基于盲源分离和多尺度熵（MSE）的滚动轴承故障诊断

发布时间：2018-07-25 13:12

【摘要】：滚动轴承是许多机械设备的重要部件之一,其能否正常运行关系到机械设备的正常与否。传统滚动轴承故障诊断方法常常忽略传感器采集的振动信号是多个源信号混合的事实,直接采用适用于平稳信号分析的傅里叶变换对非平稳振动信号进行处理,难以全面、准确地分析源信号所包含的故障类型。针对传统滚动轴承故障诊断技术的不足,本文提出基于盲源分离和多尺度熵的滚动轴承故障诊断方法。滚动轴承作为一种精密元件,当轴承某一部件出现异常时,轴承其他部件往往会产生连锁反应,传感器采集到的振动数据往往是多个部件异常振动的叠加。为了更加精准地识别各个异常情况,本文提出基于盲源分离的单通道振动信号分离方法,该方法利用极点对称模态分解将欠定盲源分离问题转换为正定盲源分离问题,然后采用基于时频分析的盲源分离方法分离源信号。仿真结果表明,该方法分离出的源信号与实际源信号相关系数分别达到0.9771、0.9784、0.9660,能够以较高的分离精度将单个多源混合信号逐一分离出来。针对分离信号的特征提取,提出采用经验模态分解和多尺度熵方法来提取分离信号的特征量。经验模态分解方法在使用过程中,常常受到端点效应的影响。针对经验模态分解方法端点效应问题,提出基于波形平均的端点效应抑制方法,根据信号自身特性来延拓信号,具有较好的自适应性,能够较好地抑制经验模态分解端点效应。为了有效识别故障类型,采用BP神经网络对故障进行辨识。实验结果表明,本文提出的滚动轴承故障诊断方法对轴承内圈故障、外圈故障以及正常状态的识别率分别达到97%、86%、90%,在一定程度上能够有效识别滚动轴承的故障类型。本文采用C#和MATLAB混合编程技术,开发了一套滚动轴承离线故障诊断分析软件。该软件通过对实际滚动轴承振动信号的分析,进一步验证了本文方法在实际应用中的有效性。
[Abstract]:Rolling bearing is one of the important parts of many mechanical equipments. The traditional fault diagnosis method of rolling bearing often ignores the fact that the vibration signal collected by the sensor is a mixture of multiple sources, so it is difficult to deal with the non-stationary vibration signal directly by Fourier transform, which is suitable for the stationary signal analysis. Accurately analyze the type of fault contained in the source signal. Aiming at the shortcomings of traditional rolling bearing fault diagnosis techniques, a fault diagnosis method for rolling bearings based on blind source separation and multi-scale entropy is proposed in this paper. As a kind of precision element, when one part of the bearing is abnormal, the other parts of the bearing often produce chain reaction, and the vibration data collected by the sensor are often the superposition of the abnormal vibration of several parts. In order to identify anomalies more accurately, a single channel vibration signal separation method based on blind source separation is proposed in this paper. In this method, the problem of under-determined blind source separation is transformed into a positive definite blind source separation problem by pole symmetric mode decomposition. Then the blind source separation method based on time frequency analysis is used to separate the source signal. The simulation results show that the correlation coefficient between the source signal and the actual source signal obtained by this method is 0.9771n0.9784n0.9660, and the single multi-source mixed signal can be separated one by one with higher separation accuracy. An empirical mode decomposition (EMD) and multi-scale entropy method is proposed to extract the characteristic quantity of the separation signal. The empirical mode decomposition (EMD) method is often affected by the endpoint effect. To solve the endpoint effect problem of empirical mode decomposition method, a waveform averaging based endpoint effect suppression method is proposed, which extends the signal according to its own characteristics. It is self-adaptive and can suppress the end-point effect of empirical mode decomposition. In order to identify the fault type effectively, BP neural network is used to identify the fault. The experimental results show that the fault diagnosis method of rolling bearing presented in this paper can identify the inner ring fault, outer ring fault and normal state of bearing, respectively, and the recognition rate is 97 / 86 / 90, which can effectively identify the fault type of rolling bearing to a certain extent. In this paper, a software for off-line fault diagnosis and analysis of rolling bearing is developed by using C # and MATLAB mixed programming technology. The software further verifies the effectiveness of this method in practical application by analyzing the vibration signals of actual rolling bearings.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TH133.33

【相似文献】