Implementation of parameter estimation in mechanistic models by dynamic optimization

The main concern of this paper is the implementation of estimation of time-varying parameters in mechanistic models. A mechanistic model is always a set of nonlinear differential algebraic equations based on basic principles of physics, chemistry, biology etc. by writing down conservation or balance equations. These models usually involve parameters which are time-varying and must be estimated from experimental data or process data. The proposed parameter estimation approach is to minimize the discrepancy between the values of the measured and model simulated outputs. The dynamic optimization is employed due to the time-varying property of estimated parameters. The sequential quadratic programming and orthogonal collocation are used to solve the nonlinear dynamic optimization problem. The approach is verified on a pilot distillation column for estimating tray efficiencies of both the stripping and the rectifying sections and the feed composition.