Parameter estimation in equivalent circuit analysis of dielectric curemonitoring signals using genetic algorithms.

摘要

This communication concerns the treatment of dielectric data obtained fromexperiments following the chemical hardening process (cure) in thermosettingresins. The aim is to follow, in real time, the evolution of the individualparameters of an equivalent electrical circuit that expresses the electricalbehavior of a curing thermoset. The article presents a methodology for thesequential inversion of impedance spectra obtained in cure monitoringexperiments. A new parameter estimation technique based on genetic algorithms isdeveloped and tested using different objective functions. The influence of theobjective functions on the modelling performance is investigated. The newtechnique models successfully spectra contaminated with high noise levels. Theintroduction of regularization in the optimization function rationalizes theeffects of outliers usually detected in cure monitoring dielectric spectra. Thetechnique was successfully applied to the analysis of a series of spectraobtained during the cure of an epoxy thermosetting resin.