复杂多变工况下装备传动部件损伤识别与剩余使用寿命预测理论方法研究

To achieve the damage recognition and remaining useful life prediction for transmission components under complex and varying operation condition, the movements and forces are analyzed based on the kinematics and dynamics. The reconstruction method of the impulse response in time series is studied based on the signal resampling and vibration separation. Combined with the time-frequency analysis method and multi-resolution analysis method which are often used in non-stationary signal processing, features used to describe the damage evolution process under varying operation are extracted. The complex mapping relationships among damage patterns, the character of the dynamic signal which can be measured, and the operation character are researched. The implicit mapping relation model among the damage state, damage features and operating condition features is built using modern intelligent machine learning methods based on their strong modeling ability. Combined with the physical model and data-driven model, the problems of the damage recognition and remaining useful life prediction in varying operation conditions are studied. The research results can provide technical supports to the applications of fault prognosis and health management techniques in equipment, and provide technical supports to the implements of the condition-based maintenance strategy.

围绕解决大幅度工况变化或者低速非整周运转条件下，装备传动部件的损伤识别与剩余使用寿命预测问题，从运动学和动力学角度出发分析传动部件的运动规律与受力特点，开展基于信号重采样与振动分离的冲击响应时序分析与重构方法研究。结合非平稳信号处理的时频分析或多分辨率分析方法，提取变工况条件下描述损伤演化过程的特征指标。揭示损伤模式、可测量动态信号特征及工况特征三者之间的复杂映射关系，借助现代智能机器学习理论的建模能力，构建损伤状态、信号特征与工况参数三者之间的隐式映射关系模型。结合物理模型和数据驱动模型探索解决变工况条件下的损伤识别以及剩余使用寿命预测难题，为故障预测与健康管理技术体系在装备中的推广应用和状态基维修理念的实施提供技术支撑。

针对变工况下装备传动部件开展了损伤识别与剩余使用寿命预测理论方法研究。从运动学和动力学角度出发分析了传动部件的运动规律和承载特性，建立了转速波动、轴承滚动体滑动、转速连续变化等多变工况下轴承和行星轮系的故障仿真模型。基于传动部件的运动几何关系和承载特点，提出了最大损伤冲击信号分离方法，提取了用于滚动轴承故障增强诊断的最大损伤冲击分离信号的包络总体平均特征和自相关总体平均特征。研究了信号增强检测的多点最佳最小熵解卷积方法，提出了集成多点最佳最小熵解卷积、阶比重采样与包络谱分析的轴承故障增强检测方案。设计了基于异常检测器的智能损伤识别方案，以行星轮系为例，分别提取了针对齿圈敏感、针对行星轮敏感、针对太阳轮敏感的分离信号特征，构建了三种损伤的异常检测器以及基于三种检测器的损伤识别模型。研究了模型驱动与数据驱动相结合的剩余使用寿命预测方法，设计了基于退化模型拟合和相关向量机相结合的剩余使用寿命预测算法。开展了变工况下滚动轴承和行星轮系的故障模拟试验，综合应用仿真数据和试验数据验证了上述理论方法的有效性。相关研究成果可为故障预测与健康管理技术体系在装备中的推广应用和状态基维修理念的实施提供技术支撑。