Development of workload models for CNC machines from 3 - Phase current consumption using ensemble method

Abstract

Increasing in the competivieness of the manufacturing industry, manufacturers have to improve productivity. Data mining is one tool that is widely applied. In injection-mold manufacturing industry, 3-phase electrical usage from CNC milling machine can be used for machinemonitoring. To reduce human error, we applied data mining technique toelectrical usage patterns for identifyingcurrent process running in CNC machines. In this paper, classifiers are created by applying 1)Naive Bayes 2) Bayes Net 3) Neural Network 4) KStar 5) Decision Table and 6) J48(C4.5) to electrical data. Later ensemble methods such as 1) AdaBoostM1, 2) Bagging, 3) Stacking, and 4) Vote are applied to each classier to create more robust models. The models are trained and tested with 10-fold cross validation. Ourpreliminary result shows that bagging ensemble of J48 classifier with no discretization in the preprocessing step gives the best AUC = 0.946.