Parameters Identification Based on Deep Neural Networks for the Microplane Material Model of UHPC Under Uni-Axial Compression

The Microplane material model is widely used to describe the complex behaviour of concrete structures. The basic parameters of the model are determined by calibration with a set of experimental data of normal strength concrete, the accuracy of simulation results heavily depends on these parameters. For the ultra-high-performance concrete (UHPC) with many properties are difference to NSC then these parameters must to be re-calibrated. This study introduce introduces a novel approach based on deep neural networks theory (DNN) for identification model parameters. The experimental data from uni-axial compression tests of UHPC specimens and Microplane-M7 Nonlinear Finite Element (NFE) simulation with huge set of model parameters are used for training model. The results show very good agreement of stress-strain curves between NFE simulation with obtained parameter set from DNN and experiment. The DNN method can be extended to apply to complex stress state of concrete.