Modeling the relationship between reliability assessment and risk predictors using Bayesian networks and a multiple logistic regression model

During development of an electronic or mechanical system, multiple sources of data are often available. Combining such data helps tackle complex problems and tradeoff decisions. Individually, these data sources provide partial information on the problems we try to solve or the decisions we want to take. In this article, we propose a methodology combining typical reliability and risk assessments used during the early investigation stages when developing electronic and mechanical systems. The proposed methodology assesses system reliability and risks in the development process. This integrated approach improves aspects of reliability assessment of a system, enable optimization of risk and reliability plans and contribute to balanced managerial decisions. While reliability assessment of a system depends on its operational stochastic behavior, risks are single events that affect the performance of a system during operation. We apply Bayesian networks and a multivariate logistic regression to model the relationship between these sources of information. The methodology is illustrated by a real case study from a company in the semiconductor business. By combining such data, we set up an infrastructure supporting effective decisions while alternative options are still available at an early stage of development.