Selection of Optimal Closure Relationships for Multiphase Flow using a Genetic Algorithm

Abstract

A genetic algorithm is used to help determine the best set of closure relationships to model multiphase flow behavior in pipes. This modeling is typically done through a mechanistic approach that uses conservation equations of mass and momentum. In addition, a set of closure relationships are required to complete the system of equations which are developed based on experimental data. Owing to a large number of possible closures, the size of the search space suffers from a combinatorial explosion problem and subject matter experts are often used to select the best solution. This paper presents a genetic algorithm, implemented in R, that is used to automate the process. Preliminary results show that it has the ability to select combinations as good as or better than those of a human expert in a reasonable amount of time.