Extreme Learning Machine for Fault Diagnosis of Rotating Machinery

Abstract:

The authors present extreme learning machine (ELM) as a novel mechanism for diagnosing the faults of rotating machinery, which is reflected from the power spectrum of the vibration signals. Extreme learning machine was originally developed for the single-hidden layer feedforward neural network (SLFN) and then extended to the generalized SLFN. We obtained the fault feature table of rotating machinery by wavelet packet analysis of the power spectrum, then trained and diagnosed the fault feature table with extreme learning machine. Diagnostic results show that the extreme learning machine method achieves higher diagnostic accuracy than the probabilistic neural network (PNN) method, exhibiting superior diagnostic performance. In addition, the diagnosis of fault feature table adding noise signal indicates the extreme learning machine method provides satisfactory generalization performance.

• Periodical: Advanced Materials Research (Volumes 960-961)
• Pages: 1400-1403
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.960-961.1400
• Online since: June 2014
• Authors: Rui Yu*, Rui Xiang, Shi Wei Yao
• Keywords: Extreme Learning Machine, Fault Diagnosis, Rotating Machinery

References

[1] K. Hornik: Neural Networks, Vol. 4 (1991), p.251.

[2] G. -B. Huang, Q. -Y. Zhu and C. -K. Siew, in: Proceedings of International Joint Conference on Neural Networks (2004), p.985.

[3] G. -B. Huang, Q. -Y. Zhu, and C. -K. Siew: Neurocomputing, Vol. 70 (2006), p.489.

[4] Yukio I, in: JSME International Journal, Series C: Dynamics, Control, Robotics, Design and Manufacturing (1994), p.237.

[5] Logan D, Mathaw J: Mechanical System and Signal Processing, Vol. 10(3) (1996), p.241.