解放军理工大学学报(自然科学版)
解放軍理工大學學報(自然科學版)
해방군리공대학학보(자연과학판)
JOURNAL OF PLA UNIVERSITY OF SCIENCE AND TECHNOLOGY(NATURAL SCIENCE EDITION)
2014年
6期
552-557
,共6页
砂土地震液化%液化势分类%Bayes判别分析
砂土地震液化%液化勢分類%Bayes判彆分析
사토지진액화%액화세분류%Bayes판별분석
seismic liquefaction of sandy soil%classification of liquefaction potential%Bayes discriminant analysis method
为了给砂土地震液化的数量化研究提供参考,基于多元统计分析理论,建立砂土地震液化判别与液化势分类的Bayes判别分析模型。模型选用震级、地面加速度最大值、标准贯入击数、比贯入阻力、相对密实度、平均粒径和地下水位等7个指标作为判别因子;将砂土液化势分为严重液化、中等液化、轻微液化和未液化4个级别,并作为Bayes判别分析的4个正态总体;以17个砂土实测数据作为训练样本,建立Bayes线性判别函数,以Bayes线性判别函数的最大值对应的总体作为样品所归属的总体;最后将建立的模型对训练样本进行回判,以回代估计误判率对模型进行检验。研究表明,对训练样本的回代误判率为0,对另外20个砂土样本的判别正确率为90%。
為瞭給砂土地震液化的數量化研究提供參攷,基于多元統計分析理論,建立砂土地震液化判彆與液化勢分類的Bayes判彆分析模型。模型選用震級、地麵加速度最大值、標準貫入擊數、比貫入阻力、相對密實度、平均粒徑和地下水位等7箇指標作為判彆因子;將砂土液化勢分為嚴重液化、中等液化、輕微液化和未液化4箇級彆,併作為Bayes判彆分析的4箇正態總體;以17箇砂土實測數據作為訓練樣本,建立Bayes線性判彆函數,以Bayes線性判彆函數的最大值對應的總體作為樣品所歸屬的總體;最後將建立的模型對訓練樣本進行迴判,以迴代估計誤判率對模型進行檢驗。研究錶明,對訓練樣本的迴代誤判率為0,對另外20箇砂土樣本的判彆正確率為90%。
위료급사토지진액화적수양화연구제공삼고,기우다원통계분석이론,건립사토지진액화판별여액화세분류적Bayes판별분석모형。모형선용진급、지면가속도최대치、표준관입격수、비관입조력、상대밀실도、평균립경화지하수위등7개지표작위판별인자;장사토액화세분위엄중액화、중등액화、경미액화화미액화4개급별,병작위Bayes판별분석적4개정태총체;이17개사토실측수거작위훈련양본,건립Bayes선성판별함수,이Bayes선성판별함수적최대치대응적총체작위양품소귀속적총체;최후장건립적모형대훈련양본진행회판,이회대고계오판솔대모형진행검험。연구표명,대훈련양본적회대오판솔위0,대령외20개사토양본적판별정학솔위90%。
To provide reference for the quantitative study on seismic liquefaction of sandy soil, Bayes discriminant model ( BDM ) for identification and classification of seismic liquefaction potential of sandy soil was established based on multivariate statistical analysis theory. According to the analysis of some influencing factors of sand lique-faction, seven parameters including earthquake magnitude, peak ground acceleration, the value of standard pene-tration test, specific penetration resistance, relative compaction, mean particle diameter, and underground water table were selected as the indexes for synthetic evaluation of the seismic liquefaction of sandy soil. The seismic liq-uefaction potential of sandy soil was divided into four grades, i. e. , serious liquefaction, medium liquefaction, slight liquefaction and non-liquefaction, and regarded as four normal collectivities in Bayes discriminant analysis. Bayes discriminant functions obtained through training 17 sandy soil samples were employed to compute the Bayes function values of samples, and the maximal function value was used to judge which collectivity the sample belongs to. The established Bayes discriminant analysis model was used to back-discriminate the training samples, and veri-fied by the ratio of mistake-discrimination. The study indicates that the ratio of mistake-discrimination of training samples is zero. The other twenty sets of sand samples regarded as testing samples were assessed by BDM and the correct rate is 90%.