Modeling Batch Crystallization
Batch Crystallization Control for Crystal Size Using Data-driven Modeling
Batch cooling crystallization is widely applied to isolate high-purity chemicals and pharmaceuticals. Control over the size of crystals produced by such operations can be critical for efficient downstream operations.
Daniel Griffin of Amgen describes a new model-based, closed-loop approach for controlling the average size of crystals produced by batch cooling crystallization. The method is distinguished in the modeling strategy employed. Rather than developing a crystallization model within the population balance framework, crystallization dynamics are viewed from a new perspective and apply a machine-learning technique to identify an empirical model of the dynamics directly from measurement data. The model is low-dimensional and can be used with dynamic programming to obtain optimal feedback control policies for producing crystals of targeted average sizes in pre-specified batch run times. Experimental results are given to demonstrate that the policies obtained in this way can be applied to produce crystals of the desired average sizes in the specified run times.
About the Presenter
Daniel Griffin
Amgen
Daniel Griffin obtained a BS from Ohio State University and a PhD at Georgia Institute of Technology, both in Chemical and Biomolecular Engineering. At Georgia Tech, Dan conducted research on the use of crystallization to remove non-radioactive salts from legacy nuclear waste under the guidance of Yoshiaki Kawajiri, Martha Grover, and Ronald Rousseau. Along the way, he became interested in the use of modern data visualization, machine learning, and dynamic programming techniques to understand and control batch cooling crystallization. The control strategies that came out of this pursuit were originally applied to control salt crystallization from nuclear waste solution simulants—work that has been recognized by the U.S. Department of Energy and received an Innovations in Fuel Cycle Technology Award. Dan is currently working as a Senior Engineer in Process Design and Development at Amgen in Thousand Oaks, CA.
References
Journal
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2016. Data-Driven Modeling and Dynamic Programming Applied to Batch Cooling Crystallization. Industrial & Engineering Chemistry Research 55, 1361-1372
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2015. Mass-Count Plots for Crystal Size Control. Chemical Engineering Science 137, 338-351
Griffin, D. J., Kawajiri, Y., Grover, M. A. and Rousseau, R. W. 2015. Feedback Control of Multicomponent Salt Crystallization. Crystal Growth & Design 15, 305-317
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2014. Robust Multicomponent IR-to-Concentration Model Regression. Chemical Engineering Science 116, 77-90
Conference Papers
Griffin, D. J., Tang, X., and Grover, M. A. Externally-Directing Self-Assembly with Dynamic Programming. 2016. Proceedings of American Control Conference (ACC), Boston, MA, pp. 3086-3091.
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2015. Combining ATR-FTIR and FBRM for Feedback on Crystal Size. American Control Conference (ACC), Chicago, IL
Griffin, D. J., Grover, M. A. K., Yoshiaki and Rousseau, R. W. 2015. Controlled Crystallization of Salts from Nuclear Waste Solutions. Waste Management (WM) Conference, Phoenix, AZ.
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2014. Supersaturation Control During Fractional Crystallization. International Symposium on Industrial Crystallization (ISIC), Toulouse, France.
Griffin, D. J., Grover, M. A., Kawajiri, Y. and Rousseau, R. W. 2014. A Methodology for Monitoring Concentrations of Complex Waste Solutions. Waste Management (WM) Conference, Phoenix, AZ.
