MODIFIED BLOCK-REPLACEMENT FOR MULTIPLE-COMPONENT SYSTEMS

This paper extends the modified block-replacement policy (MBRP) of Ber g & Epstein (1976) in two ways; we consider: 1) a discrete time framew ork which allows the use of any discrete lifetime distribution, and 2) multi-component systems, Both of these extensions increase the practi cal value of the MBRP. In MBRP, components are replaced immediately on failure, and preventive maintenance (PM) is performed at regular inte rvals. During PM every component whose age is greater than a fixed thr eshold age is replaced. Unlike many models for multicomponent systems, this policy is structured. In both the standard block-replacement pol icy (SBRP) and MBRP, PM is performed at regular intervals and so can b e planned in advance. However MBRP results in a lower average cost-rat e by replacing components selectively during PM. Our results show how successful the MBRP is in this respect by comparing its cost with the costs of good, but difficult to compute & implement, age-based PM poli cies. This paper shows that MBRP is appreciably better than SBRP and o nly slightly poorer than the ARP for multi-component systems with a va riety of lifetime distributions and a range of cost parameters. In mos t of our examples the MBRP bridges more than 2/3 of the gap between th e minimum average cost of the ARP & SBRP. Compared to the ARP, the MBR P is more convenient because PM can be planned in advance, tracking th e ages of components is unnecessary and, for multi-component systems, the optimal MBRP is easier to compute & characterize. The average cost -rate for a MBRP is not sensitive to small changes in the interval len gth or the threshold value, so the benefits of MBRP can be realized al most entirely by the proposed heuristic policy. In practice this heuri stic is adequate. Where SBRP is already used, this heuristic is especi ally attractive because the interval between applications of PM does n ot change.