Damage Localization of Stacker's Track Based on EEMD-EMD and DBSCAN Cluster Algorithms

• Published in: IEEE transactions on instrumentation and measurement Vol. 69; no. 5; pp. 1981 - 1992
• Main Authors: Li, Shupan, Qin, Na, Huang, Darong, Huang, Deqing, Ke, Lanyan
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.05.2020; The Institute of Electrical and Electronics Engineers, Inc. (IEEE)
• Subjects: Algorithms; Anomaly detection; Clustering; Clustering algorithms; Current distribution; Damage localization; Damage location; Data preprocessing; density-based spatial clustering of applications with noise (DBSCAN); empirical mode decomposition (EMD); ensemble EMD (EEMD); Feature extraction; Interpolation; Localization; Neural networks; Outliers (statistics); Signal processing; Smoothing methods; stacker; Vibrations; Wavelet transforms
• ISSN: 0018-9456; 1557-9662
• DOI: 10.1109/TIM.2019.2919375

The vibration and friction triggered by the long-term operation of stacker inevitably lead to the damage problem of stacker's track. In order to ensure the stable operation of the stacker, a novel scheme is proposed to address an accurate damage localization problem of the stacker's track in industrial environment based on ensemble empirical mode decomposition-empirical mode decomposition (EEMD-EMD) and density-based spatial clustering of applications with noise (DBSCAN). First, the original electric current signal with unbalanced distribution is processed by fixed-interval smoothing and cubic Hermite interpolation algorithms. Second, the EEMD-EMD method, which solves the mode-mixing problem, is adopted to decompose the preprocessed signal into well-defined intrinsic mode functions (IMFs). Third, based on Hilbert-Huang transform (HHT) of the IMFs, the preprocessed signal and the difference of average instantaneous amplitudes in two adjacent time points are considered as the 2-D feature vector input of the DBSCAN algorithm. As such, damage locations of the stacker's track are detected by means of the outliers that are obtained by the DBSCAN algorithm. The efficiency of the proposed localization scheme and its superiority over the existing cooperative damage localization methods, namely, the box-plot method, the pulse coupling neural network (PCNN) and wavelet transform (WT) theory method, and the box-plot and WT method, are verified through experiments using the data provided by State Grid Measuring Center of China.