Stochastic methodology for prognostics under continuously varying environmental profiles

Authors

  • Linkan Bian,

    1. Milton H. Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
    Search for more papers by this author
  • Nagi Gebraeel

    Corresponding author
    1. Milton H. Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
    • Milton H. Stewart School of Industrial and Systems Engineering, Georgia Institute of Technology, Atlanta, GA 30332, USA
    Search for more papers by this author

Abstract

We present a stochastic modeling framework for sensor-based degradation signals that predicts, in real time, the residual lifetime of individual components subjected to a time-varying environment. We investigate the future environmental profile that is deterministic and evolves continuously. Unique to our model is the union of historical data with real-time sensor-based data to update the degradation model and the residual life distribution (RLD) of the component within a Bayesian framework. The performance of our model is evaluated on the basis of degradation signals from both numerical experiments and a case study using real bearing data. The results show that our approach accurately estimates the RLD by incorporating the environmental effects and utilizing the real-time observations. © 2012 Wiley Periodicals, Inc. Statistical Analysis and Data Mining 6: 260–270, 2013

Ancillary