【作者】 刘海；

【导师】 张俊红；

【作者基本信息】 天津大学 ， 动力机械及工程， 2013， 博士

【摘要】 柴油机噪声品质直接影响着人们的身心健康与对产品的取向，是目前动力机械振动噪声控制研究领域面临的重要研究课题之一。论文开展柴油机噪声品质主客观评价、主客观相关性研究过程中的一系列基础理论研究。采用基于核的主成分分析方法提取噪声品质客观评价特征；开展噪声品质的主观评价研究；采用基于支持向量机方法进行噪声品质主客观相关性研究；利用灵敏度分析方法构建基于支持向量机的噪声品质客观特征贡献度模型；建立噪声品质仿真分析模型，为柴油机声品质设计奠定了理论基础。全文的主要研究内容如下：1、以国内某款柴油机为研究对象，通过噪声试验获取稳态工况下的辐射噪声信号，经过预处理建立噪声信号样本库；集多体动力学、有限元方法、声学边界元为一体，开展柴油机辐射噪声的仿真分析，获得噪声频域信号，进行噪声仿真与测试信号的对比，结果表明噪声仿真结果与测试结果具有较高的一致性。2、以声品质理论为研究基础，采用基于核的主成分分析方法提取与柴油机辐射噪声具有强关联性的客观评价特征，以降低客观评价维数，简化噪声品质预测模型的结构；分析主要客观评价特征随柴油机工作转速及负荷的变化规律，定性地分析影响噪声品质变化的原因。3、以样本噪声为研究对象，利用模糊聚类方法对噪声样本进行分类、降维处理；以综合满意度指标作为噪声优劣的评判标准，采用成对比较法开展噪声信号的主观评价实验研究；利用Kendall一致性方程和Bradley-Terry统计学模型对主观评价实验进行可靠性和正确性的验证。4、采用三种不同的回归分析方法，包括支持向量机、神经网络与多元线性回归方法研究柴油机噪声品质主客观评价结果之间的相关性，研究结果表明支持向量机方法在预测精度上优于多元回归、神经网络方法，其稳定性也优于神经网络方法；建立基于支持向量机的噪声品质客观评价参量的灵敏度模型，定量地计算影响柴油机噪声品质的客观评价参量贡献度大小；以噪声仿真信号作为信号源，结合物理、心理声学理论以及噪声品质预测理论，开展柴油机辐射噪声品质仿真分析；采用试验技术研究部件噪声对整机噪声品质的影响规律，提出柴油机结构改进设计方案，完成结构改进对整机辐射噪声品质影响效果的预测，从而形成预测柴油机辐射噪声品质的理论分析模型，为高声品质现代柴油机设计奠定了基础。更多还原

【Abstract】 Diesel engine noise quality directly affects people’s physical and mental health aswell as the product orientation, and it has become an important research subject inpower machine vibration and noise control research fields. Consisting of thesubjective and objective evaluation of diesel engine noise quality and correlationanalysis between subjective and objective evaluations, some basic theoreticalresearches are carried out. Kernel Principal Component Analysis (KPCA) is used toextract the main objective features of diesel engine noise quality; Research onsubjective evaluation of diesel engine noise is conducted; Sensitivity analysis methodis used to build the contribution model of objective features; Noise quality simulationmodel is built. Above all, the results provide a theoretical foundation for diesel enginenoise quality design. The main contents include:1、A domestic diesel engine is taken as a sample, engine noises generated underthe steady-state operation are collected in the engine anechoic chamber, and thesample database of diesel engine radiation noise is established after preprocessing;Noise quality simulation for diesel engine using the method of multi-body dynamics,finite element and acoustic boundary element is carried out; engine noises in thefrequency domain are obtained. Compared with the test data, simulation results areexactly consistent with the experimental results.2、Based on sound quality theory, KPCA is used to extract the main featureswhich influencing the noise quality, which can effectively reduce the featuredimension and realize the prediction model of diesel engine noise qualitysimplification; Variations of objective features along with engine speeds/loads areanalyzed to qualitatively understand the reason for the change of diesel engine noisequality.3、Noise samples are classified and reduced dimensions by Fuzzy Cluster Method(FCM). The subjective evaluation test is conducted by Paired Comparison Method,with the Integrated Satisfaction Index (ISI) as the evaluation criterion. Then thereliability and correctness are validated by Kendall’s coefficient of concordance andBradley-Terry Model.4、Three different regression analysis, including Support Vector Machine (SVM),Artificial Neural Network (ANN) model and Multiple Linear Regression (MLR) model, are used to establish the relationship between subjective and objectiveevaluations of diesel engine noise quality. The results show that the SVM model canobtain higher predicting accuracy than ANN model and MLR model, and the stabilityof the SVM model is better than the MLR model. Sensitivity model of objectivefeatures influencing the noise quality with SVM is established to quantitativelycalculate the contribution of each objective features affecting noise quality; Withsimulation noise being the signal source, noise quality simulation for diesel engine isrealized by combing the physical, psychoacoustic and sound quality theory. Theinfluences of parts on the diesel engine noise are studied to propose a scheme on partsimprovement, with the aid of SVM prediction model, the prediction of the effect ofparts improvement on the engine noise quality is realized. Thus the theoreticalanalysis model for predicting the diesel engine radiation noise quality will be formed,which lays the research foundation for modern diesel engine design with high levelnoise quality.更多还原