Monitoring machine operations using on-line sensors

Monitoring machine operations and production process conditions using on-line sensors has drawn increasing attention recently. In this paper, we discuss a situation where an on-line sensor is used to monitor a randomly deteriorating machine operation. The machine condition is described by a finite number of states, and the machine deterioration follows a Markov process. It is assumed that the sensor measurement and the true machine condition have a statistical relation. A decision is to be made on when a machine setup should be made, based on the observed sensor measurement. A Markovian model is developed by considering the cost of operating the machine and the cost of performing preventive maintenance, and a steady state threshold policy is developed by minimizing the total cost. In addition, a heuristic method based on Bayes rule is proposed. A simulation study is used to study and compare the properties of these two policies.