基于时频分析的旋转机械故障诊断方法研究与应用

发布时间：2019-03-14 08:37

【摘要】：旋转机械是生产领域中十分重要的机械设备,由于旋转机械激励源多、性质复杂,其振动信号往往是非平稳的多分量信号,其不同的非平稳特性往往对应不同的机械故障。为了更好的进行旋转机械的故障诊断,有必要对信号的时频分析方法进行研究与应用。本文根据以上的需求开展研究工作,基于维格纳-威尔分布、小波尺度谱、Hilbert时频谱等时频分析方法,结合盲分离、同步平均技术和多尺度熵方法,对旋转机械的故障诊断方法进行了较为深入的研究和应用。为了实现对旋转机械振动信号中独立源信号的提取,研究了一种基于时频分析的盲分离方法,从而更加准确的进行设备故障诊断。首先在理论上推导了该方法的实现过程；采用仿真信号分析了基于不同时频分布的盲分离的效果,并与独立分量分析的结果进行了对比；最后在转子实验台上模拟不平衡、不对中和基座松动三种故障,采用该方法较好的实现了三种故障的识别与诊断。针对旋转机械振动信号中往往夹杂着噪声干扰和具有循环平稳性的特点,结合时域同步平均可降低噪声干扰的优点和小波变换可对信号进行多尺度分析的优势,提出一种基于小波重排尺度谱的同步平均的信号分析方法。首先对信号进行连续小波变换并进行重排处理,然后对各个尺度上的信号进行时域同步平均,获得平均后的小波重排尺度谱。通过仿真分析和滚动轴承故障模拟实验检验该方法的有效性。旋转机械的Hilbert时频谱含有大量的机械设备工作状态的特征信息,然而其特征往往难于辨识,而多尺度熵可以有效的描述序列的复杂度,提出了一种基于Hilbert时频谱特征提取的设备状态识别方法。首先对信号进行希尔伯特-黄变换获得Hilbert时频谱,然后对时频谱进行区域划分和降至一维并求其多尺度熵,通过对比设备不同运行状态下的Hilbert时频谱的多尺度熵曲线,选择有效分离不同设备状态的尺度处的样本熵和时频谱的能量作为其特征向量用于设备状态识别。采用本方法对不同轴承故障状态的信号进行了特征提取,实现了轴承状态的有效识别。基于虚拟仪器开发了旋转机械振动测试与分析系统。该系统可以实现8通道的振动信号采集,并通过无线数据传输模式将测试数据传输给上位机,进行显示、分析和存储。具有常见的时域、频域等分析功能和Hilbert时频谱等时频分析方法。通过实际应用验证了系统的实用性和有效性。
[Abstract]:Rotating machinery is a very important mechanical equipment in the field of production. Because of its many excitation sources and complex properties, its vibration signal is usually a non-stationary multi-component signal, and its different non-stationary characteristics often correspond to different mechanical faults. In order to make better fault diagnosis of rotating machinery, it is necessary to study and apply the time-frequency analysis method of signal. Based on Wigner-Weill distribution, wavelet scale spectrum and Hilbert time-frequency analysis method, this paper combines blind separation, synchronous averaging technique and multi-scale entropy method. The fault diagnosis method of rotating machinery is studied and applied deeply. In order to extract the independent source signal from the vibration signal of rotating machinery, a blind separation method based on time-frequency analysis is studied, so that the equipment fault diagnosis can be carried out more accurately. Firstly, the realization process of this method is deduced theoretically, and the effect of blind separation based on different time-frequency distributions is analyzed by simulation signals, and the results are compared with the results of independent component analysis (ICA). At last, the unbalance is simulated on the rotor experimental platform, and the three faults of loose base are not neutralized. The method is used to realize the identification and diagnosis of the three kinds of faults. In view of the characteristics of noise interference and cyclic stationarity in vibration signals of rotating machinery, combined with the advantages of synchronous average in time domain and multi-scale analysis of signals by wavelet transform, the noise interference can be reduced in time domain, and the advantages of wavelet transform can be used in multi-scale analysis of signals. A synchronous average signal analysis method based on wavelet rearrangement scale spectrum is proposed. Firstly, continuous wavelet transform and rearrangement of the signal are carried out, and then the signal of each scale is averaged synchronously in time domain, and the average wavelet rearrangement scale spectrum is obtained. The effectiveness of the method is verified by simulation analysis and rolling bearing fault simulation experiment. The Hilbert spectrum of rotating machinery contains a large number of characteristic information of the working state of mechanical equipment, however, its characteristics are often difficult to identify, and multi-scale entropy can effectively describe the complexity of the sequence. In this paper, a device state recognition method based on Hilbert time spectrum feature extraction is proposed. Firstly, the Hilbert time spectrum is obtained by Hilbert-Huang transform, then the time spectrum is partitioned and reduced to one dimension and its multi-scale entropy is calculated. By comparing the multi-scale entropy curves of the Hilbert time spectrum under different operating states of the equipment, The entropy of samples and the energy of time spectrum at the scale of different equipment states are selected as their Eigenvectors for equipment state identification. Using this method, the signals of different bearing fault states are extracted, and the effective identification of bearing states is realized. The vibration test and analysis system of rotating machinery is developed based on virtual instrument. The system can realize 8-channel vibration signal acquisition and transmit the test data to the upper computer through wireless data transmission mode for display, analysis and storage. It has common time-frequency analysis functions such as time-domain and frequency-domain, and time-frequency analysis methods such as Hilbert time-frequency spectrum. The practicability and effectiveness of the system are verified by practical application.
【学位授予单位】：大连理工大学
【学位级别】：硕士
【学位授予年份】：2012
【分类号】：TH165.3

【参考文献】