November 18, 2009
Title: Mathematical Modeling of Chemical Processes - Getting the Best Model Predictions and Parameter Estimates using Limited Data
Abstract: One of the main impediments to using fundamental models for design, optimization and control of industrial processes is that it is difficult to obtain good parameter estimates that will ensure reliable model predictions. When confronted with limited experimental data and a large number of parameters to estimate, some modellers simplify their models to reduce the number of parameters. Others choose only a few parameters to adjust, and fix the remaining parameters at literature values or at reasonable guesses. This talk will describe an estimability analysis tool that can aid in selecting which parameters should be estimated using the available data, and which should be held constant. Comforting theoretical results about the consequences of model simplification and estimation of only a few parameters will also be presented.
Disciplines: Chemistry, Mathematics, Physics