Dense-Structured Network Based Bearing Remaining Useful Life Prediction System

This work is focused on developing an effective method for bearing remaining useful life predictions. The in accurately predicting the of bearings so that machine damage, production outage, and human accidents caused by unexpected failure can be prevented. study uses dataset provided FEMTO-ST Institute, Besançon, France. starts with exploration neural networks, based which biaxial vibration signals are modeled analyzed. paper introduces pre-processing signals, network model training adjustment data. trained optimizing parameters verifying its performance through cross-validation. proposed model’s superiority also confirmed a comparison other traditional models. In this study, various types data successfully predict life. algorithm achieves prediction accuracy coefficient determination as high 0.99.