边坡变形的LMD-ELM多尺度预测模型研究

Research on LMD-ELM-based multi-scale prediction model of slope deformation

为准确预测边坡变形,有效预防边坡灾害发生,提出构建基于局域均值分解(LMD)和极限学习机(ELM)的边坡变形多尺度预测模型。用LMD方法,将边坡变形时间序列分解为多尺度且相对平稳的随机项、周期项和趋势项。针对各项时间序列,分别构建基于ELM的预测模型。经累加各分项预测值,获得模型最终预测结果。以甘肃某边坡变形为案例,进行实证分析。结果表明:LMD-ELM模型能够充分挖掘数据内部隐含的变形规律,有效诠释多尺度变形与其诱发因素间复杂的响应关系,预测精度、运行速度和拟合泛化能力较其他模型有所提高。

In order to predict the deformation of slope accurately and prevent the occurrence of slope disasters effectively,a multi-scale model of slope deformation based on LMD and ELM was built. Through the LMD method,the slope deformation time series was decomposed into a random term,a periodic term and a trend term. For each time series,a prediction model was built respectively based on ELM. By accumulation of itemized predicted values,the final prediction results were obtained. A certain slope in Gansu Province was taken as an example for carrying out deformation prediction empirical analysis. The results show that the model based on LMD-ELM can reveal fully the internal rule of the deformation observation data,interpret the complex response relationships between the multi-scale deformation and its inducing factors effectively,and that the prediction accuracy,running speed and fitting generalization ability of this model are improved compared with other models.