盲信号分离算法及其在转子故障信号分离中的应用方法研究

发布时间：2018-11-13 14:44

【摘要】：在旋转机械设备状态监测和故障诊断研究中,故障的特征提取和模式识别关系到故障诊断的可靠性和准确性,也是旋转机械故障诊断研究中的关键问题。利用转子振动信号对其进行状态监测和诊断是目前旋转机械故障监测和诊断研究中常用的方法。本论文以丰富和提高机械故障诊断理论与方法为目的,用现代信号处理技术中的盲信号分离方法为工具,以机械设备中应用最广泛的旋转机械设备为研究对象,利用盲信号分离算法、中值滤波和盲信号分离相结合的方法、自适应粒子群优化的盲信号分离方法、降噪源分离方法等信号处理方法,对转子系统故障特征提取问题开展了研究工作。具体研究内容如下： (1)针对噪声干扰下的旋转机械故障特征提取问题,提出一种基于二阶盲辨识的去除干扰噪声方法。该方法利用旋转机械振动信号的非平稳性特征,将采集到的信号分成不重叠的时间窗,然后对每个时间窗内的时滞方差平均值进行估计,从而实现噪声信号与源信号的分离。这里将盲信号分离理论应用于消噪处理,其关键是分离噪声,而不是滤除噪声,因此在分离噪声时不丢失有效信号,为消噪处理提供了一种新方法。此方法通过仿真和对实际转子振动数据的处理表明,该算法可有效地分离出干扰噪声,提高采样信号的准确性。 (2)针对非线性机械故障信号分离依赖于非线性函数的选取问题,提出一种基于自适应粒子群优化的机械故障特征提取方法。该方法将采样信号的负熵做为目标函数,然后引入自适应粒子群优化的概念,通过信号的状态自适应的调整惯性因子,使其负熵最大化,从而实现各振源信号的有效分离。仿真和试验结果表明,该方法提高了分离信号的相关系数,实现了各源信号的有效分离。 (3)提出了基于降噪源分离的旋转机械故障特征提取方法。该方法是根据旋转机械振动信号的统计特征,构造降噪函数,依据降噪函数实现各分量的分离。在对仿真故障信号实验的基础上,定量比较了四种降噪函数的性能,发现基于正切降噪函数的分离结果相似系数最好,更适于混叠故障信号的分离。将基于正切降噪函数的源分离方法应用于旋转机械故障特征提取中,分析结果表明,该方法很好地从转子混叠振动信号中分离出了转子由碰摩故障引起的转子不平衡和不对中故障。 (4)针对源信号分离算法对强脉冲噪声环境下的混叠振动信号分离的失效,构建了一种基于中值滤波和盲信号分离算法相结合的方法。该方法首先通过中值滤波降噪方法对振动信号进行降噪处理,然后通过盲信号分离算法对降噪后的混叠信号进行分离。仿真和实验结果表明：在强脉冲噪声干扰下,若直接采用盲信号分离算法进行分离,其分离效果并不理想,若利用中值消噪和盲信号分离算法相结合的方法,则分离效果得到明显提升。
[Abstract]:In the research of condition monitoring and fault diagnosis of rotating machinery, fault feature extraction and pattern recognition are related to the reliability and accuracy of fault diagnosis, and are also the key problems in the research of rotating machinery fault diagnosis. It is a common method to monitor and diagnose the rotor vibration signal in the fault monitoring and diagnosis of rotating machinery at present. The purpose of this paper is to enrich and improve the theory and method of mechanical fault diagnosis, to use the blind signal separation method in modern signal processing technology as a tool, and to study the most widely used rotating machinery in mechanical equipment. Using blind signal separation algorithm, median filter and blind signal separation method, adaptive particle swarm optimization blind signal separation method, noise reduction source separation method and other signal processing methods, The fault feature extraction of rotor system is studied. The main contents are as follows: (1) aiming at the problem of fault feature extraction of rotating machinery under noise interference, a noise removal method based on second-order blind identification is proposed. Based on the non-stationary characteristic of vibration signals of rotating machinery, the collected signals are divided into non-overlapping time windows, and then the time-delay mean variance in each time window is estimated, so that the noise signal is separated from the source signal. In this paper, the blind signal separation theory is applied to de-noise processing. The key point is to separate noise, not to filter noise. Therefore, effective signals are not lost when noise is separated, which provides a new method for de-noising processing. The simulation and processing of the actual rotor vibration data show that the proposed method can effectively separate out the interference noise and improve the accuracy of the sampling signal. (2) aiming at the problem that the separation of nonlinear mechanical fault signals depends on the nonlinear function, an adaptive particle swarm optimization (APSO) based method for mechanical fault feature extraction is proposed. In this method, the negative entropy of the sampled signal is taken as the objective function, and then the concept of adaptive particle swarm optimization is introduced. The state of the signal adjusts the inertia factor adaptively to maximize the negative entropy of the signal so as to realize the effective separation of the signals. The simulation and experimental results show that the method improves the correlation coefficient of the separation signal and realizes the effective separation of each source signal. (3) A fault feature extraction method for rotating machinery based on noise reduction source separation is proposed. According to the statistical characteristics of vibration signals of rotating machinery, the noise reduction function is constructed, and the separation of components is realized according to the noise reduction function. On the basis of simulation fault signal experiments, the performance of four noise reduction functions is quantitatively compared. It is found that the separation result based on tangent denoising function has the best similarity coefficient and is more suitable for separating aliasing fault signals. The source separation method based on tangent denoising function is applied to the fault feature extraction of rotating machinery. The analysis results show that, The rotor unbalance and misalignment caused by rub-impact fault are well separated from the rotor mixed vibration signal by this method. (4) in view of the failure of source signal separation algorithm for aliasing vibration signal separation under strong impulse noise, a method based on median filter and blind signal separation algorithm is proposed. Firstly, the vibration signal is de-noised by median filtering method, and then the aliasing signal is separated by blind signal separation algorithm. The simulation and experimental results show that the separation effect is not satisfactory if the blind signal separation algorithm is used directly under the strong impulse noise interference, and if the median de-noising algorithm is combined with the blind signal separation algorithm, The separation effect was obviously improved.
【学位授予单位】：兰州理工大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TH165.3

【参考文献】