Intelligent and Sustainable Process Systems Engineering

We aim to develop enabling technologies designed to enhance automation, control, and real-time decision-making for large-scale process systems in sustainable and smart manufacturing. To that end, we leverage the power of computational multiscale modeling, control theory, network analysis, cloud computing, and machine learning combined with core chemical engineering principles. Our research focuses on developing a theoretical framework and the corresponding computational tools needed for controlling complex process networks, designing cyber-physical systems for smart process manufacturing, utilizing efficient network communication in process systems, and advancing system identification using process data analytics, with application to chemical and energy systems in the fourth industrial revolution (Industry 4.0). Our ongoing research projects have significance in solving major scientific and industrial challenges in the areas of sustainable and secure operation:

Control and automation in Process Industry 4.0

Industry 4.0 refers to the fourth industrial revolution that takes advantage of the notable technological developments on cybernetics, distributed physical devices with built-in computing and communication capabilities including the internet of things, information technology infrastructures including cloud storage and computing, new sensors technology, and new production processes including additive manufacturing. Since using the capabilities of Industry 4.0 in chemical processes and energy systems is in the initial stage of development, it is essential to clearly define the structure and methodologies that can be employed in process systems engineering and more precisely in process control.

Sustainable process control

The current paradigm for achieving the overall sustainability objectives of large-scale process systems is to combine information management and decision-making in one optimization-based control setting rather than a hierarchical separation of information and commands. The optimal control and decision-making strategies are often implemented in the industry with model predictive control (MPC), due to its relative conceptual simplicity, flexibility, performance, robustness, and its ability to efficiently handle complex multivariable systems with the hard path and terminal constraints. MPC can optimize directly in real-time the economic performance of the process, rather than tracking to desired setpoints. We develop a novel MPC framework to incorporate all quantifiable sustainability concerns in the control problem formulation.

Address the model uncertainty through online system identification using process data analysis

In general, MPC does not guarantee closed-loop stability due to considering a finite prediction horizon. However, a standard implementation of MPC using a nominal model of the system can, with slight modification, exhibit robustness to disturbances and sufficiently small modeling errors by adding terminal cost and constraints. To achieve robustness, it becomes necessary to account for all possible trajectories which could be realized by the uncertain system. Selecting an appropriate approach for the particular system involves assessing an acceptable balance between computational requirements and closed-loop performance.

Multiscale modeling and uncertainty quantification of catalytic systems for energy applications

The broad applications of reaction systems in many fields of science and technology, including process and combustion engineering, environmental science, and systems biology, necessitate developing a framework for predictive mathematical modeling and analysis which can interpret experimental data, identify the rate-controlling steps, guide experiments, and design novel materials. Our research focuses on utilizing microkinetic analysis in the predictive multiscale modeling of metal-based catalytic process systems while incorporating detailed knowledge of chemistry and experimental data. It enables the development of novel data-assisted predictive models through accounting for uncertainty and correlations in the system parameters. In addition, it rationalizes why models often under-predict reaction rates and showed that despite the uncertainty being significant, the method could, in conjunction with experimental data, identify important missing reaction pathways and provide insights into the catalyst active site and the kinetic reliability of a multiscale model.

Contact

Davood Pourkargar
Assistant Professor
785-532-2625
dbpourkargar@k-state.edu