基于模糊神经网络的泥石流危险性评价  

Assessment of debris flow hazard based on Fuzzy Neutral Network

作　　者：曹禄来 徐林荣[1] 陈舒阳[1] 接小峰 

机构地区：[1]中南大学土木工程学院,湖南长沙410075

出　　处：《水文地质工程地质》2014年第2期143-147,共5页Hydrogeology & Engineering Geology

基　　金：泥石流危害桥隧工程承灾链特征与工程易损度动态评价方法研究(41272376);高速铁路灾害防治与应急救援技术措施研究(2011G017-C)

摘　　要：将T-S模糊系统理论和人工神经网络相结合,利用模糊理论隶属度对模糊性有很强的识别精度,将泥石流危险性指标隶属度作为神经网络的激活函数输入,使用正态分布方法产生训练数据,再利用BP神经网络的误差反向传播对TS模糊系统隶属度函数等的参数进行训练调整,从而建立泥石流危险性评价的模糊神经网络模型。利用建立好的模型对云南东川八条典型泥石流沟的危险性进行评价,获得客观合理的评价结果。与刘希林的灰色聚类法和可拓物元方法的纵向对比和与线性内插产生训练数据方法横向对比表明:使用该方法能较真实地反映实际泥石流沟的危险性等级,证明模糊神经网络理论应用于泥石流危险性评价的有效性和可行性。Combined with the T -S Fuzzy System theory and Neutral Network, taken advantage of the fuzzy theory＇s high recognition accuracy towards fuzziness, this paper takes the indices membership of debris-flow hazard assessment as the input of neuronal activation function, applies the Gaussian distribution to produce the training data, employs the BP neutral network＇s error back propagation function to adjust the parameters of the T-S fuzzy system, and establishes the debris-flow hazard assessment model based on the Fuzzy Neutral Network theory. The model is employed to assess the 8 classic debris flow of the Dongchuang city, and reasonable results are obtained. After the longitudinal comparison in which the other assessment methods are used and the horizontal comparison in which the linear interpolation method is used, we come to the conclusions that the method can reflect the risk level of debris flow truthfully, and proves the effectiveness and feasibility of the fuzzy neutral network theory.

关 键 词：模糊神经网络 T—S模糊系统 泥石流 危险性评价 

分 类 号：P642.23[天文地球—工程地质学]