Adaptive Process Monitoring for Multimode Industrial Processes Through Machine Learning

In complex industrial processes, real-time monitoring of critical variables is essential for ensuring operational safety and efficiency. Traditional process monitoring models often struggle with processes characterized by multiple operating modes, leading to decreased prediction accuracy and reliability. Existing methods typically require prior knowledge of the number of operating modes and cannot adapt to new modes that emerge over time, limiting their applicability in dynamic industrial environments. To address these challenges, we propose an adaptive process monitoring framework that automatically identifies operating modes using change point detection and classifies data using Gaussian mixture models. Specialized subsoft sensor models are then constructed for each identified mode. This approach eliminates the need for prior knowledge of operating modes and enables the system to adapt to new operating conditions in real time. The effectiveness of the proposed methodology is demonstrated through a case study on the fluid catalytic cracking unit at the Parkland Refinery. The results show that our adaptive segmented model achieves a root-mean-square error (RMSE) of 2.645 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.819, significantly outperforming the nonsegmented model with an RMSE of 5.037 and a negative R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of −0.597. This adaptive framework enhances operational safety and efficiency by providing a robust and flexible monitoring solution for dynamically changing industrial processes.