Optimization of parameters of constitutive models using genetic algorithms

An Evolutionary Genetic Algorithm (EGA) is a simple and generic structure of rules and commands, which tries to simulate the evolution of 'individuals', to optimize a given problem. In the context of this paper, an individual is a given material represented by a possible set of model parameters and the problem to be minimized is the difference between the theoretical response given by the model using these parameters and the actual response obtained via laboratory tests. The individuals are classified according to a fitness measure, which is the objective function given by the error between prediction and experiments. The selection process tries to find the best individuals, which produce smaller errors, then proceeds with a series of genetic operations, such as reproduction, mutation and crossover. A generic evolutionary algorithm was coded using parallel processing. It is applied in this paper for the search of the optimum set of parameters of an elastoplastic constitutive model named sub-loading t(ij). This is quite useful since the calibration procedure of many models is not always straightforward. The simulations show that the use of AGE can drastically improve the calibration of model parameters, especially for cyclic tests