基于PSO-SVM算法的乐昌峡鹅公带滑坡体位移预测模型  

Displacement prediction model of Egongdai landslide in Lechangxia based on PSO-SVM algorithms

作　　者：徐文兵 王国河 王生 伍明兆 姚清河[1] XU Wenbing;WANG Guohe;WANG Sheng;WU Mingzhao;YAO Qinghe(Department of Applied Mechanics and Engineering,Sun Yat-sen University,Guangzhou 510006,China;China Nuclear Power Technology Research Institute Co.Ltd,Shenzhen 518040,China)

机构地区：[1]中山大学应用力学与工程系,广东广州510006 [2]中广核研究院有限公司,广东深圳518040

出　　处：《中山大学学报（自然科学版）》2020年第5期57-65,共9页Acta Scientiarum Naturalium Universitatis Sunyatseni

基　　金：广东省促进经济高质量发展专项资金海洋经济发展重大项目(GDOE[2019]A01)。

摘　　要：以乐昌峡鹅公带滑坡体作为研究对象,考虑日降雨量、渗压对边坡变形的影响,建立了BP、SVM、PSO-BP、PSO-SVM四种滑坡体变形预测模型。从乐昌峡安全检测系统导出近4年研究数据,筛选使用其中410组数据进行训练,取30组变形位移作为输出,分析后发现PSO-SVM模型为最佳模型。以PSO-SVM模型为研究对象,对粒子群算法迭代次数、种群规模、速度位置相关系数(k)等因素进行研究,得知三者分别为100、30、0.5时得到最优的PSO-SVM模型,此时的RMSE、MAPE、R^2分别为0.202 mm、0.589%、0.985。相对于大型有限元仿真软件、多元线性回归模型等传统方法,文章所提出的预测模型可以减少计算成本;在面对非线性问题时也能够获得更好的处理效果。Taking Lechangxia Egongdai landslide as the research object,the influence of daily rainfall and osmotic pressure on slope deformation is considered.By establishing BP,SVM,PSO-BP,PSOSVM four landslide body deformation prediction models,the research data of the last 4 years is derived from the Lechangxia safety inspection system,and 410 sets of data are used for training through screening,and 30 sets of deformation displacements are taken as an output,after analysis,the PSO-SVM model is found to be the accurate model.Taking the PSO-SVM model as the basic model,the factors such as the number of iterations of the particle swarm algorithm,the population size,and velocity position correlation coefficient(k)are studied,and the best PSO-SVM is obtained when the three are 100,30,and0.5,respectively.In this model,the RMSE,MAPE,and R^2 are 0.202 mm,0.589%,and 0.985,respectively.Compared with traditional methods such as large-scale finite element simulation software and multiple linear regression models,the prediction model proposed in this article can reduce the computational cost and obtain better processing results in the face of nonlinear problems.At the same time,it can reduce the lack of fitting accuracy caused by incomplete factor analysis.

关 键 词：鹅公带滑坡体 位移预测 PSO-SVM模型 参数寻优 

分 类 号：TU432[建筑科学—岩土工程]