Duct design of cross-flow fan based on parameter sensitivity and neural network

The flow mechanism of cross-flow fan is analysed qualitatively, and the key structural parameters affecting its air volume are obtained. Using the parameter sensitivity analysis method, the influence degree and positive and negative correlation of each parameter on air volume are obtained. Taking the key structural parameters as the input parameters of radial basis function (RBF) network and the air volume as the output parameter, the RBF neural network is constructed by distributing the parameter weights according to the degree of influence, and the air volume of cross-flow fan is predicted and evaluated. The results show that the key structural parameters affecting the air volume of the cross-flow fan include the inclination of the volute throat, the gap between the volute throats, the gap between the volute tongues, the suction angle, the length of the volute tongue, the diffuser angle, and the width of the air inlet. The influence degree from strong to weak is the width of the air inlet, the length of the volute tongue, the diffuser angle, the inclination of the volute throat, the gap between the volute tongues, the suction angle, the gap between the volute throats, among which the gap between the volute throats, the suction angle, the diffuser angle and the width of the air inlet are positively correlated with air volume, and the inclination of the volute throat, the gap between the volute tongues and the length of the volute tongue are negatively correlated with air volume. The maximum relative error between the predicted value and the actual value of RBF neural network is 4.56%, and the average relative error is 2.2%. Using this neural network, the air volume of cross-flow fan can be quickly evaluated.