基于蚁群算法和神经网络的位移反分析

运用蚁群算法和人工神经网络构造了位移反分析的蚁群人工神经网络模型，并基于正交试验获得的训练样本对网络进行学习，以此训练好的神经网络模型来描述岩体力学参数和位移之间的关系。该方法以神经网络为基础，用蚁群算法来学习神经网络的权系数。利用反演结果，建立快速拉格朗日快速计算法（FLAC）模型，对地表沉陷进行预测。结果表明：用蚁群算法训练神经网络，可兼有神经网络广泛映射能力和蚁群算法快速全局收敛的性能。

Displacement back analysis based on ant colony algorithm and neural network

An ACA-ANN model for displacement back analysis is founded by ant colony algorithm and artificial neural network. The network is trained with input-output data pairs obtained from numerical simulation based on the orthogonal tests. The trained network provided the relationship between mechanical parameters of the rock mass and the displacement. The method is based on the neural network, and the weighs of neural network are trained by ant colony algorithm. The inversion results were in turn used as input parameters of a FLAC model predicting the subsidence. The results show that extensive mapping ability of neural network and rapid global convergence of ant colony algorithm can be obtained by combining ant colony algorithm and neural network.