非稳态工况下的车辆声品质评价研究

The sound quality of noise for vehicles under non-stationary operation is of ever importance when assessing the quality of the product, and it is a great challenge to the NVH research and engineers. On basis of the achivements of classic psychoacoustic metrix and standardized processures of sound quality assesment, the project aimed to the invstigate the noise characteristics of vehicles at idling, accelerrating, decelerating and constant driving status, which are stationary and non-stationary conditions respectively. The interior and exterior vehicle signals will be measured for the analysis and selection of characters and for sound quality evaluating. With grouped paire-comparison method for subjective evaluation purpose, principal component analysis and prominence ratio statistical data treatment, the relationship of human subjective impression and objective psychoacoustic parameters is considered, and then a model for evaluation of non-stationary sound quality is proposed based on the support vector machine (SVM) regression for the mapping of the subjective and objective testing results. Meanwile, considering the practice of time varing unstability of vehicle driving and the uncertainty of subjective judgement, as well as the duplication of various characters, a robust control method is adapted for the construction of the prediction and the evaluation model in order to improve the stability and convergence of the evaluating method and progress. The theory and method of the project will give complementation for the total aspects of vehicle sound qulity assessment, and will surpport the sound characteristic design for vehicle development.

非稳态工况下的车辆声品质评价是NVH研究领域的挑战性问题。本项目拟在传统心理声学参数及声品质评价方法的基础上，针对怠速、等速以及加速、减速等稳态与非稳态行驶工况下的车辆噪声特点，进行系列的半消声室室内及道路测试试验，研究适用于车辆声品质评价的非平稳噪声数据采集及特征提取方法；项目结合多工况下的分组成对比较主观评价方法，利用主成分分析及显著度分析等数理统计手段，研究客观心理声学参数与主观感受之关系，通过基于支持向量机回归方法，建立非稳态工作条件下的车辆噪声客观声学参量与主观评判结果的映射关系，得到全新的车辆声品质评价模型；并针对车辆运行的不稳定性、主观判断的不确定性以及数据的多重复性等建模中的不确定性和扰动问题，提出以鲁棒控制方法提高模型的稳定性和收敛性。项目通过理论及实验研究，探索非稳态车辆噪声的声品质评价模型与方法，完善汽车的声舒适性评价体系，将为高品质车辆的声学设计提供理论与技术支撑。

非稳态工况下的车辆声品质评价是NVH研究领域的挑战性问题。本项目拟在传统心理声学参数及声品质评价方法的基础上，针对怠速、等速以及加速、减速等稳态与非稳态行驶工况下的车辆实测噪声特点，研究适用于车辆声品质评价的非平稳噪声高效评价及预测模型与方法。针对传统阈值小波去噪的不足，提出了基于遗传算法的自适应阈值去噪方法；针对非稳态车内噪声进行特征提取与分析，选取dB35小波为分解的基函数对信号进行分解，得到基于特征频带分解的小波包分解结构，模拟人耳的Bark域频率感知特性；构建声样本基于Bark尺度频带划分的特征向量和基于ERB尺度频带划分的特征向量，将特征向量作为声品质评价模型的输入因子，建立基于SVM的声品质评价模型，与传统的评价模型相对比，其预测精度更高；引入灰色关联度分析将声信号样本进行分组后建立支持向量机SVM声品质评价预测模型以实现小样本下声品质的准确预测。建立基于Moore时变响度的汽车制动工况下车内声信号响度测试系统，以某混合动力城市客车为研究对象，获得制动工况下乘员耳旁的降噪前后噪声Moore响度，较Zwicker响度模型频率分辨率更高且能量分布更为集中，更适合分析车的车内声信号特征；以乘用车驾驶员处采集的发动机启动声样本为对象，分析其时频域分布，计算样本的基本物理参数、心理声学参数和烦恼度指标，运用成对比较法进行评价得出各声样本主观偏好性得分；进行声品质应用研究，完成汽车关门声小波熵声品质评价、汽车雨刮器声品质影响因素分析、汽车喇叭声品质分析与评价以及车窗玻璃升降声测试与声品质评价等工作。项目研究成果对于汽车非稳态噪声的声品质建模、分析评价及应用具有重要的理论价值和工程意义。