Abstract

Bayesian statistics support a framework for improving the quality of product development while reducing its cycle time. This methodology is based on a model of decision making in which a design decision is a function of Ic, Ke, and Ip, where Ic is the interpretation of customer need, Ke is the design engineer's knowledge, and Ip is information about manufacturing process capability. The methodology recognizes that design engineers use Bayesian thinking during product development; that there is independence between design specifications and manufacturing tolerances; and that product manufacture moves through levels of increasing complexity. Level zero is the stage of unassembled raw materials. Level one occurs when there are one or more assemblies of raw materials. Level two is an optional stage in which assemblies have been combined into subassemblies. Level three occurs when subsystems or assemblies come together to form an integrated end product. The Bayesian methodology provides information about process potential at a given level, by gathering ever more information about the potential of the process and product. In a typical iteration at a given level, the posterior mean and variance of a quality characteristic can be estimated by bringing together a prior estimate with sample data. This article provides a example of the methodology for two iterations at level two for the size characteristic of a slot in a high precision cast assembly.