Optimal Closed-Loop Control of Step Flow Separation Using Genetic Programming

علوم کاربردی و محاسباتی در مکانیک

Articles in Press, Accepted Manuscript, Available Online from 28 December 2024   

Document Type: Original Article

DOI: 10.22067/jacsm.2024.88143.1261

Authors

siroos kasmaiee^{* 1}; Mehran Tadjfar²; saman kasmaiee³

¹Aerospace department. AmirKabir university. Tehran.Iran

²Aerospace engineering department, AmirKabir university. Tehran. Iran

³Department of Aerospace Engineering, Amirkabir University of Technology, Tehran, Iran

Abstract

In this research, novel model-free method based on machine learning for closed-loop control of flow separation behind a step is introduced. The primary goal of this study is to reduce the recirculation zone behind the step at a Reynolds number of 1350, which is achieved by a jet slot. In this study, the flow was simulated as steady and two-dimensional based on finite volume discretization. Feedback control rules have been optimized based on a cost function that includes the area of the recirculation flow and the costs associated with the injection. This optimization process was carried out using genetic programming algorithms. A tree-based genetic programming was used to construct various injection model functions and create a closed-loop control system. After evolving through 8 generations with 500 samples in each generation, the algorithm arrives at a feedback rule capable of reducing the recirculation flow area by up to 60%. This machine learning-based control system was compared with the best open-loop jet based on Kelvin-Helmholtz frequency pulsation.

Keywords

Flow control; Genetic programming; Numerical simulation; Machine learning; Computational fluid mechanics

References