Increasing complexity challenges traditional process monitoring methods. This article presents a new method based on real-time contrasts (RTC) between real-time and reference data. By assigning the reference data one class label and a window of real-time data another, the monitoring is changed into a dynamic series of classification problems. Error rates, class probability estimates , and variable contributor diagnostics are discussed, as is the imbalance that exists when the real-time window is smaller than the reference data size. The performance advantages of RTC are illustrated with experiments.
© 2001-2024 Fundación Dialnet · Todos los derechos reservados