Finite Mixture Models for Nonidentical Multiple Tooled Manufacturing

Abstract

This paper presents a novel method to assess nonidentical multiple tooled (NIMT) manufacturing processes (like multiple cavity injection molding) using finite mixture distribution (FMD) models. A stepwise methodology is presented, including supporting mathematics and statistics. The methodology is illustrated and supported by its application to two sets of real multicavity injection molding data. The method is commercially relevant and is significant in that it allows enhanced examination of the fraction of the parts nonconforming or better setting of the specification level. Included are discussions of FMD models versus normal models and novel tail probability comparison methods (ratio of tail probabilities and log PDF comparisons). The methodology is recommended for NIMT processes, and is thought to better address the adequacy evaluation of processes where there are multiple nonidentical distributions mixing in production.