Parameter optimisation of stress-strain constitutive equations using genetic algorithms

The accuracy of numerical simulations and many other material design calculations, such as the rolling force, rolling torque, etc., depends on the description of stress-strain relationship of the deformed materials. One common method of describing the stress-strain relationship is using constitutive equations, with the unknown parameters fitted by experimental data obtained via plane strain compression (PSC). Due to the highly nonlinear behaviour of the constitutive equations and the noise included in the PSC data, determination of the model parameters is difficult. In this paper, genetic algorithms were exploited to optimise parameters for the constitutive equations based on the PSC data. The original PSC data were processed to generate the stress-strain data, and data pre-processing was carried out to remove the noise contained in the original PSC data. Several genetic optimisation schemes have been investigated, with different coding schemes and different genetic operators for selection, crossover and mutation. It was found that the real value coded genetic algorithms converged much faster and were more efficient for the parameter optimisation problem.