Study On Subjective And Objective Evaluation Of Vehicle Acoustic Quality Of Pure Electric Vehicles

The medium and high frequency noise generated by the driving motor of pure electric vehicles is particularly prominent.Although the sound pressure level in the vehicle is low,the medium and high frequency noise causes the deterioration of the internal sound quality,which affects the subjective feelings of drivers and passengers.The sound pressure level alone is not enough to reflect the internal sound quality of pure electric vehicles.Therefore,it is very important to establish the interior sound quality prediction model of pure electric vehicles,aiming to accurately and scientifically reflect the real situation of the inner vehicle sound quality of pure electric vehicles.This paper combines the semantic segmentation method and grade scoring method as a subjective evaluation method,according to the characteristics of pure electric vehicle noise,select modified noise samples,the correlation test of subjective evaluation test results,eliminate the low correlation of the subjective evaluation results,to ensure the accuracy and scientificity of the subjective evaluation test results.Through the test software calculation the selected psychoacoustic objective parameters: the value of loudness,sharpness,roughness and semantic definition,with the psychoacoustic objective parameters as the input and the subjective evaluation value of pure electric vehicle as the output,the BP neural network pure electric vehicle acoustic quality prediction model is built.At the same time for the shortcomings of the model to improve the Grey Wolf Algorithm(IGWO)to optimize the model,establish IGWO-BP neural network pure electric vehicle acoustic quality prediction model,the model adopts nonlinear convergence factor,balance optimization algorithm of local and global search ability,using dynamic weight strategy,constantly update the weight coefficient in iterative learning optimization algorithm,improve the convergence speed of the prediction model and avoid the local optimal solution.The results show that the optimized IGWO-BP neural network internal vehicle acoustic quality prediction model improves the mean square error by 2.663%,7.570%,4.303%,7.303% and the average absolute percentage error by 0.671%.The convergence speed is also significantly improved,which shows that` the model is more suitable for the internal vehicle acoustic quality prediction of pure electric vehicles.