Models for optimizing component safety stock levels in large scale assembly systems

We explore the problem of establishing inventory base stock levels for products and components in an assembly environment that manages thousands of items. Customer demand for products is stochastic and procurement lead times for individual components may be lengthy. Component use in finished products is related through a single level bill of materials structure. We are interested in computationally efficient approaches for computing inventory levels of components and products that minimize the per period expected holding and backorder costs over an infinite horizon.;We examine two assembly environments in this research. In the first environment, we assume that each product has a fixed assembly lead time and assembly capacity is infinite. We construct a continuous review model with deterministic assembly lead times and a first come first served allocation rule for components and set base stock levels to minimize costs. We illustrate structural properties of the problem and offer a computationally efficient approximation for its solution. We test the accuracy of the approach in a numerical study. Additionally, we test the approach in an industrial environment and compare the approach to those currently used in practice. We find that our approximation predicts system costs over the models tested in the numerical study with an average error of 7.88% with a range of 1.3% to 13.3%. In an industrial application, our solution approach uncovered opportunities to reduce system wide inventory investment by 19% or ;In the second environment, we develop a periodic review model in which assembly capacity is finite. We analyze this environment and develop an operating policy and approach for making stocking decisions. We examine the system performance under different capacity and component allocation rules. These rules include a simplistic first come first served rule and a real time optimization rule. Our real time optimization rule for component and capacity allocation lowers costs by an average of 3.79% and reduces the average number of backorders by an average of 33.32% over the first come first served policy. Our solution approach estimates expected costs within an average error of 3.97%.