KNOWLEDGE-BASED MODELLING OF MATERIAL BEHAVIOUR WITH NEURAL NETWORKS

Abstract

To date, material modelling has involved the development of mathematical models of material behaviour derived from human observation of, and reasoning with, experimental data. An alternative, discussed in this paper, is to use a computation and knowledge representation paradigm, called neural networks, developed by researchers in connectionism (a subfield of artificial intelligence) to model material behaviour. The main benefits of using a neural-network approach are that all behaviour can be represented within a unified environment of a neural network, and that the network is built directly from experimental data using the self-organising capabilities of the neural network. Such a modelling strategy has important implications for modelling the behaviour of modern, complex material, such as composites. In this paper, the behaviour of concrete in the state of plane stress under monotonic biaxial loading and compressive uniaxial cycle loading are modelled with a back=propagation neural network.