Gain: Improved Multi-block Methodology
In the multi-block gridding method, a computational domain is decomposed into several mesh blocks. The blocks can be nested for achieving higher local resolution, or linked face to face to cover a complex three-dimensional region. Physical quantities are computed in each block by numerically solving transport equations and the information in each block is communicated to the other blocks across inter-block boundaries. Multi-block gridding accomplishes two important and separate goals: (1) the computational effort is reduced for flows in complex geometry, and (2) fine resolution can be economically achieved in localized regions.
Version 8.2 of FLOW-3D has an improved multi-block coupling scheme that makes computations more accurate, more stable, and more efficient. The method is a derivative of the Chimera domain decomposition technique originally proposed by NASA. For any boundary point connecting blocks, an eight point interpolation stencil is defined to provide data for interpolation. If there are no moving obstacles, all necessary interpolation information is calculated once and stored by the preprocessor for use by the solver. Moving obstacles require that some of this information be recomputed every time step. In any case, special considerations have been introduced to account for free surfaces and obstacles that span block boundaries in accordance with the advanced FAVOR™ and TruVOF methods used in FLOW-3D.
The efficiency of the new coupling technique is providing a natural and convenient pathway to distributed memory parallel processing, or "clustered computing." What a great example of how a small incremental change can return huge gains in accuracy, stability and speed.