Citation: Kahr, M.; Kovács, G.;
Loinig, M.; Brückl, H. Condition
Monitoring of Ball Bearings Based on
Machine Learning with Synthetically
Generated Data. Sensors 2022, 22,
2490. https://doi.org/10.3390/
s22072490
Academic Editors: Yangquan Chen,
Subhas Mukhopadhyay, Nunzio
Cennamo, M. Jamal Deen, Junseop
Lee and Simone Morais
Received: 26 February 2022
Accepted: 23 March 2022
Published: 24 March 2022
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2022 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
Article
Condition Monitoring of Ball Bearings Based on Machine
Learning with Synthetically Generated Data
Matthias Kahr
1,
* , Gabor Kovács
1
, Markus Loinig
2
and Hubert Brückl
1
1
Department for Integrated Sensor Systems, University for Continuing Education Krems,
2700 Wiener Neustadt, Austria; gabor.kovacs@donau-uni.ac.at (G.K.); hubert.brueckl@donau-uni.ac.at (H.B.)
2
Senzoro GmbH, 1030 Wien, Austria; markus@senzoro.com
* Correspondence: matthias.kahr@donau-uni.ac.at; Tel.: +43-2622-23420-59
Abstract:
Rolling element bearing faults significantly contribute to overall machine failures, which
demand different strategies for condition monitoring and failure detection. Recent advancements in
machine learning even further expedite the quest to improve accuracy in fault detection for economic
purposes by minimizing scheduled maintenance. Challenging tasks, such as the gathering of high
quality data to explicitly train an algorithm, still persist and are limited in terms of the availability
of historical data. In addition, failure data from measurements are typically valid only for the
particular machinery components and their settings. In this study, 3D multi-body simulations of a
roller bearing with different faults have been conducted to create a variety of synthetic training data
for a deep learning convolutional neural network (CNN) and, hence, to address these challenges. The
vibration data from the simulation are superimposed with noise collected from the measurement of a
healthy bearing and are subsequently converted into a 2D image via wavelet transformation before
being fed into the CNN for training. Measurements of damaged bearings are used to validate the
algorithm’s performance.
Keywords:
condition monitoring; roller bearing; fault detection; machine learning; wavelet transform;
simulated training data; 3D multi-body dynamics
1. Introduction
Reducing economic losses by observing machine degradation before it turns into
unexpected downtime is the core task of condition monitoring systems. Prediction based
maintenance especially becomes more and more reliable and accurate, thus it may replace
schedule based maintenance strategies prospectively. Rolling element bearings are one
of the crucial components in mechanical systems, where occurring bearing faults are
responsible for up to 40% of all machine failures [
1
–
3
]. Unfavorable conditions, such as
lubrication problems, contamination due to ineffective seals, misalignment and heavy
loading, can lead to progressive wear phenomena and induce faults on the bearings’ races
or rollers in form of cracks. Thus, material ablation occurs and pits are formed. These local
defects produce successive impulses which can be revealed by monitoring the vibration
signal of the bearings.
Nowadays, massive collected data, which comprise the health state of a bearing, can
be effectively analyzed by using intelligent algorithms including support vector machine
(SVM), artificial neural network (ANN) and more recently deep learning architectures.
Here, SVM and ANN methods learn patterns based on feature engineering [
4
–
6
], that
is, they rely on expert knowledge, whereas deep learning methods can automatically extract
features from raw data [
7
–
9
]. Comprehensive surveys about traditional and contemporary
condition monitoring techniques, including machine learning methods, can be found
in [10–13], respectively.
To successfully train a condition monitoring system based on a machine learning algo-
rithm, a representative quantity of training data is required. This dataset should comprise
Sensors 2022, 22, 2490. https://doi.org/10.3390/s22072490 https://www.mdpi.com/journal/sensors
