The class of 3DGRAPE codes is designed to generate volume grids using iterative methods to solve the non-linear, nonhomogeneous, elliptic partial differential heat equation with heat sources. The solution to the heat equation develops three-dimensional discrete points within a domain outside or inside a vehicle where fluid flow is simulated with Computational Fluid Dynamics (CFD). The latest version in this class of software is the 3DGRAPE/AL:V3, version 3 of the Three-Dimensional Grids about Anything by Poisson’s Equation with Upgrades from NASA Ames and Langley computer programs. The previous version, 3DGRAPE/AL:V2 [“Further Improvement in 3DGRAPE,” (LAR-16415-1) NASA Tech Briefs, Vol. 28, No. 9 (September 2004), p. 50], was advanced through the development of a new block-to-block boundary condition that guarantees C-I continuity between adjacent blocks sharing a common block face. Until this condition was developed, matching block faces were only created by projecting a straight line from one cell into each block from the matching face, thereby changing the initial grid that defines a structured multiple block system.
The new boundary condition performs a source term reconstruction based on unit vectors evaluated from one cell on either side of the interface, thereby eliminating the use of a quadratic equation solution to find the source terms that impose a C-I condition. This technique is an extension of a two-dimensional method used in the construction of block faces, originally developed with the Volume Grid Manipulator. In addition to a new boundary condition implementation, version 3 has also been coded to enable the use of parallel computing. With increased availability of multiple-core computers, this grid generation code can harness the power of the Message Passing Interface (MPI) to distribute the work of solving multiple block systems over a few CPU cores to hundreds of cores, simultaneously. The combined advances in the 3DGRAPE class of codes enables the creation of higher quality volume block-structured grids more efficiently for CFD applications, and remains the most robust elliptic partial differential equation grid generator available.
This work was done by Stephen J. Alter of NASA Langley Research Center, Victor R. Lessard of Genex System, Arthur S. Lazanoff of Computer Sciences Corporation via NASA Ames, and Elliot E. Schulman of Advanced Management Technology, Inc. LAR-17758-1