基于噪声的旋转机械故障诊断研究

发布时间：2018-05-17 13:09

本文选题：旋转机械 + 噪声信号　；参考：《中国矿业大学》2017年硕士论文

【摘要】：旋转机械是最常使用的一类机械设备,对该类设备进行故障诊断,状态监测直接关系到工厂的工作效率、经济效益以及工作人员的安全健康。然而,现今普遍使用的振动信号故障诊断方法由于其自身特点的限制已经不能满足实际的监测要求。因此,本文结合声学信号监测的非接触式测量、设备简单、信号易于测取、易于发现早期故障、无须事先粘贴传感器、可对移动目标进行在线监测等优点,对基于噪声信号的故障诊断方法进行系统研究。首先,本文对旋转机械常见故障类型及故障机理进行详细介绍。对声学的理论基础进行说明,从理论层面说明了利用噪声信号进行故障诊断的可行性。介绍了利用噪声信号进行故障诊断的方法,以及各方法的原理和特点。同时,针对旋转机械故障的噪声信号的特点进行了分析说明。其次,基于研究目的以及研究背景,结合现有实验设备设计实验内容,并且开展实验信号的采集工作。将实验分成两类,一是对定点式信号的采集,二是移动式信号的采集。主要的实验对象包括滚动轴承故障信号、转子转轴故障信号、齿轮箱齿轮啮合故障信号以及混合故障信号等。再次,利用采集的故障信号实现基于噪声信号的故障特征提取方法研究。利用短时傅里叶变换对故障特征进行提取,初步验证噪声信号中包含着故障信息成分。根据处理结果发现,噪声信号中包含更加复杂强大的背景噪声,考虑使用经验模态分解去噪,将经验模态分解与相关性分析以及峭度进行结合,实现对信号的分解与重构,进而实现对信号的去噪。对比去噪前后的频谱图,可见该去噪方法对该类信号具有较好的处理效果。因为经验模态分解频率混叠问题的存在,对基于集合经验模态分解的故障特征提取方法进行研究。然后,结合采集到的噪声信号特点,提出一种基于能量的噪声信号处理方法,并对该方法的原理以及过程进行详细的说明。最后,对旋转机械故障的噪声信号进行分类方法研究,将支持向量机应用于故障特征分类。利用交叉验证方法寻找最优模型参数,分别将定点采集信号以及移动采集信号进行分类识别,再对混合故障信号进行分类识别,观察支持向量机对故障噪声信号的分类准确度。文章最后对本次研究所做的工作进行了总结,并对相关的研究发展进行了展望。
[Abstract]:Rotating machinery is the most commonly used kind of mechanical equipment, which is a kind of fault diagnosis for this kind of equipment. The state monitoring is directly related to the efficiency of the factory, the economic benefit and the safety and health of the staff. However, the commonly used method of vibration signal fault diagnosis is unable to meet the actual monitoring because of its own characteristics. Therefore, this paper combines the non-contact measurement of acoustic signal monitoring, the equipment is simple, the signal is easy to be measured, the early fault is easy to find, the advantages of on-line monitoring of the moving target without prior sticking to the sensor, and the fault diagnosis method based on the noise signal are systematically studied. First, the common fault of the rotating machinery is found in this paper. The type and fault mechanism are introduced in detail. The theoretical basis of acoustics is explained, and the feasibility of fault diagnosis using noise signal is explained from the theoretical level. The method of fault diagnosis by using noise signal and the principle and characteristics of each method are introduced. Secondly, based on the purpose of research and research background, combining the existing experimental equipment to design experimental contents and carry out the acquisition of experimental signals, the experiment is divided into two categories, one is the collection of fixed point signals and the two is the acquisition of mobile signals. The main test objects include the rolling bearing fault signal and the rotor axis. The fault signal, the gear box gear meshing fault signal and the mixed fault signal, etc. again, the fault feature extraction method based on the noise signal is realized by the acquisition of the fault signal. The fault feature is extracted by the short time Fourier transform, and it is preliminarily verified that the noise component contains the fault information component. At present, the noise signal contains more complex and powerful background noise, considering the use of empirical mode decomposition to denoise, combining empirical mode decomposition with correlation analysis and kurtosis to realize the decomposition and reconstruction of the signal, and then realize the denoising of the signal. Because of the existence of frequency aliasing in empirical mode decomposition, the method of fault feature extraction based on set empirical mode decomposition is studied. Then, a noise signal processing method based on energy is proposed based on the characteristics of the acquired noise signal, and the principle and process of this method are described in detail. Finally, the noise signal of the rotating machinery fault is classified, and support vector machine is applied to the classification of fault features. The cross validation method is used to find the optimal model parameters, the fixed point acquisition signal and the mobile acquisition signal are classified and identified, then the mixed fault signals are classified and identified, and the support vectors are observed. Finally, the paper summarizes the work done in this research, and looks forward to the related research and development.
【学位授予单位】：中国矿业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH17

【参考文献】