A New Wavelet Feature for Fault Diagnosis of Roller Bearings Using Decision Tree

“Fault diagnosis of the roller bearings” as pattern classification problem has three main steps namely, feature extraction, feature selection, and classification. A number of machine-learning algorithms have been successfully used to solve the problem with the help of vibration signals. Feature extraction is one of the most important activities in the whole process, as the strength of the features determines the classification accuracy of the classifiers. This paper investigates the use of new discrete wavelet features for feature extraction and compares the result with that of energy definition of discrete wavelet features using Haar wavelet. The extracted features are then classified using decision tree as classifier. The study reveals that the new proposed feature performs better than that of existing energy definition features.