/OpenSTAP

OpenSTAP is a Finite Element Method (FEM) Solver based on STAP (KJ Bathe, FORTRAN IV), and modified version STAP90 (Prof. Xiong Zhang, Tsinghua University, FORTRAN 90). Basic elements, some advanced elements and methods are added and an Abaqus interface as well. Works with ParaView and Intel MKL.

Primary LanguageFortranMIT LicenseMIT

OpenSTAP

OpenSTAP is a Finite Element Method (FEM) Solver.

It is based on:

  • STAP (KJ Bathe, FORTRAN IV)
  • STAP90 (Prof. Xiong Zhang, Tsinghua University, FORTRAN 90)

It includes:

  • Basic element types:

    • Truss
    • 3T (Triangular)
    • 4Q
    • 8Q
    • 9Q
    • 4T (Tetahedral)
    • 8H (Hexahedral)
    • 20H (B Matrix Unfinished)
    • Beam
    • Timoshenko Beam
    • Plate
    • Shell
    • Plate 8Q
    • Shell 8Q
  • Advanced element types:

    • Infinite 4Q
    • Plastic truss
  • Multiple solver choices

    • Classical Skyline storage and LDLT decomposition solver
    • Intel MKL Pardiso sparse matrix solver
    • LANCZOS eigen value solver (still not completely functional)
    • MKL FEAST eigen value solver (to be tested)
  • And an ABAQUS Python plugin for converting the input file of this program.

It works with:

  • Intel MKL
  • ParaVIEW for visualizing output data (actually it converts a text output file for printing and a vtk file in v3.0 format.)

It runs fast

Minimum RAM requirement: 400MB.

Benchmark of the program on an Intel i7-4790 Processor with 16G of RAM:

  • 4087 nodes with 2884 elements
    • about 3s using classical Skyline
    • about 0.2s using Intel MKL Pardiso
  • 37067 nodes with 30370 elements
    • about 3.5s using Intel MKL Pardiso
  • 259391 nodes with 232720 elements
    • about 35s using Intel MKL Pardiso

Our program also utilized MKL's out-of-core mode which allows external storage to be used when solving large problems. An example of which is running the third benchmark on a Surface Pro 4 with 8GB of RAM. Typically the program would create a 950MB temporary file in the disk to store global stiffness matrix. This would also extend the solution time to about 23 minutes.

Plese refer to /doc for more information.