【摘要】 受岩土体性质和多种诱发因素(降雨、地震、库水位等)的影响,滑坡的位移演化必然包含着多层次的信息,难以采用单一的灰色预测、神经网络、时间序列等模型进行有效、准确的预测。本文引入信号分析中的经验模态分解(EMD)方法,同时结合对非线性函数具有很强的逼近能力的BP神经网络,建立了滑坡变形预测的EMD-BP神经网络模型,并以三峡库区树坪滑坡为例,应用该模型对ZG88监测点的累积位移进行了预测,最后将其预测值与实际监测值以及其他模型预测结果进行了对比分析,结果表明该模型的预测值与实际监测值基本一致,且模型预测精度比单一的BP神经网络和GM(1,1)模型要高,从而验证了该模型具有一定的适用性和较高的预测精度。更多还原

【Abstract】 Affected by the properties of rock and soil mass and various inducing factors(rainfall,earthquake,reservoir level,etc.),the evolution of landslide displacement implies multi-level information,so it is difficult to obtain accurate prediction results by a single grey forecasting method or time series method.This paper introduces the empirical mode decomposition(EMD) method from signal analysis,and meanwhile,combining BP neural network which has strong approximation for nonlinear function,establishes an EMD-BP neural network model to predict landslide deformation.Taking Shuping landslide in Three Gorges Reservoir area as an example,the paper predicts the deformation of ZG88 monitoring point,and compares the prediction values with the prediction results of other forecast models.The results show that the prediction results of the proposed model are basically identical with the values of actual monitoring,which improves the precision of the prediction results.更多还原