Blasting Vibration Signal Denoising based on CEEMDAN-K-means Algorithm

In view of the problem of noise and information loss in the CEEMDAN method in the denoising process of actual measurement blasting vibration signals, the clustering analysis method is considered to have good data processing ability. Based on the idea of decomposition-clustering-reconstruction, CEEMDAN-K-means algorithm for denoising of blasting vibration signals is proposed. Firstly, this method decomposes the blasting vibration signal by CEEMDAN method to obtain IMF components of different quantity levels. Then, the K-means clustering analysis algorithm is used to classify the IMF components into five different categories, and variance contribution rate verification is used. Finally, the IMF components of high frequency noise category are removed and the reconstructed pure blasting vibration signal is obtained. Taking the blasting vibration signals from an open-pit mine as example, the signal denoising performance of the CEEMDAN-K-means algorithm was evaluated by signal-to-noise ratio and root mean square error indexes. The research results show that compared with the CEEMDAN method and the EMD-wavelet threshold method, the CEEMDAN-K-means signal denoising method has the largest signal-to-noise ratio (20.06 dB), which is increased by 1.26 dB and 7.7 dB, respectively, and the smallest root mean square error (0.22 10^-3), indicating that the method not only has good denoising effect, but also has good fidelity. Through the comparison and analysis of the denoising effect of different methods, it is known that on the basis of effectively retaining the real signal component, the CEEMDAN-K-means method can effectively remove the high-frequency components contained in the measured blasting vibration signal, and has practicality and effectiveness in the field of blasting vibration signal denoising.