基于平滑先验分析和排列熵的滚动轴承故障诊断

Rolling Bearing Fault Diagnosis Based on Smoothness Priors Approach and Permutation Entropy

关键词:  航空发动机  轴承  振动  故障诊断  支持向量机

Key words:  Aero-engine  Bearing  Vibration  Fault diagnosis  Support vector machine

摘要:

由于机械系统的复杂性，滚动轴承振动信号的特征信息表现在不同尺度上，因此需要对振动信号进行多尺度分析。基于此，提出一种基于平滑先验分析（Smoothness priors approach，SPA）和排列熵（Permutation entropy，PE）的滚动轴承故障诊断方法。该方法首先采用平滑先验分析方法代替传统的时间序列分解方法对滚动轴承信号进行分解，得到轴承信号的趋势项和去趋势项；其次，分别计算趋势项和去趋势项的排列熵值；最后，将排列熵值作为特征向量，输入基于粒子群优化支持向量机建立的分类器。将该方法应用于滚动轴承实验数据并进行对比分析，结果表明，在训练样本数为每类50%的条件下，该方法的故障诊断正确率比PE和经验模态分解-PE分别高出12.5%和3.125%。

Abstract:

Due to the complexity of mechanical systems, the feature information of the rolling bearing vibration signals behave on different scales, making it necessary to analyze the vibration signal in a multi-scale way. Therefore, an approach for the fault diagnosis of rolling bearings using the permutation entropy (PE) and SPA (Smoothness Priors Approach) is proposed. Firstly, the SPA is used to decompose the rolling bearings vibration signal instead of the traditional time series decomposition method, trend component and de-trend component spanning different scales are obtained. Secondly, the permutation entropy of the trend component and de-trend component, which contain the main fault information is calculated. The permutation entropies are accordingly seen as the characteristic vector, then input to the Particle Swarm Optimization and support vector machine based classifier. Finally, the proposed method is applied to the experimental data. The analysis results show that, comparing to PE and Empirical Mode Decompose-PE, the diagnosis accuracies of the current approach increase by 12.5% and 3.125%, respectively, as the training sample size is 50% per class.