基于信息熵和α稳定分布的滚动轴承故障诊断方法

Rolling Bearing Fault Diagnosis Based on Information Entropy and Alpha Stable Distribution

为了更快、更准确的对滚动轴承的故障状态进行诊断,提出了一种结合信息熵(information entropy, IE)和α稳定分布(alpha stable distribution, ASD)参数的特征融合方法IE-ASD,基于振动信号对滚动轴承进行故障诊断。首先提取振动信号的时域、频域和时-频域的四种IE(奇异谱熵、功率谱熵、小波空间特征谱熵和小波能谱熵)和估计ASD的四个参数,构建特征向量;然后,利用粒子群优化算法(particle swarm optimization, PSO)对支持向量机(support vectors machine, SVM)参数寻优,并用所构建的特征向量对其训练和故障诊断;最后,利用凯斯西储大学轴承数据库的数据验证该方法的有效性,并与由IE、ASD各自构建特征向量的诊断结果进行对比。结果表明:所提出的方法能精确的判断滚动轴承的故障位置,且比未进行特征融合的方法有更高的准确率。

In order to diagnose the fault of rolling bearings faster and more accurately,a feature fusion method named IE-ASD(i.e.combining information entropy and alpha stable distribution parameters)is proposed to diagnose the rolling bearing fault based on vibration signals.Firstly,four information entropies(singular spectrum entropy in time domain,power spectrum entropy in frequency domain,wavelet space characteristic spectrum entropy and wavelet energy spectrum entropy in time-frequency domain)of the vibration signal are extracted and the four parameters of ASD are estimated to construct the feature vector.Subsequently,particle swarm optimization(PSO)is used to optimize support vector machine(SVM)parameters,and the constructed feature vector is used to train it and diagnose faults.The effectiveness of the method was validated on CWRU Bearing Data Set,and compared with the diagnosis results of the feature vector constructed by IE and ASD,respectively.The results demonstrate that the method can accurately determine the fault location of the rolling bearing,and has a higher accuracy than the method without feature fusion.