Optimal replacement policies for a multicomponent reliability system (Q1111464)

The authors study a discrete time multicomponent coherent system. The system as well as each of its components can be either in a functioning condition or in failed condition. If a component is functioning at the beginning of a period, then it fails during the period with some probability. Replacing a component and the system being in the failed condition at the end of a period costs some money. The problem is to find replacement policies that minimize the expected discounted total infinite-horizon cost or the long run average of the expected undiscounted cost per period. This problem is reduced to a dynamic programming model. The main result states that under the optimal policy a failed component must be replaced iff it is in a certain set (the critical component set).