Parameter identification theory of a complex model based on global optimization method

Abstract

With the development of computer technology and numerical simulation technology, computer aided engineering (CAE) technology has been widely applied to many fields. One of the main obstacles, which hinder the further application of CAE technology, is how to successfully identify the parameters of the selected model. An elementary framework for parameter identification of a complex model is provided in this paper. The framework includes the construction of objective function, the design of the optimization method and the evaluation of the identified results, etc. The parameter identification process is described in this framework, taking the parameter identification of the superplastic constitutive model considering grain growth for Ti-6Al-4V at 927°C as an example. The objective function is the weighted quadratic sums of the difference between the experimental and computational data for the stress-strain relationship and the grain growth relationship; the designed optimization method is a hybrid global optimization method, which is based on the feature of the objective function and incorporates the strengths of genetic algorithm (GA), the Levenberg-Marquardt algorithm and the augmented Gauss-Newton algorithm. The reliability evaluation of parameter identification result is made through the comparison between the calculated and experimental results and between the theoretical values of the parameters and the identified ones.