基于主成分回归的电动汽车声品质满意度预测

Satisfaction Degree Prediction of Electric Vehicle Sound Quality Based on Principal Component Regressione

为研究电动汽车车内噪声特性及声品质满意度评价方法,采用等级评分法对车内总体噪声进行主观评分,计算总体噪声A计权声压级、响度、粗糙度、尖锐度、语言清晰度指数(AI),同时对车内电机电磁噪声的声压级(SPL)、响度、尖锐度进行计算,与总体噪声客观参量共同作为原始变量,通过降维方式构建了3个主成分,并以此为自变量构建了主成分回归模型。结果表明,主成分回归模型对比多元线性回归模型平均绝对误差降低了4.3%,均方根误差降低了11.6%,可作为电动汽车声品质研究一种有效手段。

In order to study the interior noise characteristics of electric vehicle and the evaluation method of sound quality satisfaction,the overall interior noise is scored subjectively by the grade scoring method,and the A-weighting sound pressure level,loudness,roughness,sharpness and language articulation index(AI) of the overall noise are calculated,while the sound pressure level(SPL),loudness and sharpness of the interior motor electromagnetic noise are calculated.The objective parameters of electromagnetic noise and overall noise are taken as the original variables together.Three principal components are constructed by dimensionality reduction method,and the principal component regression model is built with these as the independent variables.The results show that the mean absolute error and root mean square error of the principal component regression model are 4.3% and 11.6% lower than those of the multiple linear regression model,which can be used as an effective means to study the sound quality of electric vehicle.