Development and Validation of Methodology for Fix Effectiveness Projection During Product Development

Abstract

One of the challenges that design and reliability engineers face is how to accurately project fix effectiveness during reliability planning of a product development project. All reliability projection methods currently in use require estimates of the fix effectiveness factors (FEF) in their mathematical formulation. Obviously, required test results from multiple test phases are unavailable at the onset of a project and therefore practice is to rely on engineers' subjective assessment FEFs. Such estimates are often inaccurate and mostly optimist, resulting in potentiality significant project risks in the form of delays, additional development costs, and costs associated with field failures, returns, and market position. This dissertation provides a methodology that significantly improves the accuracy of FEF estimates and also the resulting reliability metrics such as projected failures rates and MTBFs. The methodology identifies key "performance shaping factors" (PSF) that enhances or impedes an engineer's ability to "fix" a problem, and puts that information into a "causal model" via Bayesian Belief Networks (BBN) to predict FEFs. Tests and confirmation of the methodology for various products and diverse industries show a systematic error reduction in FEF estimates over the current use of unstructured subjective estimates. A second major contribution of the research is an investigation of the effect of interdependencies among various FEFs in projecting the reliability of the same product or several different products by the same organization. Independence is currently assumed by all reliability projection methods. The research (i) shows that FEFs are indeed dependent, (ii) provides a composite BBN model showing the level of dependency among two different fix activities, and (iii) quantifies the impact that fix effectiveness factors have on MTBF projections. The research therefore presents an important augmentation to the current IEC standard for reliability growth, Crow-AMSAA model, showing how to include dependent FEFs in the calculation of failure intensity.