Loaded coal-rock temperature denoising algorithm based on CEEMD and adaptive NIWT with NIWOA

Abstract

This paper examines the composition and characteristics of loaded coal-rock temperature collected directly. And a new denoising algorithm for loaded coal-rock temperature is proposed, using complete ensemble empirical mode decomposition (CEEMD) and adaptive nonlinear improved wavelet transform (NIWT) with novel improved whale optimization algorithm (NIWOA). Then the effect of the proposed algorithm’s each part is discussed. Finally, the rationale and advancement of the proposed algorithm is validated with data from uniaxial experiments. It is shown that CEEMD can entirely split the original temperature into three categories: environmental component, noise-dominated mode function, and signal-dominated mode function. And in contrast to similar, the reconstruction strategy combining CEEMD and the noise similarity based on ACF is the most effective for denoising, whose results’ SNR and RMSE are greater than 52 dB and less than 0.06, respectively. In addition, both NIWOA and NIWT have significantly improved the optimization speed, solving accuracy and denoising effect compared with the original. Their combination can effectively denoise the noise-dominated modal function, and the resulting SNR and RMSE can reach 53.6831 dB and 0.0281. Furthermore, the experimental analysis demonstrates that the proposed coal-rock temperature denoising algorithm has good performance and robustness, with results that are smooth and complete and INNR can be as low as 0.00968, which is clearly superior to other similar or current ones. The research can be useful for further studies of the evolution of the temperature field and the damage location method for loaded coal-rock fracture.