Abstract: Noise reduction processing of the draft tube pressure signals is of great significance for vibration fault identification of hydropower units; a key issue for a power plant is to process the interference component of such signals to eliminate its effect in subsequent signal analysis. To address the issue, this study used the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMD-AN) to decompose the signal into several Intrinsic Mode Functions, and used the Improved Maximum Information Coefficient (IMIC) to calculate the correlation between each intrinsic mode function and the original signal. A threshold value is determined based on the correlation coefficient calculated, and finally the intrinsic mode functions were processed and superimposed to reconstruct the pressure signal. The efficacy of this method was verified using an example analysis, and compared with Empirical Mode Decomposition and the traditional Maximum Information Coefficient method. The results show our new method IMIC-CEEMDAN has higher reliability in the purification of pressure data for hydropower units.

Key words: pressure data, complete ensemble empirical mode decomposition with adaptive noise, improved maximal information coefficient, correlation analysis, data filtering