High-accuracy classification of thread quality in tapping processes with ensembles of classifiers for imbalanced learning

Industrial threading processes that use cutting taps are in high demand. However, industrial conditions differ markedly from laboratory conditions. In this study, a machine-learning solution is presented for the correct classification of threads, based on requirements, to avoid expensive manual measurement quality indicators. First, states categorized. Second, process inputs extracted torque signals including statistical parameters. Third, different algorithms tested: base classifiers, such as decision trees and multilayer perceptrons, complex ensembles classifiers especially designed imbalanced datasets, boosting bagging decision-tree combined with SMOTE under-sampling balancing techniques. Ensembles demonstrated lowest sensitivity window sizes, highest accuracy smaller greatest learning ability small datasets. Fourth, combination models both Recall Precision resulted reliable tool, tested an extensive experimental dataset.