Investigation of Combustion Mechanism Reduction Via Intrinsic Low Dimensional Manifold Method and Neural Networks

摘要

This paper discusses the computation of the Intrinsic Low Dimensional Manifolds (ILDM's) for kinetic chemical mechanism reduction of a H{sub}2-O{sub}2, system and representing the resulting table using Neural Networks (NN). The ILDM method, results in a lookup-table representing the low dimensional manifold. A new approach of representing the lookup table using Neural Networks is discussed in this paper. Multilayered Perceptrons (MLP), NN architecture is trained to model the low dimensional manifold. It is shown that this approach can be computationally efficient for large scale problems.