Intelligent parameter optimization of Savonius rotor using Artificial Neural Network and Genetic Algorithm

Abstract

Power coefficient, the most significant criterion for evaluating the performance of Savonius rotor is a multi-dimensional function of numerous parameters like overlap ratio, number of stages, blade rotation, etc. All these parameters have been examined separately and an approximate span in which optimum performance can be attained is proposed for each one. Furthermore, neither any attempt on scrutinizing this range accurately nor any investigations on probing the probability of existence of any interacting relation among these parameters have been reported so far. Using computational intelligence, an accurate study toward this span and a probable relation among these parameters has been conducted. Power coefficient is considered as a function of six independent input parameters, according to experimental data extracted from a related paper. An Artificial Neural Network has been assigned to investigate a logical interaction among dependent and independent variables and define a cost function based on same empirical data. This function is then optimized by Genetic Algorithm and best amount for each parameter has been determined. Suggested geometry and flow field conditions have then been simulated by Computational Fluid Dynamics and acceptable agreement is detected.