Vibration signal de-noising based on improved EMD algorithm

Abstract

In order to solve the problem of poor performance of EMD (empirical mode decomposition） filter de-noising for vibration signal, an adaptive orthogonal decomposition signal de-noising method PEMD (principal empirical mode decomposition) is proposed. This algorithm combines the self-adaptability of EMD decomposition and the complete orthogonality of principal component analysis (PCA), eliminates the phenomenon of mode aliasing in the process of signal EMD decomposition, and obtains the best de-noising effect. The results showed that compared with EMD and EEMD (ensemble empirical mode decomposition）, PEMD (principal component analysis) improved 1.15 dB and 0.38 dB respectively in the simulation test, and the root-mean-square error was the smallest. In frequency domain, PEMD has the highest sensitivity to the frequency of simulation signal (30 Hz), and the noise filtering effect is the best outside 30 Hz. In the blasting vibration test, PEMD and EEMD had better performance in removing burrs, and PEMD had the best performance in preserving medium and low frequency vibration signals at 0−300 Hz, and the best performance in filtering high frequency noises above 300 Hz.