Ensembles of Ensemble Machine Learning approach for Fault detection of Bearing

Authors

  • Priyanka. S. Patil,Mahadev.S.Patil,Sunil.G.Tamhankar,Sangram.S.Patil

Abstract

Maintenance in Industry 4.0 is very important and challenging task. Ultimate goal of maintenance 4.0 is
Automated Accurate anomaly detection system for maintenance. The rolling bearing is a component of
machinery with a very important role that helps in maintaining the linear and rotational movement of the
machinery. If single bearing fails, not only machine but also the assembly line may affect or stop. For
this purpose in Industry 4.0 Machine Learning (ML) algorithm are implemented for automation of fault
detection before catastrophic failure. In this paper Ensembles of ensemble Machine Learning technique
has demonstrated for fault diagnosis of bearing such as Random Forest (RF), AdaBoost, Gradient
boosting (GB), extreme gradient boosting (XGBoost), and Extra tree classifier has utilized as based
classifier for Ensemble technique. Features are extracted from vibration data is used as input to ML
techniques and results are carried out for accuracies and performance analysis of individual base
classifier and compared with ensembles of ensemble technique. Ensembles of ensemble give highest
accuracies as compare to individual ensembles is 97.91%.

Published

2020-12-02

Issue

Section

Articles