Effect of number of features on classification of roller bearing faults using SVM and PSVM

Bearings in the machines are the major components of interest for condition monitoring. Their failure causes increase in down time and maintenance cost. A possible solution to the problem is developing an on-line condition monitoring system. The vibration characteristics can be a determining factor that will reveal the condition of the bearing parts. Visual inspection of frequency-domain features of the vibration signals may be sufficient to identify the faults, but it requires large domain knowledge and it is a function of speed. Automatic diagnostic techniques allow relatively unskilled operators to make important decisions. In this context, machine learning algorithms have been successfully used to solve the problem with the help of vibration signals. The machine learning procedure has three important phases: feature extraction, feature selection and feature classification. Feature selection involves identifying the good features that contributes greatly for classification and determining the number of such features. Often researchers overlook the later issue and arbitrarily choose the number of features. As there is no science that will tell the right number of features, for a given problem, an extensive study is needed to find the optimum number of features and this paper presents the results of such a study using SVM and PSVM classifiers for statistical and histogram features of time domain signal. The findings are very interesting and challenging; some useful conclusions were drawn and presented.

[1]  Lin Ma,et al.  Fault diagnosis of rolling element bearings using basis pursuit , 2005 .

[2]  B. Samanta,et al.  ARTIFICIAL NEURAL NETWORK BASED FAULT DIAGNOSTICS OF ROLLING ELEMENT BEARINGS USING TIME-DOMAIN FEATURES , 2003 .

[3]  S. M. Wu,et al.  On-Line Detection of Localized Defects in Bearings by Pattern Recognition Analysis , 1989 .

[4]  K. R. Al-Balushi,et al.  Artificial neural networks and support vector machines with genetic algorithm for bearing fault detection , 2003 .

[5]  K. I. Ramachandran,et al.  Feature selection using Decision Tree and classification through Proximal Support Vector Machine for fault diagnostics of roller bearing , 2007 .

[6]  P. Fabry,et al.  Principal component analysis and blind source separation of modulated sources for electro-mechanical systems diagnostic , 2005 .

[7]  Johan A. K. Suykens,et al.  A support vector machine formulation to PCA analysis and its kernel version , 2003, IEEE Trans. Neural Networks.

[8]  Vladimir N. Vapnik,et al.  The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[9]  Junyan Yang,et al.  Application Research of Support Vector Machines in Condition Trend Prediction of Mechanical Equipment , 2005, ISNN.

[10]  Christopher J. C. Burges,et al.  A Tutorial on Support Vector Machines for Pattern Recognition , 1998, Data Mining and Knowledge Discovery.

[11]  J. Ross Quinlan,et al.  Induction of Decision Trees , 1986, Machine Learning.

[12]  Peter A. Flach,et al.  Decision Tree-Based Data Characterization for Meta-Learning , 2002 .

[13]  Chih-Jen Lin,et al.  A comparison of methods for multiclass support vector machines , 2002, IEEE Trans. Neural Networks.

[14]  Asoke K. Nandi,et al.  Practical scheme for fast detection and classification of rolling-element bearing faults using support vector machines , 2006 .