Research On Denoising And Deformation Trend Extraction Of Deformation Monitoring Data Baesd On Empirical Mode Decomposition

Empirical mode decomposition does not require any prior knowledge.Based on the signal's own time feature scale,it can adaptively decompose the signal into a finite number of intrinsic modal functions with high to low frequencies.There is no wavelet basis selection and decomposition stages confirmation such as wavelet transform,having unique advantages in dealing with non-stationary complex signals.Therefore,it is of great research value and broad application prospects to apply EMD decomposition to the related processing of deformation monitoring data comprehensively affected by complex factor in the affected area.First of all,the author deeply studies the EMD decomposition algorithm model,theoretically analyzes the causes of the end-effects and modal aliasing problems in the decomposition process,and demonstrates them intuitively.At the same time,the author introduces radial basis function neural network prediction(RBF)and a complete global empirical mode decomposition(CEEMD,also known as complete EMD)to attenuate or eliminate the impact of two major problems on the decomposition results.Secondly,considering the measurement errors caused by various interference factors in the intelligent total station and GNSS deformation monitoring data,according to the distribution of noise energy in the deformation monitoring data decomposition components,the threshold processing principle of wavelet denoising is used for reference,and the improved complete EMD-threshold processing treatment is proposed.Aiming at IMF components containing more noise energy distribution,carry out threshold processing to obtain denoising sequences,make comparative analysis with two commonly used denoising results based on complete EMD decomposition to prove the improved denoising method for further improvement of the denoising effect.Finally,when the deformation monitoring data is not denoised,the Hilbert transform of the IMF component is used to obtain the Hilbert marginal spectrum of each component,and the Hilbert marginal spectrum analysis method based on the weighted energy distribution is introduced.Selecting the low-frequency IMF component with global slowly changing information for local reconstruction can achieve deformation monitoring data deformation trend extraction.By extracting the deformation trend of intelligent total station and GNSS co-point joint deformation monitoring data,the reliability and validity of the deformation trend extraction method based on EMD decomposition is verified according to the consistency of the trend curve change.In summary,the author improves the quality of decomposition results by improving or solving the main problems of EMD decomposition,proposes an improved denoising method,and introduces deformation trend extraction methods in other fields for verification data by simulation data experiments.Finally,the two methods are applied to measured data processing of both deformation monitoring of intelligent total station and the GNSS measurement,and satisfactory results are all obtained.