饱和砂土场地液化势预测的几种方法

基于地质统计学方法和神经网络技术，对饱和砂土场地液化势进行了分析和预测。预测结果表明优化的网络结构可以很好地映射场地液化势数据结构，预测精度在很大程度上取决于趋势分量的分离以及残余分量的数据结构拟合。分析结果还表明基于GRNN的场地液化势预测模型的预测精度要高于传统的Kriging估值模型的精度。提出的场地液化势几种估值模型适用于岩土参数的空间估值。

Prediction of Liquefaction Potential of Saturated Sand Sites

Based on Kriging methods and neural networks techniques, this paper analyzes and predicts the liquefaction potential of a saturated sand site. The predicted results indicate that optimal neural networks can map the data structure of liquefaction potential of sites. The prediction precision depends to a great extent on separation of trend components of data structure and fitting of the data structure of liquefaction potential. The analyzing results also indicate that the prediction precision based on general neural networks is better than that of Kriging method. The methods presented are applicable for spatial estimation of liquefaction potential or other geotechnical parameters.