Bearing fault diagnosis using hybrid genetic algorithm K-means clustering

Condition monitoring and fault diagnosis of rotating machinery are very significant and practically challenging fields in industries for reducing maintenance costs. Fault diagnosis may be interpreted as a classification problem; therefore artificial intelligence-based classifiers can be efficiently used to classify normal and faulty machine conditions. K-means clustering is one of the methods applied for this purpose. In this paper, a new fault diagnosis method is proposed by applying Genetic Algorithm (GA) to overcome the drawback of K-means which it may be get stuck in local optima. For this purpose, the best solution of GA is chosen to be the initial point for K-means clustering. The proposed method is used in fault diagnosis of the scaled rotor-bearing system experimentally. Then the result of hybrid GA-K-means clustering is compared with classic K-means clustering.