Research
Research Interests: plant-wide control of nonlinear systems; centralized and distributed economic model predictive control; operational safety and cybersecurity of closed-loop processes; digital twin development; quantum computing; image-based control
Our group's research interests are in the area of cyberphysical systems. Our objective is to integrate chemical engineering, optimization, mathematics, physics, and dynamic systems theory to develop cutting-edge solutions to important current industrial considerations, as well as to envision the future of engineering and take steps toward making it a reality. We seek to provide engineering solutions but also to mathematically characterize fundamental benefits and limitations of the solutions through rigorous theoretical analysis. Major themes of our current research are:
Developing and Evaluating Quantum Computing Algorithms for Engineering-Relevant Computations
Though next-generation manufacturing control today is implemented on classical computers, an alternative computational framework known as quantum computing is becoming more accessible to the public (for example, multi-qubit quantum computers, albeit with many faults and imperfections, are currently available for access from various companies). This raises the question of how these should impact the field of advanced manufacturing and process control. Both the lack of fault-tolerance of the current versions of these computers, as well as the fact that quantum mechanics has an inherent probabilistic character that is exploited in many algorithms for quantum computation, raises the control-theoretic question of what using such devices might mean for the safety of control systems (and when probabilistic features of a quantum algorithm or implementation might be acceptable). Dr. Durand's group is investigating this intersection of control and quantum computing from a control theory perspective. The control theory is related to algorithm design on quantum computers; this means that it is also important to investigate how current algorithms might work in solving engineering problems, and to develop new ones. Our group is working on both of these angles for evaluating quantum computing for engineering applications.
Cybersecurity of Chemical Process Control Systems
Incidents during chemical process operation can result in fatalities and injuries, property damage, and damage to the environment. The control system at a chemical plant plays an important role in incident prevention by regulating process states to their steady-state values even in the presence of disturbances. Because of the crucial role of chemical process control systems in adjusting the process states, the potential exists that they can be targeted in a cyberattack and be used to harm plant workers and residents who live or work near the plant. We are working on understanding the different cyberattacks that could be performed at a chemical plant, designing detection policies that integrate with control laws for attempting to locate attackers, and then designing the process itself to be resilient against such attacks.
Digital Twin Development for Dynamically Operated Systems
A common advanced control design used in the chemical process industries today is known as model predictive control (MPC), and it determines control actions to apply to a process by solving an optimization problem and taking predictions of the process state and constraints into account. The objective function of the optimization problem that is minimized traditionally takes its minimum at a process steady-state, so therefore a well-designed MPC should maintain the process state in a neighborhood of the steady-state. This is not necessarily the most profitable manner in which to operate a process, however. A process may be more economically operated in a time-varying fashion, due to its dynamics or changes in, for example, product and feedstock prices, than at a steady-state. A control design known as economic model predictive control (EMPC) generalizes the concept of MPC so that the controller does not necessarily operate a process at steady-state, but operates it in an economically-optimal fashion with respect to a chosen profit metric and process constraints. High-fidelity computational models (which we view as a type of ``digital twin'') can provide a framework for testing and evaluating various advanced control frameworks. Digital twins should account for all phenomena impacting the operation of a system, and developing techniques for comprehensive digital twin development is important for next-generation manufacturing. Our group works on various aspects of digital twin development, including topics such as seeking to understand what it takes to make more physics-based process models automatically from operating data, or developing frameworks for testing image-based control designs.
Accounting for Material Dynamics in Computational Materials Development and Materials Manufacturing/Behavior Control
Materials are key to our capabilities to perform engineering feats. They are key to effectively accomplishing things as mundane as dish-washing to things as complex as flight. Yet materials pose many difficulties for us in their design and manufacture, particularly for advanced materials such as nanomaterials or stimuli-responsive materials. A key aspect of materials and materials manufacturing is dynamic behavior on a molecular level. Our group is working to develop computationally-tractable optimization-based design and control strategies for materials development and manufacture, with a focus on advanced materials and control of material behavior during their use.
