Neural network based material description of uncured rubber for use in finite element simulation

The finite element method (FEM) is widely used for structural analysis in engineering. In order to predict the behaviour of structures realistically, it is important to understand and to describe the material behaviour. Therefore, extensive material tests have to be conducted. For highly inelastic materials, such as uncured rubber, the characterisation of the behaviour requires a quite complex rheology. Rheological models are used to describe time-dependent mechanical material behaviour (stress-strain-time dependencies). The mapping of the real material behaviour by such models is only possible with restrictions. However, the evaluation of these models at each integration point within the FEM needs time consuming internal iterations in most cases. In order to describe the material behaviour without model restrictions and to reduce computational cost, the aim of this work is the development of a procedure which enables structural analyses without a specific constitutive material model. In this paper, a neural network is used in order to describe uncured rubber behaviour as a model-free approach