旋转机械的非线性故障检测
发布时间：2018-07-29 10:04
【摘要】：旋转机械运行状态的好坏,会直接影响系统的工作性能。本文对旋转机械的故障检测技术和方法进行研究,针对振动信号存在的非线性特性,研究了非线性评价指标;探讨了信号分解对降低非线性程度的影响,及降低非线性程度的方法;综合运用信号分解、时间序列建模、隐马尔科夫等理论,构建故障检测模型,对旋转机械故障做出精确判断,为确保旋转机械正常工作具有重要意义。主要研究内容如下:研究了振动信号的非线性特性,确定了嵌入维数、延迟时间两个重要参数。采用混沌与分形理论,对非线性评价指标进行研究,给出最大Lyapunov指数、柯尔莫哥洛夫熵、关联维数、盒维数计算方法。采用信号分解手段降低非线性程度,比较了信号经小波分解和集成经验模态分解后的非线性度强弱,提出二者结合的振动信号去噪方法。构建振动信号故障检测模型,采用时间序列建模的方法,精确提取能够表征故障的特征。振动信号经过集成经验模态分解后,计算混沌与分形的数值特征,根据非线性强弱评价指标,判断分解信号的非线性程度。针对分解后的线性分量,建立线性模型,提取线性模型参数;针对分解后的非线性分量,构建Volterra模型,提取Volterra模型参数。在深入研究非线性特征提取的基础上,探讨了HMM技术的实现方法,提出采用H MM模型进行故障识别。对旋转机械的轴承信号进行实验分析,将信号分解后提取的线性、非线性特征量输入到HMM模型中,对正常、内环故障、外环故障、滚动体故障这四种信号进行模式识别,实验结果表明该模型能够准确识别旋转机械故障,且识别率高。
[Abstract]:The running state of rotating machinery will directly affect the working performance of the system. In this paper, the fault detection techniques and methods of rotating machinery are studied, and the nonlinear evaluation index of vibration signal is studied, and the influence of signal decomposition on the reduction of nonlinear degree is discussed. And the methods of reducing nonlinear degree, synthetically using the theory of signal decomposition, time series modeling, hidden Markov and so on, to construct the fault detection model, and to make accurate judgment on the fault of rotating machinery. In order to ensure the normal operation of rotating machinery has important significance. The main contents are as follows: the nonlinear characteristics of vibration signal are studied, and two important parameters, embedding dimension and delay time, are determined. The nonlinear evaluation index is studied by using chaos and fractal theory. The calculation methods of maximum Lyapunov exponent, Kolmogorov entropy, correlation dimension and box dimension are given. Using signal decomposition to reduce the degree of nonlinearity, the degree of nonlinearity after wavelet decomposition and integrated empirical mode decomposition is compared, and a method of vibration signal denoising is proposed. The fault detection model of vibration signal is constructed, and the time series modeling method is used to accurately extract the characteristics that can represent the fault. After integrated empirical mode decomposition, the numerical characteristics of chaos and fractal are calculated, and the degree of nonlinearity of the decomposed signal is judged according to the evaluation index of nonlinear intensity. For the decomposed linear component, the linear model is established to extract the parameters of the linear model, and for the decomposed nonlinear component, the Volterra model is constructed to extract the parameters of the Volterra model. Based on the research of nonlinear feature extraction, the realization method of HMM technology is discussed, and the method of fault identification based on hmm model is proposed. Through the experimental analysis of bearing signals of rotating machinery, the extracted linear and nonlinear eigenvalues are input into the HMM model after signal decomposition, and the four signals, namely normal, inner, outer and rolling faults, are recognized by pattern recognition. The experimental results show that the model can accurately identify the rotating machinery faults, and the recognition rate is high.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17
[Abstract]:The running state of rotating machinery will directly affect the working performance of the system. In this paper, the fault detection techniques and methods of rotating machinery are studied, and the nonlinear evaluation index of vibration signal is studied, and the influence of signal decomposition on the reduction of nonlinear degree is discussed. And the methods of reducing nonlinear degree, synthetically using the theory of signal decomposition, time series modeling, hidden Markov and so on, to construct the fault detection model, and to make accurate judgment on the fault of rotating machinery. In order to ensure the normal operation of rotating machinery has important significance. The main contents are as follows: the nonlinear characteristics of vibration signal are studied, and two important parameters, embedding dimension and delay time, are determined. The nonlinear evaluation index is studied by using chaos and fractal theory. The calculation methods of maximum Lyapunov exponent, Kolmogorov entropy, correlation dimension and box dimension are given. Using signal decomposition to reduce the degree of nonlinearity, the degree of nonlinearity after wavelet decomposition and integrated empirical mode decomposition is compared, and a method of vibration signal denoising is proposed. The fault detection model of vibration signal is constructed, and the time series modeling method is used to accurately extract the characteristics that can represent the fault. After integrated empirical mode decomposition, the numerical characteristics of chaos and fractal are calculated, and the degree of nonlinearity of the decomposed signal is judged according to the evaluation index of nonlinear intensity. For the decomposed linear component, the linear model is established to extract the parameters of the linear model, and for the decomposed nonlinear component, the Volterra model is constructed to extract the parameters of the Volterra model. Based on the research of nonlinear feature extraction, the realization method of HMM technology is discussed, and the method of fault identification based on hmm model is proposed. Through the experimental analysis of bearing signals of rotating machinery, the extracted linear and nonlinear eigenvalues are input into the HMM model after signal decomposition, and the four signals, namely normal, inner, outer and rolling faults, are recognized by pattern recognition. The experimental results show that the model can accurately identify the rotating machinery faults, and the recognition rate is high.
【学位授予单位】：天津理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17
【参考文献】
相关期刊论文 前6条
1 张永宏;陶润U，
本文编号：2152329
本文链接：https://www.wllwen.com/jixiegongchenglunwen/2152329.html
