基于局域均值分解和 BP 神经网络的大坝变形预测

摘要

从时频联合域分析角度出发，探讨基于局域均值分解（LMD）和BP神经网络模型的大坝变形预测方法；利用局域均值分解把时间序列分解为多个具有不同尺度的PF分量，然后再用BP神经网络对其进行预测并将结果进行叠加重构合成。实例结果表明，基于LMD-BP的大坝变形预测方法的预测精度高于多项式曲线拟合预测模型，其预测结果与实际监测值相比具有较高的一致性。%A new model based on LMD-BP network modeling for dam deformation prediction is presented . Firstly ,in view of time-frequency analysis ,through the local mean decomposition(LMD) ,deformation time series are decomposed into the components of several different frequency PF components and then the network method is applied to the predicted results .Therefore the predicted value of the forecast horizontal displacements of dam is obtained .T he experimental results indicate that this model has advantage over polynomial curve fitting model and the predicted results can be highly consistent with the actual monitoring and measuring results .