基于BIC-PCA和LMD的朔黄铁路边坡变形预测方法

Prediction method for slope deformation of Shuohuang railway based on BIC-PCA and LMD

针对全球卫星导航系统(GNSS)监测数据处理中噪声抑制和变形信息提取精度不高等问题,提出一种联合使用贝叶斯信息准则(BIC)-主成分分析(PCA)和局部均值分解(LMD)的GNSS铁路边坡变形数据处理及信息提取方法:考虑PCA主分量个数确定,将贝叶斯信息准则引入PCA建立BIC-PCA模型;进而利用BIC-PCA对变形监测数据进行分析,实现噪声抑制;然后利用LMD算法对噪声抑制后的监测数据进行分析,从中提取周期项、趋势项和波动项等隐含的变形信息;最后建立支持向量回归(SVR)模型,对未来变形趋势进行预测。实验结果表明,所提方法预测精度较高且噪声稳健性较强,预测结果的均方根(RMS)误差和平均预测误差(APRE)分别为6.30和7.26,远小于反向传播(BP)神经网络和灰色GM(1,1)模型。

Aiming at the challenges of noise suppression and deformation information extraction in the processing of global navigation satellite system(GNSS)monitoring data,the paper proposed a method for data processing and information extraction of GNSS railway slope deformation using a combination of Bayesian information criterion(BIC)-principal component analysis(PCA)and local mean decomposition(LMD)algorithms:considering determining the number of principal components in PCA,the BIC was introduced into PCA to establish a BIC-PCA model;and the BIC-PCA was used to analyze the deformation monitoring data to achieve noise suppression;then,the LMD algorithm was used to analyze the monitoring data after noise suppression,and the implicit deformation information such as periodic terms,trend terms and fluctuation terms were extracted;finally,a support vector regression(SVR)model was established to predict future deformation trends.Experimental results showed that the proposed method would have high prediction accuracy and strong noise robustness,the root mean square(RMS)error and average prediction error(APRE)of the prediction results could be 6.30 and 7.26,respectively,which are much smaller than those of BP and GM(1,1)methods.