摘要
采用多层前馈神经网络与渗流有限元数值计算相结合的方法,反演了西大洋水库主坝(0+875断面)的渗透参数,并基于反分析的结果对该坝段在高库水位下的稳定渗流和非稳定渗流情况进行了分析与安全评价.此外,分析了观测量(观测项目、测点数量和测点位置等)对反分析结果的影响;提出了对神经网络BP算法的改进措施;在训练实例选取上采用了均匀设计法,有效减少了训练实例数.
The paper combines the multilayere feedforward neural network with seepage finite element numerical calculation method to calculate the seepage coefficient of main dam of Xidayang Reservoir(0+875section).Based on the result of inverse analysis,the steady seepage and unsteady seepage of this dam section under high water level are analyzed and assessed.In addition,the influence of observed quantity(observation items ,the number of measuring point and the location of measuring point )on inverse analysis is analyzed and improvement to BP arithmetic is put forward.In order to reduce effectively training examples,the uniformization design method is adopted.
出处
《水利水电技术》
CSCD
北大核心
2003年第6期53-56,共4页
Water Resources and Hydropower Engineering
基金
河北省自然科学基金资助项目(501175).
关键词
土坝渗流
反分析
人工神经网络
渗透系数
安全评价
西大洋水库
earth dam seepage
inverse analysis
artificial neural network
seepage coefficient
safety assessment
Xidayang Reservoir
