The Support Vector Machine (SVM) i s a machine learning algorithm based on the Statistical Learning Theory (SLT), w hich can get good classification effects even with a few learning samples. SVM r epresents a new approach to pattern classification and has been shown to be part icularly successful in many fields such as image identification and face recogni tion. It also provides us with a new method to develop intelligent fault diagnos is. This paper presents a SVM-based approach for fault diagnosis of rolling bea rings. Experimentation with vibration signals of bearings is conducted. The vibr ation signals acquired from the bearings are used directly in the calculating wi thout the preprocessing of extracting its features. Compared with the methods ba sed on Artificial Neural Network (ANN), the SVM-based method has desirable adva ntages. It is applicable for on-line diagnosis of mechanical systems. The Support Vector Machine (SVM) isa machine learning algorithm based on the Statistical Learning Theory (SLT), w hich can get good classification effects even with a few learning samples. SVM r epresents a new approach to pattern classification and has been shown to be part icularly successful in many fields such as image identification and face recognition. It also provides us with a new method to develop intelligent fault diagnos is. Experiments with vibration signals Compared with the methods ba sed on Artificial Neural Network (ANN), the SVM-based method has desirable adva ntages. It is applicable for on-line diagnosis of mechanical systems.