Search

Due to uncertainties in material properties and use conditions, reliability predictions are often subject to considerable error. Such inaccurate predictions lead to maintenance decisions that are expensive or not able to prevent failures. This article proposes a Bayesian approach for failure time prediction of degrading components from condition data that accounts for uncertainties and incorporates prior information on degradation behavior via prior distributions. The approach is based on posterior sampling to handle more general statistical models. Simulation examples are presented to show that by incorporating condition monitoring data or including prior knowledge the effectiveness of maintenance decisions can be significantly improved. Application of the approach in a real setting is illustrated using data from the literature. [ABSTRACT FROM AUTHOR]

In Robust Parameter Design (RPD) the means and the covariances of noise variables, commonly assumed as known, are estimated from operating or historical data and hence can involve considerable sampling variability. In addition, for cases where there are noise factors that are measurable or with strong autocorrelation a more effective control strategy is to update the estimates of noise factor as the production takes place. This article presents a Bayesian approach to online RPD that accounts for uncertainty in noise factor and response models and allows the user to update the model estimates with production data and achieve more effective control performance. The proposed method is compared to existing dual response and certainty equivalence control approaches from the literature. Simulation examples and a case study that uses real manufacturing data from an injection molding process are used to demonstrate the proposed method. [ABSTRACT FROM AUTHOR]

Two new Bayesian approaches to Robust Parameter Design (RPD) are presented that recompute the optimal control factor settings based on on-line measurements of the noise factors. A dual response model approach to RPD is taken. The first method uses the posterior predictive density of the responses to determine the optimal control factor settings. A second method uses in addition the predictive density of the noise factors. The control factor settings obtained are thus robust not only against on-line variability of the noise factors but also against the uncertainty in the response model parameters. On-line controllable and off-line controllable factors are treated in a unified manner through a quadratic cost function. Both single and multiple-response processes are considered and closed-form robust control laws are provided. Two simulation examples and an example taken from the literature are used to compare the proposed methods with existing RPD approaches that are based on similar models and cost functions. [Supplementary materials are available for this article. Go to the publisher's online edition of IIE Transactions for the following free supplemental resource: Appendix] [ABSTRACT FROM AUTHOR]