Multi-scale identification of the elastic properties variability for composite materials through a hybrid optimisation strategy. (1st November 2019)

• Record Type: Journal Article
• Authors: Cappelli, Lorenzo; Balokas, Georgios; Montemurro, Marco; Dau, Frédéric; Guillaumat, Laurent
• Is Part Of: Composites. Number 176(2019)
• Publication Date: 2019-11-01
• Subjects: Composite materials -- Homogenisation -- Buckling -- Uncertainty quantification -- Surrogate model -- Inverse problems -- Optimisation; Composite materials -- Periodicals; Materials science -- Periodicals; Composite materials; Periodicals; Electronic journals; 620.118
• DOI: 10.1016/j.compositesb.2019.107193
• Languages: English
• ISSNs: 1359-8368

Abstract

Abstract: The problem of the identification of the variability characterising the elastic properties of the constitutive phases of a composite (at the microscopic scale) is addressed in this work. To this purpose, the information contained into the probability distribution of the first buckling load of a macroscopic composite specimen is considered, in order to develop a multi-scale identification strategy (MSIS). The goal of the proposed MSIS is achieved by solving an inverse problem: the minimisation of the distance between the numerical and the reference buckling response of the plate, at the macroscopic scale, in terms of statistical moments. Furthermore, thermodynamic constraints are considered to ensure the positive definiteness of the stiffness tensor of each constituent of the composite. The proposed strategy relies on: (a) a semi-analytical homogenisation method, to perform the microscopic / mesoscopic scale transition; (b) the Monte-Carlo technique and an Artificial Neural Network to determine the material properties variability; (c) a general hybrid optimisation algorithm able to deal with optimisation problems defined over a domain of variable dimension to perform the solution search. The effectiveness of the MSIS is proven through two meaningful benchmarks.