1. PREDICTION OF BEARING REMAINING USEFUL LIFE BASED ON EUCLIDEAN DISTANCE USING AN ARTIFICIAL NEURAL NETWORK APPROACH
Authors: DALIA M. AMMAR , ELSAYED S. ELSAYED , MOHAMMAD A. YOUNES
Abstract: Accurate prediction of Remaining Useful Life (RUL) of machines and machine components is very important for reliability evaluation. This paper proposes an Artificial Neural Network (ANN) as a method for accurate prediction of RUL of bearings based on vibration measurements during an accelerated life test. The input features to the neuro-predictor are: the vibration signal in the time domain, the dominant harmonics of the bearing vibration signal expressed in a set of selected coefficients of the discrete cosine transforms (DCT), and the main harmonics of the vibration signal as expressed by Fast Fourier Transform (FFT).The Euclidean distance which is a measure of time to failure based on RMS value is used as the figure of merit for the validation of the ANN. Henceforth; the RUL of the bearing can be predicted as the output of the neuro-predictor. The results prove that the suggested methodology can successfully be applied for prediction of bearing RUL.
Keywords: Remaining Useful Life (RUL), Reliability evaluation, Artificial Neural Network (ANN), Discrete Cosine Transform (DCT), Machine Tool Dynamics.