(204w) A Systematic Approach to Model Development and Calibration

In this work, a systematic procedure is proposed for the development and validation of dynamic models of processes in chemical engineering, where the entire process is first divided in different units, which are based on a combination of different phenomena's described by different more or less complex rigorous models. The selection of these elementary models is mostly done without the knowledge of which combination of all models is the best. It might be possible that the selected models explain a single phenomenon’s very well, but when they are linked to describe the whole unit they may fail. Furthermore, it is not known then if a combination of less complex and more robust models will be able to describe the unit with the same accuracy and less effort. For this purpose, a solution strategy is presented and discussed. Furthermore, to calibrate the obtained models against experimental data, nonlinear parameter estimation problems have to be solved. The estimates are though random quantities. Their uncertainty depends on the measurement errors and the way how the experiments are performed. To obtain parameter estimates with maximum reliability, an experimental design is computed by minimizing a functional of the variance-covariance matrix subject to the model equations, constraints, experimental cost and operability. This task leads however to intricate non-standard optimal control problems which are solved by tailored numerical methods presented in this work. Thus, optimal multi-experiments can be designed in parallel, sequentially, or online. The developed solution strategies within a proposed systematic framework will be demonstrated with applications of diverse complexity.