CAJ | 학술논문

成都地铁1，2号线盾构在砂卵石地层施工诱发多次地面塌陷事故，其地面塌陷变形曲线受多种因素影响，且各因素对地面塌陷变形曲线的影响表现出非线性特性，因此地面塌陷变形曲线很难用显示的数学表达式进行求解。最小二乘支持向量机是基于统计学习理论的机器学习方法，该方法能避免传统神经网络诸多缺陷，能够分析复杂因素对结果影响的潜在规律，据此引入最小二乘支持向量机，以地层物理力学参数、盾构埋深和地层损失数量为输入参数，建立地面塌陷变形曲线预测模型。经过样本检验，预测模型具有较强的泛化能力，预测结果精度和可靠性较高。
성도지철1，2호선순구재사란석지층시공유발다차지면탑함사고，기지면탑함변형곡선수다충인소영향，차각인소대지면탑함변형곡선적영향표현출비선성특성，인차지면탑함변형곡선흔난용현시적수학표체식진행구해。최소이승지지향량궤시기우통계학습이론적궤기학습방법，해방법능피면전통신경망락제다결함，능구분석복잡인소대결과영향적잠재규률，거차인입최소이승지지향량궤，이지층물리역학삼수、순구매심화지층손실수량위수입삼수，건입지면탑함변형곡선예측모형。경과양본검험，예측모형구유교강적범화능력，예측결과정도화가고성교고。
When shield crossed the sandy cobble stratum in Chengdu metro No.1 and No.2 lines,the induced surface subsidence reached as high as dozens of times. Surface collapse deformation curve is influenced by many factors,and the influences of the factors on surface collapse deformation curve show the nonlinear characteristics. So the surface collapse deformation curve is difficult to solve with mathematical formulas. Least squares support vector is a machine learning method based on the statistical learning theory. It can avoid shortcomings of traditional neural network and analyze influencing rule on the result with complicated factors. Thereby least squares support vector machine method was introduced to establish prediction model for surface collapse deformation. Prediction model took physico-mechanical parameters of stratum,buried depth of shield and ground loss value as input parameters. Though testing sample data,the prediction mode has strong generalization ability, and its prediction result has high accuracy and reliability.