Automatic quality control systems based on pattern recognition and sparse signal processing

Abstract

This paper presents the application of two methods for automatic detection and classification of abnormal quality control patterns. A generator of synthetic concurrent control charts was implemented to create mixtures from patterns described in the literature. In order to obtain features for the classification step, the generated charts were initially processed via sparse regression using the Least Absolute Shrinkage and Selection Operator (LASSO) method. Then, we assess the performance of the classifier which was founded on an artificial neural network (ANN) on two different situations: i) with inputs given by the observed (raw) data and ii) with inputs given by the features generated by the LASSO method. ANN fed with sparse inputs performed extremely close to the ANN fed with raw data, using considerably less inputs.