大坝变形的去噪傅里叶模型预测

Prediction of dam deformation based on de-noising Fourier model

针对大坝变形预测中非平稳性且含噪声的数据处理问题,该文提出一种基于剔除含噪声信号的大坝变形傅里叶(Fourier)预测新算法。首先利用经验模态分解(EMD)法将变形时间序列分解成具有不同尺度特征的固有模态函数(IMF)分量,并计算出各分量与原始信号的相关系数;然后根据相关系数剔除含噪声的IMF分量,并采用游程判定法对剩余的波动程度相似或相同的分量进行信号重构;最后利用傅里叶函数对重构后的分量进行曲线拟合,并据此构建大坝变形预测模型,对出现的模型系数采用最小二乘算法求解。经算例验证,并与GM(1,1)、BP神经网络和傅里叶模型对比分析,结果表明该文算法预测精度较高,可用于大坝的变形预测。

To solve the problem of nonstationary and noise data processing in dam deformation prediction,a new algorithm for dam deformation prediction based on de-noising Fourier model was proposed.First,the empirical model decomposition(EMD)method was used to decompose the deformation time series into intrinsic mode function(IMF)components with different scale characteristics and calculate the correlation coefficient between each component and the original signal.Then,the IMF component containing noise was removed according to the correlation coefficient,the reconstructed components were curve fitted by Fourier function,and the dam deformation prediction model was constructed based on the Fourier transform function.The coefficients of the model were resolved by least squares algorithm.Comparing with GM(1,1),back propagation(BP)neural network and Fourier model,the experimental results showed that the algorithm had high accuracy and could be used in dam deformation prediction.