Parameter estimation in chemical engineering ; a case study for resin production
In this report we present a study on parameter estimation in the field of resin production. The mathematical model of the chemical process contains a set of 12 differential algebraic equations (DAEs) and 16 unknown parameters; 8 series of measurements are available, performed under different initial conditions and at different temperatures. To estimate the unknown parameters we solve the system of model equations and tune the model by varying the parameters in order to fit the solution of the DAEs with the measurements. The differential equations are solved by the BDF method. As a fitness criterion we use the sum of the squared weighted residuals, which is minimised by a Levenberg-Marquardt algorithm. Not only the optimal parameter values are determined, but also their reliability is investigated in combination with the feasibility of the mathematical model. With the available measured data 12 of the 16 unknown parameters could be estimated within acceptable statistical bounds.
|Department of Numerical Mathematics [NM]|
Stortelder, W.J.H. (1996). Parameter estimation in chemical engineering ; a case study for resin production. Department of Numerical Mathematics [NM]. CWI.