基于语义细分法的某汽车传动系声品质评价

Sound Quality Evaluation of Driveline System of Certain Automobile Based on Semantic Differential Method

针对中国某汽车企业传动系声品质评价问题,开展了该传动系声品质的主观评价,建立了与主观评价相统一的声品质客观评价模型。传动系噪声在半消声室试验台架上测试获得,测试工况包括全工作转速范围的匀加速和部分定速工况,通过测试数据的一致性分析确定用于声品质分析的噪声样本;综合考虑该传动系噪声的物理特征和人耳的听觉特性,并依据声品质主观评价方法的内涵属性,选择语义细分法作为声品质主观评价方法。研究结果表明:从强度、音调、音色等多个维度出发,建立汉语背景下语义细分法的语义空间,运用专家咨询法、项目区分度、主成分分析和因子分析方法,可以确定语义空间中适用于汽车传动系声品质评价的评价指标,并验证了其有效性;选取评价主体完成该传动系声品质的主观听音试验,获得主观评价结果;选取响度、尖锐度、语言清晰度等指标作为该传动系噪声客观评价指标并计算出结果;以主观评价结果为应变量,客观指标计算结果为自变量,通过一元回归分析,发现响度、尖锐度、语言清晰度与主观评价结果相关性较好;以主观评价结果为响应,多个客观指标计算结果为输入,运用多元回归分析方法及支持向量机智能建模方法,建立了该传动系啸叫客观评价的回归模型和人工智能模型,并进行了验证,实现了该汽车传动系声品质的客观量化评价。

In allusion to the problem of the sound quality evaluation of an automobile driveline system, the driveline system was assessed subjectively. Meanwhile, the objective evaluation model corresponding to the subjective evaluation was established. Driveline noise was measured on a semi-anechoic chamber, with acceleration conditions over a full operating speed range as well as some conditions of constant speed. Noise samples for sound quality analysis were determined by consistency analysis of the test data. Considering the physical characteristics of the driveline system noise and the auditory characteristics of human ear, the semantic differential method was selected as subjective evaluation method of the sound quality in terms of its intension. The resultsshow that the semantic space of semantic differential method under Chinese background is established from the perspectives of sound intensity, tone, timbre and so on. And the evaluation indexes for driveline system of sound quality in the semantic space are obtained and verified by expert consultation method, item discrimination, principal component analysis and factor analysis. The subjective evaluation results are obtained through listening test of the sound quality of the driveline system. The results are calculated by selecting loudness, sharpness, speech intelligibility and so on as evaluation indexes. Based on the single regression method, subjective evaluation results are closely related to objective metrics, such as loudness, sharpness and speech intelligibility, with subjective evaluation result as an independent variable and objective indicators as a dependent variable. Taking subjective evaluation result as a response and objective indicators as input, the regression model and artificial intelligence model are established and verified via the multiple regression method and the support vector machine, with the realization of objective and quantitative evaluation of the sound quality of the driveline system.