Abstract
This paper examines the computation of optimal preventive maintenance policies for systems which are viewed over a horizon whose length is modeled as being random. Preventive maintenance is modeled as a new mode of failure which has a different distribution of time required to make repair.
For the simple case of a system made up of one component, general conditions are developed which assure optimality of the preventive maintenance policy. Preventive maintenance for a two component-one cold standby system is also analyzed to illustrate the application of our methodology on more complex systems. Computational examples are included to emphasize the generality and tractability of this approach to reliability modeling.