Study of Graphene Loudspeaker Based on GA-BP Neural Network Model
In order to solve the problem that the traditional physical model of graphene sound generators requires a large amount of calcu-lation and has low accuracy,a BP neural network model based on GA genetic algorithm is proposed,which has higher accuracy and adapta-bility.Firstly,the principle and the experimental design of graphene thermoacoustic are introduced and GA-BP neural network model is es-tablished.Secondly,the parameters of the model are adjusted,and three different neural network models based on dropout,regularization,both regularization and GA genetic algorithms are compared.Thirdly,sinusoidal excitation amplitude,frequency and measurement distance of the graphene loudspeaker are inputted and GA genetic algorithm is used to globally optimize the weight and bias of the hidden layer.Fi-nally the optimization results are brought into the BP neural network to predict the sound pressure level.Under the condition that regulari-zation is used,the prediction accuracy of BP neural network is 98.05%and the mean square error is 0.23.At the same time,the prediction accuracy of GA-BP neural network reaches 98.62%,and the mean square error is only 0.14.After optimization,the accuracy is increased by 0.57%and the mean square error is reduced by 41.36%,showing better accuracy and adaptability.This work provides a high precision and high adaptability scheme for predicting the nonlinear output of multi-class feature sensors.
万方数据
维普
