SUMMARY
We develop a machine monitoring system which utilizes proportional hazard modelling to optimally determine when a machine failure appears to be pending, such that the machine should be scheduled for repair. The model considers repair lead time and the stochastic history of the sensor measurements being monitored in making these decisions. This research provides a significant contribution to the area of automated machine diagnostics and monitoring systems, which are becoming a critical competitive technology for capital intensive industries. The predictive model demonstrates good performance under both simulated and actual machine failure conditions, and is robust across a variety of failure characteristics.
Notes
∗To whom correspondence should be addressed.