Questions about collision model

[chz31]

Hi,

I'm trying to land a plate to a fractured orbital floor. I tried to disable the tetrahedron model for the orbital floor and fix it first. However, the collision behaved a bit strange.

The plate did penetrate the orbit entirely, but a close look showed that the collision model appeared to be activated at certain regions, and those regions were actually deformed (though they still penetrated the orbit model).

Here are how I set up the plate and orbit model
Plate with tetrahedron:

    plateNode = root.addChild("plate")
   # Creating the falling plateNode object
    plateNode.addObject('EulerImplicitSolver', name='odesolver')
    plateNode.addObject('CGLinearSolver', name='Solver', iterations=25, tolerance=1e-05, threshold=1e-05)
    
    #tetrahedron topology
    plateNode.addObject('TetrahedronSetTopologyContainer', name='topo', src=meshLoaderCoarse.getLinkPath())
    plateNode.addObject('TetrahedronSetGeometryAlgorithms', template='Vec3d', name='GeomAlgo')

    plateNode.addObject('MechanicalObject', template='Vec3d', name='MechanicalModel', showObject=True)
    plateNode.addObject('TetrahedronFEMForceField', name='FEM', youngModulus=10000, poissonRatio=0.4, method='large')
    plateNode.addObject('UncoupledConstraintCorrection', defaultCompliance=1e-9) #need for gravity filed #LinearSolverConstraint
    plateNode.addObject('MeshMatrixMass', massDensity=10, topology='@topo')

    collision = plateNode.addChild('Collision')
    collision.addObject('MeshTopology', src=meshLoaderFine.getLinkPath())
    collision.addObject('MechanicalObject', name='CollisionMO')
    collision.addObject('TriangleCollisionModel', name='CollisionModel', contactStiffness='100')
    collision.addObject('BarycentricMapping', name='CollisionMapping', input='@../MechanicalModel', output='@CollisionMO')

    # Fine visual model for plate with tetrahedron
    fineVisualModel = plateNode.addChild('FineVisualModel')
    fineVisualModel.addObject('OglModel', name='VisualModel', src=meshLoaderFine.getLinkPath())
    fineVisualModel.addObject('BarycentricMapping', name='Mapping', input='@../MechanicalModel', output='@VisualModel')

Here is how I set up the orbit model without tetrahedron:

    # Creating the orbit object
    orbitNode = root.addChild("Orbit")
    orbitNode.addObject('MechanicalObject', name="mstate", template="Rigid3", showObject=True, translation2=[850, -150.0, 250], rotation2=[0, 0, -90])
    orbitNode.addObject('UniformMass', name="mass", vertexMass=[totalMass, volume, inertiaMatrix[:]])

    #### Collision subnode for the orbit
    orbitCollis = orbitNode.addChild('collision_orbit')
    orbitCollis.addObject('MeshTopology', src=orbitLoaderFine.getLinkPath())
    orbitCollis.addObject('MechanicalObject', name='CollisionMO')
    orbitCollis.addObject('TriangleCollisionModel', moving=False, simulated=False)
    orbitCollis.addObject('LineCollisionModel', moving=False, simulated=False)
    orbitCollis.addObject('PointCollisionModel', moving=False, simulated=False)
    orbitCollis.addObject('RigidMapping')

    # Fine visual model for plate with tetrahedron
    orbitVisu = orbitNode.addChild('FineVisualModel')
    orbitVisu.addObject('OglModel', name='VisualModel', src=orbitLoaderFine.getLinkPath())
    orbitVisu.addObject('RigidMapping')

The full script orbit_with_plate_tetrahedron.py and testing models are here: https://drive.google.com/drive/folders/1I6t7hrlMDNoies4Azd0gR5-LQjyJW0t5?usp=sharing

The only way I could make the collision work so that the plate did not penetrate the orbit is to disable the tetrahedron in both the plate and orbit.

Can anyone give me some suggestions? Thanks!

[chz31]

The collision detection somehow worked after adding'LineCollisionModel'and PointCollisionModel to the collision model of the plateNode in addition to only using TriangleCollisionModel! I also used LinearSolverConstraintCorrection and SparseLDLSolver. I also had to set up the Youngs Modulus super high for the plate as 1.05e11.

[bakpaul]

Congrat for this solution !

It makes sense, as I've told you the UncoupledConstraintCorrection is not mechanically sound and for complex collision problems with high coupling between contact points (which is the case here), you'll get a very bad behaviour.