© J.C. Baltzer AG, Science Publishers Throughput-dependent periodic maintenance policies for general production units Eleftherios Iakovou a , Chi M. Ip b and Christos Koulamas c a Department of Industrial Engineering, University of Miami, P.O. Box 248294, Coral Gables, FL 33124, USA E-mail: lefteris@eng.miami.edu b Autonation, Inc., The 110 Tower-110 S.E. 6th St., P.O. Box 22776, Fort Lauderdale, FL 33301, USA c Department of Decision Sciences and Information Systems, Florida International University, Miami, FL 33199, USA A Markov decision model is presented to determine the optimal maintenance policy for a general production unit. The unique characteristic of the proposed model is the inclusion of the unit’s throughput rate as a decision variable. We show that under appropriate assumptions, the generator matrix of the underlying Markov chain that models the unit’s deterioration process exhibits a separability property which significantly facilitates the solution procedure. Keywords: Markov processes, phase-type distribution, maintenance, inspection, repairs 1. Introduction The majority of equipment used in the production of goods and delivery of services is subject to deterioration with usage and age. This deterioration results in higher production costs and can be slowed down by the use of preventive maintenance. Preventive maintenance is usually less costly than failure replacement and prolongs the useful life of the production equipment. The literature on preventive maintenance includes numerous papers on determining what preventive maintenance action to take and the scheduling of preventive maintenance. However, to our knowledge, none of these papers links the rate of deterioration of the production equipment to its through- put. The objective of this paper is to address the above issue by determining the optimal maintenance policy and throughput for a production facility which can be represented as a single unit. A distinguishing feature of our model is that we allow the production unit to operate at various throughput rates. Our approach is to propose a Markov decision model which utilizes information about the condition of the Annals of Operations Research 91(1999)41–47 41