Predicting residual fault times in reliability growth management under limited failure data set

Upon failure discovery, redesign or corrective measures are always implemented to eliminate the defects and improve system reliability in reliability growth management, in which developmental and field data are always utilized to capture the health status with the aid of the reliability growth model. In the presence of partial failure information or limited failure data set, however, it has been more and more challenging to confidently obtain accurate parameter estimators and reliability growth prediction results with an explicit reliability growth model. In this research, taking the S-shaped growth trend and saturation characteristics of reliability growth data into consideration, we investigate a two-stage reliability growth management procedure in predicting residual failure times and fault numbers. Specifically, Stage 1 is to fit the limited reliability characteristic quantity with the grey Verhulst model and to predict the subsequent reliability values within the range of system reliability requirement; whilst Stage 2 is to correspond the predicted reliability values with the Gompertz growth curve and to output the residual failure times of the reliability growth test. An illustrative example shows that integration of the Gompertz reliability growth curve and the grey Verhulst model is not only capable to effectively fit with the S-shaped failure data set, but also enables to precisely predict the residual failure times of the highly reliable product.