I.C. Schick; "Bayes-Optimal Reactive and Proactive...", Dec. 3, 13:40
  • FENS
  • I.C. Schick; "Bayes-Optimal Reactive and Proactive...", Dec. 3, 13:40

You are here

Faculty of Engineering and Natural Sciences

 

FENS IE SEMINARS

 

 

 

 

Bayes-Optimal Reactive and Proactive Maintenance Control Policy for a Deteriorating Machine

 

 

Irvin Cemil Schick

 

Massachusetts Institute of Technology

 

 

Abstract

 

This talk will address the problem of optimal decision-making for performing maintenance on a deteriorating machine from the viewpoint of Bayes-optimality. Given an observation

 

window containing recent product quality, and the time elapsed since the most recent maintenance, we seek the policy that minimizes the expected total cost over a given optimization horizon. The deteriorating machine is modeled as a partially observable Markov process, and dynamic programming is used to formulate the optimal control policy. Knowledge of the time elapsed since the most recent maintenance results in a more sensitive (and realistic) control strategy, and makes it possible to schedule future maintenance when maintenance is not needed immediately. Furthermore, it helps mitigate the influence of delays when a significant amount of time elapses between the production and inspection of parts---because, for example, a long physical distance separates the machine from the inspection station that assesses the quality of its products. Thus, the method is applicable not only to machines in isolation, but also to machines embedded in a production line. This talk covers joint work with Gianpietro Bertuglia and Stanley B. Gershwin.

 

 

Biography

 

Irvin Cemil Schick was born in Istanbul. He received his B.S., M.S., and Ph.D. from the Massachusetts Institute of Technology, where he is currently a research scientist. He has worked in industry, as well as teaching at Harvard University, M.I.T., and Boston University. His main research interests are estimation, stochastic systems, and statistical decision theory, applied to manufacturing processes and complex data networks.

 

 

 

Wednesday, 3 December 2008, 13:40-14:30, FENS G035