Essays on product design, and manufacturing decisions
Abstract
Part One proposes an application for experimental design in the design centering process. Design centering is commonly used to find optimal design parameter levels for products such as integrated circuits, and mechanical assemblies that have correlated performance characteristics. The most computationally expensive step in the design centering process involves the use of simulation for estimating yield. We propose a less computationally expensive strategy where simulation is replaced by experimental design. An additional contribution of this research is the derivation of experimental design levels based on the beta distribution. Part Two presents a computationally efficient strategy for allocating tolerances that only requires information of the means, and variances of the design parameter distributions. After parameter design, the next step in the product design process is tolerance allocation. Often tolerances are allocated among design parameters in an ad-hoc manner. Part Three presents a mathematical programming model for allocating continuous improvement targets among suppliers. The allocation is done using a minimum cost rationale, where the cost of the allocation decision to the supplier, manufacturer, and consumer is considered. For some test problems, the solution using a minimum cost rationale is contrasted with an ad-hoc policy where all suppliers are allocated the same rate of continuous improvement. Part Four presents a combinatorial optimization model that can be used to improve product quality when it is not possible, in the short run, to alter the settings of raw material design parameters. For many manufacturers, variation in the quality of incoming raw materials is the most significant factor affecting outgoing product quality. Since the combinatorial optimization model is NP-complete for most realistic problems, a heuristic called simulated annealing is used to find good solutions to the raw material matching problem. We demonstrate the applicability of the model by testing it on some real data provided by a pharmaceutical manufacturer.
Degree
Ph.D.
Advisors
Moskowitz, Purdue University.
Subject Area
Management
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.