next up previous contents
Next: Optimal control problems Up: PurposeProblem Descriptions, and Previous: Dynamic model

Parameter estimation problems

The problem in parameter identification is to estimate unknown parameters tex2html_wrap_inline1233 of the dynamic model. (inverse problem). Measurement values tex2html_wrap_inline1235 of an experiment are given which have been obtained at the l times tex2html_wrap_inline1239 tex2html_wrap_inline1241 where tex2html_wrap_inline1243 .

Measurement values at times tex2html_wrap_inline1239 are quantities depending on

The efficiency of the numerical computations can be improved if the case of directly measured state variables is treated separately from the case that only functions of them have been measured.

Positive real constants tex2html_wrap_inline1253 can be used in order to weight the deviations from the tex2html_wrap_inline1235 -values in the nonlinear least squares objective


The parameters tex2html_wrap_inline1233 have to minimize tex2html_wrap_inline1197 subject to the differential-algebraic equations (1), the boundary conditions (2), and the inequality constraints (3).

The resulting nonlinear least squares problem with nonlinear constraints can be solved by either a generalized Gauss-Newton or a Sequential Quadratic Programming Method.

Oskar von Stryk
Tue Feb 1 13:50:42 CET 2000