SVD降噪与排列熵方法在齿轮故障诊断中的应用

发布时间：2018-08-13 08:21

【摘要】：齿轮是机械设备中的一种通用传动零部件,广泛应用于现代机械中,但由于本身结构复杂,工作环境恶劣等原因,齿轮极易出现故障,是一种易损件。在有齿轮装置的传动机械中80%的故障与齿轮的故障有关,在齿轮箱的各零件中,齿轮本身的故障比例最大,其故障率达60%以上。一旦机械设备出现齿轮故障,就可能中断生产,给企业造成巨大的经济损失,甚至带来危害生命等灾难性后果;所以针对齿轮的机械故障诊断方法进一步深入的研究具有重要的实际意义。本文为了降低噪声对齿轮故障诊断的干扰影响,采用SVD降噪方法来提高故障诊断的准确性。首先通过阐述基于Hankel矩阵SVD降噪方法的理论基础,采用常用的三种奇异值阈值处理方法,在旋转机械振动故障试验平台上进行实验测试,获取齿轮故障振动信号,并通过三种方法信噪比与均方根误差以及时频域分析结果对比,从中发现奇异值中值方法降噪的效果比其它两种方法更显著,验证了SVD降噪方法在齿轮故障诊断中的可行性及有效性。其次,采用计算简捷、快速的排列熵方法对SVD降噪后齿轮故障的特征信息进行提取,即将排列熵方法引入齿轮的故障诊断中,详细阐述了排列熵算法及过程,分析了排列熵算法的特性,应用MATLAB软件对实验测试信号进行相空间重构,并将计算获得的排列熵值作为齿轮故障的特征向量,具有较好的抗噪性和突变检测效果,从而验证了排列熵方法可作为齿轮故障状态变化的定量依据。最后,采用支持向量机理论(SVM)作为智能诊断方法,研究了多类SVM分类器,利用常用的几种核函数分别构建了多类SVM分类器,使用齿轮不同状态的排列熵特征向量,分别与这些多类SVM分类器相结合进行训练,获得齿轮典型的故障模式,从而可实现对齿轮故障进行诊断及分类。通过上述对比分析,可以发现采用径向基核函数构建的多类SVM分类器的分类效果优于其它两种核函数;同时将排列熵与神经网络相结合以及排列熵与支持向量机(SVM)相结合分别对齿轮故障进行诊断,从对比分析结果,可以证明排列熵与支持向量机(SVM)相结合的方法对齿轮故障诊断的准确度更高。
[Abstract]:Gear is a kind of universal transmission parts in mechanical equipment, which is widely used in modern machinery. However, because of its complicated structure and bad working environment, gears are prone to malfunction and are easily damaged. In the transmission machinery with gear device, 80% of the faults are related to the fault of the gear. Among the parts of the gear box, the proportion of the faults of the gear itself is the largest, and the failure rate of the gear itself is more than 60%. Once there is gear failure in mechanical equipment, the production may be interrupted, which will cause huge economic losses to enterprises, and even bring disastrous consequences such as endangering life and so on. Therefore, it is of great practical significance to further study the mechanical fault diagnosis method of gear. In order to reduce the influence of noise on gear fault diagnosis, SVD noise reduction method is used to improve the accuracy of fault diagnosis. In this paper, the theoretical foundation of SVD denoising method based on Hankel matrix is introduced, and three kinds of singular value threshold processing methods are used to test the vibration fault of rotating machinery on the platform of vibration test, and obtain the vibration signal of gear fault. By comparing the SNR of three methods with the root mean square error in time and frequency domain, it is found that the noise reduction effect of the median singular value method is more significant than that of the other two methods. The feasibility and effectiveness of SVD noise reduction method in gear fault diagnosis are verified. Secondly, a simple and fast permutation entropy method is used to extract the characteristic information of gear fault after SVD noise reduction. The permutation entropy method is introduced into gear fault diagnosis, and the algorithm and process of permutation entropy are described in detail. The characteristic of permutation entropy algorithm is analyzed, and the phase space reconstruction of experimental test signal is carried out by using MATLAB software, and the calculated permutation entropy value is taken as the characteristic vector of gear fault, which has good anti-noise and abrupt detection effect. It is verified that the permutation entropy method can be used as the quantitative basis of gear fault state change. Finally, the support vector machine (SVM) theory is used as the intelligent diagnosis method, and the multi-class SVM classifier is studied. The multi-class SVM classifier is constructed by using several common kernel functions, and the permutation entropy eigenvector of different gear states is used. Combined with these kinds of SVM classifiers, the typical fault modes of gears can be obtained, and the fault diagnosis and classification of gears can be realized. Through the above comparative analysis, it can be found that the classification effect of the multi-class SVM classifier constructed by radial basis function kernel function is better than that of the other two kernel functions. At the same time, combining permutation entropy with neural network and permutation entropy with support vector machine (SVM), the gear fault diagnosis is carried out, and the results are compared and analyzed. It can be proved that the combination of permutation entropy and support vector machine (SVM) is more accurate in gear fault diagnosis.
【学位授予单位】：辽宁科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH132.41

【参考文献】