ASPECT

This is the list of changes made after the release of Aspect version 1.4.0 for version 1.5.0. All entries are signed with the names of the author.
New: Blankenbach benchmarks were added.
(Timo Heister, 2017/02/16)
New: The particle property interpolation interface can now be queried for only a part (or only one) of the particle properties. This makes interpolation more efficient, because the compositional fields are solved onebyone and therefore oftentimes only one of the properties is actually needed. Interpolating all needed properties in one call is still more efficient in the general case.
(Rene Gassmoeller, 2017/02/15)
Removed: The material model interface contained the functions 'reference_density', 'reference_thermal_expansion_coefficient', 'viscosity_ratio', and 'thermodynamic_phase'. Additionally many material models implemented functions like 'reference_cp', and 'reference_thermal_diffusivity'. These functions were not used in any place and could be inconsistent with the model properties, because they had no input to figure out the real model conditions. Therefore they were removed. Additionally, the postprocessors visualizing the 'viscosity_ratio' and 'thermodynamic_phase' functions were removed, because there is no material model in the current version that actually implements these functions, and they are not tested. The new recommended way to determine a reference material property for a given model is to evaluate the material model at reference conditions (given by the adiabatic profile) as implemented in the 'basic statistics' postprocessor. Existing user plugins that implement these funtions will continue to work, as long as the functions are only used within that plugin. User plugins that called these interface functions from outside can either cast the material model reference to a particular material model that implements these functions, or use the recommended approach to determine reference properties described above.
(Rene Gassmoeller, 2017/02/09)
Changed: The 'basic statistics' postprocessor now uses a better defined reference state to compute properties like the Rayleigh number. In particular, it computes material properties for the prescribed adiabatic temperature and pressure conditions at the surface, instead of using the deprecated reference_property() functions of the material model interface. This way it can also be used for other material models than the 'simple' model, and also highlights possible inconsistencies between the reference profile and reference temperatures assumed in the matherial model. During the rework the multiplicative output while doing initial adaptive refinement steps was also removed.
(Rene Gassmoeller, 2017/02/08)
Removed: The 'steinberger' material model contained an option to use it for incompressible models. This option included a somewhat complicated density scaling, and relied on the availability of a compressible adiabatic profile in incompressible models. This profile is not longer available with the recent changes to the adiabatic reference profile, and the feature is too obscure to justify a fix. Therefore, it was removed and the 'steinberger' material model now only works for compressible formulations.
(Rene Gassmoeller, 2017/02/07)
New: A cookbook for continental extension was added, which uses uses the visco plastic material to simulate the thermalmechanical evolution of a continental plate extending at a constant velocity.
(John Naliboff, 2017/01/20)
Changed: The finite strain cookbook used a way to compute the finite strain that was only appropriate for a small amount of strain. This was fixed to a proper finite strain computation. Added tests and benchmarks for the new formulation, and updated the according particle property in the same way. Also updated and extended the cookbook description in the manual.
(Rene Gassmoeller, 2017/01/19)
New: Added a "heat flux densities" postprocessor.
(Timo Heister, 2016/12/26)
New: ASPECT now allows querying its own version number as well as the version of all of the underlying libraries using the –version
or v
command line flags.
Similarly, using –help
or h
allows querying command line usage of ASPECT.
(Wolfgang Bangerth, 2016/12/1931)
New: A material model for incompressible (using the Boussinesq approximation) and compressible computations (with ALA or TALA) for a nondimensionalized problem. This can be used for several benchmark problems like Blankenbach, King, etc..
(Timo Heister, Juliane Dannberg, Rene Gassmoeller, 2016/12/18)
New: ASPECT now supports the choice between different formulations for the governing equations including boussinesq and anelastic liquid approximation. For this, the adiabatic conditions have been extended to provide values and gradients of the reference density. Several benchmarks for these formulations have been added.
(Juliane Dannberg, Rene Gassmoeller, Timo Heister, 2016/12/14)
New: ASPECT now also generates a output/particles.visit
file that allows Visit to read in all files from all processors and all time steps that contain particle data.
(Wolfgang Bangerth, 2016/12/02)
Changed: The 'Stokes only' nonlinear solver scheme now uses the nonlinear solver tolerance parameter (previously it was hard coded to 1e8), and it computes the nonlinear residual as current residual divided by initial residual, consistent with the 'iterated Stokes' solver scheme.
(Juliane Dannberg, Rene Gassmoeller, 2016/11/22)
Changed: The adiabatic profile now contains a reference density profile, and the derivative of this reference density profile. The InitialProfile adiabatic profile now relies on the adiabatic heating to decide if the temperature increases with depth. Additionally, an offbyone bug was fixed in InitialProfile leading to minor changes in the adiabatic profiles (relative change of ~1e4), changing models that rely on the profile. Models that rely on the adiabatic profile might be changed by this PR if they are compressible, but do not use adiabatic heating, or vice versa.
(Timo Heister, Juliane Dannberg, Rene Gassmoeller, 2016/11/20)
New: The visco plastic material model now includes an option for strainweakening of cohesion and the internal angle of friction. Strainweakeing of these properties is commonly used to help localize deformation in regions undergoing plastic failure. Note that using this option requires tracking the finite strain tensor through particles or compositional fields.
(John Naliboff, 2016/11/10)
New: Two particle generators were added. One, generates particles at the quadrature points for each active cell in the triangulation. Two, generates a uniform distribution of particles in the unit cell and transforms each of the particles back to real region in the model domain for each active cell in the triangulation.
(Harsha Lokavarapu, 2016/11/10)
Changed: The exchange of ghost particles that was introduced lately can be quite expensive for models with many particles, and is often unnecessary if the particles are used as passive tracers. Therefore, a new input parameter 'Update ghost particles' controls this exchange, and its default is set to 'false'. Model parameter files using active particles will need to be changed accordingly.
(Rene Gassmoeller, 2016/10/18)
Improved: The matrix assembly of Stokes and Advection systems has been optimized, by assembling less (only the relevant) DoFs, and by optimizing calls to deal.II functions. The overall speedup for box models is between 20 and 40% of the assembly time, likely somewhat less for curved geometries. This change will require changes in user written assembler plugins, because the Stokes system assembly now only loops over Stokes degrees of freedom.
(Rene Gassmoeller, 2016/10/17)
Improved: Box models without deformed mesh now use a MappingCartesian, which assumes all mesh cells are aligned with cartesian coordinate axes. Matrix assembly and particle transport in such mappings is around 20 % faster compared to a general MappingQ1 for other box models.
(Rene Gassmoeller, 2016/10/14)
Changed: HDF5 particle output files are now named 'particles...' instead of 'particle...' to be consistent with the vtu output. Also particle properties with more than one component are now correctly split into scalar fields in the output files, if they have more or less components than the number of spatial dimensions in the model.
(Rene Gassmoeller, 2016/09/20)
New: Multiple particle properties can be intialized by specifying multiple particle property function components as opposed to one particle property.
(Harsha Lokavarapu, Gerry Puckett, 2016/09/20)
Changed: The timestep entry in the statistics file has been moved to column 3 and is now the timestep used for the timestep corresponding to the current row.
(Jonathan Robey, 2016/09/16)
Changed: The 'cell average' particle interpolator is now more tolerant against cells without particles by interpolating properties from neighboring cells. This is necessary, because during refinement even children of cells with a reasonable number of particles can be void of particles.
(Rene Gassmoeller, Jonathan PerryHouts, 2016/08/31)
Changed: Particle properties should now declare which solution properties they need to update themselves. The particle world then only computes values and gradients of the solution at the particle positions if necessary, which can reduce the computational cost of the particle update for simple particle properties.
(Rene Gassmoeller, 2016/08/30)
New: .visit output files now also contain information about the model time, as long as ASPECT was build with at least deal.II 8.5.0.pre. Previously, this information was only available in the Paraview .pvd files.
(Rene Gassmoeller, Juliane Dannberg, 2016/08/24)
New: There is now an initial topography plugin that returns initial topography values based on an ascii data file.
(Anne Glerum, 2016/08/22)
Fixed: The point value postprocessor forgot to take into account the mapping we use when describing curved boundaries.
(Rene Gassmoeller, Wolfgang Bangerth, 2016/08/16)
Changed: Particles now also store their location in the coordinate system of their current cell. This decreases the number of times this location has to be computed by inverting the mapping for the current cell, which is expensive. On average this change will save 4050% of the overall particle computing time, while increasing the particle memory footprint (which is usually small compared to the system matrix).
(Rene Gassmoeller, 2016/08/12)
Fixed: The chunk geometry pull back function now returns a corrected longitude value when 180 hemisphere is crossed.
(Anne Glerum, 2016/08/09)
Changed: It is now possible to read in ascii data files of which the coordinates are not equally spaced.
(Anne Glerum, 2016/08/05)
New: It is now possible to create compositional fields that are not advected by a field method, but interpolated from particle properties. This simplifies the process of using the particles as 'active' particles that carry information that influences the solution. E.g. the material model can access the compositional field that is interpolated from particle properties and use this as 'composition' information.
(Rene Gassmoeller, 2016/08/02)
New: There is now an initial topography plugin which reads from the prm file polygon definitions and set the initial topography to be constant within those polygons.
(Menno Fraters, 2016/07/26)
Changed: Particle initialization no longer routinely computes the solution at the particle positions, since it is usually not needed and complicates the initialization process. Instead it evaluates the initial conditions at the particle positions. It is still possible to access the solution by evaluating it manually inside of particle property plugins. Additionally the 'initial composition' property now utilizes the userprovided names of the compositional fields to identify its particle properties (they are now named 'initial [field_name]', where '[field_name]' is replaced by the user provided name).
(Rene Gassmoeller, 2016/07/18)
New: Added parameter “Adapt by fraction of cells” to switch between refining a certain number of cells based on the fraction of total error (default behaviour) and the fraction of total number of cells
(Lev Karatun, 2016/07/20)
Changed: It is now possible to set the gravity to a negative value in order to calculate backward advection.
(Jacky Austermann, 2016/07/13)
New: There is a new postprocessor that outputs statistics to the screen about the memory usage and nonzero entries of matrices. It can be called with the name 'matrix statistics'.
(Sam Cox, 2016/07/01)
New: There is now a plugin structure to add initial topography to geometry models.
(Menno Fraters and Anne Glerum, 2016/07/01)
New: There is a new postprocessor that outputs statistics about the memory usage at each timestep. It can be called with the name 'memory statistics'. This replaces the helper function 'output_program_stats()', which has been removed, along with the variable 'output_parallel_statistics'.
(Sam Cox, 2016/06/30)
Changed: In input files, lines that end in a backslash now require a space before the backslash if the two lines should not be conjoined immediately, because leading spaces on the continuing line is ignored. See the Section "The structure of parameter files" in the manual.
(Jonathan Robey, 2016/06/30)
Changed: The default for "Initial adaptive refinement" cycles is now 0.
(Timo Heister, 2016/06/28)
New: There is a new parameter gamma in the entropy viscosity stabilization that allows to scale the stabilization with the strain rate (in addition to the velocity).
(Juliane Dannberg, 2016/06/28)
New: There is now a postprocessor that outputs the heatflux density at each boundary face into a text file and a postprocessor that outputs the heatflux density at each boundary face for visualization.
(Jacky Austermann, 2016/06/28)
New: There is a new function Utilities::real_spherical_harmonic which calculates the values of a fully normalized real spherical harmonic.
(Ian Rose, 2016/06/28)
Changed: The files handling the free surface implementation have been renamed to free_surface.h and free_surface.cc for better consistency with the rest of the file names.
(Ian Rose, 2016/06/27)
Changed: The previouslydeprecated functions 'composition' and 'temperature' have now been removed. Their replacements are 'boundary_composition' and 'boundary_temperature'.
(Sam Cox, 2016/06/27)
Changed: The free surface implementation now represents mesh deformation using a vector of displacements for the mesh vertices, using a MappingQ1Eulerian to transform from the reference cell to the physical cell.
(Ian Rose, 2016/06/26)
Changed: The visualization postprocessor now writes every initial refinement stage if 'Run postprocessors on initial refinement' is set to true.
(Rene Gassmoeller, 2016/06/25)
New: There is now a postprocessor that outputs statistics about the number of particles per cell in the model domain. It outputs global maximum, minimum and average number of particles per cell.
(Rene Gassmoeller, 2016/06/24)
New: There is now the option to model melt transport (twophase flow). This core of the implementations includes additional variables in the solution vector, a new assembler with an additional equation that will be solved and a modified advection equation for the porosity field, a new preconditioner for models with melt transport, and additional melt outputs for the material model.
(Juliane Dannberg, Timo Heister, 2016/06/24)
New: Particles can now carry the integrated strain they have experienced over the course of the model. They store all components of the symmetric strain tensor, which can be converted into the invariants or investigated individually.
(Rene Gassmoeller, 2016/06/10)
New: There is a new optional feature for the discontinuous temperature and compositional solutions. After solving the advection equation, a "bound preserving limiter" working as a correction procedure is applied to the discontinuous advection fields. The limiter will stabilize the discontinuous advection solutions and keep it in the range of user defined global maximum/minimum values. Whether or not the limiter is used is determined by an entry to the parameter file.
(Ying He, 2016/06/02)
New: Tests can now be marked that they are expected to fail by the EXPECT FAILURE keyword in the .prm.
(Timo Heister, 2016/06/01)
New: There is a new boundary traction model "ascii data" that prescribes the boundary traction according to pressure values read from an ascii data file.
(Juliane Dannberg, 2016/05/24)
New: A material model plugin for viscoplastic rheologies, which combines diffusion, dislocation or composite viscous creep with a Drucker Prager yield criterion.
(John Naliboff, 2016/05/19)
Changed: The traction boundary conditions now use a new interface that includes the boundary indicator (analogous to the velocity boundary conditions).
(Juliane Dannberg, 2016/05/19)
New: There is now a visualization plugin to visualize the maximum horizontal component of the compressive stress.
(Wolfgang Bangerth, D. Sarah Stamps, 2016/05/12)
New: There is a new visualization postprocessor "artificial viscosity
composition" to visualize the artificial viscosity of a compositional field.
(Juliane Dannberg, Timo Heister, 2016/05/03)
New: Mesh refinement strategies based on artificial viscosity, composition gradients, or composition threshold.
(Juliane Dannberg, Timo Heister, 2016/05/03)