A displacement field approach based on FEM-ANN and experiments for identification of elastic properties of composites

Abstract

An inverse approach combining numerical and experimental results with full-field displacement measurements will allow the identification of all the in-plane elastic properties from experimental tests. Instead of the use of a non-destructive technique, an alternative hybrid approach is proposed to obtain the field displacement. The experimental displacement field was replaced by the nodal displacement values numerically determined using the finite element method (FEM). The use of a surrogate model based on artificial neural network (ANN) enables to establish the relationship between the elastic properties and the displacement field avoiding the exhaustive calculations based on FEM. The Uniform Design Method is used to select the input values for ANN learning procedure. The optimal estimation of the model parameters is performed by minimizing an error functional defined as the difference between the experimental measurements and the simulated output results from ANN approximation model. Two examples were chosen to validate the proposed approach. The first one uses the off-axis tensile test in order to calibrate the numerical method eight-harness satin weave glass fiber reinforced phenolic composite. The second example is a hinged cross-ply laminated plate supporting a vertical load. The numerical results show strong agreement with the experimental values.