The Application Of Neural Network Finite Element Method In Geotechnical Engineering

The numerical analysis method, especially the finite element method (FEM) has become the primary tool of numerical analysis and optimization design in geotechnical engineering. However, due to the large amount of fuzzy and uncertain factors in geotechnical engineering, many limilations have been brought to the existent numerical analysis methods.In order to improve the accuracy of the numerical analysis, based on the examinations of a great lot of references and the strong ability of mapping high-nonlinearly and self-learning, a neural network model was set up to model the constitutive relation of soils, which was applied to artifical neural network finite element methods analysis (ANN-FEM). at the same time, the algorithm of ANN-FEM and whose application in pile foundations were study.There are three parts in this paper. First, introduce the basic characteristic of neural network, the structure of the BP and its classic calculate methods and then take discussion for minish error of the network and the process of implement. Second, a neural network soil consititutive relation model was incorporated into FEM by parameter method, and the program written in MATLAB is compiled to learn and train neural network. Finally, some examples were analysed by FEM and ANN-FEM respectively.The results indicate that the trendlines have not obvious differences between them as a whole, ANN-FEM has better fault-tolerance and its results are more reasonable, therefore, the method has certain significance in practical engineering.