Research On Accelerated Interior Noise Quality Based On GA-BP Neural Network

The vehicle interior noise quality directly affects people's physical and mental health and the product preference,which make it become one of the most important research related to vehicle performance optimization and control.This paper focused on a series of basic theoretical research including the acceleration noise of objective and subjective,their relevance and sound quality optimization and control.The content of this research is outlined as follow:Taken 18 domestic passenger cars as the research object,according to vehicle noise signal acquisition experiment of GB/T 18697-2002 design of full throttle acceleration,the paper reprocessed noise signal to obtain 30 vehicle interior noise samples of acceleration using Landes multi-channel data collection front-end.Using time-varying algorithm of objective parameters,we explored 5 objective parameters(linearity,loudness,sharpness,roughness,articulation index).Furthermore,the subjective evaluation test of 30 noise samples was carried out by grading evaluation method,and subjective evaluation value of each noise sample was acquired after eliminating unstable evaluation results.Multiple linear regression analysis was utilized to establish a mathematical model describing subjective satisfaction with the objective parameter.The result showed that sound quality of vehicle acceleration is primarily affected by the degree of linearity,loudness,roughness.Meanwhile,nonlinear prediction model of interior noise condition was built and validated by taking the objective parameters as input,subjective satisfaction as output using GA-BP neural network theory.Based on GA-BP neural network model and multiple linear regression model,sound quality of 10 noise samples were predicted and analyzed.The result indicated that precision and performance of GA-BP neural network prediction model was better than multiple linear regression prediction model,demonstrating the superiority of the GA-BP neural network model.The research of acceleration noise optimization was done based on GA-BP neural network nonlinear model.Sound quality optimization research of a car was carried out because of obvious roar produced from vehicle acceleration.We reduced the lug dynamic stiffness optimization scheme of exhaust system,recorded interior noise before and after the experiment and analyzed sound quality objective parameters,which showed the improvement of each objective parameter value.Applied objective analysis results to the mathematical model of GA-BP,vehicle interior sound quality was improved by nearly 20%,which further confirmed the correctness of the established nonlinear prediction model of accelerated sound quality.