Analysis of vibration signal's time-frequency patterns for prediction of bearing's remaining useful life

In this research, the frequency features of vibration signal are chosen to analyze a bearings' vibration characteristics under unbalanced load in common operation conditions. The development process of an unbalanced fault was identified by a set of time-based vibration frequency spectrum. Based on the time-frequency features, the bearing's remaining useful life (RUL) can be predicted. A 2-layer neural network is designed to recognize and track the fault's feature patterns contained in the vibration signal. This research provides tools to analyze the features of a bearing vibration signal and provides effective pattern recognition techniques for bearing health diagnosis and RUL prediction.