In theory, identification of material properties of microscopic materials, such as thin film or single crystal, could be carried out with physical experimentation followed by simulation and optimization to fit the simulation result to the experimental data. However, the optimization with a number of finite element simulations tends to be computationally expensive. This paper proposes an identification methodology based on nanoindentation that aims at achieving a small number of finite element simulations. The methodology is based on the construction of a surrogate model using artificial neural-networks. A sampling scheme is proposed to improve the quality of the surrogate model. In addition, the differential evolution algorithm is applied to identify the material parameters that match the surrogate model with the experimental data.The proposed methodology is demonstrated with the nanoindentation of an aluminum matrix in a die cast aluminum alloy. The result indicates that the methodology has good computational efficiency and accuracy.
Abstract-Identification of material properties involves physical experimentation followed by modeling, simulation and manual optimization. However, the last step tends to be computational expensive. This paper investigates an artificial neural network (ANN) surrogate model for identifying material parameters. The proposed approach is illustrated with a case study based on a nano-indentation test.
IndexTerms-Surrogate models, optimization, metal-mechanic properties, infill sampling, inverse analysis.
Abstract-Surrogate models can be used to replace expensive computer simulations for the purposes of optimization. In this paper, we propose an optimization approach based on artificial neural network (ANN) surrogate models and infill sampling criteria (ISC) strategy to evaluate design variables. The criterion for infill sample selection is a function which aims at identify design that offer potential improvement. We employ four widely used analytical benchmark problems to test the proposed approach. Our results show that a more accurate surrogate model obtained with fewer points is obtained when one includes the infill sample criterion to an ANN-based optimization.Index Terms-Surrogate model, design variables, artificial neural network, infill sampling criteria, optimization, benchmark function.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.