The software system solves constrained nonlinear parameter estimation problems. Synonyms are data fitting, nonlinear regression, parameter identification, curve and surface fitting. The numerical methods are based on highly efficient Gauss-Newton-SQP algorithms.
Model functions are defined in a modeling language called PCOMP and are interpreted and evaluated during runtime.
General: Arbitrary fitting criteria depending on parameters to be estimated, additional nonlinear equality or inequality constraints, and optionally on a second independent model variable
Numerical routines: Gauss-Newton-type SQP methods, automatic differentiation, alternative norms (sum of absolute values, maximum of absolute values), confidence intervals for estimated parameters, correlation and covariance matrix, identification of significance levels for estimated parameters
User interface: 'Find' and 'Filter' commands, extensive report and graphics facilities, also for Gnuplot, 3D plots for two model variables (surface fitting), in- and output of data from and to text or Excel files
Limitations: 30 fitting criteria and data sets, 100 model variables, 2,000 time values, 3,000 experimental data, 200 linear or nonlinear constraints
Modeling language (PCOMP) : Highly flexible statements depending on parameters to be estimated, tabulated data, splines, linear and piecewise constant interpolation
Implementation: Numerical routines in Fortran executable outside of the GUI, GUI in VBA, Microsoft Access 2007 run-time version included, database with 240 examples, Windows XP or Vista (contact the author for Win 2K version)