(147ae) Data-Driven Process Monitoring and Control for Smart Manufacturing

Smart manufacturing (SM) techniques involve using information from equipment and sensors to make decisions that improve productivity, flexibility, and cost of running manufacturing processes. Attaining SM requires advanced monitoring and control systems to analyze real-time data and make decisions that help drive the manufacturing process's overall goal.

Our research has been focused on developing advanced algorithms for process monitoring and control using statistical analysis and Artificial Intelligence (AI). We employed data-driven methods for fault detection and diagnosis (FDD) using the Tennessee Eastman Process (TEP) and an air separation unit (ASU). We developed novel AI-based models, including the Probabilistic bidirectional recurrent network (PBRN) and the Shared Parameter Network (SPN) for process monitoring using actual plant data. We showed that these models outperform state-of-the-art fault detection and process prediction applications.

Many FDD models have been proposed in the research community. Unfortunately, only a few of these models have been deployed for practical applications. Through tutorials and practical applications, our research has also demonstrated how FDD models can be deployed for real-time fault detection using heat exchanger equipment and a Smart Manufacturing Innovation Platform (SMIP).

The successful use of Reinforcement Learning (RL) for decision-making in the AI community has provided a good opportunity for improving feedback control techniques such as model predictive control (MPC). Limitations of using these RL methods include large data requirements, stability guarantees, and handling state constraints. Model-based and hybrid RL algorithms offer opportunities for tackling these limitations. Our research has developed model-based RL algorithms and successfully applied them to control the van de vusse reactor and the quadruple tank system.

Research Interests

My interest is in Control Systems, Optimization, and Artificial Intelligence (AI) for solving complex problems in manufacturing processes. I’m particularly interested in developing optimal control techniques using Model Predictive Control and Reinforcement Learning. I am also interested in using advanced AI models, such as probabilistic deep learning models, for modeling, fault detection, and predictions of hard-to-measure process variables. I seek industry opportunities to apply my knowledge to problem-solving and value creation.