Prediction of Constitutive Laws of Sea Ice Based on Long Short Term Memory Neural Network
In order to establish the constitutive relationship of sea ice accurately,an efficient and accurate method to predict the constitutive relationship of sea ice is sought,aiming at the disadvantages and low efficiency of the two traditional methods of physical test and numerical simulation test.The sea ice constitutive relationship is predicted based on long short-term memory(LSTM)neural network,and the sea ice triaxial compression test is numerically simulated by discrete element method(DEM)bonding element model.Stress-strain curves of samples under different groups of confining pressure and loading rate are obtained as data sources,and multi-layer LSTM network is built.The stress-strain curves of 5 groups of confining pressure and 3 groups of strain rate are used as training data to train the neural network to predict the stress-strain relationship of sea ice under the remaining 1 group of confining pressure and 1 group of strain rate,and the mean square error is used as an evaluation index to evaluate the prediction results.The results indicate that the LSTM neural network data-driven model can effectively predict the constitutive law of sea ice materials,and this method can provide reference for the design of polar marine structures and the study of sea ice mechanical behavior.
sea iceconstitutive lawlong short term memory(LSTM)neural networkdiscrete element method(DEM)
NETL
NSTL
万方数据
