Associate Professor Evanghelos Zafiriou

Joint Appointment: Institute for Systems Research (ISR)
Ph.D., California Institute of Technology, 1987

Room 1208A Chemical and Nuclear Engineering Building
E-mail: zafiriou@umd.edu
Phone: (301) 405-6625

Honors and Awards

Recipient of an NSF Presidential Young Investigator Award (1990)

Current Research Interests

The past decade has witnessed a resurgence of interest and activity in process control. While economic plant optimization translated into highly complex and integrated process designs, tighter product specifications and narrower profit margins increased the performance requirements of the control system. The difficulties that were encountered when attempting to meet the new challenges through traditional methods, have led to the emergence of a new point of view, where control is not an isolated task to be carried out at the very end; instead it interacts with the design and modeling tasks to become part of a larger systems engineering problem. Work in our group studies the interacting tasks of process control, optimization and model identification, and focuses on applications from the oil, chemical and, most recently, the semiconductor manufacturing industry.

Model Predictive Control

MPC algorithms use a linear or nonlinear process model to predict future values of the process outputs. The fact that modeling error is always present creates the need for designing controllers which are robust with respect to model-plant mismatch. Optimization is used in real-time to obtain appropriate input values. The presence of constraints as physical bounds as well as performance and safety specifications result in complex control laws. We study the design of MPC algorithms and the use of CAD tools for tuning the nonlinear control system for stability and performance in the presence of modeling error.

Integration of Process Identification/ Modeling and Control

The degree of accuracy required of a process model used for designing a control system depends on the closed-loop performance and stability specifications for the controller. We study the use of closed-loop control criteria during the identification of linear and nonlinear input/output models, in particular linear models used for constrained MPC, and nonlinear Volterra series models. Our work also examines the use of control requirements in reducing the size of first principles process models that will be used for on-line control, focusing on rapid thermal chemical vapor deposition.

Optimal Operation of Batch Processes

Many polymerization and biochemical processes are operated in a batch or semi-batch manner. This is also the case for most semiconductor manufacturing processes, including single-wafer processing in batches (runs) of one wafer each. The problem of determining the optimal operating profile during each batch is very important. Our work aims at the development of methods that improve the performance of the process by modifying the operation from batch to batch (run to run), as well as in real-time during the batch. Our current application focus is on rapid thermal processing, chemical vapor deposition, and plasma etching as used in semiconductor manufacturing.

Selected Publications

Robust Process Control, M. Morari and E. Zafiriou, Prentice-Hall, Englewood Cliffs, NJ (1989).

Integration of Process Design and Control, E. Zafiriou, ed., Elsevier Science Inc., Oxford, England (1994).

"Robust Model Predictive Control of Processes with Hard Constraints," E. Zafiriou, Comput. and Chem. Eng.,14, 359-371 (1990).

"Optimal Control of Semi-Batch Processes in the Presence of Modeling Error," E. Zafiriou and J. M. Zhu, Proc. Amer. Control Conf.,pp. 1644-1649, San Diego, CA (1990).

"Stability of SISO Quadratic Dynamic Matrix Control with Hard Output Constraints," A. L. Marchal and E. Zafiriou, AIChE J.,37, 1550-1560 (1991).

"Nonlinear Quadratic Dynamic Matrix Control with State Estimation," G. Gattu and E. Zafiriou, Ind. and Eng. Chem. Res.,31, 1096-1104 (1992).

"Optimization-based Tuning of Nonlinear Model Predictive Control with State Estimation," E. Ali and E. Zafiriou, J. Proc. Control., 3, 97-107 (1993).

"An Approach to Run-to-Run Control for Rapid Thermal Processing," E. Zafiriou, R. A. Adomaitis and G. Gattu, Proc. Amer. Control Conf., 1286-1288, Seattle, WA (1995).

"Observer Based Nonlinear Quadratic Dynamic Matrix Control for State Space and Input/Output Models," G. Gattu and E. Zafiriou, Can. J. Chem. Eng., 73, 883-895 (1995).

"On the Dynamic Resiliency of Contrained Processes," E. Zafiriou and H.W. Chiou, Comput. and Chem. Eng., 20, 347-355 (1996).

"Bath to Batch Optimization Using Neural Network Models," D. Dong, T.J. McAvoy and E. Zafiriou, Ind. and Eng. Chem. Res., 35, 2269-2276 (1996).

"Identification of MIMO Volterra Series and Application to FCC Unit, Proc. 13th World Congress of the Int. Federation of Autom. Control, Volume M, 217-222, San Francisco, CA (1996).