Wavelet-based features for prognosis of degradation in rolling element bearing with non-linear autoregressive neural network

In rotary machine, the frequently failure component is the rolling element bearing (REBs).The timely identifying the potential fault can prevent the breakdown and failure of rotary machine. A health assessment model based on Non-Linear Autoregressive Neural Network along with exponential value of Health Indicator (HIE) is proposed in this paper. The prognoses of bearing degradation estimates as: In first step, the friction torque transducer used to acquire vibration signal over the lifetime of bearing. After that, in second step, useful features are extracted after the processing of signals through continues wavelet transform (CWT). Then, exponential method is applied to remove the shortcomings into feature vectors. Finally, HIE features as input to the network; the prediction of bearing degradation is performed using optimal NAR and NARX network. An experimental result clarified that the performance of NARX network is better for the prediction of degradation. Outperform application of proposed method and conformation of degree of accuracy; it is compared with the published literature. The proposed methodology is more effective for a dataset of seeded defect and accelerated life test. The results indicate this methodology signifies the actual situation well and is able to precisely and efficiently predict the bearing degradation.