Design for Variation

Abstract

ABSTRACT Statistical engineering is the study of how to best utilize and integrate statistical methods with other engineering disciplines and information technology to solve high-level, complex, multidisciplinary problems. Engineering disciplines apply the principles or “building blocks” of the sciences to design something new or to improve something already in existence, frequently a machine, tool, or structure (Hoerl and Snee 2010b). Statistical engineering involves the application of the principles of probability and statistics, in conjunction with the work of other engineers and scientists, to quantify and control variability, uncertainty, and risk in order to improve a process, product, or service. The design for variation (DFV) process developed at Pratt & Whitney and described in this article is shown to possess all attributes of a statistical engineering application. Design for variation, a strategic initiative at Pratt & Whitney, represents a large-scale improvement of engineering design and analysis processes in order to quantify and reduce variability, uncertainty, and risk, requiring the efforts of engineers, statisticians, and information technology specialists. This article also presents examples of the DFV process as applied to turbine airfoil cooling, high cycle fatigue, hollow fan blade producibility, and optimization of thermal management systems and overall system performance. The new ‘DFV-enabled’ processes help engineers to actively manage variability, uncertainty, and risk directly via design, materials, and process changes.