复杂机电设备微弱特征提取与早期故障诊断方法研究

发布时间：2018-01-13 09:16

本文关键词：复杂机电设备微弱特征提取与早期故障诊断方法研究　出处：《北京工业大学》2011年硕士论文　论文类型：学位论文

【摘要】：在机电设备故障诊断中,由于背景噪声、振动信号传播路径等因素的影响,所测信号中有效故障信息往往被强大的背景噪声所淹没。尤其是在机电设备早期故障中,微弱特征信号完全被噪淹没,从而无法获得设备运行状态信息。本文以机电设备为对象,研究了机电设备微弱特征提取方法和早期故障诊断实用技术。 (1)阐述了差分振子的数学模型和微弱信号检测基本原理,探讨了系统参数对差分振子检测性能的影响及其选取原则,并首次分析了差分振子检测微弱信号的敏感特性和容错特性。指出若增大系统参数,差分振子的收敛速度将会增加,检测敏感性增强,容错特性降低;反之,检测的容错性增强,敏感性降低。利用该方法分析了轧机齿轮和滚动轴承的故障早期数据,成功地提取出了设备早期故障特征,验证了差分振子对设备早期故障微弱特征提取的有效性。 (2)首次提出了利用差分振子相图大小表征微弱信号相对幅值的机理和方法,给出了差分振子相图收敛状态的自动辨识方法。理论分析和仿真实验均证明了在相同的系统参数下,差分振子相图大小与信号幅值存在单调递增的线性比例关系,故可利用相图大小表征信号幅值的相对大小。利用该方法成功揭示了高线轧机轴承故障发生和发展的劣化过程,验证了方法的有效性。 (3)首次分析了差分振子系统参数对检测带宽的影响,提出了基于差分振子阵列的复合频率信号检测方法,实现了复合频率信号无噪声频谱复现。指出小系统参数对应较大的检测带宽,而大系统参数对应较窄的检测带宽,进而利用差分振子相图所表征的频率信息和幅值信息,实现了复杂信号的频谱复现。 (4)首次将差分振子与随机共振理论有机融合,实现了微弱信号的精确检测。针对随机共振频谱中的奇倍频虚假频率现象,利用差分振子的选频特性对随机共振产生的谱峰进行检测以去除虚假频率,实现微弱信号的精确检测。利用该方法提取了棒材轧机故障早期数据中的微弱特征,可望实现设备早期故障诊断。 (5)针对目测观察混沌振子相变的主观性问题,提出利用Hu氏不变矩定量描述和刻画混沌振子相图的敛散性和对称性,实现了混沌振子检测微弱信号时临界阈值和相图状态的自动辨识。高线轧机轴承点蚀故障的实例分析证明了该方法的有效性。上述研究成果在复杂机电设备微弱特征提取中取得了较好的效果,在早期故障诊断领域具有广泛的应用前景。
[Abstract]:In the fault diagnosis of electromechanical equipment, due to background noise, vibration signal propagation path and other factors. The effective fault information of the measured signal is often submerged by strong background noise, especially in the early fault of electromechanical equipment, the weak characteristic signal is completely submerged by noise. In this paper, the weak feature extraction method of electromechanical equipment and the practical technique of early fault diagnosis are studied. 1) the mathematical model of differential oscillator and the basic principle of weak signal detection are expounded, and the influence of system parameters on the detection performance of differential oscillator and its selection principle are discussed. The sensitivity and fault tolerance of differential oscillator for weak signal detection are analyzed for the first time. It is pointed out that if the system parameters are increased, the convergence rate of differential oscillator will increase, the sensitivity of detection will increase, and the fault-tolerant characteristic will decrease. On the other hand, the fault tolerance and sensitivity of the detection are enhanced, and the early fault data of rolling bearing and gear are analyzed by this method, and the early fault characteristics of the equipment are extracted successfully. The effectiveness of differential oscillator for weak feature extraction of equipment early fault is verified. (2) the mechanism and method of using differential oscillator phase diagram to characterize the relative amplitude of weak signal are proposed for the first time. An automatic identification method for the convergent state of the differential oscillator phase diagram is presented. The theoretical analysis and simulation results show that the convergent state of the differential oscillator is under the same system parameters. There is a monotone increasing linear relation between the size of the phase diagram and the amplitude of the signal. Therefore, the relative magnitude of the signal amplitude can be characterized by the size of the phase diagram, and the deterioration process of bearing fault occurrence and development in high wire rolling mill is successfully revealed by using this method, and the validity of the method is verified. The influence of the differential oscillator system parameters on the detection bandwidth is analyzed for the first time, and the detection method of the composite frequency signal based on the differential oscillator array is proposed. The noiseless spectrum reproduction of the composite frequency signal is realized. It is pointed out that the small system parameters correspond to a larger detection bandwidth, while the large system parameters correspond to a narrower detection bandwidth. Furthermore, the frequency and amplitude information represented by the differential oscillator phase diagram are used to realize the spectrum reproduction of the complex signal. 4) the differential oscillator and the stochastic resonance theory are combined for the first time to realize the accurate detection of the weak signal, aiming at the false frequency phenomenon of odd frequency doubling in the stochastic resonance spectrum. The spectral peak generated by stochastic resonance is detected by using the frequency selection characteristic of differential oscillator to remove false frequency and to detect the weak signal accurately. The weak features of the early fault data of bar mill are extracted by this method. It is expected to realize early fault diagnosis of equipment. 5) aiming at the subjective problem of visual observation of chaotic oscillator phase transition, Hu's invariant moment is used to quantitatively describe and depict the convergence and divergence and symmetry of chaotic oscillator phase diagram. The critical threshold and phase diagram state of chaotic oscillator for weak signal detection are automatically identified. The effectiveness of this method is proved by an example of pitting fault of bearings in high wire rolling mill. The above research results have achieved good results in the weak feature extraction of complex electromechanical equipment and have a wide application prospect in the field of early fault diagnosis.
【学位授予单位】：北京工业大学
【学位级别】：硕士
【学位授予年份】：2011
【分类号】：TH165.3

【参考文献】