Development of an artificial neural network (ANN) constitutive model for mechanical metamaterials

Metamaterials are a group of materials with artificial engineered structures that exhibits customized properties which are not naturally available in other materials. To accelerate the computational analysis of components made from metamaterials that helps novel engineering product design process, it is crucial to develop an accurate and robust model for these materials in macroscale. The classical approach to drive a material model in continuum level is based on development of a phenomenological model to represent the physical behaviour of the material. However, this approach has specific limitations in including the effect of tailoring design parameters in the model which is a key element for metamaterials. In this study, we have proposed an artificial neural network (ANN) constitutive model to represent the macroscale mechanical behaviour of metamaterials in threedimensional domain. Because of its extraordinary capabilities to stimulate computational performance in identifying and constructing prospective microstructure model for mechanical metamaterials, the proposed ANN constitutive model provides intriguing advantages over conventional models. The ANN constitutive model has been trained based on strain-stress data which is obtained from microscale simulation of 3D cubic lattice structure under various loading conditions. The trained material model is then validated by measuring the accuracy of material behaviour prediction.