Research On Rotating Part Sound Quality Of Gasoline Engine Based On Improved Tonality Model And BP Neural Networks

At present,noise pollution has been listed as the four major pollution sources in the world together with water pollution,air pollution and solid waste pollution.With the amount of car ownership keeping growing,the proportion of traffic noise in the urban ambient noise is on the rise,which brings hearing impairment to people and affects people’s physical and mental health more seriously.Therefore,controlling the automotive NVH strictly,reducing noise level and improving the quality of the noise has become the consensus of the automotive industry,but also one of the strategic advantages that the major automakers has been pursuing.Therefore,the study on the acoustic quality of car noise based on the human auditory system has a strong engineering and practical significance for optimizing the design of automotive components.The radiation noise of rotating parts in gasoline engine was taken as the research object and carried out the radiation noise test strictly accordance with the national standards in this paper.The effective noise samples with the best acoustic characteristics were screened out to establish the database.Then we carried out the subjective and objective evaluation experiment to get the subjective preference and psychoacoustic objective parameters of each noise sample.According to the structural characteristics of rotating mechanical parts to analyze its radiation noise’s acoustic characteristics,it was found that there was the single-frequency noise or the order noise,which has the most pronounced tonal characteristics.The most important is the fact that compared with the broadband noise,the tonal noise is more likely to cause people’s subjective "bored".However,the correlation between Aure’s tonality and subjective preference was too low to fully reflect the influence of tonal characteristics on subjective feelings.In this paper,the calculation model of Aure’s tonality was analyzed in detail to find the cause of decreasing correlation.Based on the tone-to-noise ratio(TNR),which can better express the energy ratio of tonal components in noise and have a higher correlation with the subjective feelings,an improved tonality evaluation model was proposed.To better solve the complex non-linear problem between the subjective sound quality evaluation results and objective psychoacoustics parameters,and achieve the mapping of sound quality from objective parameters to subjective evaluation,a back propagation neural network(BPNN)prediction model was established based on neural network.After analyzing its training and predictive performances,the network optimization algorithm: Genetic Algorithms(GA)and Particle Swarm Optimization(PSO)were introduced to optimize its deficiencies.Then based on the BPNN network model,two optimization models: the genetic algorithm-back-propagation neural network(GA-BPNN)model and the PSO-back-propagation neural network(PSO-BPNN)model,which can optimize the initial weights and thresholds were established.Finally,this paper analyzed the prediction performance of the three models.The results show that the PSO-BPNN model can achieve convergence more quickly and improve the prediction accuracy of sound quality,which can further lay a foundation for the research on noise quality evaluation of rotating parts.