Implementation of constitutive models in computational intelligence framework

An appropriate constitutive model embedded in a finite element engine is the key to the successful prediction of the observed behaviour of geotechnical structures. However, to capture the behaviour of geomaterials accurately, the constitutive models have to be complex involving a large number of material parameters and constants. This paper presents a methodology for converting or recasting complex constitutive models for geomaterials developed based on any constitutive theory into a fully trained Artificial Neural Network (ANN), which is then embedded in an appropriate finite element solution code. The length of strain trajectory traced by a material point, also called 'intrinsic time' is used as an additional input parameter in training. For the purpose of illustration, the two-surface deviatoric hardening model in the multilaminate framework developed by Lee and Pande (2004) have been cast in the form of an ANN. Computational efficiency is perceived to be the main advantage of this methodology.