Multiblock implementation strategy for a 3-D pressure-based flow and heat transfer solver

This article reports on a multiblock implementation of a general three-dimensional single-block computational fluid dynamics code, which is developed in a nonorthogonal, structured, collocated finite-volume grid system, and incorporates a range of turbulence models. The multiblock implementation is essentially block-unstructured, each block having its own local coordinate system unrelated to those of its neighbors. Any of the blocks may interface with more than one neighbor along any block face. Interblock communication is handled by inner-boundary connection information (receive and send point index arrays) and effected via two-layer dummy cells along interblock boundaries. This communication procedure is easy to extend to parallel computation. The implementation of the algorithm, which takes the advantage of Fortran 90, employs a method to keep most of the single-block code unchanged. Two cases are presented to validate the implementation, and another case with a block number ranging from 1 to 160 blocks is presented for test of the influence of the multiblocking on the convergence rate and execution time.