基于非线性匹配追踪数据驱动时频分析的变工况机械故障诊断方法研究

Nonlinear matching pursuit data-driven time–frequency analysis (NMPDTA) is a novel sparse representation method for analyzing non-stationary signal proposed by Thomas Y H. Based on empirical mode decomposition and compressed sensing thoery, the NMPDTA method combines the advantages of both. The NMPDTA method not only can extract each signal component from the non-stationary signal rapidly, and compute the instantaneous frequency of each signal component accurately, also has strong anti-noise ability and signal processing self-adaptivity. The vibration signal of mechanical equipment under variable working condition is always nonlinear and non-stationary , and it contains time-varying operation information and fault information of mechanical equipment. This project intend to introduce the NMPDTA method into fault diagnosis for mechanical equipment under variable working condition, and use it to separate fault component from vibration signal of faulted mechanical equipment and extract fault characteristic at the same time. This project will build complete and systematic fault diagnosis methods for mechanical equipment under variable working condition based on NMPDTA. The fault characteristic of mechanical equipment under variable working condition is usually related to instantaneous frequency, while the NMPDTA method can estimate instantaneous frequency accurately, therefore, the research achievements in this project can be applied in extracting fault characteristics and diagnosing fault for mechanical equipment under variable working condition.

非线性匹配追踪数据驱动时频分析方法是Thomas Y H等人近年提出的一种非平稳信号稀疏表示方法，该方法是在经验模态分解和压缩感知的基础上提出的，兼具了两者的优点。该方法不仅可快速分离非平稳信号中的各信号成分，并精确估计信号成分的瞬时频率，且具有较强的抗噪能力和信号分析自适应性。机械设备变工况运行时，其振动信号为非线性非平稳信号，且信号中包含了时变的设备运行信息和设备故障信息。本项目拟将非线性匹配追踪数据驱动时频分析方法用于变工况下的机械故障诊断，研究用该方法分离变工况机械设备故障振动信号中的故障成分并提取故障特征，在此基础上建立完整、系统的基于非线性匹配追踪数据驱动时频分析的变工况机械故障诊断方法。变工况机械设备的故障特征通常与瞬时频率相关，而非线性匹配追踪数据驱动时频分析方法则具有精确的瞬时频率估计能力，因此，本项目研究成果能有效应用于变工况机械设备故障振动信号的特征提取与诊断。

工程实际中机械设备常处于变工况运行，易导致突发设备故障。变工况下机械设备振动信号多为非平稳信号，且信号中包含了时变的设备运行信息和设备故障信息，研究适合变工况下机械设备的信号处理方法对变工况机械设备的故障诊断具有极为重要的意义。本项目在对非线性匹配追踪数据驱动时频分析（简称DDTFA）方法进行理论研究和改进完善的基础上，首先对其参数选取、适应性、局限性，重点对初始相位函数优化选取进行了研究，分别提出了MSCSD-DDTFA、VMD-DDTFA、ZPDF-DDTFA等方法及其一种改进的DDTFA方法，然后对上述方法在变工况机械设备微弱故障、复合故障诊断中的应用展开了深入研究，仿真分析与应用实例验证了MSCSD-DDTFA、VMD-DDTFA、ZPDF-DDTFA及改进的DDTFA方法应用于变转速齿轮断齿故障诊断、VMD-DDTFA应用于变转速齿轮裂纹故障诊断、ZPDF-DDTFA应用于恒定工况滚动轴承裂纹故障诊断、改进DDTFA方法应用于齿轮箱复合故障（断齿+轴承外圈裂纹）和行星齿轮箱太阳轮断齿故障诊断的有效性，并针对上述方法对变工况机械设备微弱故障、复合故障诊断的局限性进行了仿真分析与实例验证，最后建立了完整、系统的基于DDTFA方法的变工况和恒定工况的机械故障诊断方法。项目研究成果能为变工况和恒定工况下机械设备故障振动信号的特征提取与诊断提供了新的有效途径与技术手段，具有重要的理论意义与应用价值。