Learning-Derived Cost Evolution in Materials Selection

Abstract

Materials selection is a complex, but important, problem for manufacturing firms. Poor material choices can negatively affect the firm's market share or profits. In the face of this complexity, most selection methods make a number of simplifications, including limiting problem scope to selection for a single product or application, and assuming material properties and design criteria are constant over the problem's time horizon. Such assumptions, however, do not always apply, especially when material preference is based on the materials' "emergent properties," the values of which are context-dependent. Consequently, these properties can evolve with changes in context and potentially alter the preferred material identified by the selection method. This thesis investigates the impact of considering cost evolution on a firm's materials selection decision, and seeks to identify strategies the firm can adopt when introducing new materials to its products. To that end, a framework for incorporating cost evolution, specifically from learning, into the materials selection process is proposed and demonstrated using single-product and multi-product automotive case studies. In the single-product method, material options are ranked by their respective manufacturing costs. The multi-product problem is more complex and requires an analytical framework that combines an integer linear program and a genetic algorithm to select materials for any number of products over a specified time horizon. Case study results indicate that when selection problem scope is limited to a single product, accounting for learning in the decision process has minimal impact on the preferred material. When several products are included in the problem scope, however, the firm is able to leverage "shared learning" so that experience gained from manufacturing one product can be applied to lower the costs of other products that share a common resource, such as a manufacturing process line, with the initial product. Not only does the consideration of shared learning impact the preferred materials that are suggested by the selection framework, it also helps to better characterize the circumstances under which the firm should introduce a new material on a test bed. Additionally, the case study results emphasize the use of one material across multiple applications and indicate that this approach helps the firm cope with uncertainty in selection criteria.