Design and Operation at Unstable Steady States

Many processes have been overdesigned because engineers are reluctant to design near or within regimes of complex operations, where they are often economically optimal.  This is prevalent in processes with exothermic and autocatalytic reactions and where phases appear and disappear, especially in the critical region. We are developing designs that operate more economically closer to these nonlinear regimes, which are characterized by multiple steady states, periodic and even chaotic operation, often exhibiting inverse response.  Improved control strategies are being developed to permit reliable operation near these regimes.

Nonlinear Model-based Controller Design for Non-minimum-phase Processes

To achieve more effective process designs, nonlinear model-based controllers are being
developed for processes with a non-minimum-phase, delay-free part.  The controllers are derived by exploiting the connections between model-predictive and input-output, linearization controllers, and are designed to satisfy input constraints.  The performance of the control laws is illustrated using numerical simulation and real-time experiments.  Special challenges arise when the process exhibits both non-minimum-phase and unstable steady states.  In one approach, we seek to alter the design to modify the process dynamics.  Using bifurcation analysis, optimization algorithms, and the addition of sensors and actuators,  proposed design changes are explored more effectively.

Dynamic Risk Assessment of Inherently Safe Chemical Processes

To obtain inherently safer plant designs, we are experimenting with game theory to solve the multiobjective optimization problem that involves tradeoffs between profitability, controllability, safety and/or product quality, and flexibility. Then, given more optimal designs, that are inherently safer, we are developing methods for plant-specific, dynamic risk assessment using accident precursor data; that is, data recorded when abnormal events occur. Our models estimate the failure probabilities of various critical accident scenarios, associated with a process unit after the occurrence of an abnormal event, using Bayesian analysis and copulas. In future work, we seek to extend these methods for plant-wide processing systems, and to better understand the interactions between equipment- and human-related safety systems.

Semicontinuous Distillation with Reaction in a Middle Vessel (SDRMV)  

We are currently laying the foundations for this novel concept. Semicontinuous Distillation with Reaction in a Middle Vessel (SDRMV) combines reaction and distillation in a manner that overcomes the difficulties that makes typical reactive distillation cost prohibitive. Based on the semicontinuous distillation methods recently developed by our group, SDRMV combines ordinary distillation with tank reactors used as middle vessels. By operating in a complex sequence of changing operating modes, the SDRMV systems we are studing achieve reaction and separation goals by using process equipment for multiple purposes while keeping the scale of the system small enough to satisfy the demands of the fine chemicals industry.  In future work, we seek to optimize this hybrid system using novel stochastic optimization methods that navigate through narrow flooding and weeping constraints. 

Optimization and Control of Discrete Systems in Materials Processing

This project, in collaboration with Professor Talid Sinno, builds upon our previous work in the control of Czochralski crystallization processes. Its focus is on multi-scale analysis of materials processes to aid in their design optimization and control. Given the results of molecular dynamics calculations showing the aggregation of silicon vacancy clusters at very small length and time scales, we are developing lattice models using kinetic Monte-Carlo methods that permit calculations at much larger length and time scales. These models are being experimented with to develop nonlinear model predictive control techniques that permit single-crystal growth with well-distributed vacancy cluster aggregates.