Author: A. Musaev, A. Makshanov and D. Grigoriev
Title of host publication: Proceedings - 2024 International Russian Automation Conference (RusAutoCon), IEEE, 2024
Abstract: The task of proactive stabilization of technological processes (TP), the evolution of the state of which is described by non-stationary random processes, is considered. Metric precedent analysis technologies, which belong to the class of machine learning tasks, are proposed as a mathematical tool. The training data used is a data set containing large arrays of retrospective data obtained during the monitoring of the state of the TP during its previous operation. Basic mathematical models of ongoing processes and the algorithmic apparatus of precedent data analysis used are presented. The results of numerical studies of technologies for forecasting the state of non-stationary TPs and the effectiveness of proactive stabilization algorithms built on their basis are given.
Title of host publication: Proceedings - 2024 International Russian Automation Conference (RusAutoCon), IEEE, 2024
Abstract: The task of proactive stabilization of technological processes (TP), the evolution of the state of which is described by non-stationary random processes, is considered. Metric precedent analysis technologies, which belong to the class of machine learning tasks, are proposed as a mathematical tool. The training data used is a data set containing large arrays of retrospective data obtained during the monitoring of the state of the TP during its previous operation. Basic mathematical models of ongoing processes and the algorithmic apparatus of precedent data analysis used are presented. The results of numerical studies of technologies for forecasting the state of non-stationary TPs and the effectiveness of proactive stabilization algorithms built on their basis are given.