基于阵列全息的车辆振源识别与贡献排序方法研究

Making a diagnosis and identification reasonably on the main vibration source for vehicle is the premise of optimizing the vibration response. In this project, based on the theory of array beamforming and holography, the methods of identifying vibration sources and their characteristics are studied according to the exciting position, transmission path, frequency, phase, and the number of test sensors, its position etc. First, assuming that the excitation source is independent and is in the same plane with the test sensor, the transmission path between the target response and exciting is limited, based on the amplitude and phase obtained from the array sensors, the array method, including the array beamforming and time delaying algorithm, shall be designed to identify the source location. Simultaneously, the best selecting principles of the number and position of test sensors are determined; when the excitation source is more than one with relavant features, combining with blind source separation to de-relate, the source location can be identified and the contribution in classified frequency band can be sorted. Further more, the designed method of identifying sources is extended to the three-dimensional vibration system. With modal characteristics of the vehicle’ some structure got from the experiment, the whole frequency band is divided into the resonant and non-resonant area, applied by an array of source identification methods. The proposed method can be used to be real-time monitoring of vehicles running and while abnormal peak coming and directing fault location point. In terms of vehicle performance, it can also give a more precise location of the vibration source, thus to facilitate the vehicle vibration optimization.

对车辆振动主要激励源进行合理的诊断和识别是振动响应优化的前提。本项目基于阵列波束形成与全息理论，从振动激励的位置、传递路径、频率、相位以及测试传感器数量、位置选择等方面进行源贡献的识别和特征研究。首先假设激励源独立，至目标响应的传递路径有限，与测试位置处于同平面，基于阵列传感器获得的幅值和相位，进行波束形成和时延法的阵列算法设计，应用于源位置识别，并给出测试传感器数量与位置最佳的选取原则；在激励源多于1个，且源具有相关特征时，结合盲源分离进行解相关，在分类频段，选择频段进行源位置识别与贡献排序；进而推广应用于三维空间的振动系统。试验获得车辆部分结构的模态特征，划分为共振区和非共振区，进行阵列方法的源识别。本项目形成的方法可以实时的监测车辆运行状态，在特殊峰值超过允许量级时指向故障位置，在车辆振动性能方面对主要振源进行位置识别与能量排序，便于整车振动优化。

针对车辆振动的源识别问题，基于阵列传感器测试手段，发展现代振源识别方法，提高识别效率和节约成本。具体内容如下：.（1）基于单三角阵列的振源识别理论，发展了单三角形阵列对短距离的振源识别、等腰直角三角形阵列对中远距离振源识别，进而提出了线性阵列、米字阵列。米字阵列算法包罗了单三角形和等腰直角三角形的算法，同时也推导了非直角等腰三角形识别振源的算法。在一阶米字阵列算法的基础上提出了二阶米字阵列。实验证明，二阶米阵列算法精度可以实现振源识别精度82%以上。.（2）项目研究了振动测试信号处理方法，有效提高源识别的信号基础精度。研究了经验模式的分解特性，提出了增加大幅值的高频分量和叠加多次有色噪声，有效改善EMD虚假分量和模式混叠。提出在LMS算法迭代过程中以比较阶数和步长变化时的最小均方误差期望为收敛方向，联合经验模式分解和最小均方算法（LMS）发展了一种自适应去噪方法，有效避免LMS算法对宽频信号的不稳定性、EMD分解的不唯一性和去噪中阈值的选择问题。.研究了模糊传递路径的振源识别，通过对源观测信号进行四阶累积量计算，对形成的多个切片进行近似对角化处理，进而以延迟点为中心进行对角化矩阵的平均，改进联合对角化盲分离算法。针对具有确定相关性的源识别，通过模态特征获取，对响应观测信号进行小波包子带分解，以高阶累积量互信息量（HMI）表征子带间的相关性，选择最为独立的子带作为分解信号，进而联合新对角化方法进行分离，实现源特征提取和识别。.针对车辆悬挂，推导了车辆实际运行时的悬置传递特性，结合试验设计悬置件隔振特性的计算流程，提出了以独立动力系统激励试验、独立路面激励试验、实际稳态车速试验获得振源间的干涉成分、共频成分和相应比例；最后，通过实车试验测试与数据处理，获得动力系统悬置的时域隔振特性。.（3）联合matlab和Labview编制实时测量与分析软件，基于NI硬件平台搭建采集框架，选用轻型车进行方法的应用，实现整车状态的实时监控与分析。