Denoising Method for Deformation Monitoring Data Based on ICEEMD-ICA and MDP Principle

  Objectives  Considering the inaccurate separation of signal and noise of empirical mode decomposition (EMD) method and the uncertainty of independent component analysis (ICA), a new method for denoising deformation data with improved complete ensemble empirical mode decomposition (ICEEMD), independent component analysis (ICA) and minimal distortion principle (MDP) is proposed.

  Methods  Firstly, ICEEMD method is used to decompose the deformation monitoring data effectively, and the virtual noise signal is constructed. Secondly, ICEEMD decomposition of virtual noise is carried out to extract twice virtual noise signal which is closer to real noise. The input observation channel is composed of twice virtual noise and original deformation data and processed by ICA. Then, by calculating the correlation coefficient between the independent components and the input signal after ICA processing, the sorting uncertainty and phase uncertainty of independent components can be solved. Finally, the MDP criterion is used to effectively solve the amplitude uncertainty of independent components.

  Results  Through the detailed analysis of noisy simulation data and actual bridge GNSS deformation monitoring data, the results show that the proposed method has achieved good denoising effect and can effectively improve the performance of denoising.

  Conclusions  It also fully verified the feasibility and effectiveness of the proposed method indenoising of deformation monitoring data.