基于MESIHRD-CNN的驱动桥冲击异响主客观评价方法研究OA

Aiming at the problem of impact-induced noise in a specific type of drive axle,the method for predicting such noise is studied.At first,the noise signal from drive axle was collected,and the noise characteristics were thoroughly analyzed to clarify its underlying generation mechanism.Subsequently,a subjective evaluation test was conducted on 86 groups of noise samples using the rank scoring method,which laid a foundation for the subjective evaluation of the follow-up study.To tackle the problem of con-volutional interference between gear mesh impacts and natural frequency convolution of axle housing,a no-vel Minimum Envelope Spectrum Interference-to-Harmonic Ratio Deconvolution(MESIHRD)is pro-posed.This method eliminates the effects resulting from the convolution of gear mesh impacts and the nat-ural frequency convolution of axle housing.Following this,multi-dimensional objective parameters are calculated and subjected to correlation analysis,which verified the effectiveness of the proposed deconvolu-tion method in identifying gear meshing impacts.Finally,a subjective and objective evaluation model is constructed by integrating MESIHRD with a convolutional neural network(CNN).Comparisons were made at the feature level between deconvolved features and original feature parameters,and at the model level against Backpropagation Neural Network(BPNN),Support Vector Regression(SVR),and Long Short-Term Memory(LSTM)networks.The results demonstrate that the combination of characteristic parameters processed by the proposed deconvolution method and the CNN model yields the best perform-ance in predicting impact-induced noise in the drive axle,showing significant potential for engineering ap-plications.