A machine produces an item at a constant rate, which is assumed to be great
er than the demand rate, and the demand is assumed to be known and constant
. While operating, the machine can fail, and upon failure it requires servi
ce. The machine times-to-failure and repair times are random, and during re
pairs, demand is backordered as long as the backordering level does not exc
eed a prescribed amount, after which demand is lost. By considering time to
be of discrete units and the times-to-failure and repair times to be geome
trically distributed, we model the production-inventory system as a Markov
chain and develop an efficient algorithm to compute the potentials that are
used to formulate the cost function. The model results are then compared t
o simulation results where time is treated as a continuous parameter. (C) 2
001 Elsevier Science Ltd. All rights reserved.