Accelerating the fatigue analysis based on strain signal using Hilbert–Huang transform

Abstract

Purpose The purpose of this paper is to present the application of Hilbert–Huang transform (HHT) for fatigue damage feature characterisation in the time–frequency domain based on strain signals obtained from the automotive coil springs. Design/methodology/approach HHT was employed to detect the temporary changes in frequency characteristics of the vibration response of the signals. The extraction successfully reduced the length of the original signal to 40 per cent, whereas the fatigue damage was retained. The analysis process for this work is divided into three stages: signal characterisation with the application of fatigue data editing (FDE) for fatigue life assessment, empirical mode decomposition with Hilbert transform, an energy–time–frequency distribution analysis of each intrinsic mode function (IMF). Findings The edited signal had a time length of 72.5 s, which was 40 per cent lower than the original signal. Both signals were retained statistically with close mean, root-mean-square and kurtosis value. FDE improved the fatigue life, and the extraction did not affect the content and behaviour of the original signal because the editing technique only removed the minimal fatigue damage potential. HHT helped to remove unnecessary noise in the recorded signals. EMD produced sets of IMFs that indicated the differences between the original signal and mean of the signal to produce new components. The low-frequency energy was expected to cause large damage, whereas the high-frequency energy will cause small damage. Originality/value HHT and EMD can be used in the strain data signal analysis of the automotive component of a suspension system. This is to improve the fatigue life, where the extraction did not affect the content and behaviour of the original signal because the editing technique only removed the minimal fatigue damage potential.

Fault Analysis of Wind Power Rolling Bearing Based on EMD Feature Extraction

“…Decomposing the data used in this article into EMD will get several IMF components and then use the kurtosis principle to select the appropriate IMF components. First, the original signal of bearing vibration is decomposed by EMD to obtain several IMFs [21][22][23]. Second, the kurtosis value is calculated.…”

Common Crossing Structural Health Analysis with Track-Side Monitoring

“…The EMD provides many advantages compared to the short-time Fourier transform (STFT) and wavelet transform (WT): analysis of nonlinear and nonstationary signals, better time and frequency resolution. Numerous recent papers approve the successful application of the EMD in mechanical and civil engineering [27][28] for fault detection and prediction. In the present paper the Ensemble Empirical Mode Decomposition (EEMD) is used, that has no drawback of the mode mixing of EMD.…”

A Dynamic Maintenance Strategy for Multi-Component Systems Using a Genetic Algorithm

“…In this context, prognostics and health management (PHM) is playing an increasingly important role. PHM can be combined with multiple disciplines, such as machine learning [4,5] sensor technology [6,7], fault diagnosis [8], statistics [9][10][11][12] and reliability engineering [13][14][15][16], to realize an online evaluation of system health status and to predict the state of the system based on the current information. PHM converts signals and data detected by sensors into health status information about the system, and it alerts users so that the system can be maintained in a timely manner.…”