(127a) A Comprehensive Framework for Estimation and Dynamic Optimization of Chemical Reaction Systems

The first step after the chemist conceptualizes a new reaction route is experimentation. Following the first seed experiments, in the process of maximizing the reaction performance, the questions that arise are:

• How should the next experiments be designed?
• What is the most probable reaction mechanism and kinetic model?
• What is the potential of this chemistry to achieve an economically viable yield?

The major tasks involved in answering the questions above are:

• Mechanism and Kinetic Model Discovery
• Experiment Design
• Reactor Performance Optimization

When the tasks above are addressed quantitatively, they involve the development and solution of parameter estimation models (or design optimization models) with a system of nonlinear differential algebraic equations. The burden associated with this effort has resulted in qualitative reasoning being the prevalent method of addressing the tasks above.

In this work, we have developed a comprehensive software framework that allows researchers to address the tasks above quantitatively, while at the same time allowing them to focus on yield improvement, rather than dealing with the difficulties of modeling nonlinear differential algebraic systems.

A major aspect of this work has been to develop reliable methods for the solution of dynamic optimization models. We use orthogonal collocation on finite elements for the efficient solution of the differential algebraic models, with much emphasis on resolving the typical difficulties, for example by implementing methods for:

Efficient Initialization of the composition profiles

Modeling of reaction systems with rates of significantly varying orders of magnitude

Solution of higher index systems

Useful diagnostics that catch poorly specified systems

We highlight projects involving particularly difficult reaction/reactor systems and the techniques employed to solve them. The framework has been implemented in a corporate research environment, with a knowledge management component to archive established kinetics, and a project execution component, to perform the tasks of kinetic estimation, virtual experimentation and reactor optimization. With this framework, we have been successful in analyzing a wide range of reaction systems involving petrochemicals, polymers and pharmaceuticals, resulting in improved experiment designs and reactor yields. Customizable and easy visualization of the model results is one of the strengths of the framework, allowing the researcher to effectively screen their hypotheses of the reaction mechanism and kinetics.