(617bm) Simultaneous Reaction Identification and Parameter Estimation

Understanding the chemical mechanism through which a reaction occurs is crucial to the design, optimization, and control of a reactive system. Elucidating a possible reaction path, and identifying the associated kinetic rate parameters, routinely relies on the expertise of an individual to identify possible reaction pathways and intermediate species. A curve-fitting procedure is then used to identify kinetic rate parameters, and quantify uncertainty in their estimation [1]. Consideration of alternative reactions or mechanisms is then performed in an exhaustive fashion.

We present an optimization-based approach that utilizes a mixed-integer non-linear programming (MINLP) formulation to identify a set of reactions that describe the chemistry observed in a steady-state continuous stirred-tank reactor, and estimate the associated kinetic rate parameters. This approach utilizes a superset of chemical reactions, which can contain different mechanistic pathways for the same reaction. A curve-fitting procedure, similar to one described previously for subset selection in multiple linear regression [2], is then used to minimize an objective that balances the complexity of the model with its fit in order to identify a probable subset of reactions. The utility of this methodology is demonstrated on a number of case studies.

References

[1] Warren E. Stewart and Michael Caracotsios. 2008. Computer-Aided Modeling of Reactive Systems. Wiley-Interscience, New York, NY, USA.

[2] Cozad, A., N. V. Sahinidis, and D. C. Miller, Automatic learning of algebraic models for optimization, AIChE Journal, 60, 2211-2227, 2014.