next up previous contents
Next: 4.2.2 A Levenberg-Marquardt Optimization Up: 4.2 Optimization Experiments Previous: 4.2 Optimization Experiments


4.2.1 Global Optimization

Figure 4.2 depicts a DONOPT based optimization allowing for global optimization with bound constraints. This optimization tool is minimizing the value of the target quantity, which must be a single floating point number. Additionally, it supports the introduction of optimization constraints. The description for an optimization based on DONOPT is defined in Description 4.3.


\begin{Figure}
% latex2html id marker 4432\centering
\includegraphics{fig/exp/optimization}\caption{
A \textsf{DONOPT}{} based optimization.}
\end{Figure}


\begin{Modeldesc}
% latex2html id marker 4675
[t]\caption{
The \textsf{DONOPT}{}...
...me}\dq{} )\quad ] \\
)\end{tabbing}\end{minipage}\end{flushleft}\end{Modeldesc}

The equality-constraint demands that the quantity at a selected port must not differ from the value specified in the description by more than a certain tolerance, whereas the inequality-constraint requires that the selected quantity must not be negative during the optimization procedure. DONOPT keeps the settings of the selected ports to be optimized between their bounds. Therefore, each model port which can be varied by DONOPT must be of the type bound-float.



Rudi Strasser
1999-05-27