19)

New Features

Added MUSCL material model for stabilizing advection in porous flow problems for use with CVFEM.
Added relative mass flux porous flow option to stabilize advection for problems with mesh motion, currently only stabilized with CVFEM.
Linear system re-initialization no longer happens if contact mapping is either unchanged or a subset of the prior mapping. This can substantially improve performance for some configurations on the GPU in particular.
CHEMEQ models can now be deactivated using user-defined conditions instead of just temperature and temperature rate of change. This feature uses user-defined functions from the material definition to define a more flexible deactivation condition that the user may choose - for example based on reaction rates or other custom conditions evaluated per block.
The gradient of user fields is now calculated automatically, so the gradient of a user field named myfield could be used elsewhere in Aria as grad_myfield.
Added diagnostic mode for MPMD SPARC/Aria coupling - activating it with --tftklog transfer will output detailed information about where the mesh gap exceeds a threshold gap size based on mesh size. The user can also set a maximum gap size with MAXIMUM TRANSFER CLIPPING DISTANCE = X to enable an error for mismatch over that length.
Added a non-dimensional residual scaling option based on the diagonal inverse of the Jacobian and maximum absolute values of the degrees of freedom.
Disabled specifying both advection (ADV) and diffusion (DIFF) terms for non-species mass balance in order to prevent double counting of the Darcy mass flux. Either ADV or DIFF models can be used but not both.
Add automated teko_multiphysics preset solver based on heuristic methods described in Teko Heuristic Multiphysics Preconditioner.
Add contact search result caching to avoid un-needed linear system re-initialization calls.
Integrate against new Trilinos version with Kokkos 4.5, leading to as much as a 2.8x speedup in MueLu solve times.

Bug Fixes

Fixed bugs with the “per model” postprocessor related to postprocessing fluxes on enclosures with bulk nodes and using per-model postprocessed quantities in other places (e.g. in string functions in other postprocessors).
Fixed bug with “using P0” postprocess where if multiple models were present (e.g. model A, model B), only contributions from the first model provided were used.
Fixed bugs with the “ideal_gas” density model - was incorrectly using SPECIES instead of SPECIES_FRACTION to calculate mixture molecular weight, and had a sensitivity error.
Fixed sensitivity error in the gradient of the foam_time_temp density model.
Fixed issue where block names with “.” and chemeq models were creating global variable names that were not allowed (replaced “.” with “_” in the global variable name).
Fixed a bug with the ray tracer flux model “leaking” through to the wrong side on double-sided shells.
Fixed a bug where transition layer elements were not being properly refined in simulations with dynamic mesh adaptivity.
Resolve bug in Teko setup for solving linear systems with constraints.