Feature extraction and fault severity classification in ball bearings

The present study attempts to diagnose severity of faults in ball bearings using various machine learning techniques, like support vector machine (SVM) and artificial neural network (ANN). Various features are extracted from raw vibration signals which include statistical features such as skewness,...

Full description

Saved in:
Bibliographic Details
Published in:Journal of vibration and control 2016-01, Vol.22 (1), p.176-192
Main Authors: Sharma, Aditya, Amarnath, M, Kankar, PK
Format: Article
Language:English
Subjects:
Artificial intelligence
Ball bearings
Classification
Entropy
Faults
Feature extraction
Methodology
Neural networks
Uncertainty
Vibration
Vibration analysis
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The present study attempts to diagnose severity of faults in ball bearings using various machine learning techniques, like support vector machine (SVM) and artificial neural network (ANN). Various features are extracted from raw vibration signals which include statistical features such as skewness, kurtosis, standard deviation and measures of uncertainty such as Shannon entropy, log energy entropy, sure entropy, etc. The calculated features are examined for their sensitivity towards fault of different severity in bearings. The proposed methodology incorporates extraction of most appropriate features from raw vibration signals. Results revealed that apart from statistical features uncertainty measures like log energy entropy and sure entropy are also good indicators of variation in fault severity. This work attempts to classify faults of different severity level in each bearing component which is not considered in most of the previous studies. Classification efficiency achieved by proposed methodology is compared to the other methodologies available in the literature. Comparative study shows the potential application of proposed methodology with machine learning techniques for the development of real time system to diagnose fault and it’s severity in ball bearings.
ISSN:1077-5463
1741-2986
DOI:10.1177/1077546314528021