Multi-objective Optimization for Axial Flow Fan Based on BP Neural Network and Genetic Algorithm

On account of the high non-linear relationship between the combinational of structural parameters and the performance parameters of axial flow fan, to predict and optimize the performance of axial flow fan is a challenge problem. In this study, the back propagation neural network (BP) and the genetic algorithm (GA) would be applied to optimize the structural parameters combination and make the axial fan with the best performance, based on the non-linear mapping properties of BP and the parallel processing, stochastic, and self-adapting search abilities of GA. Firstly, the 3- dimension model of the axial flow fan is set up, and the samples database could be gained by the Computational Fluid Dynamics (CFD). Then, the non-linear mapping relationship between the structure parameters and the function parameters of axial flow fan is established by BP neutral network, and the results predicted by BP network and the outcomes simulated by CFD are compared to make an error analysis, which could demonstrate the BP network is stable and reliable. The trained network would be applied to GA algorithm to make the global optimization to purse a combination of structure parameters which could make the jet range and efficiency of axial fan with the optimal performance. With the same driving power, the CFD simulation shows that the model based on the optimal combination of structure parameters of axial flow fan can improve the range by 7.2m and the efficiency by 10.24% compared with the original one. Moreover, this optimization scheme provides guidance for the design of axial flow fan's structure parameters in the future.