PROCESS CHARACTERIZATION AND OPTIMIZATION-BASED ON CENSORED-DATA FROMHIGHLY FRACTIONATED EXPERIMENTS

Censored data resulting from life-test of durable products, coupled wi th complicated structures of screening experiments, makes process char acterization very difficult. Existing methods can be inadequate for mo deling such data because important effects and factor levels might be identified wrongly. This article presents an expectation-modeling-maxi mization (EMM) algorithm, where censored data are imputed as pseudo-co mplete samples and a forward regression is used to compare all main ef fects and 2-factor interactions for process characterization. Then, th e best combination of controllable variables is determined in order to optimize predictions from the final model. A sensitivity study of the selected models, with changes of imputation and parameter estimation methods, shows the importance of using appropriate models and estimati on methods in EMM. Our analysis of the Specht (1985) heat-exchanger li fe-test data indicates that E, EG, EH in the wall data and A, K, D, DJ in the corner data are the dominating factors. However, in finding th e best process recipe, one might use a model with a few additional ter ms, which leads to more accurate predictions for better process optimi zation.