改进的进化遗传算法在软弱结构面力学参数选取中的应用

A RESEARCH ON EVOLUTION GENETIC ALGORITHM OF EVALUATING THE SOIL FRICTION PARAMETERS

对进化遗传算法进行了改进 ,提出了新的交叉算子和变异算子 ,使得改进后的算法具有更好的全局收敛能力 ,加快了网络的学习速度 ,提高了预测精度。利用野外对结构面的地质描述和室内试验获得的软弱层带的物理性质 ,将该算法应用于溪落渡水电站坝区软弱结构面力学参数的获取 ,将 9个试验点的预测值与实测值比较 ,误差小于 4.6 % ,表明具有较高的可信度 ,是软弱结构面力学性质预测的一种新方法。

Some improvements of the evolution genetic algorithm (EGA) are presented in this paper. The new operators of crossover and mutation can quicken the learning speed of network. The new algorithm has better performance to obtain the global optimization solution and higher predicting precision than traditional EAG. Evaluation models of friction parameters based on the modified EGA are established which is function of some physical properties of soil in structures. The Xiluodu Hydroelectric Power Station of Jinsha River is going to be another giant project following the Three Gorges Water Conservancy of the Yangtze River. Investigations show that interlayer and intrastratal faults existing in basalt are main structures in the dam site. They will affect the deformation and stability of sliding of high arch dam abutments, dam foundation's rock masses, and surrounding rocks of underground plants etc. The study of these structures' strength is very important to this project. The model is used to predict the friction parameters of soft structure in the dam site. The studies of 9 cases show that the relative error is less than 4.6%, the model is performed well and can be used to predict the friction parameters of soft structure in different areas and under different conditions.