Joint finite element: Artificial Neural Network numerical analysis of multilevel composites

In this paper Artificial Neural Network with hidden layers is applied to define a constitutive relation sigma-epsilon in an incremental form. Then ANN is used to describe dependence of effective material parameters of a composite on physical and geometrical characteristics of micro components at the micro-structural level. Using classical tools of homogenisation it is impossible to define such dependences in symbolic manner. These applications are illustrated with an example of superconducting strand for nuclear fusion device.