Analysis of microchannel resistance factor based on automated simulation framework and BP neural network

In this paper, self-design automated simulation and artificial neural network (ANN) model were developed to analyze and estimate the resistance factor in rectangle cross-section microchannels. The main purpose is to obtain a universal solution method through numerical simulation which can solve the resistance factor problem for invariant cross-section microchannels. Through Python language, the automatic coalescent of preprocessing Gambit, computing software CFD and post-processing Tecplot make the simulation framework realize the automatic acquisition of microchannel resistance factor samples. Then, 100 simulation samples with different aspect ratios for Reynolds numbers ranging from 50 to 500 were obtained. After validation, the width and height of microchannels were applied as input data set of the ANN model, and the resistance factor was determined as the target data. In order to improve BP algorithm for training ANN, a new swarm evolution algorithm was realized by combining the strong point of gradient descent method, genetic algorithm and particle swarm optimization, which is called particle swarm evolution algorithm. Finally, the result of resistance factor model was established and verified by several existing measurement value of pressure drop from remarkable experimental.