A comparative Study for Ball Bearing Fault Classification Using Kernel-SVM with Kullback Leibler Divergence Selected Features

摘要: Bearing early fault detection and diagnosis (classification, estimation, …) is a key issue in Condition Monitoring (CM) of rotating machinery. In this context, we propose paper multi-fault classification comparison between traditional Support Vector Machine (SVM) solutions wavelet SVM (WSVM). For work several kernel functions were considered the Kullback Leibler Divergence (KLD) framework. First, Empirical Mode Decomposition (EMD) employed to preprocess vibration signals acquired from rolling bearings elements. Second, specific statistical analysis study performed select most salient components different obtained Intrinsic Functions (IMFs). Then, KLD retained IMFs calculated carry out three bearing ball severities for operating conditions. Thanks four criteria, namely accuracy rate average $(ARA)$ , support vector $(SVA)$ training time $(T_{r}t)$ testing $(T_{s}t)$ our results are derived highlight technique allowing obtain better results.