Hybrid Models

There are many ways that a Hybrid Model can be assembled.  This will be influenced by the type of process and the kind of partial knowledge we have at hand.  Quite often, the main reason we are not able to write a detailed knowledge-driven model describing the inner workings of a process is our lack of understanding about the rate processes taking place.  Such rate processes are reaction rates and rates of mass and energy transfer.

In an example of such a hybrid model, we have utilized the data for the incoming monomer flow rate and the reactor temperature measurements to back-calculate the polymerization rate data at several time instants and each experiment.  These back-calculated data were used to estimate a DRSM model for the dependence of the polymerization rate on the operating conditions of the reactor.  The resultant hybrid model was used to optimize the operation of the process, significantly increasing the productivity of the existing reactors and thus avoiding the need to build a second reactor.  For more details, please see the following publication.

Bardooli, A., Y.C. Dong, and C. Georgakis, (2022) “Mass and energy balance-assisted data-driven modeling and optimization of batch processes: The case of a batch polymerization process”.  Computers & Chemical Engineering, 160, 107701.