| dc.description | This thesis investigates the current inventory model used in the Canadian Armed Forces (Army) to field spare parts to various self-contained organization (units) to sustain operational equipments for fixed cycle periods. Since the current model determines inventory levels for each space j, j = 1,...,J up to a specified fixed availability service measure which is the same for each item, we propose and analyze two distinct models in which the objective sought is to determine the optimal number of spares S$\sb{\rm j}$, j = 1,...,J (considered consumables or throw-away modules subject to Poisson demands) required at the beginning of the period in order to optimize system performance by either (1) maximizing system availability A$\sb{\rm S}$ and/or minimizing total expected system backorders BO, constrained to the specified budget. Both models result in optimized stock levels $\rm\{S\sb{j},$ j = 1,...,J$\}$ that are approximately the same. Various solution methods for solving this non-linear integer optimization type of problem such as dynamic programming, marginal and Lagrange analysis are investigated and compare both models vs the current military model; simple heuristics are developed to improve near optimal solutions. We also analyze the link between both performance measures and develop a more appropriate measure of system performance: the expected number (and proportion) of equipments still operational at the end of the period or AA$\sb{\rm S},$ with and without part failure dependencies. Variants and extensions to multiple location and indentured types of systems are also discussed. We include randomly generated numerical test problems, whose results significantly underscore the usefulness of the proposed procedures across all measures of system performance A$\sb{\rm S},$ and BO and AA$\sb{\rm S}.$ | en_US |
