机械非线性多故障模式多源动态特征辨识及自适应诊断研究

Multi-source dynamic feature extraction and recognition for the mechanical nonlinear multi-fault mode are the difficult technique and problem that the fault diagnosis is applied in the industrial process. It not only extracts the time-frequency features from the single signal, but also ensures the corresponding signal-frequency-space relations among the nonlinear variables, multi-fault modes and the faulty features from the multiple sources. The research is characteristic of the theoretical innovation and valuable applications with the challenges. In terms of the urgent requirements by the condition monitoring and fault diagnosis of the automatic industrial process, the research on the multi-source dynamic feature recognition for the nonlinear multi-fault mode and adaptive diagnosis is done. The mechanism about the generation of the dynamic features in the multiple sources under the nonlinear multi-fault modes is analyzed. The research on the theory about the multi-scale parallel factor decomposition is done. The optimization algorithm on the feature factor selection is developed. The analysis precision of the three-dimensional signal-frequency-space feature signal is improved. The research on the optimization algorithm about the decomposition routine of the feature factors with the characteristic correspondence and consistency is done to solve the confliction between the higher efficiency and higher precision. The reconstruction algorithm of the three-dimensional model about the multi-source feature factors is proposed. The research on the algorithm and theory of the multi-dimensional dynamic feature extraction, recognition and decision making, as well as the methodology of the adaptive diagnosis is beneficial for the applications of the mechanical fault diagnosis during the industrial process and improvement in the precision and intelligent capability of the mechanical fault diagnosis.

机械非线性多故障模式多源动态特征辨识是故障诊断在流程工业生产线中应用所遇到的技术瓶颈和难题，它不仅提取单源故障信号时频特征, 而且要确保特征提取后非线性变量及多故障模式和多源故障特征在时间、频率和空间上的对应关系, 它的研究具有理论创新和实用价值及挑战性。针对全自动流程工业生产线安全监控与故障诊断的迫切需求，开展机械非线性多故障模式多源动态特征辨识及自适应诊断研究。分析非线性多故障模式下多源信号动态特征形成机理；研究多尺度平行因子分解理论，发展特征因子筛选优化算法，提高三维时频空特征信号分析精度；研究特征因子分解路径具有对应性和整体一致性的优化算法，调和特征辨识高效高精度与计算复杂性之间的矛盾；提出多源特征因子的三维时频空模型重构算法；研究基于多维动态特征提取辨识和决策算法理论的自适应诊断方法，促进机械故障诊断在流程工业生产过程中的应用，提高机械故障检测精准度和智能化水平。

本项目主要研究了机械非线性多故障模式多源动态特征辨识及自适应诊断。针对机械非线性多故障模式多源动态特征辨识及自适应诊断中存在的一些基础性和共性科学问题,进行了理论创新和实验论证研究,按计划完成了项目任务,主要研究成果为: 分析了非线性多故障模式下多源信号动态特征形成机理; 研究和提出了面向非线性故障模式下复杂机械系统动态特征的多尺度平行因子分解理论，提出了特征因子筛选优化算法和方法；研究了特征信号分解路径对应性和整体一致性理论,提出复杂机械系统特征因子传播路径的识别方法和优化理论算法,发展了基于非线性系统识别理论的特征因子三维时频空模型特征；研究了基于多维动态特征提取辨识和决策算法理论的自适应诊断方法，本项目提出的机械非线性多故障模式自适应诊断算法已在流程过程机械诊断、石油管道等相关领域安全检测中取得了应用，提高了机械故障检测精准度和智能化水平。