An operating condition classified prognostics approach for Remaining Useful Life estimation

Abstract

This paper presents a prognostics approach based on operating condition for estimating the Remaining Useful Life (RUL). Operating condition is used to describe the state or environment of a system. This approach is suit for the dataset that contains sensor measurements and operational settings. Predicting RUL contains two stages: modeling stage using the training dataset and predicting stage using the result of modeling and testing dataset. This approach can increase available information in modeling stage and simulate the actual work situation of the test unit in the predicting stage. The performance of this approach was tested by the dataset from 2008 PHM Data Challenge Competition where sensor measurements and operational settings were provided. The task of the competition was to estimate the RUL of an unspecified system. The results showed that this prognostic method could get accurate predictions in most situations and had a good rank in all competition results.