Bearing Fault Diagnosis Based on Statistical Feature Extraction in Time and Frequency Domain and Neural Network

Bearing is an important component of almost every mechanical system used in industrial environment. Hence the defect in bearing must be detected in advance to avoid catastrophic failure. This paper aims to diagnose the defect in bearing automatically using machine intelligence. A condition monitoring setup is designed for analyzing the defects in outer race, inner race and rolling element of bearing. MATLAB is used for feature extraction and neural network is used for diagnosis. It is found that the amplitude at defect frequencies may not always clearly indicate the increment; hence statistical analysis of bearing signature is a better alternative. The work presents an experimental investigation carried out on an experimental set-up for the study of bearing fault at same angular speed and load. This paper proposes an approach of damage detection in which defects in bearing are accurately analysed using vibration signal and neural network.