Friction Signal Denoising Using Complete Ensemble EMD with Adaptive Noise and Mutual Information

Abstract

During the measurement of friction force, the measured signal generally contains noise. To remove the noise and preserve the important features of the signal, a hybrid filtering method is introduced that uses the mutual information and a new waveform. This new waveform is the difference between the original signal and the sum of intrinsic mode functions (IMFs), which are obtained by empirical mode decomposition (EMD) or its improved versions. To evaluate the filter performance for the friction signal, ensemble EMD (EEMD), complementary ensemble EMD (CEEMD), and complete ensemble EMD with adaptive noise (CEEMDAN) are employed in combination with the proposed filtering method. The combination is used to filter the synthesizing signals at first. For the filtering of the simulation signal, the filtering effect is compared under conditions of different ensemble number, sampling frequency, and the input signal-noise ratio, respectively. Results show that CEEMDAN outperforms other signal filtering methods. In particular, this method is successful in filtering the friction signal as evaluated by the de-trended fluctuation analysis (DFA) algorithm.

The Evaluation of Noise Spectroscopy Tests

“…The behavior of power spectral density P r /∆f [Ws]: (a) the frequency range of f~0.1 MHz-1000 THz and temperatures between 10 and 1000 K, according to relation (9); (b) in the frequency range of f~100 MHz-10 GHz and at temperatures of T = 100-1000 K, from Planck's law of a black body, relation (9).…”

“…For higher frequencies, a more complex definition is used [14]: However, if the relationship between the power spectral density Pr/∆f and the frequency f has to be captured as either the broadband task or individual analyses, then, given that the frequency range f extends over several decades or reaches above 3 GHz, we can observe a difference between the simplified Expressions (12) and (13) and the one derived from the radiation of hemisphere (9). By extension, when comparing the spectral power densities represented in Figures 6-8, we can also notice marked differences in the behavior of the functions.…”

“…This condition is caused by the broadband signal and its interaction with the measured element. Several related papers [4][5][6][7][8][9] introduce and discuss various applications of noise spectroscopy, utilizing tests of periodic materials and metamaterials. The research performed to date has nevertheless shown that, for evaluating and comparing the quantitative values monitored during the assessment of properties of periodic structures via noise spectroscopy, the approach based on the description of the frequency spectrum is considerably ineffective and unclear.…”

“…The actual processing is then performed with powerful tools of filtering as a means of signal modification to facilitate the retrieval of a minor portion of parameters of the large-scale system being examined or described. In this connection, the technique designed by Li et al [9], who exploit hybrid filtering, appears to be a promising approach that eliminates spurious effects in friction signals. The discussion presented herein is aimed to contribute, at least partially and complementarily, to the current understanding of UWB signals and spectroscopic tools; the authors intend to demonstrate and summarize the essential physical structures for UWB systems with quantitative description (quantities across the frequency spectrum), thus exposing the properties of the entire signal spectrum propagation in the nonlinear system of the investigated task.…”

Research of Planetary Gear Fault Diagnosis Based on Permutation Entropy of CEEMDAN and ANFIS

“…Since then, scholars have studied CEEMDAN, and achieved good results. Li [10] combined the EMD, EEMD, CEEMDAN and hybrid filtering methods to denoise the friction signal. The results showed that CEEMDAN can better preserve the signal features than EMD and EEMD.…”