【作者】 景亚兵；

【导师】 刘昌文； 毕凤荣；

【作者基本信息】 天津大学 ， 动力机械及工程， 2017， 博士

【摘要】 内燃机驱动的发电机组广泛应用于军用和民用的多个领域,而发电机组运行噪声过大,不仅严重影响附近人员的工作生活,而且会影响装备使用作业的安全性及可靠性,因此,降低发电机组噪声的需求十分迫切。物理声源识别及灵敏度分析对于发电机组噪声控制具有重要意义,但由于受传递路径、机械结构、工作循环及转速波动等影响,发电机组物理声源与辐射噪声之间的映射关系具有高度非线性,其系统方程非常复杂,灵敏度指标无法直接计算得到。因此,识别发电机组的主要噪声源,研究发电机组噪声源灵敏度及低噪声结构设计对于发电机组噪声控制有着重要的理论意义和巨大的工程实用价值。结合小型发电机组的噪声信号特点,开展了发电机组的表面噪声源和物理噪声源识别研究,在物理声源识别的基础上,将支持向量机回归理论应用到噪声预测中,分析了主要物理噪声源的灵敏度。最后开展了低噪声结构设计研究,针对主要物理声源——配气机构噪声,研究了主要结构参数对其动力学特性的影响;针对传递路径——隔声罩,研究隔声罩进、排气口位置、隔声罩内声源位置以及温度场、流场对隔声罩外部声场的影响,并进行了小型发电机组低噪声结构设计。全文的主要研究成果如下:第一、以某国产小型发电机组为研究对象,利用声成像技术,根据近场声全息测试结果重建声源表面的声强分布并生成噪声成像图,根据声强分布识别得到了封闭状态发电机组表面主要声源的位置在隔声罩进排气通风口;同时指出由于测试阵列空间分辨率不足、发电机表面形状复杂以及零部件相互遮挡反射的影响,噪声成像分析不能精确识别开架状态小型发电机组的表面辐射声源。第二、针对盲源分离时经验模态分解(EMD)不能确保IMF分量完全正交,独立分量分析(ICA)不能实现单通道信号输入的问题,研究利用EMD、ICA协同分析方法对噪声信号进行盲源分离,仿真分析结果表明,EMD-ICA协同分析法可通过单个观测信号分离出多个独立的源信号,提供了分离结果用于独立噪声源识别的前提条件。第三、在研究小波分析理论的基础上,对盲源分离得到的独立分量进行连续小波变换,结合内燃机噪声信号时域、频域的先验知识,对小波时频图进行解读,识别出了主要独立分量所表征的发电机组物理噪声源,确定该小型发电机组辐射噪声主要由气门开启、落座冲击噪声、齿轮啮合噪声和进气噪声构成。第四、分别采用支持向量机、多元线性回归及神经网络三种方法研究小型发电机组物理声源与测点声压级之间的映射关系,研究结果表明支持向量机在预测精度及预测结果稳定性上优于多元线性回归和神经网络。推导了基于支持向量机的物理声源灵敏度模型,定量计算了小型发电机组主要物理声源的灵敏度。第五、针对隔声罩低噪声结构设计,研究得到隔声罩进排气口位置、隔声罩内声源位置、隔声罩内外温度及冷却风流速对隔声罩外部声场的影响,指导了仿真模型简化及隔声罩进排气口位置、声源位置的优化;将上述结果应用于小型发电机组隔声罩设计优化,同时考虑声、热、流多物理场的影响,经仿真分析验证,优化后该发电机组辐射噪声的7米处声压级低于66dB(A)。第六、建立整个进气机构的多质量动力学模型,模拟系统的实际边界条件对配气机构动力学特性进行充分的仿真分析,研究了零件刚度、气门弹簧刚度及预紧力、气门间隙对配气机构动力学特性的影响规律。研究结果可指导配气机构低噪声结构设计。更多还原

【Abstract】 Internal combustion engine driven generator sets are widely used in many fields both in military and civilian,while they can cause serious noise problem,not only affecting the normal work and life of nearby personnel,but also damage the equipment itself.Low-noise generator sets development is an essential work.Physical sound source identification and sensitivity analysis are important for the noise control of generator sets.However,the mapping relationship between physical sound source and radiated noise is highly nonlinear due to the influence of composite factors such as transmission path,mechanical structure,engine working cycle,speed fluctuation,etc.Sensitivity index cannot be directly calculated due to the complexity of the system equation.Accordingly,it is valuable to identify the main noise source of generator sets,to develop its sensitivity and low-noise structure in noise control of generator sets.In this paper,physical surface noise identification and physical source identification were conducted.Based on the identification of physical source,support vector machine(SVM)regression theory was applied to noise prediction,and noise source sensitivity was analyzed.Then,the design of low-noise structure was studied: for the main sound source from valve mechanism noise,the influence of main structural parameters on the dynamic character was analyzed;for the transmission path-acoustic enclosure,the influence of inlet and outlet position,sound source position,temperature field and flow field of circulating air on the external sound field was analyzed.Finally,low-noise structure design of the generator set was developed.The main research results are as follows:Taking a small generator set as the research object,near field acoustic holographic test was conducted with photoacoustic tomography technology.And sound intensity distribution and noise imaging was obtained by reconstructing the sound source surface according to the test result.The position of the main sound source on the surface of the generator set was centralized in the inlet and outlet vents area in accordance with sound intensity distribution.At the same time,noise imaging analysis cannot accurately identify the surface radiated sound,due to the lack of spatial resolution of the test array,the complex shape of generator surface and the mutual reflection of the parts.The empirical mode decomposition(EMD)for blind source separation cannot ensure that the IMF components are completely orthogonal.And independent component analysis(ICA)cannot be realized by using single channel signal input.To solve these problems,the EMD and ICA cooperative analysis were used to blind source separation.Simulation results show that the EMD-ICA collaborative analysis method can separate multiple independent source signals with single channel input,which provided a prerequisite for independent noise source identification.Based on the theory of wavelet analysis,the independent component of blind source separation was decomposed by continuous wavelet transform(CWT).Comparing to the prior knowledge on internal combustion engine noise signal in time and frequency domain,wavelet time-frequency diagram was analyzed to identify the main physical noise source represented by independent component.Results show that the radiated noise of the small generator set is mainly composed of valve opening and seating impact noise,gear meshing noise and intake noise.The support vector machine(SVM),multiple linear regression and neural network were respectively used to study the mapping relationship between the physical sound source and the sound pressure level of the measuring field.Results show that SVM is superior to multiple linear regression and neural network in predicting accuracy and stability.Sensitivity model of the physical sound source was build based on SVM,and sensitivity value was calculated using this model.For the transmission path control,this thesis aims at the low-noise structure design of the sound proof hood,the influence of inlet and outlet position,sound source position,temperature field and flow field of circulating air on the external sound field was analyzed.Simulation model was simplified according to the analysis,and the location of the air vents and sound source were optimized.Simulation results show that the noise level of the generator set is less than 66 dB(A)at 7m,after optimizing the sound proofhood structure under the consideration onto the influence of sound,heating and flow fields.For the main physical sound source control,dynamical model of the valvetrain multibody system was established.Dynamic characters of the model were simulated and analyzed by the actual boundary conditions.The effects of the stiffness of the parts,the rigidity and preload of the valve spring,and the valve clearance on the dynamic characters were simulated.Simulation results can guide low-noise structure design of the valve train.更多还原