A Diagnostic System for Remaining Useful Life of Ball Bearings

Abstract

This document describes estimation and approximation process of bearings Remaining Useful Life from 2012 Data Challenge. Data were received with the help of PRONOSTIA Platform, constructed for the needs of IEEE Data Challenge. This paper shows different methods of data processing to approximate Remaining Useful Life, which is very important in industries and it is the main part of maintenance. Tests were made on learning set provided by Data Challenge. Raw Data were extracted, filtered and analyzed using algorithms implemented in Matlab. During tests there were used kurtosis, root mean square algorithms and moving average, which helped to process data to be useful for next tests. Remaining useful life was approximated using exponential fitting and different length of original data. Process of analysis boils down to determine at what point it is possible to correctly determinate the bearing remaining useful life.