Nature Inspired Curve Fitting Strategies for Viscoelastic Materials Mechanical Properties

Abstract

Viscoelastic models have been used to solve problems in different areas, as in structural damping problems or to simulate the mechanical behavior of biological tissues.
In order to properly simulate the behavior of these materials, the frequency dependency of theirs mechanical properties must be taken into account. In that way, one of the first steps to achieve a fair viscoelastic model is to curve fit mechanical parameters, adjusting experimental data from characterization tests. Traditionally this curve fitting procedure is made through minimum least squares methodologies.
It this work, three alternative curve fitting strategies for viscoelastic materials are studied: Artificial Neural Networks, Genetic Algorithms and Particle Swarm Optimization. These strategies are analyzed and the quality of each curve fitting procedure, based on real experimental data, is evaluated pointing the advantages and disadvantages of each methodology.