The Rawlings Research Group -- Walter R. Witkowski

James B. Rawlings Research Group

Walter R. Witkowski
Ph.D., University of Texas at Austin, 1990

Publications

1: James B. Rawlings, Stephen M. Miller, and Walter R. Witkowski.
Model identification and control of solution crystallization processes: a review.
Ind. Eng. Chem. Res., 32(7):1275-1296, July 1993.
2: James B. Rawlings, Walter R. Witkowski, and John W. Eaton.
Modelling and control of crystallizers.
Powder Tech., 69(1):3-9, 1992.
3: Walter R. Witkowski.
Model Identification and Parameter Estimation of Crystallization Processes.
PhD thesis, The University of Texas at Austin, June 1990.
4: Walter R. Witkowski and James B. Rawlings.
The interpretation of particle size distribution measurement and its impact on process identification.
In Proceedings of Second World Congress on Particle Technology, pages 486-493, 1990.
5: Walter R. Witkowski, Stephen M. Miller, and James B. Rawlings.
Light scattering measurements to estimate kinetic parameters of crystallization.
In Allan S. Myerson and Ken Toyokura, editors, Crystallization as a Separations Process, pages 102-114. American Chemical Society, 1990.
6: James B. Rawlings, Walter R. Witkowski, and John W. Eaton.
Control issues arising in population balance models.
In Proceedings of the 1989 American Control Conference, pages 677-682, 1989.
7: Walter R. Witkowski, Stephen M. Miller, and James B. Rawlings.
Kinetic parameter estimation of crystallization processes.
Annual AIChE Meeting, San Francisco, California, November 1989.
8: Walter R. Witkowski and James B. Rawlings.
Kinetic parameter estimation of naphthalene/toluene crystallization.
National AIChE Meeting, Houston, Texas, April 1989.
9: Walter R. Witkowski and James B. Rawlings.
Modelling and control of crystallizers.
In Proceedings of the 1987 American Control Conference, pages 1400-1405, 1987.

Thesis Abstract

Model Identification and Parameter Estimation of Crystallization Processes

This research investigates model identification for crystallization processes. Specifically, the kinetic parameter estimation of batch crystallizers using nonlinear optimization techniques is studied. The seeded, isothermal batch crystallization of naphthalene from a naphthalene-toluene solution is analyzed to verify the parameter estimation schemes.

The model identification technique presented is flexible in model formulation and inclusion of available experimental measurements. In contrast to methods used in the literature, this approach allows the assessment of the parameter estimate reliability, providing a measure of the accuracy and validity of a model.

Accurate model solution methods are necessary for efficient and reliable parameter identification. Three different formulations of the method of weighted residual approach are used. The results using Galerkin's technique with Laguerre polynomials, orthogonal collocation, and the standard finite element method are compared in two crystallization case studies. Issues such as numerical stability, efficiency, and solution convergence are discussed.

A lack of sufficient process measurements is the primary problem to be overcome in developing a reliable model identification and verification scheme. A measure of the solute concentration is available. To provide additional process information, a Malvern 3600Ec Particle Sizer, which is based on Fraunhofer light scattering theory, is used to measure the crystal size distribution (CSD).

A numerical study is performed to determine if the parameter estimation problem using the available concentration and CSD information is possible. Using pseudo-experimental data, generated by solving the model and adding random noise, reliable process measurements are produced. It is shown that while solute concentration data is sufficient to estimate the kinetic growth parameters, additional process information is necessary to evaluate the nucleation parameters. The Malvern's obscuration measurement is the most accurate concerning the CSD. Using this measurement along with concentration allowed accurate identification of all of the kinetic parameters. Complete identification is not achieved using only the concentration data along with the CSD size class information. These conclusions involving the effectiveness of various measurements are verified using the experimental data.

The classical model formulation is found to be adequate to describe the process dynamics at different operating conditions. Operating conditions that are tested include seed loading, seed size and stirrer speed.

University of Wisconsin
Department of Chemical Engineering
Madison WI 53706