Study on two-step mechanical fault diagnosis integrating symbolisation method based on the division of probability density space and support vector machine

ABSTRACT

In the traditional mechanical fault diagnosis, the number of normal samples and various fault samples is the same, which shields the adverse effects of sample imbalance on the fault diagnosis results. However, in actual industrial applications, the difference between the number of normal samples and the number of various fault samples is very different, so the problem of sample imbalance is unavoidable.The imbalance of the samples will lead to the tilt of the classification hyperplane of the support vector machine. There are similar problems in other classification methods, resulting in the generalisation ability of the training model is not strong.This paper proposes a two-step mechanical fault diagnosis integrating symbolisation method based on the division of probability density space and support vector machine. First, the normal samples are stripped out by the symbolisation method based on the division of probability density space; then, the support vector machine is applied to classify the fault samples. This method is applied for bearing fault diagnosis. The diagnostic accuracy increases by 3.34% compared to the single-step diagnosis with support vector machine.

KEYWORDS:

• Symbolisation
• division of probability density space
• mechanical fault diagnosis

Acknowledgments

The authors gratefully acknowledge the supports of all the fundings.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

This research was funded by the National Key Research and Development Program of China [Grant number 2016YFC0600900], the Yue Qi Distinguished Scholar Project of China University of Mining & Technology (Beijing) [Grant number 800015Z1145] and the National Natural Science Foundation of China [Grant number U1361127].

Notes on contributors

Xiao Yajing

Xiao Yajing, Doctoral Student of Engineering. Studing in ChinaUniversity of Mining & Technology (Beijing) .Her research interests include mechanical fault diagnosis and state prediction.