融合疲劳现象学与奇异谱分解的起重机损伤识别及寿命预测研究

The estimation model of crack propagation corrected by multiple factors for random load is established based on fatigue phenomenological theory by considering the residual stress redistribution, the structural factors and the load effect. The quantitative prediction research of crack propagation for crane structure is studied, according to the screening model of main crack means to determine the most harmful crack under crack coexist circumstances. The optimization singular spectral decomposition method combining with adaptive boundary continuation, nonlinear minimization optimization and optimal iterative control is proposed, and the state parameters of crack propagation is extracted according to the signal characteristic of strain, vibration and acoustic emission, based on which the crack propagation mechanism model is modified by the weighting coefficient. At the same time, the damage of crane structure is detected according to the sensitive characteristic extracted by the optimization singular spectral decomposition method. The state parameter of crack propagation is predicted by using the mechanism model, the optimization method of singular spectral decomposition and bayesian intelligent algorithm based on the state parameters of crane operation. Further, the assessment method of running condition for crane based fusion of crack propagation is established, and the injury classification strategy are given by using the signal characteristics extracted according to the theory of the singular spectrum decomposition and the fatigue limit stress characteristic Smith - Watson - Topper (SWT), based on which the the remaining life of crane is predicted by using the algorithm of best stress concentration factor.

基于疲劳现象学理论综合考虑残余应力重分布、结构因素及载荷对裂纹扩展的影响，建立随机载荷下多因素修正裂纹扩展机理模型，提出多裂纹同时存在时确定危害最大裂纹的主干裂纹筛选模型，实现起重机结构裂纹扩展定量预测研究。提出自适应边界延拓、非线性最小化优化及最优迭代控制等的优化奇异谱分解方法，提取应变/振动/声发射信号中裂纹扩展状态特征参数，采用加权系数对初始裂纹扩展机理模型修正，同时根据优化奇异谱分解提取的敏感特征量对起重机结构损伤进行检测。根据实测起重机运行状态参数，运用机理模型与优化奇异谱分解及贝叶斯智能算法对裂纹扩展的状态参数进行预测，提出基于裂纹扩展信息融合的起重机运行状态评估方法，同时将基于优化奇异谱分解理论提取的特征与Smith-Watson-Topper (SWT)疲劳应力极限特征相结合，给出损伤分类策略，根据不同损伤阶段运用最佳应力集中因子算法对起重机剩余寿命进行预测。

起重机服役过程中受到疲劳、磨损、冲击等因素影响，不可避免地发生性能退化、结构损伤、故障甚至失效等。因此，对其进行损伤识别及关键参数的在线监测，并在此基础上进行故障诊断和剩余寿命预测是目前工程机械智能维护急需解决的关键难题，也是学术界研究的热点，具有重大的理论意义和实用价值。在面上基金项目的资助下，本项目紧密围绕起重机及其关键基础部件的裂纹扩展，损伤特征提取，以及剩余寿命预测，开展了一系列理论及工程应用研究。研究成果主要分成以下三部分：1）综合考虑多因素对裂纹扩展的影响，建立横幅、随机载荷历程下裂纹扩展模型，提出多裂纹同时存在情形下确定危害最大裂纹的方法即主干裂纹筛选模型，实现起重机结构裂纹扩展的定量预测；2）将奇异谱分解（SSD）引入起重机主梁损伤特征提取中，深入研究其分解机理及其重构算法，提出改进SSD算法提取复杂工况下起重机振动数据的敏感损伤特征，同时基于自编码算法实现起重机及其基础部件损伤及故障的精确诊断；3）研制了基于Web端及Android/IOS平台的起重机运行状态参数的在线监测系统。基于提出的裂纹扩展模型、SSD、贝叶斯方法，进一步建立基于裂纹扩展信息融合的起重机运行状态评估方法，同时提出一种基于深度学习的机械健康指标，实现起重机关键部件性能退化的准确评估，并根据Bootstrap方法获取剩余寿命的置信区间从而对剩余寿命进行精确的区间预测。4）提出了基于深度学习的起重机关键基础部件全寿命健康指标与剩余寿命预测方法。研究成果发表在Mechanical Systems and Signal Processing, IEEE/ASME Transactions on Mechatronics等国际权威期刊，其中发表SCI论文24篇。授权相关发明专利6件，转让1件，软件著作权1件，具有显著工程应用价值。