Human-in-the-loop and graph theoretic aspects of a smart control room

Abstract

Industry 4.0 and Smart manufacturing paradigms aim to make manufacturing processes flexible, enabling the human-in-the-loop to make better decisions and develop the workforce. The dissertation presents the concept of a Smart Control Room (SCR) that is envisioned to play a key role in manufacturing plants and organizations. Three major aspects of the SCR are presented in detail. A key role of the SCR is to use high-fidelity plantwide models to assist operators in better decision making and forecasting. The development of a rigorous plantwide flowsheet of the Vinyl Acetate Monomer process in Aspen Dynamics and MATLAB is presented. Key improvements of the MATLAB nonlinear model, specifically for the distillation column are discussed. A simulated study of the distillation column startup is also presented. Operator training for various levels of expertise during complicated plant operations is an area where automation algorithms can play a key role. To explore this, the development of a Human-in-The-Loop Supervisory Model Predictive Control (HiTL MPC) algorithm is presented. Using a single-input single output example, it is shown that the algorithm is able to accept human suggested inputs and based on degrees of cooperation, the final control output is computed. The successful performance of the algorithm with a multi-input multi-output example using the VAM upstream flowsheet is also demonstrated, with switching of manual and automatic input pairs occurring during the process. The data arising out of a process plant is hierarchical and networked in nature. To understand this, a novel graph theoretic and Graph Signal Processing based approach is discussed in detail. The filtering and predictive performance of the tool is also presented with the use of industrial data of a cryogenic air separation unit.