摘要
泥石流平均流速是防治工程设计的重要参数,其设计预测精度直接影响工程投资.应用模糊优选神经网络模型,以模糊可变识别模型当优化准则参数等于2,距离参数等于1的特例为神经网络激励函数,研究隐含层在不同隐节点数情况下的泥石流平均流速预测精度,并以精度最高的隐节点数构建神经网络的拓扑结构,对云南蒋家沟黏性泥石流平均流速进行预测.研究结果表明,预测精度较高,有参考应用价值.
Debris flow average velocity is one of the basic parameters in civil engineering design, and its forecasting precision directly influences the investment. Fuzzy optimization neural network （FONN） model is applied to debris flow average velocity＇s forecasting. That is using the fuzzy variable recognition model in which the parameter of optimization criterion is equal to 2 and parameter of distance is equal to 1 as the activation function of neural network. Based on the study of forecasting precision in different numbers of the hidden node, and the topological structure with highest precision is selected for viscous debris flow average velocity＇s forecasting in Jiangjia Ravine. Research results show that the precisions of fitting and forecasting are satisfactory and can be useful for design.
出处
《大连理工大学学报》
EI
CAS
CSCD
北大核心
2008年第6期887-891,共5页
Journal of Dalian University of Technology
基金
国家自然科学基金资助项目(50779005)
水利部科技创新资助项目(SCXC2005-01)
关键词
模糊优选
神经网络
预测
泥石流
平均流速
fuzzy optimization
neural network
forecasting
debris flow
average velocity
