A novel usage of neural network in optimization and implementation to the internal flow systems

Abstract

PurposeTo propose a robust and more effective algorithm for aerodynamic design optimization problem by using neural network.Design/methodology/approachNeural network and genetic algorithm (GA) are hybridized in a new way, and quasi one‐dimensional Euler equations are solved for internal flow in the nozzle.FindingsThe results indicate that the nozzle design can be performed successfully and quickly by using the implemented algorithm. It is observed that using the method decreased CFD solver calls about 21 and 46 per cent for transonic and supersonic flow, respectively.Originality/valueIt is the first time that the neural network is used for the candidate solution in the GA.