一种联合π定理和神经网络的试验数据挖掘模式  

A pattern of experiment data mining based on π theorem combined with ANN

作　　者：顾正华[1] 唐洪武[2] 段子冰[3] 

机构地区：[1]浙江大学建筑工程学院,浙江杭州310028 [2]河海大学水电学院,江苏南京210098 [3]南京水利科学研究院,江苏南京210024

出　　处：《海洋工程》2006年第3期85-90,94,共7页The Ocean Engineering

基　　金：水利部科技创新项目(XDS2004-04);南京水利科学研究院开放流动研究基金项目(Yk90501)

摘　　要：为了解决试验数据分析手段的不足和单一问题,联合人工神经网络理论和π定理,提出一种试验数据的挖掘模式,并把它应用于快速船伴流场特性分析中,结果表明该数据挖掘模式不但能有效地克服前述问题,还能降低物理模型试验对试验组次的要求,得到量纲和谐的、具有统一形式的非线性公式,有助于提升物理模型试验的知识发现层次。In the past artificial analysis of experimental data, those rich measured data were not utilized enough and effectively because of relatively unadvanced and simple analysis means in two aspects： 1） some physical knowledges which can be discovered from the data were not found adequately; 2） some physical rules were misapprehensively judged out due to the influence of a few inaccurate data, To solve those problems, a pattern of experiment data mitring is presented in this paper by combining artificial neural networks theory with π theorem. It was applied to natural analysis of greyhound＇ s wake field, and results indicate that the pattern can not only solve the above problems but also reduce the need of physical model experiment to group test number, and gain an unlinear formula which has harmonious dimension and uniform format. The pattem will improve the level of physical model experiment in KDD （ Knowledge Discovery from Database）.

关 键 词：数据挖掘 π定理 人工神经网络 物理模型试验 快速船伴流场 

分 类 号：TV131[水利工程—水力学及河流动力学] TP18[自动化与计算机技术—控制理论与控制工程]