CAJ | 학술논문

结合南水北调中线工程某高填方渠道工程现场观测数据，利用 BP 神经网络中多层前馈神经网络及其反传算法，建立了渠道施工沉降和土体力学参数的多元输入输出连续映射分析模型，从而对岩土参数进行位移反演分析，得到填筑渠道土体的力学参数值。反演结果表明，土体凝聚力和内摩擦角反演相对误差较小，而压缩模量的反演误差随着填筑高程的升高有变大的趋势。所得研究结论为计算和分析填方渠道的有效沉降提供了理论支持。
결합남수북조중선공정모고전방거도공정현장관측수거，이용 BP 신경망락중다층전궤신경망락급기반전산법，건립료거도시공침강화토체역학삼수적다원수입수출련속영사분석모형，종이대암토삼수진행위이반연분석，득도전축거도토체적역학삼수치。반연결과표명，토체응취력화내마찰각반연상대오차교소，이압축모량적반연오차수착전축고정적승고유변대적추세。소득연구결론위계산화분석전방거도적유효침강제공료이론지지。
Based on in - situ nonitoring data of a high filling channel in Middle Route Project of South - to - North Water Di-version,a nodel with continuous napping between input and output for channel settlenent and soil nechanical paraneters is es-tablished by using nultilayer feed - forward neural networks and back analysis of BP neural network. The nodel is applied to cal-culation of soil nechanical paraneters by defornation back analysis. The results show that the relative error of cohesion and fric-tion angle is slight;however the conversion error of conpression nodule is increasing with filling elevation. The conclusions pro-vide theoretical support for settlenent calculation of high filling channel.