基于隐马尔可夫模型与信息融合的设备故障诊断与性能退化评估研究

发布时间：2018-07-05 19:53

本文选题：状态监测 + 轴承　；参考：《上海交通大学》2014年博士论文

【摘要】：随着科学技术的进步和生产效率的提高，机械设备不断向高速、高精度、重载和高可靠性的方向发展，设备的结构也日趋复杂化。在生产过程中，机械故障不但影响工作效率，并且可能引起严重的安全问题。由于机械设备的性能和状态在使用过程中总会随着持续运行而逐渐恶化，因此开展机械设备故障诊断技术的研究对于维护设备安全，提高生产的效率和可靠性具有重要意义。轴承和齿轮作为机械设备的关键零部件之一，其工作状态的好坏严重影响着设备性能的变化。因此对轴承和齿轮的故障诊断和性能退化评估一直是设备故障诊断的研究重点。本文在分析轴承故障机理的基础上，，提出了基于频带熵的自适应滤波器方法并用于轴承微弱故障的特征提取。由于设备在工作过程中总会经历由正常到退化到最终失效的过程，如果能够获得设备的实时健康信息，对于维护策略的制定、降低维护成本和生产损失有着积极的意义。本文利用耦合隐马尔可夫模型的多通道信息融合能力，深入讨论了耦合隐马尔可夫模型在轴承故障诊断和性能退化评估中的应用。主要包括以下几个方面的内容： (1)结合机械设备状态监测的理论基础和实际工程应用需求，阐述了论文选题的背景和研究意义。回顾和分析了国内外在轴承特征提取、信息融合、故障诊断和性能退化评估与预测方法的研究热点和现状，确立了本文的研究内容和技术框架。 (2)介绍了滚动轴承的结构和运动特征，通过轴承的点蚀故障模型说明了轴承故障原理和各个特征频率计算方法。利用滚动轴承特征频率调制的规律，结合振动信号的时频分布特点和信息熵理论，提出了一种基于频带熵的自适应滤波器设计方法来提取滚动轴承的微弱故障信号。 (3)介绍了轴承故障诊断中常用的时域和频域指标以及特征约减算法在故障诊断中的应用。给出了一种使用正交基的局部保持投影降维方法，研究了如何利用类内类间距离指标来优化邻接图构造参数的选择。 (4)介绍了马尔可夫链和隐马尔可夫算法的基本概念和算法，讨论了隐马尔可夫模型的评估问题、解码问题和学习问题及基本算法。通过试验证明了特征约减和隐马尔可夫模型在轴承故障诊断中的可行性和有效性。 (5)针对单链隐马尔可夫模型在多通道数据故障诊断中的局限性，讨论了基于LPP的特征层信息融合与隐马尔可夫模型结合的诊断方法、基于隐马尔可夫模型和D-S的决策层信息融合方法以及耦合隐马尔可夫模型的多通道信息融合方法在轴承故障诊断中的应用。研究了耦合隐马尔可夫模型的概率推导和故障诊断建模算法，利用轴承人工故障试验证明了基于耦合隐马尔可夫模型的故障诊断能获得更好的诊断精度。通过与其余特征提取和诊断方法的比较，进一步证明了耦合隐马尔可夫模型的诊断准确性。 (6)研究了利用耦合隐马尔可夫模型对多通道监测数据进行性能退化评估建模和性能指标计算的方法。利用性能指标给出了自适应报警限的计算方法。最后对齿轮自然失效试验数据、滚动轴承加速疲劳试验数据和滚柱轴承自然疲劳试验数据的分析验证了耦合隐马尔可夫模型对完备和非完备数据进行性能退化评估的有效性，结果证明了所选的性能指标能够定量的反映出轴承的性能退化程度。
[Abstract]:With the progress of science and technology and the improvement of production efficiency, mechanical equipment is constantly developing towards high speed, high precision, heavy load and high reliability, and the structure of equipment is becoming more and more complex. In the process of production, mechanical failures not only affect the efficiency of the work, but also cause serious safety problems. Because of the performance and state of the mechanical equipment, In the process of use, it will always deteriorate with the continuous operation. Therefore, it is of great significance to carry out the research on the fault diagnosis technology of mechanical equipment to maintain the safety of the equipment and improve the efficiency and reliability of production.
As one of the key parts of the mechanical equipment, bearing and gear have a serious influence on the change of equipment performance. Therefore, the fault diagnosis and performance degradation assessment of bearing and gear are always the focus of research on equipment fault diagnosis. On the basis of analyzing the mechanism of bearing fault, this paper puts forward the self - frequency entropy based on the frequency band. The adaptive filter method is used to extract the characteristics of the weak fault of the bearing. Because the equipment will always experience the process from normal to final failure during the working process, it has positive significance for the maintenance strategy formulation and the reduction of the maintenance cost and the loss of production if the equipment is able to obtain the real-time health information. The multi-channel information fusion ability of Markov model is used to discuss the application of the coupled hidden Markov model in bearing fault diagnosis and performance degradation evaluation.
(1) combining the theoretical basis of state monitoring of mechanical equipment and the requirement of practical engineering application, the background and research significance of the thesis are expounded. The research heat point and current situation of bearing feature extraction, information fusion, fault diagnosis and performance degradation assessment and prediction methods are reviewed and analyzed, and the research content and technical frame of this paper are established. Frame.
(2) the structure and motion characteristics of the rolling bearing are introduced. The principle of bearing fault and the calculation method of each characteristic frequency are explained by the fault model of the pitting corrosion of the bearing. A adaptive filter based on the frequency band entropy is proposed, which is based on the characteristic frequency modulation of the rolling bearing and the time frequency distribution characteristics of the vibration signal and the information entropy theory. The design method is used to extract the weak fault signals of rolling bearings.
(3) the application of time domain and frequency domain index in fault diagnosis of bearing and the application of feature reduction algorithm in fault diagnosis are introduced. A local maintenance projection reduction method using orthogonal basis is given, and how to optimize the selection of construction parameters of adjacency graph by using intra class distance index is studied.
(4) the basic concepts and algorithms of Markov chain and hidden Markov algorithm are introduced. The evaluation of hidden Markov models, decoding problems, learning problems and basic algorithms are discussed. The feasibility and effectiveness of feature reduction and hidden Markov model in bearing fault diagnosis are proved by experiments.
(5) in view of the limitation of single chain hidden Markov model in multi channel data fault diagnosis, the diagnosis method based on LPP based feature layer information fusion and hidden Markov model is discussed. The method of information fusion based on Hidden Markov model and D-S decision layer and multi channel information fusion method of coupled hidden Markov model are discussed. The application of the bearing fault diagnosis is studied. The probability deduction and fault diagnosis modeling algorithm of the coupled hidden Markov model are studied. The fault diagnosis based on the coupled hidden Markov model can obtain better diagnostic accuracy by using the bearing artificial fault test. The comparison with the other features and the diagnosis methods is further proved. The diagnostic accuracy of the coupled hidden Markov model (HMM).
(6) the method of performance degradation assessment modeling and performance index calculation of multi-channel monitoring data using coupled hidden Markov model is studied. The calculation method of adaptive alarm limit is given by using performance index. Finally, the natural fatigue test data of gear, the data of rolling bearing accelerated fatigue test and the natural fatigue test of roller bearing are given. The analysis of the experimental data validates the effectiveness of the coupled hidden Markov model (HMM) for performance degradation assessment of complete and incomplete data. The results show that the selected performance indicators can quantitatively reflect the performance degradation of the bearings.
【学位授予单位】：上海交通大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH165.3;TP202

【参考文献】