Processes, Journal Year: 2025, Volume and Issue: 13(4), P. 1043 - 1043

Published: March 31, 2025

Aiming at the strong spatio-temporal coupling relationship between data in actual industrial production process, which leads to problem of insufficient reliability and poor timeliness traditional process anomaly monitoring methods, a time series model based on graph similarity network with multi-scale features is proposed, can react anomalies timely effective manner guarantee safety. First, graph-building method for spatio-temporally coupled time-series using multidimensional time-varying feature map embedding designed capture dependence time, while topology utilized learn spatial data; second, similarity-based strategy innovatively proposed measure difference degree index standard normal data. Finally, validated operating condition Tennessee-Eastman (TE) as well fault The experimental results show that identify more quickly accurately than other typical significantly improves monitoring.

Language: Английский