Spinning Wheel-GPU

[Mohammad Wasfi]

Hello there, 

I am trying to perform a simulation using the gpu module where I want to import a cylindrical screw (picture attached) into a bed on granular media and then have the screw spin on top of the bed and then obtain the forces/torques on the screw. 

I am not sure about what would be the best way to import the screw in a way that it's set on the bed at the start of the simulation. Also, I am not sure how I can impose a type of motion that would force the screw to spin on top of that bed since I am importing it as an .obj file. Finally, due to the shape of the screw, I assume that when the screw moves it will tend to turn to the left/right. Is there a way to restrict the motion of the screw to be just straight while spinning? I was wondering if there are any demos regarding this or if someone has done something similar before. Any help will be appreciated. 

Thank you so much in advance, 

[Ruochun Zhang]

Hi Mohammad,

I think that demo_GPU_mixer.cpp demo does exactly what you needed. While you are looking into this demo, I just want to mention one thing to avoid confusion: in terms of controlling the mesh motion in C::GPU, the Transform call before AddMesh should only be used to "align" the mesh with its CoM before the simulation starts, so that when the mesh is loaded it is expressed in its centroid and principal frame; it should not be used to set the initial location of the mesh. Instead, like in the demo, use ApplyMeshMotion to control the mesh motion throughout the simulation.

Hope it helps!

Ruochun

[Mohammad Wasfi]

Hi Ruochun,

Thank you so much for your help. This actually helped a lot. I have some questions that I hope you could answer. I am trying to control the mesh motion using  ApplyMeshMotion as you mentioned. However, I feel like I do not fully understand everything related to that so I wanted to ask some questions and hopefully, this will help others as well. I have looked at the documentation and they are really brief so I hope you could answer some of these questions here. I will use examples from my code(attached here). I might be using some stuff incorrectly so I think it would be a good way to see if what I am doing is correct or not. 

1- gpu_sys.AddMesh("../Screw_CM.obj", ChVector<float>(0,0,-90), ChMatrix33<float>(1), screw_mass);

                  -  ChVector<float>(0,0,-90): does this represent the initial position where the mesh is initiated, and does it affect other parts of the                            simulation?
                   - ChMatrix33<float>(1): what does this represent? I looked through the documentation and it says that this is the rotscale. Could you                           give a little more explanation as to what that is?

2- ChVector<> ball_initial_pos(0, 0, 0);      

std::shared_ptr<ChBody> screw_body(sys_screw.NewBody());

        screw_body->SetMass(screw_mass);
        screw_body->SetInertiaXX(ChVector<>(screw_inertia_x, screw_inertia_y, screw_inertia_z));
        screw_body->SetPos(ball_initial_pos);

        screw_body->SetBodyFixed(false);
        sys_screw.AddBody(screw_body);

           - When creating a system using the imported mesh, what does  screw_body->SetPos(ball_initial_pos) do? Clearly, it is a way to set the                         position of the mesh. However, since the  ChVector<float>(0,0,-90) was used before when adding the mesh, is there a need to use  SetPos?

3-     ChVector<float> mesh_pos(0, 0, 0);      

        float rev_per_sec = 10.f;
        float ang_vel_Z = rev_per_sec * 2 * (float)CH_C_PI;
        ChVector<> mesh_lin_vel(0);
        ChQuaternion<> mesh_rot = Q_from_AngZ(t * ang_vel_Z);
        ChVector<> mesh_ang_vel(0, 0, ang_vel_Z);
        gpu_sys.ApplyMeshMotion(0, mesh_pos, mesh_rot, mesh_lin_vel, mesh_ang_vel);

                -  mesh_pos(0, 0, 0): to me, it seems that we are setting the position of the mesh a third time. could you explain how this affects the                             simulation? 

                 -   mesh_lin_vel(0): does this represent the initial velocity of the mesh? or is it applying a constant velocity throughout the entire                                    simulation?

                   -   mesh_rot : Could you explain what this exactly is and how changing it affects the simulation?

                  -   mesh_ang_vel(0, 0, ang_vel_Z): is this the initial rotational speed or it is a constant rotational speed that the mesh will have                                        throughout the entire simulation? 

Thank you so much in advance for taking the time to help us with everything, 

[Ruochun Zhang]

Hi Mohammad,

1. The prerequisite is to know that C::GPU works in center-of-mass frames, like Chrono in general. The accompanying vector and matrix that you specify during AddMesh are intended to be used as some sort of pre-processing for your mesh file. To cut to the chase, if your mesh file is created such that the mesh is already in its center-of-mass frame, then for all purposes, you should always call gpu_sys.AddMesh("../Screw_CM.obj", ChVector<float>(0,0,0), ChMatrix33<float>(1), screw_mass);

If your mesh is not already in its center-of-mass frame, or you want to scale it before loading it, then you can supply those arguments. ChMatrix33<float>(1) would create an identity matrix, and this matrix will be applied to every point in your mesh (as a side note, the matrix is applied before the vector is also applied to every point in your mesh), as a part of the pre-processing. Obviously, the identity matrix does nothing to your mesh. If it was a diagonal matrix, it would "scale" your mesh; if it was a unitary matrix, it would rotate your mesh; if it was a general matrix the it would "rotate and scale" your mesh, hence its name.

Because of what I said, I do not recommend using this ChVector argument to set the initial CoM (I just want to be more specific; you can call it initial position, too) of the mesh in simulation, it is against its design purpose. I said not recommend, but it can functionally work as a initial position specifier: think about it and you'll figure out why. Also, while other things may work fine, you will get strange torque readings from the mesh if you do that: again you can figure out why.

In principal, use ApplyMeshMotion to control mesh position/velocity in simulation, including the initial position.

2. As you can see, this is managing a ChBody and has nothing to do with the mesh in your gpu_sys. SetPos just sets the position of that ChBody. Because it has nothing to do with gpu_sys, if you need it to be at a location, you need to set it separately. In this co-simulation example, the mesh you loaded is the proxy in DEM simulation for this ChBody. The gpu_sys will collect the contact force on the mesh, feed this info to the ChBody, and the ChBody will use this force info to update this body's location, and then the new location is used to update the mesh location in gpu_sys. ChBody can manage more complex rigid body motions such as those involving joints, so it is used to help manage external objects in the DEM simulation.

3. If you now understand the above, then you can see you are really just setting the mesh position (in gpu_sys) for the very first time here. ApplyMeshMotion is called at every time step, so every quantity here is not just some initial value, but the values throughout the simulation. Some of the values are changing every step (because t changes), such as mesh_rot; others such as velocities happen to be constants in this simulation. mesh_rot is a quaternion which represents the rotation status of the mesh (or specifically, the mesh's local coordinate system). If that sounds unfamiliar then it is probably helpful to read a bit more about quaternions.

Hope it answers your questions!

Ruochun

[Mohammad Wasfi]

Thank you so much, this is really helpful!