人民黄河
人民黃河
인민황하
Yellow River
2014年
1期
109-110,114
,共3页
BP神经网络%强度折减法%极限平衡法%反演%强度参数
BP神經網絡%彊度摺減法%極限平衡法%反縯%彊度參數
BP신경망락%강도절감법%겁한평형법%반연%강도삼수
BP neural network%strength reduction method%limit equilibrium%inverse analysis%strength parameters
金川水电站泄洪洞进口自然边坡高陡，岩层发生明显倾倒变形，岩体质量差，对边坡的稳定不利，采用BP神经网络方法对其强度参数进行反演分析。根据边坡的重要性和规范确定边坡的最低安全系数，由边坡目前的变形情况确定反演工况和滑带。利用极限平衡法组织神经网络训练样本，将安全系数作为网络输入，黏聚力和内摩擦角作为网络的输出，并将其代入极限平衡和有限元强度折减模型进行检验，结果表明：BP神经网络与极限平衡法计算的安全系数的误差在1％以内，有限元强度折减法计算边坡浅层变形体的平均抗剪强度参数和安全系数，也与BP神经网络计算得到的结果相符，说明BP神经网络进行参数反演是可靠的。
金川水電站洩洪洞進口自然邊坡高陡，巖層髮生明顯傾倒變形，巖體質量差，對邊坡的穩定不利，採用BP神經網絡方法對其彊度參數進行反縯分析。根據邊坡的重要性和規範確定邊坡的最低安全繫數，由邊坡目前的變形情況確定反縯工況和滑帶。利用極限平衡法組織神經網絡訓練樣本，將安全繫數作為網絡輸入，黏聚力和內摩抆角作為網絡的輸齣，併將其代入極限平衡和有限元彊度摺減模型進行檢驗，結果錶明：BP神經網絡與極限平衡法計算的安全繫數的誤差在1％以內，有限元彊度摺減法計算邊坡淺層變形體的平均抗剪彊度參數和安全繫數，也與BP神經網絡計算得到的結果相符，說明BP神經網絡進行參數反縯是可靠的。
금천수전참설홍동진구자연변파고두，암층발생명현경도변형，암체질량차，대변파적은정불리，채용BP신경망락방법대기강도삼수진행반연분석。근거변파적중요성화규범학정변파적최저안전계수，유변파목전적변형정황학정반연공황화활대。이용겁한평형법조직신경망락훈련양본，장안전계수작위망락수입，점취력화내마찰각작위망락적수출，병장기대입겁한평형화유한원강도절감모형진행검험，결과표명：BP신경망락여겁한평형법계산적안전계수적오차재1％이내，유한원강도절감법계산변파천층변형체적평균항전강도삼수화안전계수，야여BP신경망락계산득도적결과상부，설명BP신경망락진행삼수반연시가고적。
The characteristics of the inlet of flood discharging tunnel of Jinchuan Hydraulic Power Plant are natural high and steep slope,obvious inclined deformation of rock stratum and poor quality of rock mass,which is adverse to the stability of the slope. It chose BP neural network method to inverse analysis of the intensity parameter. Determining the minimum safety factor of slope according to the importance of the slope and norms, and deciding the inversion conditions and slip zone on the base of slope deformation. Neural network training samples were organized by limit equi-librium,the safety factor as the network input,C andφas the output,and taking them as the limit equilibrium and finite element strength reduc-tion calculation parameters to test. The results show that the error of BP neural network and limit equilibrium methods is within 1%,and finite ele-ment strength reduction calculating the average shear strength parameters and safety factor of shallow deformation are also consistent with BP neural network calculation results ,which indicates that the BP neural network to parameter inverse analysis is reliable.
