Abstract: On the Repairman Problem with Controlled Servers and Switch-over Costs

This paper considers a maintenance system in which the queue size at the repair facility c an be controlled by changing the repair time distribution. The production line consists of a f inite number of identical and independent machines working in parallel; they are subject to fa ilure with an exponential time-to-failure distribution. Each breakdown is repaired at a single server repair facility using one of two possible repair time distributions whose choice is ba sed on the number of machines waiting to be repaired. The cost structure includes a holding co st, a repair cost, and a fixed switch-over cost when the repair time distribution is changed f rom one distribution to the other. The control problem is represented by a semi-Markov decisio n model in which the decision epochs are the repair completion epochs. The optimality criterio n to be considered is the long-run average cost per unit time. A policy-iteration algorithm is used to compute the optimal stationary policy within a class of two-parameter policies. Numer ical results are reported using an exponential distribution for the repair time.