基于时间序列标度分析的旋转机械故障诊断方法研究

发布时间：2018-09-04 07:18

【摘要】：机械故障诊断的关键问题是故障特征提取。机械故障信号通常具有强烈的非平稳和非线性特征，本文在总结现有机械故障诊断方法优缺点的基础上，采用统计物理学上的标度分析方法来研究复杂机械故障信号的波动状况，提出了基于时间序列标度分析的旋转机械故障诊断方法。本文从一个新的角度来研究机械故障诊断问题，形成了具有学科交叉特色的机械故障诊断方法。本文的研究工作主要包括以下六个部分： (1)受自然界大量存在的标度曲线转折现象的启发，本文将时间序列的多标度指数作为机械故障信号的故障特征，提出了基于时间序列多标度指数特征的机械故障特征提取方法。利用齿轮箱和滚动轴承故障数据对该方法的性能进行了验证，，结果证明了该方法的有效性。 (2)针对原始序列标度曲线的特征参数难以提取的问题，本文采用增量序列的波动特征来表达机械系统的动力学行为，提出了基于增量序列标度特征的机械故障诊断方法。利用齿轮箱和滚动轴承故障数据对该方法的性能进行了验证，结果证明了该方法的有效性。 (3)通过分析从增量序列标度曲线上提取的数据点在坐标图上的分布特征，本文发现故障状态所对应的数据点可以近似拟合为一条直线，而正常状态所对应的数据点则明显地偏离这条直线。为了描述这种有趣的现象，提出了“故障线”的概念，随后探讨了“故障线”的成因。 (4)针对齿轮箱故障信号所具有的非平稳和非线性特点，本文提出了基于时间序列多重分形特征的齿轮箱故障特征提取方法。该方法采用MF-DFA计算齿轮箱故障信号的多重分形谱，然后利用从多重分形谱上提取的特征参数对齿轮箱进行故障诊断。利用齿轮箱故障数据对该方法的性能进行了验证，结果证明了该方法的有效性。 (5)为了解决滚动轴承的故障类型和损伤程度难以识别的问题，本文提出了基于MF-DFA和马氏距离判别法的滚动轴承故障诊断方法。该方法利用MF-DFA计算轴承故障信号的多重分形谱，从多重分形谱上提取特征参数，然后采用马氏距离判别法对这些特征参数进行分类。利用滚动轴承故障数据对该方法的性能进行了验证，结果证明了该方法的有效性。 (6)本文对旋转机械振动信号出现多重分形的原因进行了研究。通过比较原始数据及其重排数据和替代数据的广义Hurst指数曲线，本文确定了数据波动的内在长程相关性是导致齿轮箱和滚动轴承振动信号出现多重分形的主要原因。
[Abstract]:The key problem of mechanical fault diagnosis is fault feature extraction. Mechanical fault signals usually have strong nonstationary and nonlinear characteristics. This paper summarizes the advantages and disadvantages of existing mechanical fault diagnosis methods. In this paper, the scale analysis method in statistical physics is used to study the fluctuation of complex mechanical fault signals, and a fault diagnosis method for rotating machinery based on time series scale analysis is proposed. In this paper, the problem of mechanical fault diagnosis is studied from a new point of view, and a method of mechanical fault diagnosis with interdisciplinary characteristics is formed. The research work of this paper mainly includes the following six parts: (1) inspired by the phenomenon of scale curve turning in nature, the multi-scale exponent of time series is regarded as the fault feature of mechanical fault signal. A method of mechanical fault feature extraction based on multi-scale exponential feature of time series is proposed. The performance of the method is verified by using the gearbox and rolling bearing fault data. The results show that the method is effective. (2) the characteristic parameters of the original series scale curve are difficult to extract. In this paper, the dynamic behavior of mechanical systems is expressed by the fluctuation characteristics of incremental sequences, and a mechanical fault diagnosis method based on the scaling features of incremental sequences is proposed. The performance of the method is verified by using gearbox and rolling bearing fault data. The results show that the method is effective. (3) the distribution characteristics of the data points extracted from the incremental series scale curve on the coordinate diagram are analyzed. In this paper, it is found that the data points corresponding to the fault state can be approximately fitted into a straight line, while the data points corresponding to the normal state deviate from the line obviously. In order to describe this interesting phenomenon, the concept of "fault line" is proposed, and the causes of "fault line" are discussed. (4) aiming at the non-stationary and nonlinear characteristics of gearbox fault signal, In this paper, a method of gearbox fault feature extraction based on time series multifractal feature is proposed. This method uses MF-DFA to calculate the multifractal spectrum of the gearbox fault signal, and then uses the characteristic parameters extracted from the multifractal spectrum to diagnose the gearbox fault. The performance of the method is verified by using gearbox fault data. The results show that the method is effective. (5) in order to solve the problem that the fault type and damage degree of rolling bearing are difficult to identify, This paper presents a fault diagnosis method for rolling bearings based on MF-DFA and Markov distance discrimination. In this method, the multifractal spectrum of bearing fault signal is calculated by MF-DFA, and the characteristic parameters are extracted from the multifractal spectrum, and then these parameters are classified by Markov distance discriminant method. The performance of the method is verified by the rolling bearing fault data. The results show that the method is effective. (6) the reason of multifractal of vibration signal of rotating machinery is studied in this paper. By comparing the generalized Hurst exponent curves of raw data, rearrangement data and substitute data, this paper determines that the inherent long range correlation of data fluctuation is the main cause of multifractal of vibration signals of gearbox and rolling bearing.
【学位授予单位】：南京航空航天大学
【学位级别】：博士
【学位授予年份】：2013
【分类号】：TH165.3

【参考文献】