The neural network approach in plasticity and fracture mechanics

The Sections devoted to the applications of neural networks in plasticity and fracture mechanics cover three topics. The first one is associated with the implementation of hybrid programs in which neural procedures are used for the analysis of elastoplastic constitutive equations by means of back-propagation neural networks. The first program corresponds to the bending analysis of elastoplastic beams. The second program deals with the analysis of elastoplastic plane stress problem. The second topic is related to the so-called Panagiotopoulos approach. The approach depends on the formulation of the Quadratic Programming Problems and then analyzing them by the Hopfield-Tank network. This approach was used successfully for the analysis of unconstrained and constrained QPPs associated with the classical crack problem and the analysis of elastoplastic structures. The third topic corresponds to the parameter identification problem. This problem is analyzed by means of two neural networks. The supervised learning of a simple backpropagation neural network interacts with the analysis of subsidiary equations by means of the Hopfield-Tank network.