Eventos Anais de eventos
COBEM 2019
25th International Congress of Mechanical Engineering
A Simulation-Driven Deep Learning Approach for Condition Monitoring of Hydrodynamic Journal Bearings. Part I: Diagnostics of Wear Faults
Submission Author:
Stephen Ekwaro-Osire , Texas , United States
Co-Authors:
Diogo Stuani Alves, Tiago Machado, Katia Lucchesi Cavalca Dedini, Ozhan Gecgel, João Paulo Dias, Stephen Ekwaro-Osire
Presenter: Stephen Ekwaro-Osire
doi://10.26678/ABCM.COBEM2019.COB2019-0707
Abstract
Early diagnostics of wear faults in hydrodynamic bearings using advanced condition monitoring is essential to avoid sudden failures during operation of rotating machines. Although an increasing number data-driven condition monitoring approaches for rotating machines have been proposed in the past decade, they are mainly focused on rolling element bearings and rely on the collection of large amounts of training data from physical experiments or from the field. This work proposes a simulation-driven framework based on a deep learning algorithm to automatically extract features and classify wear faults in hydrodynamic journal bearings using simulated vibrations signals. A model to calculate the dynamic response of a hydrodynamic journal bearing system considering the existence of localized wear in the bearing walls was developed. A large database of simulated vibration signals for several bearing operation conditions, wear severities and artificially added noise levels was created. A deep convolutional neural network algorithm was developed to classify the bearing wear severity using the simulated vibration signal database as training/testing data. Preliminary results confirm the potential capability of using simulated data to train machine learning algorithms to be later used on real applications to predictions faults in rotating machinery.
Keywords
journal bearings, condition monitoring, wear faults, simulation-driven deep learning
