Title: Feature selection and fault classification for an induction motor bearing data by using random forest classifier

Authors: Raj Kumar Patel; Sudhir Agrawal; Anshul Sharma

Addresses: Rajkiya Engineering College, Sonbhadra, UP-231 206, India ' BIT Gorakhpur, UP-273 209, India ' NIT, Hamirpur, HP 177 005, India

Abstract: In induction machine, faults can be avoided if it is detected early. Bearing failure is the major cause and accounts for up to 41% of the faults that occur in rotating machines. This fault is distinguished by vibration signals whose amplitude is however low and linked with frequencies. A study is accomplished to magnify the diagnostic pertinence of the measurement of vibration using the recognition of indicative features for damage classification, location, and severity. To accomplish the task, a composite feature pool is developed by calculating features from different domains, i.e., time, frequency, and time-frequency. In totality, 46 features are calculated and out of that most desirable features are selected by applying principal component analysis (PCA) and minimum redundancy maximum relevance (mRMR) technique. The characteristic features are given as input to the radial basis function neural network (RBFNN), support vector machine (SVM), and random forest (RF) classifier, and finally, the realisation of these classifiers is correlated. The proposed feature selection method along with the random forest classifier gives the better results.

Keywords: mRMR; PCA; SVM; random forest; statistical feature.

DOI: 10.1504/IJQET.2023.134886

International Journal of Quality Engineering and Technology, 2023 Vol.9 No.4, pp.299 - 320

Received: 05 Jan 2023
Accepted: 10 Aug 2023
Published online: 15 Nov 2023 *

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