Abstract
The fault diagnosis of hydraulic pumps is currently important and significant to ensure the normal operation of the entire hydraulic system. Considering the nonlinear characteristics of hydraulic-pump vibration signals and the mode mixing problem of the original Empirical Mode Decomposition (EMD) method, first, we use the Complete Ensemble EMD (CEEMD) method to decompose the signals. Second, the time-frequency analysis methods, which include the Short-Time Fourier Transform (STFT) and time-frequency entropy calculation, are applied to realize the robust feature extraction. Third, the multiclass Support Vector Machine (SVM) classifier is introduced to automatically classify the fault mode in this paper. An actual hydraulic-pump experiment demonstrates the procedure with a complete feature extraction and accurate mode classification.
Highlights
Hydraulic systems have been widely used in aeronautics, astronautics, automobiles, shipping, and so on
Complete Ensemble EMD (CEEMD) is selected to adaptively decompose signals into a small number of intrinsic mode functions (IMFs) or modes, and the Short-Time Fourier Transform (STFT) algorithm and time-frequency entropy analysis method are simultaneously used to obtain the fault feature vectors composed by multiscale time-frequency entropy
This paper is organized as follows: Section 2 introduces the relevant feature extraction and mode classification methodology, which includes the CEEMD, STFT, timefrequency entropy, and multiclass Support Vector Machine (SVM) method; Section 3 describes the case study to validate the entire method; Section 4 presents the conclusions of this paper
Summary
Hydraulic systems have been widely used in aeronautics, astronautics, automobiles, shipping, and so on. The reconstructed signal contains residual noise, and different realizations of signal plus noise may produce different numbers of modes To overcome these difficulties, another EMD method has been proposed and successfully applied to vibration signal analysis, complete EEMD (CEEMD), which provides an exact reconstruction. CEEMD is selected to adaptively decompose signals into a small number of IMFs or modes, and the Short-Time Fourier Transform (STFT) algorithm and time-frequency entropy analysis method are simultaneously used to obtain the fault feature vectors composed by multiscale time-frequency entropy. This feature extraction method is defined as the CEEMD-STFT time-frequency entropy method. This paper is organized as follows: Section 2 introduces the relevant feature extraction and mode classification methodology, which includes the CEEMD, STFT, timefrequency entropy, and multiclass SVM method; Section 3 describes the case study to validate the entire method; Section 4 presents the conclusions of this paper
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
