Common plane higher order quadrature#208
Conversation
srwopschall
left a comment
srwopschall
left a comment
There was a problem hiding this comment.
@ebchin - a lot of comments, but only a handful of things that really need to be ironed out. Please review. In the meantime I need to look at the entirety of CommonPlane.cpp and verify the code and when/where average face coordinates are used vs. current configuration and possibly warped face coordinates. I want to be sure we don't have any inconsistencies.
| { | ||
| constexpr int numVerts = 4; | ||
|
|
||
| RealT x1[numVerts] = { 0.0, 1.0, 1.0, 0.0 }; |
There was a problem hiding this comment.
Minor: considering adding a comment about the configuration of the two blocks including the gap.
| tribol::ExecutionMode::Sequential ); | ||
|
|
||
| tribol::setPenaltyOptions( 0, tribol::KINEMATIC, tribol::KINEMATIC_CONSTANT ); | ||
| tribol::setCommonPlaneIntegrationOptions( 0, rule, 3 ); |
There was a problem hiding this comment.
Consider clarifying what 0 and 3 are since someone may come to the tests to see how this is used.
| return result; | ||
| } | ||
|
|
||
| struct EdgeLocalContactResult { |
There was a problem hiding this comment.
Is this struct intended to hold a force result. Consider consistent naming similar to the WarpedQuadForceResult.
| { | ||
| constexpr int numVerts = 2; | ||
|
|
||
| RealT x1[numVerts] = { 1.0, 0.0 }; |
There was a problem hiding this comment.
Minor: consider adding comment about configuration of the two edges. Is this an interpen interaction or full overlap?
| tribol::ExecutionMode::Sequential ); | ||
|
|
||
| tribol::setPenaltyOptions( 0, tribol::KINEMATIC, tribol::KINEMATIC_CONSTANT ); | ||
| tribol::setCommonPlaneIntegrationOptions( 0, rule, 4 ); |
There was a problem hiding this comment.
Clarify what 0 and 4 are.
| const RealT length = magnitude( x1 - x0, y1 - y0 ); | ||
|
|
||
| for ( int qp = 0; qp < num_qpts; ++qp ) { | ||
| const RealT s = segment_rule_coords[qp]; |
There was a problem hiding this comment.
Minor: consider for clarity adding comment about what this mapping is.
| const bool mapped_face2 = | ||
| EvalLinearEdgeAtProjectedPoint( actual_xf2, x_q, overlapNormal, x_qf2, phi_q2, use_rate ? dim : 0, | ||
| use_rate ? &actual_vf2[0] : nullptr, use_rate ? vel_q2 : nullptr ); | ||
| if ( !mapped_face1 || !mapped_face2 ) { |
There was a problem hiding this comment.
Same comment. Add debug statement here? Is this bad if we get in this if-block?
|
|
||
| has_contact = true; | ||
|
|
||
| RealT local_pressure = local_gap * penalty_stiff_per_area; |
There was a problem hiding this comment.
Can you clarify whether this routine before computed the pressure? As a sanity check is this line inside ComputeGapRatePressure? Shouldn't this only compute the pressure as done on line 675? Or did the original version of this routine compute it all?
| // contact overlap patch // | ||
| //////////////////////////////////////////////////////////////////////// | ||
| EvalWeakFormIntegral<COMMON_PLANE, SINGLE_POINT>( cntctElem, phi1, phi2 ); | ||
| EvalWeakFormIntegralCommonPlane( cntctElem, pen_enfrc_options.common_plane_rule, |
There was a problem hiding this comment.
I think here we need to clarify which set of face coordinates cntctElem has and how they are used in this routine and whether that conflicts with multipoint integration using current configuration physical coords.
| // Mortar retains the historical triangle rule for now. | ||
| GaussPolyIntTri( elem, integ, 3, TRI_RULE_LEGACY ); | ||
|
|
||
| // call Taylor-Wingate-Bos integation rule. NOTE: this is not |
There was a problem hiding this comment.
Since we are touching integration rules, the TWB integration rule never worked and was experimental at the time. It is not consistent with the underlying finite element basis functions. Let's consider removing.
2 participants
No description provided.