The MAST source code comes with a large number of demonstration problems for conduction, structural, thermoelastic, fluid analysis, fluid-structure interaction analysis and optimization problems. A detailed listing of these example problems is provided here.


Running examples

Following compilation of the example_driver executable these examples are run from the command line by specifying a run_case argument to the executable. Running this executable without any options lists all the available examples.

Simple execution

A typical execution is:

 $ mpirun -np <N> example_driver run_case=plate_bending -ksp_type preonly -pc_type lu 

Here, <N> is replaced with the number of parallel MPI processes to be launched for this run case.

MAST uses the command-line options to initialize itself and libMesh, which further passes the options to SLEPc and PETSc. Therefore, configuration of eigensolver, nonlinear and linear solvers can be completely chosen through command-line options. In the example above, a direct LU-decomposition solver is specified for all linear solves. It is noted that the direct solver implementation in PETSc only supports a single-processor run. However, with interfaces to packages like MUMPS and SuperLU_dist, PETSc is able to provide support for distributed-memory direct solvers. If PETSc is configured with MUMPS, then it will automatically use this solver for N>1 in the above example. Additional solver options can be found in the PETSc user manual

Complex solver setup

For multiphysics solutions, solver configuration can be specified for each discipline by using the option --solver_system_names which instructs MAST and libMesh to prepend the name of libMesh::System to each solver object. Then, PETSc and SLEPc associate command-line options with a solver only if the solver name is included in the option. For example, the following execution command specifies different configurations for the structural (structural_) and fluid (fluid_complex_) solvers:

$ mpirun -np 20 example_driver --solver_system_names
run_case=beam_fsi_flutter_analysis
-structural_st_pc_factor_mat_solver_package mumps
-structural_st_ksp_type preonly 
-fluid_complex_ksp_type gmres -fluid_complex_pc_type asm
-fluid_complex_pc_asm_overlap 1 
-fluid_complex_ksp_sub_pc_type ilu
-fluid_complex_sub_pc_factor_levels 2 
-fluid_complex_ksp_gmres_restart 100 -fluid_complex_ksp_atol
1.e-4 -fluid_complex_ksp_rtol 1.e-8 

This will launch the two-dimensional panel flutter example on 20 CPUs with a direct LU-decomposition solver for solution of the structural eigenproblem, and a GMRES solver with a level-1 overlap additive Schwarz method (ASM) preconditioner for solution of the complex fluid system-of-equations while using an ILU(2) preconditioner for local solves. Absolute and relative convergence tolerances are also specified to be \(10^{-4}\) and \(10^{-8}\), respectively and a GMRES restart after 100 iterations is requested.