Deformation Data Analysis And Prediction Research Based On Empirical Wavelet Transform

High-precision deformation data processing is a complex and important task,and the reliability of the processing results will have a huge impact on the analysis of the operating state of the deformed body.Traditional deformation data processing methods such as wavelet and empirical modal decomposition have problems of modal aliasing and end-point effects during processing,which seriously affect the accuracy of the processing results and are not conducive to obtaining the true and accurate operating status information of the building.Based on the signal processing method—Empirical Wavelet Transform(EWT),this paper studies a more effective deformation data processing method.The research contents and experimental results are as follows:(1)Verify the decomposition ability of EWT,the results show that EWT smaller wave and empirical mode decomposition have better ability to separate mixed signals and have stronger anti-modal aliasing ability;EWT has spectrum when processing complex deformation data Over-segmentation problem,this paper improves the EWT spectrum division method,studies an improved algorithm IEWT(Improved Empirical Wavelet Transform,IEWT),and gives a plan to determine the number of IEWT spectrum division,and uses engineering actual deformation data to IEWT The decomposition effect of the algorithm was verified.(2)Study the application of IEWT method in the preprocessing of deformation data.Combining IEWT and Inter Quartile Range(IQR),a new method of gross error detection based on IEWT-IQR is studied,and the effectiveness of the method is verified by simulation experiments and engineering experiments;IEWT and wavelet The transformation method is combined to study a new denoising method based on IEWT-WV.The feasibility and effectiveness of the IEWT-WV method for denoising deformation data are verified through simulation experiments and engineering experiments;the IEWT-WV method is applied to The multi-path observation data is denoised and error corrected.The correction results show that the multi-path error correction results based on the IEWT-WV method in the three directions of X,Y,and H are better than the db4 wavelet and EMD threshold methods,which confirms that the IEWT-WV The method is used to weaken the feasibility and effectiveness of multipath effect errors.(3)According to the changing characteristics of surface subsidence in the mining area,IEWT fusion adaptive kernel learning based relevance vector machine(aRVM),autoregressive integrated moving average model(ARIMA),Construct IEWT-aRVM-ARIMA multi-scale combined deformation prediction model.The model uses IEWT to decompose the subsidence data into two time scales with different physical meanings,namely trend and random fluctuations,which makes the change characteristics of each component more obvious,and then uses the aRVM method and ARIMA method to predict the time series.The model is used to predict surface settlement in Xihaozhuang mining area,and compared with linear regression,BP neural network and aRVM.The experimental results show that the IEWT-aRVM-ARIMA model can obtain relatively better prediction accuracy and is a better method for predicting settlement deformation.