【作者】 杨川；

【导师】 于德介；

【作者基本信息】 湖南大学 ， 机械工程， 2012， 博士

【摘要】 随着汽车行业的高速发展，人民生活水平的不断提高，顾客对汽车的舒适性要求也随之提高。而噪声作为评价汽车品质的主要因素之一，也越来越受到重视。近年来，人们发现传统的噪声控制已经不能很好地满足市场需求,于是，声品质（Sound Quality）的研究与应用成为噪声控制领域的研究热点，越来越受到汽车企业的重视。声品质已成为产品竞争力的重要因素。汽车关门的声音不只具有噪声特性，更多的还是具有品质的特性。顾客对汽车的首要印象不仅来源于汽车的外观，也来源于汽车关门声品质。若汽车关门声音的品质差，将直接让顾客决定放弃该车，从而影响汽车企业的竞争力。因此，汽车关门声品质研究是汽车声品质研究的重要内容之一。论文选题源于国家“跃升计划”专项——中国高水平汽车自主创新能力建设等项目，结合某企业开发的一款样车中的车门系统项目，对汽车关门声品质客观评价参数、汽车关门声品质预测方法以及汽车关门声品质优化设计方法展开深入系统地研究。论文主要研究内容有：（1）针对目前声品质预测模型的预测精度不高问题，将EMD分解引入汽车关门声品质预测，研究建立了基于EMD分解的BP神经网络、支持向量机预测模型；将临界频率带解析小波分解引入汽车关门声品质预测，研究建立了基于临界频率带解析小波分解的BP神经网络、支持向量机预测模型。对比分析了基于心理学参数的传统声品质预测模型和基于时频信号处理方法的声品质预测模型的预测精度。（2）针对现有的声品质评价参数不能全面准确评价汽车关门声品质的问题，将现代时频分析方法引入汽车关门声品质客观评价，研究利用时频信号处理技术提取汽车关门声信号声品质特征，在此基础上建立新的基于时频信号处理技术的汽车关门声品质客观评价参数；提出了基于伪Wigner-Ville分布的汽车关门声品质评价参数SMPWVD(Sound Metric Based On Pseudo Wigner-Ville Distribution)以及基于临界频率带小波分解的汽车关门声品质评价SMCBWD(Sound Metric BasedOn Critical Band Wavelet Decomposition)；对比分析了SMPWVD和SMCBWD与汽车关门声品质主观评价绩效值之间的相关性。（3）针对汽车关门声品质分析离不开实际样车的问题，将普遍用于汽安全碰撞分析的仿真方法应用于关门碰撞仿真分析，并结合试验设计方法和响应面法建立了关门碰撞优化问题的近似模型；基于遗传算法，以减小车门系统能量为目的，对近似模型进行了优化；为了验证车门系统经优化设计改进后汽车关门声品质得到改善，分别采用仿真和实验方法对改进效果进行验证。建立了车门系统的声辐射仿真分析模型，计算了优化前后车门辐射出的声压值及传统汽车关门声品质评价参数（响度、尖锐度），并进行了对比分析；基于汽车关门声品质的客观评价试验和主观评价试验方法，对修改前样车(OTS3)、修改后样车(OTS4)及一辆关门声品质较好的标杆车进行了关门声品质客观评价和主观评价，得出了响度、尖锐度、SMPWVD、SMCBWD以及主观偏好性绩效值，并进行了对比分析。本文对汽车关门声品质预测模型、汽车关门声品质评价参数以及汽车关门声品质优化设计方法进行深入研究。建立了基于时频信号处理方法的汽车关门声品质预测模型，有效提高了汽车关门声品质预测模型的预测精度；提出了基于伪Wigner-Ville分布的汽车关门声品质评价参数SMPWVD以及基于临界频率带小波分解的汽车关门声品质评价SMCBWD，为汽车关门声品质的客观评价提供了准确有效的评价参数；建立了一套汽车关门声品质优化设计流程，通过优化汽车关门碰撞能量，成功改善了汽车关门声品质。论文研究成果能为汽车关门声品质预测、评价以及改善提供有效途径，具有重要的理论意义及工程应用价值。更多还原

【Abstract】 With the development of vehicle industry and the improvement of people’s livingstandard, the customer’s requirement for the comfort of the car is improved. As one ofthe main factors of automobile quality, noise has attracted more and more attention. Inrecent years, people has found that the traditional noise control cannot meet themarket demand, so the research and application of sound quality has become a focusin the field of noise control, and is attracted more and more attention by the carcompanies. Sound quality has become the important factors of productcompetitiveness.The sound of door-slamming has not only the characteristics of noise, but alsothe characteristics of quality. The customer’s first impression of a car not only camefrom the appearance of a car, but also came from the sound quality of door-slamming.If the sound quality of door-slamming is poor, the car will be given up by customersand the competition ability of the enterprise will be influenced directly. Therefore, theresearch of the sound quality of door-slamming is one of the important contents invehicle quality. The dissertation sponsored by the project “Yue Sheng Program” andthe independent innovation capabilities of high-level car construction in China. Thesound quality prediction model, the sound quality evaluation parameters and thesound quality optimization methods of door-slamming were studied in thisdissertation.The main research contents of this dissertation were(1) Aiming at the problem that the precision of the sound quality predictionmodel of door-slamming is poor, EMD (Empirical Mode Decomposition) method wasintroduced to predict the sound quality of door-slamming. Based on EMD, BP (BackPropagation) neural network prediction model and SVM (support vector machines)prediction model were established. Furthermore, Critical Band WaveletDecomposition was also introduced to predict the sound quality of door-slamming.Based on Critical Band Wavelet Decomposition, BP (Back Propagation) neuralnetwork prediction model and SVM (support vector machines) prediction model wereestablished. The comparative analysis has been done between the traditionalprediction model based on psychology parameters and the new prediction model basedon time-frequency analysis methods. (2) Aiming at the problem that the existed sound quality evaluation parameterscannot evaluate the sound quality of door-slamming fully and accurately, thetime-frequency analysis methods were introduced into the objective evaluation ofsound quality of door-slamming. The characteristics of the sound quality wereextracted by time-frequency analysis methods, and based on it new sound qualityevaluation parameters were established. Two new sound metrics called S ound Metricbased on Pseudo Wigner-Ville Distribution (SMPWVD) and Sound Metric based onCritical Band Wavelet Decomposition (SWCBWD) were developed. The comparativeanalysis of the correlation coefficients between the subjective sound qualityperformance value of door-slamming and SMPWVD or SMPWVD has been done.(3) Aiming at the problem that the sound quality of door-slamming can’t beanalyzed without actual vehicles, the explicit finite element method widely used invehicle safety impact analysis was introduced into the analysis of door-slamming. Bycombining the explicit finite element method with the experimental design methodand the response surface method, an approximation model of door-slammingoptimization was established. To reduce the vibration energy of door-slamming, theapproximate model was optimized by using the genetic algorithm. The simulation andexperimental results before and after optimization verified that the sound quality ofdoor-slamming of the optimized model was improved. Furthermore, the soundradiation simulation model of the door system was established. By comparing thesound pressure and traditional sound quality evaluation parameters (loudness,sharpness) of the sound radiation simulation model before and after the optimization,the sound quality of door-slamming of the optimized model was improved. Based onthe objective and subjective evaluation experiments, the loudness, sharpness,SMPWVD, SMCBWD and the subjective sound quality preference of the car beforemodified (OTS3), the modified car (OTS4) and a benchmarking car was test andanalyzed.This dissertation made in-depth researches on the sound quality prediction model,the sound quality evaluation parameters and the sound quality optimization methodsof door-slamming. The new prediction model based on time-frequency analysismethods was established. The prediction precision of the new prediction model wasimproved compared with the traditional prediction model. Two new sound metricscalled SMPWVD and SWCBWD which can provide the accurate and effectiveevaluation parameters were developed. What’s more, the sound quality optimizationprocess of door-slamming was established. By optimizing the energy of door-slamming, the sound quality of door-slamming was improved. The researchresults of this dissertation can provide the effective way to predict, evaluate andimprove the sound quality, and have an important theoretical significance andapplication value.更多还原