Bearing Useful Life Prediction based on Vibration Analysis with Empirical Mode Decomposition and Artificial Neural Network Methods

In this study, a method based on signal processing and artificial intelligence was used to estimate the remaining life of the bearing. To this end, the vibration signals from run-to-failure condition was utilized. The signals was processed by Empirical Mode Decomposition method. Then 10 features were extracted from each signals, in order to model the life estimation. At last, artificial neural network with Levenberg-Marquardt training algorithm was used to estimate the remaining useful life of the bearing. The experimental results showed that the amplitude of the vibration signals increased over time with a very small slope before the failure. But with approaching the failure time, the vibration amplitude increased with a sharp slope. The results of neural network modeling with whole features extracted by EMD method showed that the correlation coefficient between the value predicted by the neural network and the actual value was 0.9260. But using the neural network trained by the best features of EMD, the correlation coefficient between the actual value and the predicted value was obtained 0.94307. Hence, based on the obtained results, the proposed method can effectively be used to estimate the bearing remaining useful life.