【作者】 倪广健；

【导师】 舒歌群；

【作者基本信息】 天津大学 ， 动力机械及工程， 2012， 博士

【摘要】 本文以柴油机的表面辐射噪声问题为研究对象,进行了柴油机的噪声源识别研究,采用虚拟技术对柴油机辐射噪声进行预测和结构优化设计;建立了一个基于生理学功能特性和解剖学结构特性的人耳耳蜗模型,结合心理声学理论,对柴油机表面辐射噪声的声音品质进行了主客观的评价研究。在柴油机的噪声源识别研究中,通过试验获得了柴油机表面辐射噪声在各个转速下的频谱特性。采用层次分析法研究了柴油机不同转速、不同频率范围对噪声的贡献度;给定转速下,各部件对不同频率范围的噪声的贡献大小。以权重值的形式对影响柴油机辐射噪声的因素进行了定量表示,确定了影响柴油机辐射噪声的关键因素和柴油机的主要部件,为柴油机结构优化设计提供依据。结合有限元法和柔性多体动力学方法建立了柴油机虚拟样机。通过对比实验模态分析和有限元模态分析结果,验证了有限元模型的正确性。对柴油机虚拟样机计算得到的柴油机表面振动速度分布情况和试验测得的柴油机辐射噪声分布情况进行定性比较,良好的一致性表明该虚拟样机模型可以用来预测柴油机的表面辐射噪声分布情况。以低噪声柴油机设计为出发点,对机体裙部、肋板、曲轴箱结构进行了优化设计和优化效果分析。针对额定转速下辐射噪声贡献最大的部件油底壳进行了减振降噪研究。通过增加加强筋、改变壁厚等方法来提高油底壳结构刚度,分析了不同方案对柴油机表面振动速度的影响。区别于采用滤波器组建立的人耳外周听觉模型,从波的角度建立了耳蜗的传递波模型。通过与小鼠耳蜗试验结果的对比,定性说明耳蜗模型的正确性。采用波有限元方法求解得到了耳蜗内声波的传播情况,结合WKB方法计算得到了耳蜗在外界刺激下所产生的基底膜响应。针对柴油机噪声测试试验中获取的噪声样本进行了声品质主观评价研究。选取不同频率段的噪声样本作为评审测试中所使用的声音样本,邀请不同年龄、性别、背景的人作为评审团,分别采用成对比较法和打分法两种主观评价方法对噪声样本进行了评价,获得了噪声主观满意度。通过计算1 kHz,40 dB的纯音激励产生的响度的算例来定量说明该耳蜗模型可以用于声品质客观物理参数的预测计算。以建立的耳蜗模型为基础,结合心理声学理论,计算了各噪声样本的响度、尖锐度等心理声学客观物理参数,并分析了客观参数与主观评价结果的相关性。在此基础之上,引用人工神经网络技术对声品质的客观评价物理参数与表面辐射噪声主观舒适度之间的非线性关联进行了预测。更多还原

【Abstract】 The diesel engine radiated nosie was taken as the subject of this research. The radiated noise sources were identified using experimental methods and virtual techniques. Virtual techniques such as finite element method and multi-body dyamic method were applied to predict the surface vibration and radiated noise from the diesel engine, and optimize engine structure. The sound quality of the radiated nosie was evaluated subjectively based on a cochlear model built from anatomical data and physiological characteristics, and objectively based on paired comparison and marking studies.The spectrum of the diesel engine radiated noise was obtained from experiments as well as the analytic hierarchy process, AHP, which was used to calculate contributions of different engine speed and frequency range to the overall nosie level. Next, the contributions of different engine components to the overall radiated noise at different frequency ranges at a given engine speed were also calculated using the AHP. The influence factors of the radiated noise were given in terms of weight factors which can show the quantitative importance of each engine component, providing guidance to structural optimazation. The virtual prototype of the diesel engine was built using the finite element method and the multi-body dynamic method. Comparison of the results of modal analysis between the simulation and experiments shows that the computational model is accurate. The predicted surface vibration level of the engine has a good correlation with the radiated nosie distribution obtained from the experiments, indicating that this virtual prototype can be used to predict the radiated nosie level of the engine. In order to reduce vibration and noise of the diesel engine, structural optimizations were carried out for engine body, skirt and crankcase. The oilpan was also optimized due to its contribution to the overall radiated nosie is maximal at the rated speed, 2200 rpm. Several actions, such as adding strengthening ribs and changing thickness, were applied to stiffen the oilpan.The traveling wave model of the cochlea was built in contrast with those formed based on filter band. Comparison between predicted results and experimental data of the chinchilla cochlea qualitively shows that this traveling model of the cochlea is reasonable. The wave finite element method was used to calculate wave propergation in the cochlea and the response of the basilar membrane was calculating using the WKB approximation for a stimulus at a given frequency.Noise samples at different frequency ranges were chosen based on engine experiments for the subjective evaluation. People of different age, background and gender were invited for the jury test. Paired comparison and marking were used as the jury test to evaluate the noise samples subjectively and subjective satisfaction was obtained. The loudness value due to a 1 kHz, 40 dB tone was first calculated using the aformentioned cochlear model to show its accuracy. The phyco-acoustic parameters such as loudness and sharpness were calculated using the cochlear model and the correlation between those objective parameters and subjective satisfaction was calculated to show the contribution of each phyco-acoustic parameter to subjective satisfaction. Finally, a neural network model was built and trained to show the nonlinear realation between objective phyco-acoustic parameters and subjective satisfaction.更多还原