Improvement And Application Of In Situ Adaptive Tabulation

It is very computationally expensive to apply large-scaled detailed chemical kinetics with strong stiffness in combustion simulation.With the rapid development of computational technology and computational mathematics, numerous reduction and acceleration methods have been proposed in the past three decades. In situ adaptive tabulation (ISAT), as one of storage retrieval methods, has been widely used in combustion simulations with detailed chemical kinetics, in which data are stored in a form of a binary tree.The basic theory of ISAT was introduced and three improved algorithms of IS AT focus on the improvement of search procedure were proposed in this study, within the range of error allowed. The improved algorithms were tested in the pairwise mixing stirred reactor (PMSR) and were compared with the original ISAT algorithm. The computational results show that two of the improved methods significantly reduce computational time. The first enhanced method with the index of recent nodes was first proposed. Although this method is effective for query points sequence of which are adjacent to and that in the same region of accuracy (EOA), and computational time can be saved to some extent, it spends more time in the process of searching. The second improved algorithm introduced the maximum half-length of the principal axes of the EOA of a record point in the search procedure, saving20%time compared with the original ISAT algorithm and reducing memory usage. The third improved method is a modification of the growth procedure, in which the EOA of a record point in the binary tree only grows for the current query point, and the sequential queries are still based on previous EOA in order to reduce errors caused by the growth operations. The results indicated that16%computing time can be saved compared with the original ISAT algorithm. For all the points within the EOA, it can ensure that the error caused by the approximation calculation was less than the given error tolerance. Finally, the acceleration effect of the above algorithms were.discussed based on statistical analysis. It indicates that, for the calculation of the speed-up factor, growth number should be considered, while the retrieve number can be neglected.