Application of differential evolution algorithm for identification of experimental data

Abstract

In the paper, the authors present the approach to modelling of austenitic steel hardening basing on the Frederick-Armstrong’s rule and Chaboche elastic-plastic material model with mixed hardening. Non-linear uniaxial constitutive equations are derived from more general relations with the assumption of an appropriate evolution of back stress. The aim of the paper is to propose a robust and efficient identification method of a well known material model. A typical LCF strain-controlled test was conducted for selected amplitudes of total strain. Continuous measurements of instant stress and total strain values were performed. Life time of a specimen, signals amplitudes and load frequency were also recorded. Based on the measurement, identification of constitutive equation parameters was performed. The goal was to obtain a model that describes, including hardening phenomenon, a material behaviour during the experiment until the material failure. As a criterion of optimisation of the model least square projection accuracy of the material response was selected. Several optimisation methods were examined. Finally, the differential evolution method was selected as the most efficient one. The method was compared to standard optimisation methods available in the MATLAB environment. Significant decrease of computation time was achieved as all the optimisation procedures were run parallel on a computer cluster.