基于神经网络和元胞自动机的滑坡运动过程模拟研究  

Simulation of Landslide Sliding Process Based on Neural Network and Cellular Automata

作　　者：黎华[1] 闵祥强[1] 付俊[1] 李方[1] 黄海峰[2] 

机构地区：[1]武汉理工大学资源与环境工程学院,武汉430070 [2]三峡大学防灾减灾湖北省重点实验室,宜昌443002

出　　处：《武汉理工大学学报》2017年第1期37-41,47,共6页Journal of Wuhan University of Technology

基　　金：国家自然科学基金青年基金(41301588);国家自然科学基金面上项目(41571514);湖北省重点实验室(三峡大学)开放基金(2016KJZ05)

摘　　要：集成GIS、元胞自动机和神经网络模型对滑坡滑动路径和范围进行模拟研究。采用神经网络模型挖掘出滑坡滑动规律,获取滑坡元胞转换规则,再耦合元胞自动机对滑坡进行模拟。首先确定滑坡关键影响因子,之后利用GIS空间分析能力,结合高精度DEM数据生成地形变量如坡度、坡向、曲率、地形湿度指数等,然后结合BPNN和CA构建滑坡预测模型,最后利用BPNN-CA模型对杉树槽滑坡进行实例分析和模拟,模拟精度较为理想。提出的模型能够为滑坡综合防治和管理提供科学依据,同时也为滑坡的综合研究提供新的视角。The geological hazard is a result of multiple factors, which is represented and modeled by using partial differential e- quations. The geological hazard undeniably possesses complex systems behavior, so geological hazard processes can be adequately modeled and analyzed by complex systems theory and cellular automata. This study integrated GIS, cellular automata and artificial neural network to simulate the flow paths and areas of landslide. The artificial neural network was applied to excavate the slid rules of landslide,and it was also coupled cellular automata to simulate landslide. First of all, we determined the key influencing factors of landslide. After using the GIS spatial analysis ability, high resolution digital elevation model was used to calculate topo- graphic variables such as slope, aspect, curvature and topographic wetness index, etc. Based on BPNN and CA, the landslide predic- tion model was constructed. Finally the developed BPNN-CA simulation model was tested on Shanshucao landslide, and the result of simulation was fairly ideal. The developed model provides scientific basis on comprehensive control and management of land- slide,and provides new thought for the study of landslide.

关 键 词：元胞自动机 神经网络 GIS 滑坡 模拟 

分 类 号：P642[天文地球—工程地质学] TP389[天文地球—地质矿产勘探]