基于电流与振动信息融合的转子系统典型故障诊断方法研究

发布时间：2018-08-09 08:40

【摘要】：随着科技水平的发展,现在工业生产中对旋转机械的安全可靠性方面的要求越来越高,并且转子系统是旋转机械的核心部件,一旦出现故障,就会导致设备丧失工作性能。这些故障会引起连锁反应,导致整个生产系统不能正常工作,造成巨大的经济损失,甚至引发重大的灾难性事故。目前,采用单一信号诊断方法比较片面,故障诊断准确率比较低,已经无法满足需求,因此研究基于电机电流及振动信号融合的故障诊断方法势在必行。本文主要针对转子系统的转子不平衡、偏角不对中、平行不对中这三种典型故障进行诊断,研究了这三种故障的故障机理,建立了系统的故障动力学方程。通过研究方程发现,这三种故障都会引起水平x向和竖直y向振动加速度的变化;并且会引起轴扭矩变化,进而改变电磁扭矩,导致磁通量的变化,最终改变电机电流。接着,设计搭建了转子系统故障诊断试验台,拟定了三种故障的试验方案,并完成了试验研究。首先,通过搭建的试验台采集故障和无故障状态的电流和振动信号,并进行了信号的预处理。对振动信号采用小波软阈值消噪法进行消噪处理,信号时域图上大部分不规则毛刺基本消失,凸显出信号的故障特征。对电流信号采用陷波滤波法进行去工频处理,50Hz工频成分基本被去除,凸显出信号的故障特征。然后,提取了电流信号和振动信号的时域、频域及小波包能量特征。研究发现,频域特征的故障敏感度比时域特征的故障敏感度高,振动信号特征的故障敏感度比电流信号特征的故障敏感度高。考虑到不同传感器测出的信号单位不同,各特征所占比重也不同,对特征向量进行了标准化处理。由于特征向量维数较高,运用主元分析法对特征向量进行了降维处理。最后,设计了用于转子系统典型故障诊断的贝叶斯网络,分别基于振动信号、电机电流信号、融合信息进行了故障诊断。结果表明,融合信息故障诊断准确率最高,振动信号次之,电流信号最差。验证了贝叶斯网络融合信息诊断方法的正确性和有效性。
[Abstract]:With the development of science and technology, the requirements for the safety and reliability of rotating machinery are becoming higher and higher in industrial production, and the rotor system is the core component of rotating machinery. Once failure occurs, the equipment will lose its working performance. These failures will cause chain reaction, cause the whole production system not to work properly, cause huge economic losses, and even lead to major catastrophic accidents. At present, the single signal diagnosis method is one-sided and the accuracy of fault diagnosis is low, so it is imperative to study the fault diagnosis method based on motor current and vibration signal fusion. In this paper, the rotor unbalance, deviation angle misalignment and parallel misalignment of the rotor system are mainly used to diagnose the three typical faults. The fault mechanism of the three faults is studied, and the dynamic equation of the system fault is established. Through the study of the equation, it is found that these three faults will cause the horizontal x and vertical y direction vibration acceleration change, and cause the axis torque change, and then change the electromagnetic torque, lead to the change of magnetic flux, and finally change the motor current. Then, the fault diagnosis test-bed of rotor system is designed and built, three kinds of fault test schemes are worked out, and the experimental research is completed. First, the current and vibration signals of fault and no fault state are collected by the test bed, and the signal preprocessing is carried out. The wavelet soft threshold de-noising method is used for the vibration signal, and most of the irregular burrs in the signal time domain are basically disappeared, which highlights the fault characteristics of the signal. The 50 Hz power frequency component of the current signal is basically removed by notch filtering method, which highlights the fault characteristics of the signal. Then, the time domain, frequency domain and wavelet packet energy characteristics of current signal and vibration signal are extracted. It is found that the fault sensitivity of the frequency domain feature is higher than that of the time domain feature, and the fault sensitivity of the vibration signal feature is higher than that of the current signal feature. Considering that the signal units measured by different sensors are different and the proportion of each feature is different, the characteristic vectors are standardized. Because of the high dimension of the feature vector, the principal component analysis method is used to reduce the dimension of the feature vector. Finally, a Bayesian network for rotor system fault diagnosis is designed, which is based on vibration signal, motor current signal and fusion information respectively. The results show that the fault diagnosis accuracy of fusion information is the highest, the vibration signal is the second, and the current signal is the worst. The correctness and effectiveness of Bayesian network fusion information diagnosis method are verified.
【学位授予单位】：太原理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17

【参考文献】