Title: Machine vibration prediction using ANFIS and wavelet packet decomposition

Authors: Youliang Yang; Qing Zhao

Addresses: Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta T6G 2V4, Canada. ' Department of Electrical and Computer Engineering, University of Alberta, Edmonton, Alberta T6G 2V4, Canada

Abstract: This paper proposes a new method for building time series model to predict machine vibration values. Instead of building a time series model based on the raw machine vibration signal, the vibration signal will be first decomposed into different levels using wavelet packet decomposition (WPD). Sub-signals can be reconstructed from those wavelet packet coefficients. Time series model is built for each of those sub-signals, using adaptive neuro-fuzzy inference system (ANFIS). The final prediction value is the sum of the prediction values of all the models. Comparing to the other two methods, which are building ANFIS model based on the raw vibration signal and building ANFIS models based on the sub-signals generated with discrete wavelet decomposition, experimental results show that the method using ANFIS and WPD outperforms the other two methods.

Keywords: machine vibration monitoring; time series modelling; wavelet transforms; adaptive neuro-fuzzy inference system; ANFIS; neural networks; fuzzy logic; wavelet packet decomposition.

DOI: 10.1504/IJMIC.2012.045693

International Journal of Modelling, Identification and Control, 2012 Vol.15 No.3, pp.219 - 226

Published online: 29 Nov 2014 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article