[GeoMechanicsApplication] Add a utility function for calculating the internal force vector #12669
Conversation
If not, an exception is thrown.
If they are, an exception is thrown.
| GeoEquationOfMotionUtilities::CalculateInternalForceVector(b_matrices, stress_vectors, integration_coefficients), | ||
| "Cannot calculate the internal force vector: input vectors have different sizes") | ||
| } | ||
|
|
There was a problem hiding this comment.
All Matrices in the b_matrices should have the same sizes. This also hold for the vectors in stress vectors.
Further the stress_vector size should be equal to the first size of a b matrix
There was a problem hiding this comment.
Good catch! I have added checks for these preconditions that are only carried out by debug builds.
…DEBUG_ERROR_IF` The tests that cover these checks are now conditionally built and run, too.
| } | ||
|
|
||
| // The following tests only raise errors when using debug builds | ||
| #ifdef KRATOS_DEBUG |
There was a problem hiding this comment.
Do we think these checks would ever be helpful for cases which are run in release? I like that this way it doesn't impact release performance, but I'm just putting the question here.
There was a problem hiding this comment.
Since the input for GeoEquationOfMotionUtilities::CalculateInternalForceVector is prepared by us (the programmers) and any inconsistencies in the input data are expected to affect previous operations, @WPK4FEM and myself decided that a debug check is adequate here. It's more like an assertion: this is never supposed to happen, but we do want to catch it at debug-time if it ever happens. As you wrote yourself, the end-user (who uses a release build) doesn't pay the price of the additional checks, which benefits general performance.
| const auto integration_coefficients = std::vector<double>{0.25, 0.4}; | ||
|
|
||
| KRATOS_EXPECT_EXCEPTION_IS_THROWN( | ||
| GeoEquationOfMotionUtilities::CalculateInternalForceVector(b_matrices, stress_vectors, integration_coefficients), "Cannot calculate the internal force vector: matrix-vector product cannot be calculated due to size mismatch") |
There was a problem hiding this comment.
This is a very long line, I expect it needs to still be formatted?
There was a problem hiding this comment.
Hmm... no. For some unclear reason, clang-format doesn't break this long line. When I manually insert a line break between the function call and the expected error text, clang-format will revert it to the original long line...
|
|
||
| auto result = Vector{ZeroVector{rBs.front().size2()}}; | ||
| for (auto i = std::size_t{0}; i < rBs.size(); ++i) { | ||
| result += prod(trans(rBs[i]), rStressVectors[i]) * rIntegrationCoefficients[i]; |
There was a problem hiding this comment.
i think that with noalias(result) += should be faster
📝 Description
Added a utility function for calculating the internal force vector. As input, it requires vectors of B-matrices, stress vectors, and integration coefficients (one for each integration point). When the input vectors have different sizes or when they are empty, an error is raised.
The internal force vector is computed as follows: