Fault Classification of Ball Bearing by Rotation Forest Technique

Bearing failure is one of the most common causes of breakdown in rotating machines. The machine learning techniques such as Support vector machines (SVM), Artificial neural network (ANN) are widely used for fault classification. These methods are slow and sometime give inaccurate results. Therefore, the search for new classifier techniques is a necessity to increase the classification efficiency with less computation time. In this study, a classifier ensemble is used for fault classification called Rotation forest. Data obtained from Case Western Reserve University have been used to extract time-based statistical features. In all κ subsets are formed by randomly bifurcating the feature set. Principal Component Analysis (PCA) is used on each subset. All principal components are saved to preserve the transformation in the data. The novel features are calculated using κ axis rotations. This results in improved efficiency of fault classification.