Linear Dislocated Time Series - Deep Neural Network for Bearing Fault Classification

Bearing is a critical component in an induction motor. Diagnosing bearing fault is one of important task in maintenance activities. Deep learning has been applied for diagnosing bearing fault. This paper compares several dislocating methods as the input for Deep Neural Network (DNN) which will be used for classifying electric motor bearing fault. Linear Dislocated Time Series (LDTS) is proposed and evaluated for feeding (input) of Deep Neural Network. The result shows that LDTS gives the better accuracy (97.29%) compared to other dislocating methods.