Abstract：We propose a random noise suppression method CEEMD-PCA based on principal component analysis (PCA) optimized complementary ensemble empirical mode decomposition (CEEMD) for DSQ water tube tiltmeter signal. The method incorporates eight IMF component quality evaluation indexes, such as correlation coefficient, distribution entropy, MSE, R2, SSE, RMSE, MAE, MAPE; it implements dimensionality reduction and compression of the index value matrix with the help of principal component analysis to transform it into a new parameter that can represent the characteristics of all different types of indexes, and constructs a comprehensive IMF component quality evaluation function to complete the original noise-containing signal according to the score ranking results. We complete the linear reconstruction of the original noisy signal according to the score ranking results. The results of both simulated and measured signal denoising experiments show that the CEEMD-PCA model outperforms the classical models such as Kalman filter, 70 th-order low-pass FIR filter, etc., improves the signal-to-noise ratio of the original signal, and accurately completes the signal reconstruction, which can better retain the effective components.
GUO Xiaofei,OU Tonggeng,LIU Tianlong. A Random Noise Suppression Method for DSQ Water Tube Tiltmeter Signals Based on CEEMD and PCA[J]. jgg, 2024, 44(9): 978-984.
