A Novel Feature Extraction Method Using Ant Colony Clustering Analysis

Abstract:

A novel feature extraction method is presented by combining wavelet packet transform with ant colony clustering analysis in this paper. Vibration signals acquired from equipments are decomposed by wavelet packet transform, after which frequency bands of signals are clustered by ant colony algorithm, and each cluster as a set of data is analyzed in frequency-domain for extracting intrinsic features reflecting operating condition of machinery. Furthermore, the robust ant colony clustering algorithm is proposed by adjusting comparing probability dynamically. Finally, effectiveness and feasibility of the proposed method are verified by vibration signals acquired from a rotor test bed.

• Periodical: Applied Mechanics and Materials (Volumes 37-38)
• Pages: 32-35
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.37-38.32
• Online since: November 2010
• Authors: De Bin Zhao, Ji Hong Yan
• Keywords: Ant Colony Clustering Analysis, Feature Extraction, Wavelet Packet Transform

References

[1] N.G. Prelcic, F.P. González and M.E.D. Jiménez: Signal Processing, Vol. 81 (2001) No. 8, pp.1641-1662.

[2] A. Ahmad and L. Dey: Data & Knowledge Engineering, Vol. 63 (2007) No. 2, pp.503-527.

[3] Y. Yang and M. Kamel: Proceedings of the IEEE Swarm Intelligence Symposium (SIS), Indiana, USA, (2003), pp.65-71.

[4] U. Boryczka: Applied Soft Computing, Vol. 9 (2009) No. 1, pp.61-70.