Associate Chair and Associate Professor of Chemical Engineering
Office: 359 Benson
- B. Chem. Eng., University of Minnesota, 1979.
- Ph.D., University of Wisconsin, 1984.
- Process Control and Optimization
- Sensor Interpretation
Process Design and Control
Process control represents an area of growing importance. The advent of microcomputers has made it possible to design controllers of unprecedented power and utility, while the increased integration and complexity of today's chemical processes has made the application of such controllers a necessity. We are interested in a wide range of problems involving control and design. In particular our research has specifically focused on developing robust controllers, exploring the use of nonlinear controllers, and examining the application of novel techniques to process control and the interpretation of sensor data.
One important issue in controller theory and application is to design robust controllers and to understand the interaction between the process and the controller. A robust controller is one which performs well even if the process model it is based on is inaccurate or changing with time. The approach used by most researchers in the field of robust controller has been based on extensions of previous control ideas which are often restricted to linear problems. While we are interested in such work our main focus has taken a somewhat different approach. It is based on combining the applicable control theory with an optimization or nonlinear programming approach. This means the problem of designing a robust controller is formulated as an optimization problem with the goal of designing the controller which performs the best on the disturbances and uncertainties which are the most difficult to control. By combining ideas from linear control theory, nonlinear control theory, nonlinear dynamic programming, and game theory, we hope to be able to gain additional insight into the design of nonlinear controllers as well as to understand how the models and their uncertainties influence the design of the controller.
Where conventional control related approaches fail, it may be possible to use unconventional techniques. We are interested in those techniques that can improve our ability to model or interpret the data being sent the model from sensors. One unconventional approach is based on the use of neural networks. A neural network is an abstraction of how the brain operates developed as a computer algorithm. By training the neural network it is possible, for example, to have a neural network interpret a spectrum to determine the concentration of a species in a mixture or to represent the dynamic behavior of a system. We are developing these approaches for the use of these networks and exploring their limitations.
- Wise, B.M., Holt, B.R., Gallagher, N.B., et al. A Comparison of Neural Networks, Non-linear Biased Regression and a Genetic Algorithm for Dynamic Model Identification. Chemometrics and Intelligent Laboratory Systems 1995, 30(1), pp. 81-89.
- Jayaraman, B. Holt, B. R. An Optimization Approach to Robust Nonlinear Control Design. Int. J. Control 1994, 59(3).
- Sekulic, S., Holt, B. R., et. al. Nonlinear Multivariate Calibration Methods in Analytical Chemistry. Analytical Chemistry 1993, 63.
- Soucy, K. A., Holt, B. R. State Estimation and Feedback Control for the Processing of Polymer Composite Systems. Proceedings of the American Control Conference June 1992.
- Jayaraman, B., Holt, B. R. Worst Case Scheduled Controller Design for Nonlinear Systems. Proceedings of the American Control Conference June 1992.