陕南地区土岩接触带滑坡地质灾害预测模型构建  

Establishment of Landslide Geological Disaster Prediction Model in Soil-rock Contact Zone in Southern Shaanxi

作　　者：牛洪涛[1] NIU Hongtao(Ankang University,Ankang 725000,China)

机构地区：[1]安康学院

出　　处：《华南地震》2019年第4期19-24,共6页South China Journal of Seismology

基　　金：陕西省教育厅项目：陕南地区土岩接触带滑坡地质灾害预测与治理研究（18JK0019)

摘　　要：陕南地区多为黄土泥岩土质,土岩接触带极易发生滑坡地质灾害,亟需构建有效的滑坡预测模型。BP神经网络模型预测滑坡过程中,神经网络的初始权值随机选择性强、缺乏足够科学依据,导致网络结构不稳定、收敛效果差,预测误差大。为此采用全局搜索性能优的遗传算法优化神经网络初始权值,基于BP算法训练神经网络,构建精度高、效率高的BP神经网络滑坡预测模型;模型将体现陕南地区土岩接触带特点的接触带岩性、孔隙比、含水率、液性指数、坡度、坡高因子作为输入信号,输出结果为(0,1)、(1,0),分别表示发生滑坡与不发生滑坡。经过仿真验证,模型预测结果与实际结果一致,可用于陕南地区土岩接触带滑坡地质灾害的实际预测。Most of southern Shaanxi is loess,mudstone and soil.Landslide geological hazards easily occur in the soil-rock contact zone.It is urgent to build an effective landslide prediction model.In the process of landslide prediction by traditional BP neural network model,the initial weights of the neural network have strong random selectivity and lack of sufficient scientific basis,resulting in unstable network structure,poor convergence effect and large prediction error.In order to optimize the initial weights of the neural network,the genetic algorithm with good global search performance is used to train the neural network based on BP algorithm,and a BP neural network landslide prediction model with high accuracy and efficiency is constructed.The model takes the lithology,void ratio,water content,liquid index,slope and slope height factors of the contact zone reflecting the characteristics of the soil-rock contact zone in southern Shaanxi as input signals,and the output results are(0,1)and(1,0),which indicates that landslides occur and no landslides occur respectively.The simulation results show that the prediction results of the model are consistent with the actual results,and can be used for the actual prediction of landslide geological hazards in the soil-rock contact zone in southern Shaanxi.

关 键 词：陕南地区 土岩接触带 滑坡地质灾害 BP神经网络 初始权值 预测模型 

分 类 号：P694[天文地球—地质学]