction of the smallest cell size, and it still throws an error at t=0).

Any help or insight you fine folks have would be greatly appreciated.

Best,
Mark Ploeger

Edit: I've attached a project file. This is for a purely rotating paddle, but it has the same issues. If the "Wall 2" BC (for the surface of the paddle) is set to uw=0, it resolves ok, and the rotation of the paddle is correct. However, if that is set to "moving wall" with the derivative of x and y position in the ux and uy fields, it has an error on the first timestep. I realize this particular problem could be done in the rotating machinery module, but this is a stepping stone towards having rotation and translation, so I want to be able to do it within the deformed geometry model.