Statistical material parameters identification based on artificial neural networks for stochastic computations

A general methodology to obtain statistical material model parameters is presented. The procedure is based on the coupling of a stochastic simulation and an artificial neural network. The identification parameters play the role of basic random variables with a scatter reflecting the physical range of possible values. The efficient small-sample simulation method Latin Hypercube Sampling is used for the stochastic preparation of the training set utilized in training the neural network. Once the network has been trained, it represents an approximation consequently utilized in a following way: To provide the best possible set of model parameters for the given experimental data. The paper focuses the attention on the statistical inverse analysis of material model parameters where statistical moments (usually means and standard deviations) of input parameters have to be identified based on experimental data. A hierarchical statistical parameters database within the framework of reliability software is presented. The efficiency of the approach is verifiedusing numerical example of fracture-mechanical parameters determination of fiber reinforced and plain concretes.