ML-Based Proactive Control of Industrial Processes.

This paper discusses the use of optimal control for improving the performance of industrial processes. Industry 4.0 technologies play a crucial role in this approach by providing real-time data from physical devices. Additionally, simulations and virtual sensors allow for proactive control of the process by predicting potential issues and taking measures to prevent them. The paper proposes a new methodology for proactive control based on machine learning techniques that combines physical and virtual sensor data obtained from a simulation model. A deep temporal clustering algorithm is used to identify the process stage, and a control scheme directly dependent on this stage is used to determine the appropriate control actions to be taken. The control scheme is created by an expert human, based on the best industrial practices, making the whole process fully interpretable. The performance of the developed solution is demonstrated using a case study of gas production from an underground reservoir. The results show that the proposed algorithm can provide proactive control, reducing downtime, increasing process reliability, and improving performance.