GitLab

This may have been related to the delaunay optimality condition for triangles near the don't-care condition. There was a negative tolerance allowing sub-optimal elements to appear in the mesh which violated some assumptions and led to mesh divergence issues

Fixes for C++14 compatibility
Adds header files in Matrix project that were missing from previous commits

TODO: Mesh creation fails with non-convex domains

    Torque calculation from body forces is complicated by discontinuities in field gradients across material boundaries

Fix for Mesh::insert_internal_boundaries with small airgaps
    This method is probably not 100% fixed, but it is more stable than before (SalientPoleSynchrel regression tests run with CylindricalAirgap<2>)

Initial implementation of CylindricalAirgap class which automates creation of multiple annuli between two radii

Problems with discontinuous material boundaries have fields which suffer from jump discontinuities
This means the divergence of the Maxwell stress tensor is impulsive at material interfaces
Evaluating net force/torque using body forces requires both volumetric and surface integrals in these cases
This is a nontrivial excercise

In the meantime, the next step will be to add a special calculation method for cylindrical airgaps

Implements SpatialForcing which allows, e.g., spatially varying current densities defined by lambda functions (e.g. sinusoids)
Implements tests for SpatialForcing using sinusoidal functions in test_Square and test_Circle
Begins implementation of tests on two circular annuli (test_Annuli)

Mesh::mesh_data() now returns a MeshData object containing a vector of Triangles instead of nested vectors
Added SquareOrder2 tests of second order triangular elements

Reorganize some code into separate shared libraries

Store FiniteElementMesh DiscreteRegions in std::map with Contour as key

Changes the way number of elements and unknowns are calculated in Physics
    This still needs some work
Adds FunctionArgument class for passing arguments to Forcing objects with multiple parameters
Adds test demonstrating parallelism using OpenMP

FiniteElementMesh remains templated to facilitate storage of variable order elements

This eliminates all Boost library dependencies

    Part of this involves using more shared pointers and vectors (as opposed to arrays) as certain member variables
    Some of that has been done, some work may (probably) still needs to be done
    It's been a few months since this has been looked at so it will require some review

This is a bit of a mess right now

Work on rotational motion and related boundary refinement
Output for vector plots
Starts on nonlinear material properties
Begin encapsulating solver routines

Mesh::smooth() is based on satisfaction of a Laplacian condition

MappedBoundaryPair guarantees correct behavior of PeriodicBoundaryConditions when the UniformDiscretization option is set to false

Implement more rotational motion tests

Implement more rotational motion tests

Changes a DiscreteBoundary<2> to a DiscontinuousBoundary<2> This will be used for implementing relative motion in a user friendly way

This is a new route for implementation of relative motion. Renames Boundary to DiscreteBoundary (to differentiate from Sketch::ContinuousBoundary)

Fix minor inconsistency with antiperiodic boundary conditions

Begin implementation of PeriodicBoundaryCondition construction from PeriodicBoundaryPairs

Still need to work on special boundary types (e.g. sliding interfacing and periodic boundaries)

Expanded on magnetostatic testing with a uniform current density on a circular domain

Nodes, Triangles, Regions are complete (except material properties). Still need to work on Boundaries.