Multivariate Process Control Chart Pattern Classification Using Multi-Channel Deep Convolutional Neural Networks

Statistical process control (SPC) charts are commonly used to monitor quality characteristics in manufacturing processes. When monitoring two or more related simultaneously, multivariate T2 often employed. Like univariate charts, chart pattern recognition (CCPR) plays a crucial role SPC. The presence of non-random patterns indicates that is influenced by one assignable causes and corrective actions should be taken. In this study, we developed deep learning-based classification model for recognizing To address the problem insufficient representation one-dimensional (1D) data, explore advantages using two-dimensional (2D) image data obtained from threshold-free recurrence plot. A multi-channel convolutional neural network (MCDCNN) was incorporate both 1D 2D representations data. This tested on processes with different covariance matrices compared other traditional algorithms. Moreover, effects imbalanced datasets dataset size performance were analyzed. Simulation studies revealed MCDCNN outperforms techniques identifying patterns. For most significant one, our proposed method achieved 10% improvement over methods. overall results suggest can beneficial intelligent