中文摘要：

以九江绕城高速公路桩网复合地基加固软基试验段为工程背景,利用智能反演方法、正交试验设计和有限元数值方法相结合,对软基土体参数进行反演和工后沉降预测。研究结果表明,采用有限元和正交试验设计方法相结合,可以为BP神经网络和遗传算法参数反演模型提供大量的训练样本,能够确保参数反演精度;工程应用证明,BP神经网络和遗传算法与ADINA有限元程序相结合对软基工后沉降进行计算和预测是可行的,BP神经网络反演方法计算的软基沉降最大误差为5.26%,遗传算法计算的软基沉降最大误差为3.1%,因此,遗传算法在桩网复合地基软基沉降预测中具有更高的预测精度。

英文摘要：

Taking Jinjiang ring expressway pile-net composite foundation test section as the project background, the soil pa- rameters of soft soil foundation were inversed and the post-construction settlement of soft foundation was predicted by using the intelligent inversion method, orthogonal experimental design and finite element method. The results show that a large number of training samples for BP neural network and genetic algorithm parameter inversion model can be provided by com- bining with finite element and orthogonal experimental design method, which can ensure the accuracy of parameter inver- sion. It is feasible that the post-construction settlement of soft foundation is predicted by using the BP neural network and genetic algorithm combined with ADINA finite element program through the engineering application. Compared with the measured value, the maximum error of soft foundation settlement calculated by using the BP neural network inversion method is 5.26% ,while the genetic algorithm inversion method is 3.1%. Therefore,the settlement prediction accuracy of pile-net composite foundation is higher by using the genetic algorithm inversion method.