基于PSO-RBF的边坡安全系数预测  

Forecasting of the Slope Safety Factor Based on PSO-RBF

作　　者：胡昊[1] 魏继红[1] 

机构地区：[1]河海大学地球科学与工程学院,南京211100

出　　处：《河南科学》2016年第5期747-751,共5页Henan Science

基　　金：国家自然科学基金项目(41102162)

摘　　要：传统的径向基函数(RBF)神经网络在边坡稳定性预测中已经得到了广泛的应用,但由于其在预测中易陷入局部最优且参数选取不当会对收敛性产生影响.故引入粒子群算法(PSO)对RBF神经网络进行优化,利用其全局搜索能力对RBF神经网络的隐含层基函数中心值、宽度以及隐含层至输出层的连接权值进行参数寻优,建立了基于PSO-RBF的边坡安全系数预测模型.以114组边坡数据为训练样本,8组边坡数据为测试样本,结果显示基于PSO-RBF网络预测结果的最大误差为7.36%、最小为0.18%、平均误差为3.77%,而基于单纯RBF网络的预测结果的相应误差分析别为11.04%、1.34%、6.19%.可以看出,前者的预测结果明显优于后者,表明经粒子群算法优化后的RBF在预测精度上有了明显的提高.The traditional radial basis function（RBF）neural network has been widely used in the forecasting ofslope stability. However,it＇s easy to fall into local optimality and the improper parameter will have a bad impact onconvergence. Therefore,the particle swarm optimization（PSO）is introduced to optimize the RBF neural networkoptimize,using PSO＇s global search ability to optimize three parameters of RBF：central values of hidden layer basefunction,width and connection weights between hidden layer and output layer. Then the PSO- RBF model isestablished to forecast slope safety factor. In this paper,use 114 slope data as training samples and 8 slope data astest samples. The results show that the maximum error,minimum error and average error based on PSO-RBF are7.36%,0.18% and 3.77% respectively,while the corresponding datum for RBF are 11.04%,1.34% and 6.19%respectively. It can be seen clearly that the former＇s prediction results are better than those of the latter,indicatingthat PSO-RBF algorithm has been improved significantly in the prediction accuracy.

关 键 词：粒子群算法 RBF神经网络 边坡安全系数 预测 

分 类 号：TU457[建筑科学—岩土工程]