(236c) On the Model-Free Optimization of Batch and Simi-Batch Processes through the Design of Dynamic Experiments: The Case of Batch Reactors

The power and utility of the recently introduced methodology of the Design of Dynamic Experiments (Georgakis, 2008, 2009) is explained and its utility demonstrated by three simulated case studies of batch reactors. The DoDE methodology provides a way to optimize the operation of a variety of batch processes (chemical, pharmaceutical, food processing, etc.) when the most desirable profile of at least one time-varying operating decision variable needs to be selected. This methodology calculates the optimal operation without the use of an a priori model that describes in some accuracy the internal process characteristics. It systematically designs experiments that explore a number of dynamic signatures in the time variation of the unknown decision variable(s). Constrained optimization of the interpolated response surface model, calculated from the performance of the experiments, leads to the selection of the DoDE-optimal operating conditions. These are compared to the optimal operation that is obtained by the use of an accurate process model and a model-based optimization using sequential quadratic programming on a discretized version of the reactor model by Radau collocation on finite elements.

It is shown that each of the reactor examples considered can be quickly optimized, by the DoDE approach. Dynamic inputs examined included reactor temperature and fed-batch flow rates of co-reactant. Several kinds of reaction stoichiometry were simulated with varied degrees of kinetic complexity. In one of the cases only four dynamic experiments enable the definition of a DoDE-optimum operation yielding(Georgakis, 2008, 2009) a reactant conversion that is only just 1.43% less that the true model-based optimum. In the case of a semi-batch reactor, the DoDE experiments led to an optimal operation that was 20% better than the one obtained by the classical DoE approach. Measurement error of 1%-10% does not reduce the power of the approach; the uncertainty of the predicted optimum is less than the respective measurement error.

__________________________________

Georgakis, C. (2008). Dynamic Design of Experiments for the Modeling and Optimization of Batch Process -- Patent Pending.

Georgakis, C. (2009, July 12-15). A Model-Free Methodology for the Optimization of Batch Processes: Design of Dynamic Experiments. Paper presented at the IFAC Symposium on Advanced Control of Chemical Processes 2009, Istanbul, Turkey.