基于广义回归神经网络的无黏性土管涌判定研究  

Study on judgment for piping in non-cohesive soil based on generalized regression neural network

作　　者：薛新华[1,2] 杨兴国[1,2] 

机构地区：[1]四川大学水力学与山区河流开发保护国家重点实验室,四川成都610065 [2]四川大学水利水电学院,四川成都610065

出　　处：《人民长江》2012年第1期42-44,90,共4页Yangtze River

摘　　要：分析了现在广泛采用的判定管涌破坏手段的不足之处。在分析广义回归神经网络的基本原理和算法基础上,建立了无黏性土管涌判别的广义回归神经网络模型。以前人试验结果作为对比,采用特征粒径和孔隙率作为判别指标,对土样的渗透破坏形式进行判别。计算结果表明,该模型的管涌渗流破坏形式判定结果与前人试验结果完全一致,该方法为无黏性土管涌渗流破坏形式的判别提供了新的研究思路。We demonstrate the disadvantages of current judging standard for piping types. A generalized regression neural net- work model for evaluating the piping in non - cohesive soil is established, on the basis of the analysis of fundamental theory and algorithm of generalized regression neural network. Taking the previous studies as contrast examples, we judge the piping types in non -cohesive soil by adopting characteristic particle size and void ratio as evaluation indexes, and the results are in conformity with previous studies.

关 键 词：管涌 广义回归神经网络 无黏性土 流土 

分 类 号：TU441[建筑科学—岩土工程]