A Study of Progressive Solution to Data Distribution Maturity Problem at Product Inspection Stage of Manufacturing

Production lines are alive because the important factors of production such as man, machine, material, method, and environment (4M+E for short) change suddenly and continuously. The change causes serious obstacles to achieving the requirements of quality, cost, and delivery (QCD for short). In this article, a progressive solution to the production change problem is proposed and discussed. The target is product inspection of production process. The problem to be solved here is to acquire the inspection model and criteria from multivariate data set extracted from sensing data. The change of 4M + E and QCD in production is reflected in the data distribution; i.e., sophistication of production process is strongly connected with data distribution maturity. The proposed method is to fuse one class SVM as a non-parametric method and MTS as a parametric method according to the data distribution maturity. Experiments were conducted to evaluate performance of the proposed method.