See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/336251910

IoT based Condition Monitoring System Design for Investigation of Non-Oil

Ball Bearing in terms of Vibration, Temperature, Acoustic Emission, Current
and Revolution Parameters

Conference Paper · October 2019

CITATIONS READS

6 4,315

3 authors:

Mustafa Çakır Mehmet Ali Güvenç

ISKENDERUN TECHNICAL UNIVERSITY ISKENDERUN TECHNICAL UNIVERSITY
48 PUBLICATIONS   119 CITATIONS    14 PUBLICATIONS   145 CITATIONS   

SEE PROFILE SEE PROFILE

Selcuk Mistikoglu
Iskenderun Technical University
29 PUBLICATIONS   758 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wave Energy Converter Control View project

Talaşlı İmalatta Titreşim Analizi View project

All content following this page was uploaded by Mustafa Çakır on 04 October 2019.

The user has requested enhancement of the downloaded file.

 Proceedings of 10th International Symposium on
Intelligent Manufacturing and Service Systems
Pages 1059-1068

IoT based Condition Monitoring System Design for

Investigation of Non-Oil Ball Bearing in terms of Vibration,
Temperature, Acoustic Emission, Current and Revolution
Parameters

Mustafa ÇAKIR1, Mehmet Ali GÜVENÇ2 and Selçuk MISTIKOĞLU3

1,2,3 Iskenderun Technical University/Mechanical Engineering Department, Hatay, TURKEY

[email protected], [email protected], [email protected]

KEYWORDS - Internet of Things, Condition Monitoring, Machine Learning, Bearing, Virtuino,

Microcontroller.

ABSTRACT
The Internet of Things (IoT) concept facilitates our life in many areas. One of the facilities
provided in this area is undoubtedly condition monitoring. Unlike regular maintenance, IoT systems
that perform continuous control operations can provide great advantages to the company with a
warning that a serious failure will occur. It is a vital importance to determine defective bearings
during the rotation of the power generating and power consuming machines without reaching the
critical level.
In this study; an experimental setup was created, and a Condition Monitoring Application
(CMA) was performed in the laboratory environment which could monitor vibration, temperature,
acoustic emission (AE), current and revolution measurements for ORS 6203 type defective bearing
detection. In the designed test setup, a microcontroller platform used as a system brain, a direct
current machine with a shaft mounted bearing, and miscellaneous sensors used such as 3-axis
accelerometer that senses vibrations, a hall sensor module for revolution measure, a microphone
module for AE detection, current and voltage sensor modules for power consumption measurement,
contactless temperature sensor module for temperature measurement. CMA has been designed for
monitoring of system via Bluetooth module. With the help of the microcontroller platform, the data
collected from the sensors is transmitted wirelessly to the Android smartphone to be plotted and
recorded. After the data exceeds certain thresholds, the control system is able to provide SMS
and/or e-mail notification to the related maintenance team with the Android smartphone. Linear
Discriminant Analysis (LDA), Random Forest (RF), Support Vector Machine (SVM), k-Nearest
Neighbor (kNN), and Decision Tree (DT) algorithms applied to the collected data and models were
created. As a result of accuracy analysis over %96 success was achieved for all models.
1 INTRODUCTION
Many of the products we use in our daily lives go through industrial production processes.
The continuation of these processes without interruption or with minimal interruption is directly
related to the maintenance of the equipment used in the production stages. Possible stop of the plant
due to unexpected faults cause production delay, work accident, loss of production and product cost
increase [1]. Maintenance costs have an important role in the total operating cost. Especially in an
environment where global competition is increasing day by day, to present the best product with the
lowest cost is among the main targets of the producers. One of the most critical elements of the
production stages is undoubtedly electric motors.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1059-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

Electric motors are all around us. We see them in many areas form industry to daily life.
People use them in pumps, manufacturing floors, elevators, wind turbines, HVAC, etc. It is seen
that electric motors are used in almost every field in order to meet operational requirements. The
health status of the engines must be checked because of the critical duty at the place of operation.
According to the literature review [2] motor failures are basically divided into two groups; electrical
and mechanical faults especially bearing faults. As reported by [3], bearing failures account for half
of the total motor failures.
Bearings are one of the most critical components in motors because they connect the motor
stator and motor rotor [4]. Bearing defects for small and medium-sized machines account for the
majority of all defects [5]. Preventing possible failures is very important in terms of reducing
economic losses and the safe operation of industrial facilities. That’s why condition monitoring of
the motors is crucial.
According to [6], condition monitoring techniques are divided into two categories: offline and
online monitoring. Offline condition monitoring is also called a periodic condition monitoring
system. In this method, the reading from the sensors on the machine is performed within a certain
time interval. Analysis of the read sensor data is performed in a laboratory environment away from
the operating machine. Another method; online condition monitoring is called continuous condition
monitoring. In this system, data is continuously collected by sensors from the running machine and
compared to an acceptable value or in other words a threshold value. Recently online systems are
much popular due to the development of IoT technologies.
IoT concept facilitates our life in many areas. One of the facilities provided in this area is
undoubtedly condition monitoring. Unlike regular maintenance, IoT systems that perform
continuous control operations can provide great advantages to the company with a warning that a
serious failure will occur. It is a vital importance to determine defective bearings during the rotation
of the power generating and power consuming machines without reaching the critical level.
The size of the data transmitted to the cloud system, that is associated with IoT technology, is
quite high. The analysis of these huge data blocks, which is also called Big Data, can be analyzed
with Artificial Intelligence (AI) approaches in a shorter period of time compared to classical
approaches. There is much different AI software to develop solutions to real-life problems. But
some of them need a license and some of them are open-source.
The use of open-source software is becoming more common. The lack of license fees for that
kind of software offers many advantages to the user. In addition, many libraries developed by users
are open-source. The developer can contribute to the development of libraries with the code he
developed. Machine Learning (ML), which is a subfield of AI, is very popular in this issue. ML tool
kits serve with licensed programs as well as free open-source software tool kits [7] such as Weka
and R. Open-source R, which is an environmental for statistic computing and a programming
language, is widely used in data analysis [8]. Since the R programming language interface is not
user-friendly, code development is usually performed by most programmers in the RStudio
environment. Detection of bearing damage requires extensive research. For this reason, researchers
analyze data with ML tool kits.
Many methods have been developed based on the ML system such as Artificial Neural
Network (ANN), Genetic Algorithm (GA), Ant Colony Optimization (ACO), Fuzzy Logic (FL),
Hidden Markov Model (HMM), Random Forest (RF), Support Vector Machine (SVM), etc. The
use of SVM for bearing condition monitoring and diagnostics has continued to increase in recent
years. SVM has excellent performance in the generalization process, so the machine can provide
high accuracy in classification for condition monitoring [9].
Although the analyzing of vibration is the most commonly used method to detect bearing
damage, monitoring of other parameters such as temperature [10], aquatic emission [11], current
[12], and revolution with vibration will increase the accuracy of the damage detection.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1060-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

2 MATERIALS AND METHODS

Condition Monitoring System consists of four parts; first one is setup for the experiment, the second
one is the CMA that runs on Android system to visualize the data, the third one is cloud system that
used for storing the data to analyze and the last one is PLX-DAQ that allows data transfer to
computer environment. The test setup consists of a microcontroller unit and peripheral units. CMA
was created with a program called Virtuino. The electronic connection scheme of the setup, is
shown in Figure 1, was created with Fritzing which is the open-source
software.

Figure 1: Fritzing schematic of electronics control system of CMA

A detailed description of the hardware and software materials can be found in the following
sections.
2.1 Hardware Materials

Figure 2: Hardware materials

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1061-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

2.1.1 Microcontroller Platform (Arduino Uno)

It is an open-source microcontroller platform based on the ATmega328P microcontroller. It

has 14 digital Input/Output (I/O) pins, 6 analog inputs, a 16 MHz quartz crystal and 8-bit
microcontroller. It includes everything needed to make the microcontroller ready for operation;
simply plug the USB cable into a computer, or run on an AC-DC adapter or battery to start. [13].
Microcontroller platform that used in setup is shown in Figure 2(a). Microcontroller platform is
used as a brain of the system to collect sensors data and communicate with the designed CMA that
runs on a mobile device over Bluetooth module.

2.1.2 Bluetooth Module (HC-05)

HC-05 Bluetooth module, shown in Figure 2(b), is a key part of the CMA because of
providing communication between microcontroller and smartphone. It operates over unlicensed
ISM (Industrial-Scientific-Medical) band, globally available frequency of 2.4GHz. It can link
digital devices up to 10m.

2.1.3 WiFi Module (NodeMCU)

ESP8266 based Wi-Fi module, shown in Figure 2(c), is a very popular electronic part because
of its use for IoT based projects [14]. This platform is designed for low-cost internet applications
and has a structure similar to the Arduino boards such as having Input / Output units, PWM outputs
and I2C, SPI communication support. In addition to the Arduino Uno platform, Wi-Fi is available.

2.1.4 Accelerometer (ADXL345)

The ADXL345, which is shown in Figure 2(d), has 3-axis accelerometer with 13-bit
resolution. It supports up to ±16 g acceleration and easy to use with the microcontroller over Serial
Peripheral Interface (SPI) or Inter-Integrated Circuit (I2C) communication protocol [15]. It is ideal
for vibration measurement on bearing housing for CMA because of its features such as cheap,
small, thin and has ultra-low power consumption as low as 23 µA in measurement mode and 0.1 µA
in standby mode.
One of the most important dynamic problems of design engineering to be solved is
undoubtedly mechanical vibrations [16]. There are some negative effects of mechanical vibrations.
These are material fatigue, abrasion, high stresses, and noise. Systems exposed to mechanical
vibrations due to these effects can take serious damage. If the necessary precautions aren't taken, the
system may be damaged. Therefore, it is vital to carry out mechanical vibration measurements.
Mechanical vibrations can be axial, radial and torsional. Hence, the measurements of vibration
should be collected from 3-axis.
2.1.5 Current Sensor Module (ACS712)

The ACS712, which is shown in Figure 2(e), is fully integrated, nearly zero magnetic hystereses,
hall effect-based linear current sensor (AC/DC) with 5V single supply operation and a low-
resistance current conductor [17].

2.1.6 Linear Magnetic Hall Sensor Module

The module, which is shown in Figure 2(f), has A3141 chipset which detects the magnetic
field and LM393 operational amplifier on the board. The potentiometer can control the sensitivity
of the sensor. If a magnetic field is detected by the sensor, a signal sends over D0 pin. Neodymium
magnets have been placed on the motor shaft to measure revolution value as shown in Figure 3.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1062-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

Figure 3: Hall sensor placement with neodymium magnets for revolution measurement
2.1.7 Contactless Temperature Sensor Module (MLX90614)

The MLX90614, which is shown in Figure 2(g), is an Infra-Red thermometer and uses for
non-contact temperature measurements. It has a wide range and high accuracy temperature measure,
from -40 to 125 ˚C for sensor temperature and form -70 to 380 ˚C for object temperature [18]. And
also, it is tiny and has a low cost. Because of these features, it is ideal for industrial temperature
control of moving parts.
As design criteria, bearing metal temperatures should not exceed specified operating
conditions. The rising temperature of the bearing may be due to loss of grease or bearing failure. It
should be provided bearing temperature alarm and shutdown limits. Otherwise, the bearings’ fault
may cause serious damage to machines. Therefore, it would be useful to monitor the temperature of
bearing for the health of machines.
2.1.8 Voltage Sensor Module

The analog input of the microcontroller platform used in the designed system is limited to 5
Volt DC. If higher voltage values are to be measured, it is necessary to use a voltage divider. The
direct current machine uses 220AC to 24V DC power supply. In the design of the electronic system
was used a 5:1 voltage divider with a 4XΩ and an XΩ resistor. This type of design is restricted to
voltages that are less than 25V. More than that voltage will exceed the voltage limit of analog input
and cause damage on the microcontroller platform. Microcontroller which is used in the setup has
10-bit Analog Digital Converter (ADC) so the resolution is 0.00489 V (5V/1023). The voltage
divider circuit used in the setup is shown in Figure 4.

Figure 4: Voltage divider for less than 25V supply voltage measurement
2.2 Software Materials

2.2.1 Condition Monitoring Application (CMA)

CMA is a monitoring application to view sensor values on the screen of the android based
mobile device. Not only it has monitoring feature but it also has an alarming feature with sound
over the buzzer of the smartphone, notification feature with e-mail/SMS if threshold values exceed
and exporting data to excel format etc. It can communicate with the microcontroller platform over
the Bluetooth module. Hence, the microcontroller platform sends the data comes from sensors to the

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1063-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

mobile device. The CMA has analog displays to show the instant value of sensors and charts to
show continuous value.
The CMA used for this study has been designed by Virtuino app because it has a user-friendly
interface. It can communicate efficiently with control boards such as Arduino and ESP, to send a
control signal to the boards and to monitor the condition of the designed system with mobile
devices wirelessly The Virtuino app has been used for different applications such as Smart Home
Automation [19], Classroom Monitoring [20], Smart Agriculture [21], and Intelligent Healthcare
Monitoring [22].
2.3 Method
The all system (CMA and Setup) designed step by step. This section explains all these design
steps. The setup has an electric motor, healthy and damaged ORS 6203 type bearings, an Arduino
UNO, an ESP8266 Wi-Fi module, an HC-05 Bluetooth module, and miscellaneous sensors. The
sensors are an accelerometer, a hall sensor, a current sensor, a voltage sensor, a microphone, and a
contactless temperature sensor. The top view of the setup used for experiments is shown in Figure
5.

Figure 5: Top view of Setup

The first step of designing the system was to make connections to the peripheral units
(sensors and communication modules) to the microcontroller platform. These connections had been
shown in Figure 1 before. The second step was to calibrate all the sensors with measurement
devices which have high accuracy. And the third step was to do experiment with healthy and non-
oil ball bearings and collect data from the test setup to the PC via PLX-DAQ [23] (Parallax Data
Acquisition Tool Software). The collected data were also transferred to the cloud over the ESP8266
Wi-Fi module. The general structure of the system is basically shown in Figure 6.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1064-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

Figure 6: General structure of the designed system

At the fourth step, the CMA interface was designed with the Virtuino App. The collected data
was transferred to the smartphone screen over the Bluetooth module. Thus, the data were visualized
with charts and analog instruments on the smartphone display. In the designed CMA, all sensor data
are displayed continuously with charts and instantly with analog displays. The CMA’s interface
with 5 different panels is shown in Figure 7. The first four panels show respectively ambient and
bearing temperature, x-y-z vibration level, revolution and AE, current and voltage at chart widgets.
The last panel shows all sensor value on analog instruments.

Figure 7: CMA interface for displaying all sensor values on analog instruments and charts

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1065-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

The next step was to analyze the data with Decision Tree (DT) algorithm, popular ML tool
kit, to determine the threshold values for parameters of vibration, temperature, AE, current and
revolution. Then, the alarms were set to the experimental system parameters due to the result of DT.
Microcontroller platform is used as a brain of the system to collect sensors data and
communicate with CMA that runs on a mobile device over Bluetooth module. If the measured value
of sensors is above the acceptable values, this means that the bearing is damaged and a warning is
issued. After the oil-free ball bearing is mounted on the motor shaft, the vibration level exceeds the
threshold. In this case, the CMA sends e-mail and SMS notification. This is illustrated in Figure 8.

Figure 8: SMS and e-mail alarm notification from the CMA to the maintenance team
The last step was to analyze all data with ML tool kits. Models were created from the CART,
LDA, kNN, SVM, and RF algorithms. Then accuracy analysis was performed on these models.

3 RESULTS

The experiments in this study consist of two parts. The first test with healthy ball bearing
mounted on the motor shaft and the second test with non-oil ball bearing. Observation data obtained
from the first test were classified as "0-Healthy Bearing" and observation data obtained from the
second experiment were classified as "1-Damaged Bearing". The data obtained data consists of 8
different parameters. These parameters are, Gx (x-axis vibration), Gy (y-axis vibration), Gz (z-axis
vibration), Temperature (ambient and bearing), I (Current), AE and Revolution. The size of all data
has 12264 rows and 8 columns. This data was transferred to RStudio environment in csv format.
The data, 70 % of all, was used for training purposes and the remaining 30% was used for test
analysis. CART model was used to determine the threshold of parameters. Parameters based
prediction results of CART model is shown in Figure 9.
Other popular models were created as LDA, kNN, SVM, and RF from the library named
“caret”, “kernlab”, “caTools”, “rpart”, and “ROCR” in RStudio for accuracy analysis. As a result of
accuracy analysis, 96.13% for LDA, 97.48% for kNN, 98.25% for SVM and 98.11% success was
achieved for RF models.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1066-
Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

Figure 9 Parameters based prediction results of CART models

4 CONCLUSION
Although the analyzing of vibration is the most commonly used method to detect faulted
bearing, monitoring of other parameters such as temperature, AE, current and revolution with
vibration increased the accuracy rate of the damage detection. In the estimation made by vibration
analysis, the analysis of the data obtained from 3-axes increases the accuracy of the estimation. An
8% difference was observed between the estimation based on the vibration data obtained from the
single axes and the estimation based on the vibration data obtained from the 3-axes. If AE and
current data are added in addition to the vibration data, the accuracy rate increases by less than 1%.
On the other hand, if revolution data is added, it increases nearly 5% and finally, if temperature data
is added, a 5% increase is observed.

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1067-
 Cakir et al. Iot Based Condition Monitoring System Design For Investigation Of...

REFERENCES
[1] E. C. Fitch, Proactive Maintenance for Mechanical Systems, no. 6. Elsevier Science Ltd, 1992.
[2] S. Nandi, H. A. Toliyat, and X. Li, “Condition monitoring and fault diagnosis of electrical motors - A
review,” IEEE Trans. Energy Convers., vol. 20, no. 4, pp. 719–729, 2005.
[3] P. O’Donnell, “Report of Large Motor Reliability Survey of Industrial and Commercial Installations,
Part I,” 1985.
[4] S. Lu, P. Zhou, X. Wang, Y. Liu, F. Liu, and J. Zhao, “Condition monitoring and fault diagnosis of
motor bearings using undersampled vibration signals from a wireless sensor network,” J. Sound Vib.,
vol. 414, pp. 81–96, 2018.
[5] A. Bellini, F. Immovilli, R. Rubini, and C. Tassoni, “Diagnosis of Bearing Faults of Induction
Machines by Vibration or Current Signals: A Critical Comparison,” Ind. Appl. Soc. Annu. Meet.
2008. IAS ’08. IEEE, vol. 46, no. 4, pp. 1–8, 2008.
[6] C. Malla and I. Panigrahi, “Review of Condition Monitoring of Rolling Element Bearing Using
Vibration Analysis and Other Techniques,” J. Vib. Eng. Technol., no. 0123456789, 2019.
[7] S. Landset, T. M. Khoshgoftaar, A. N. Richter, and T. Hasanin, “A survey of open source tools for
machine learning with big data in the Hadoop ecosystem,” J. Big Data, vol. 2, no. 1, pp. 1–36, 2015.
[8] E. Kaya, M. Agca, F. Adiguzel, and M. Cetin, “Spatial data analysis with R programming for
environment,” Hum. Ecol. Risk Assess., vol. 0, no. 0, pp. 1–10, 2018.
[9] A. Widodo and B. S. Yang, “Support vector machine in machine condition monitoring and fault
diagnosis,” Mech. Syst. Signal Process., vol. 21, no. 6, pp. 2560–2574, 2007.
[10] J. A. Henao-Sepulveda, M. Toledo-Quinones, and Y. Jia, “Contactless Monitoring of Ball Bearing
Temperature,” no. May, pp. 1571–1573, 2006.
[11] R. B. W. Heng and M. J. M. Nor, “Statistical analysis of sound and vibration signals for monitoring
rolling element bearing condition,” Appl. Acoust., vol. 53, no. 1–3, pp. 211–226, 1998.
[12] S. Liling and X. Boqiang, “An improvement of stator current based detection of bearing fault in
induction motors,” Conf. Rec. - IAS Annu. Meet. (IEEE Ind. Appl. Soc., no. 50407016, pp. 2277–
2281, 2007.
[13] M. Banzi and M. Shiloh, Getting started with Arduino: the open source electronics prototyping
platform. Maker Media, Inc., 2014.
[14] O. Oral and M. Çakır, “Nesnelerin İnterneti Kavramı ve Örnek Bir Prototipin Oluşturulması,”
Mehmet Akif Ersoy Üniversitesi Fen Bilim. Enstitüsü Derg., vol. 177, no. 1, pp. 172–177, 2017.
[15] Analog Devices, “Digital Accelorometer Data Sheet.” p. 40, 2009.
[16] S. Mıstıkoğlu, “Predictive Maintenance,” Miller Magazine, vol. 12, no. 101, pp. 60–66, 2018.
[17] Inc Allegro MycroSystems, “Hall effect-based linear current sensor ACS712,” pp. 1–14.
[18] Melexis, MLX90614 - Single and Dual Zone Infrared Thermometer in TO-39. 2006.
[19] W. A. Jabbar, M. H. Alsibai, N. S. S. Amran, and S. K. Mahayadin, “Automation System for Smart
Home,” 2018 Int. Symp. Networks, Comput. Commun., pp. 1–6, 2018.
[20] J. Shekhar, D. Zerihun, and M. Haile Abebe, “Automated Classroom Monitoring with IoT and
Virtuino App,” Int. J. Adv. Res. IT Eng., no. April, 2019.
[21] S. Bhowmick, B. Biswas, M. Biswas, A. Dey, S. Roy, and S. K. Sarkar, “Application of IoT-Enabled
Smart Agriculture in Vertical Farming Sutanni,” Adv. Commun. Devices Netw., vol. 537, pp. 521–
528, 2019.
[22] K. P and A. S. P, “Intelligent Healthcare Monitoring in IoT,” Int. J. Adv. Eng. Manag. Sci., vol. 4, no.
6, pp. 441–445, 2018.
[23] “PLX-DAQ | Parallax Inc.” [Online]. Available: https://www.parallax.com/downloads/plx-daq.
[Accessed: 09-Jul-2019].

IMSS’19 Sakarya University - Sakarya/Turkey, 9-11 September 2019, pp. 1059-1068 -1068-

View publication stats