摘要
为了解决滑坡地质灾害传统预测方法中出现的综合性、实用性不强等问题,本文研究用基于优化参数设置的BP神经网络模型来预测滑坡地质灾害.该方法基于BP神经网络,顾及与滑坡地质灾害产生紧密相关的地质条件和环境因素,对BP神经网络的输入层、隐含层、输出层的参数进行优化;再由历史的经验数据通过训练、泛化建立基于BP神经网络的地质灾害预测模型;最后,按照0和1的组合结果对滑坡地质灾害进行预测.本文利用该模型对汶川地震诱发的滑坡地质灾害进行分析预测,结果表明：该模型的预测结果与实际结果吻合度达到86％～90％,预测精度较高,验证了基于改进的BP神经网络预测滑坡地质灾害的方法是实际可行的.
In order to solve the problems in the traditional forecasting methods such as the weak comprehensiveness and practicality, BP neural network algorithm is used to predict landslides based on optimization of parameter settings. Considering the geological conditions and environmental factors which are closely related with the landslide, the parameters of the BP neural network input layer, hidden layer and output layer are optimized. Then the prediction model of geological disasters is established based on BP neural network through training, generalizing the historical experience data, so as to predict landslides according to the combination of 0 and 1. The model has been used to analyze and forecast landslides which were induced by Wenchuan earthquake. Results show that the goodness of fit between model predictions and actual results reached 86%. The high prediction accuracy proves that the forecasting method of landslides based on improved BP neural network is practical.
出处
《工程勘察》
2014年第8期55-60,共6页
Geotechnical Investigation & Surveying
基金
江苏省高校自然科学研究面上项目(12KJB420003)
2012年苏州市发改委科技服务专项
关键词
滑坡灾害
BP神经网络
参数设置
影响因子
预测方法
landslide
BP neural network
parameter settings
influencing factors
forecasting methods
