Incremental Identification of Complex Reaction Systems

Abstract:
To identify adequate models for the description of complex reaction systems from experimental data, an incremental identification framework is presented in this work. The dynamic identification problem is decomposed into a sequence of simpler subproblems, rendering the technique very efficient and robust and supporting the modeling process by revealing novel information in each step. The approach is universally applicable to arbitrarily complex reaction systems, including multi-phase, non-isothermal and distributed parameter systems. Both structured and unstructured models are flexibly integrated in the identification process, depending on the available knowledge. To model unstructured reaction kinetics, a new adaptive function estimation approach based on sparse grids is analyzed. Incremental identification is fully integrable into an iterative model building strategy with experiments designed for high information content. An efficient design criterion is derived to identify unstructured model parts. Numerous examples illustrate the theory throughout the text.