Performance of different optimization concepts for reactive flow systems based on combined CFD and response surface methods

Abstract

CFD-based optimization provides an efficient tool to improve systems in which conversion processes depend on a strong interaction between flow field and chemical reactions. The present work develops an improved multi-parameter and multi-objective optimization concept for reactive flow systems and demonstrates this concept for a new quench reactor. Response surface methods (metamodels) are used within the metamodel-based optimization process to accelerate the optimization rapidly. Different metamodels such as polynomials, K-nearest, radial basis functions, anisotropic Kriging, smoothing spline analysis of variance, and artificial neural networks are applied and carefully analyzed. It can be shown that radial basis function metamodels in combination with a certain number of CFD calculations provide a robust and efficient optimization scheme. The computational effort can be reduced by a factor of 17, which provides a reliable basis to optimize more complex reactive flow systems, e.g. the high pressure partial oxidation of natural gas or crude oil.