基于人工神经网络的液化震陷预估方法

采用人工神经网络的较强的非线性映射能力和学习能力，提出了一种基于对角递归网络的液化震陷预估方法。本方法由于可以直接从已知震陷资料出发，直接基于震陷资料样本建模，因而具有很强的客观性，避免了以往震陷预估方法由于人为引入的土的变形假设与实验所造成的误差，因而具有广泛的工程实用价值。

A NEURAL NETWORKS-BASED METHOD FOR EVALUATING BUILDING SETTLEMENTS DUE TO EARTHQUAKE LIQUEFACTION

A new method for evaluating building settlements due to earthquake liquefaction based on diagonal recurrent neural nerworks is presented in this paper, in which the strong nonlinear mapping and learning ability of neural networks is adopted. Since the modeling of this method is directly based on real measured seismic settlement samples, it has the advantage of stonger objectiveness, and it can avoide the mistakes due to factitious soil-deforming assumption and experiments used in the previous methods, so the method proposed in this paper has widely practical engineering value.