Influence of objective function on parameter identification result

With the development of the computer technology and the numerical technology, inverse analysis method has been widely applied to identify the mode parameter. The parameter identification methodology mainly includes the selection of the appropriate experimental data, the definition of the objective function and the optimization algorithm. However, little study has been reported on how to define the appropriate objective function for the goodness-of-fit of a model to a set of experimental data till now. Taking the parameter identification of a macro-micro coupled superplastic constitutive model as example, the influence of the selected objective function form on the parameter identification result is studied. The selected objective function form includes hetereoscedastic maximum likelihood error estimator (HMLE) and improved mean squared error estimator (S-RMSE), which the data of every sample is normalized with the according average value. The study shows that the selection of the objective function has important influence on the parameter identification result. For the identification of the macro-micro coupled superplastic constitutive model, the identification result is better, when the S RMSE is adopted. It may be due to that it couple the strength of HMLE and that of MSE.