Accelerometer Sensor Data Analysis of Bridge

Structural Health Monitoring System

1st Anindya Ananda Hapsari
Eko Supriyanto, Abu Hasan, Amin Suharjono
School of Graduate Studies
Departement of Electrical engineering
Management and Science University
Politeknik Negeri Semarang
Shah Alam, Malaysia
Semarang, Indonesia
[email protected]
[email protected]

XXX-X-XXXX-XXXX-X/XX/$XX.00 ©20XX IEEE

 Abstract— The use of Wireless Sensor Network (WSN) for II. RELATED WORKS
monitoring systems continues to grow. one of them is the to
monitor a building, by given the importance of security to the
On research about Structural Health Monitoring System
quality of the building, especially buildings for public facilities on Single Degree of Freedom Bridge [6]. Deals with
such as bridges. but for a good monitoring system it is monitoring bridges with sensors and accelerometer sensors ,
necessary to analyze the sensor nodes used. In this study a and wight in motion. This research is already using the
bridge monitoring system prototype was built by integrating method of WSN. And the data is displayed in some form,
software and hardware with microcontroller and sensor namely time and frequency. But the accelerometer data is not
devices. Sensor nodes generate data and are processed with a yet in shows on the internet. Wireless Sensor Networks
microcontroller, then immediately sent to the cloud by the Technology for Bridge Health Monitoring by [7] describes
internet network using MQTT. This system is built with a the research on monitoring the bridge with WSN. This
firmware-based Internet of Things (IoT) and this is verified research data further discuss the power consumption and
using an accelerometer sensor. Data from the accelerometer power on the system. But data from the sensors are not
sensor is then processed and analyzed in the time domain and shown in this study. And just discuss power supply is needed
frequency domain using FFT. And in this paper the author on WSN system.
analyzes the raw data from the ADXL345 accelerometer
output sensor and is applied to monitor the bridge. From the Next research by [8], already using WSN method and use
experiment, an accelerometer output sensor can vary multiple sensors. This research uses a sensor device is
depending on the sensitivity of the values. relatively expensive and not yet explained how the data is
analyzed. The sensor data is still in the form of graphs in the
Keywords—accelerometer, fast fourier transform, internet time domain. And not explained how the generated sensor
of things, monitoring, mqtt, shms, wireless sensor network data, and how to process data from the sensor.
Research about Structural Health Monitoring
I. INTRODUCTION (HEADING 1) Measurements under Train Speed Changing [9]. In this
The bridge is one of the very important infrastructures for research are described and illustrated the influence of train
road transport. Along with the progress of the times, bridge speeds towards the railway bridge structure with
construction built increasingly long and pretentious. accelerometer sensor. The device can monitor displacement
However, the bridge looks strong, sturdy and stately could when trains pass. The data in this paper is a comparison of
just become very fragile and dangerous for the users. the data fast and slow trains train when passing the bridge.
Therefore there are required monitoring system to find out Already using WSN method and a few sensors deployed
the cause of the damage to the structure of the bridge which directly on the bridge but the data haven't been uploaded or
can cause the collapse [1]. Used of Structural Health displayed directly on the internet.
Monitoring System (SHMS), structure analysis can be Based on related works that are already described in this
performed to detect and identify the presence of damage or research will be designed a system with sensor and
degradation of the structure of the bridge so that it could
accelerometer node MCU ESP8266 which is the firmware
prolong the service bridge because it can find out the cause
for the application of the technology of the Internet of
of the damage [2]. Factors cause damage and a reduction in
the strength of the structure of the bridge such as the Things on a prototype monitoring system health structures
presence of overload, earthquake, vibrations, collisions, and of the bridge. The data on this system can be seen anywhere
the degradation of the quality of construction of the bridge by internet network as it uses MQTT. So it will be a frugal
[3]. By placing a sensor node on the bridge can be known of and efficient system. The resulting data will be uploaded
the condition of the bridge and the observable parameters directly to the cloud using the MQTT protocol, saved by the
must not exceed the limits specified service (Serviceability database using node .js and the data saved in the database.
Limit State) [4]. Several other studies on Structural Health And data can be viewed by subscribers wherever and
Monitoring System to observe communication system and whenever.
find out the most appropriate topology used for monitoring
bridge structure when used on large bridges [5]. For the III. SYSTEM DESCRIPTION
analysis of data generated by the sensor on the system can On this research built a prototype of health monitoring
use a Fast Fourier Transform algorithm and wavelet system bridge with wireless sensor network. This study uses
transform to produce clearer data for further analysis [6]. SDLC (Software Development Life Circle) method, and
Research on Structural Health Monitoring System using develop several interconnected sensors node can sending
Wireless Sensor Network mode has been there one of them data from the accelerometer sensor reading of results. Draft
using Arduino, ATmega32, and Zigbee is quite expensive and a general overview of the health monitoring system of
sensors are on each node. Because of it, the alternative and the bridge structure is shown in Figure 1.
more efficient system is needed.
Based on the reasons in the background. The author
conducting research on Structural Health Monitoring System
with the design of a more simple hardware and system with
cheap and efficient technology. We propose Accelerometer
Sensor Data Analysis of Bridge Structural Health Monitoring
System, the prototype of the system develop with Internet of
Things that can display the data directly from the sensor to
the internet with low cost hardware, and the authors perform
analysis at the sensor used in this system.
 Fig. 1. Architechture design of the SHM system Fig. 2. The hardware of Wireless Sensor Node

§ B. Software
The workings of this system, starting with the The sensor nodes are configured with the program so
determination of the reference or indication that will be that the sensor can work and produce data. Used the
applied. The sensor used in these tools namely Arduino IDE software to upload programs to the node MCU
accelerometer and sensors will be installed in the body ESP 8266. Upload the program to the node MCU esp8266
section of the prototype structure of the bridge. Data from with the Arduino IDE that has been previously installed
the accelerometer sensor can be analyzed and compared the library from node MCU ESP8266 version
results. Several nodes comprising sensors, accelerometer
and sensor node MCU will be installed at some point and
can monitor the condition of the bridge simultaneously.
With each node connected with Wi-Fi networks that are
generated by the node MCU ESP 8266, connected with a
Wi-Fi modem that is connected to the internet. Data can be
displayed in the form of a real-time graph. The system is
designed with a highly efficient. The creation of this system
is divided into two parts, namely the manufacture of
hardware and software.
A. Hardware
Wireless sensor nodes are composed of four
components, a processing unit, a transceiver unit, a battery
unit, and a sensor unit. This research uses the node MCU
ESP 8266 as data processing. And node MCU ESP 8266
already there is a wireless module with Wi-Fi 802.11 b/ g/ n,
so it can instantly transmit wireless data. Node MCU ESP
8266 working on the resolution 32 bit [7]. In this paper, the
structural vibration is sampled by an accelerometer sensor.
Accelerometer sensor used is the ADXL 345 type that has a
pretty cheap price. Not only low-cost but also have low-
power 3-axis accelerometer with a measurement range of
±16 g. Node MCU and accelerometer sensors work on the
voltage of 3.3 volts. Sensor output data has ADXL 345 with
the digital format. The minimum resolution that can be read
by the accelerometer sensor is of 4mg/ LSB.

Fig. 3. Flow diagram of the program for the node

sensor
 The configuration for Wi-Fi and wireless connections, as Next is by laying the same sensor that is horizontally
well as on the program via Arduino IDE. And then aligned right at the Earth's surface and produce a value of 0
configure the program for the connection with MQTT. The g on the X-axis, 0 g on the Y-axis, and 1 g on the Z axis, that
flow of the program can be seen in Figure 3.2. In this can be seen in Table 1.
system using five sensors to sample the real data from the
board. The system used MCU with a wireless module to
send the data sampled from the bridge prototype. All of the
node sensors sends the sampled data to the cloud using
MQTT. The data that appear and then saved into a database
using the node.js program. The data acquired and stored in
the database is subsequently carried out further analysis. It
also carried the serial data retrieval, to do an analysis of the
data of the accelerometer sensor is still in the form of raw
data.
C. Digital Accelerometer Sensor Data
The experiment can be seen in the comparison of the
data from the accelerometer sensor by performing different
sensitivity settings and compare with the datasheet from the
sensor being used. Accelerometer digital has raw data Fig. 4. Live data sensor with X, Y, Z axis from
output data that can be converted to mg or g [8]. Sensitivity accelerometer
on digital sensors related to the value of the bit of the
sensor. The value of g from digital accelerometer sensor can
be determined based on the value of the LSB of the data that
can be spelled out with formula one, and the value of the
LSB can be searched with the formula two [9].

(1)

It can be viewed and compared with the results of

existing data on the datasheet the sensor to be used. The
value of LSB vary depending on the value of bit sensors are
used. Digital accelerometer has different bits of resolution
depending on the type.

Fig. 5. Experimet set up on bridge model

(2)
Table I Data values to g accelerometer sensors with
different sensitivity
IV. EXPERIMENT RESULTS
This study observed how the data generated by the Sensitivity x (g) y (g) z (g) LSB
accelerometer sensors on a prototype bridge. previously seen 2 -0,0585 0,0507 1,0491 3,9
how the data generated by the sensor node. Data over time
4 -0,0624 0,0546 1,0452 7,8
can be seen live with MQTT applications and can also be
stored in the database is next on to draw in the shape of the 8 -0,0624 0,0468 1,0296 15,6
graph. The first experiment to see if the sensor has been
16 -0,0630 0,0315 1,0395 31,5
successfully transmitted data directly with MQTT. And the
result can be seen on the graph of fig 4. Displaying sensor
data sent wireless and online using MQTT. Then do the experiment by giving objects to move above the
The next test is taking the sampling data resulting from a bridge, it generates data read out sensor as shown in Figure
reading of the accelerometer sensor node in place on the 6. It can be seen from the graph that there are changes to the
body of the prototype model of the bridge. experiment with reading of data sensors accelerometer in the time domain and
changing the values of the sensor sensitivity accelerometer then do the conversion with FFT.
ADXL 345 and compare with the datasheet. the raw data On Figure 6 there are shows the graph of the
generated from the sensor output accelerometer sensitivity experimental data, and describe the form of the output signal
differently when in settings with different values. but from from the sensor is x, y, z-axis. Figure a shows the output
the resulting data would be worth approaching when signal which is read by a sensor node and later changed into
computed using the formula (1). the form of FFT on the graph b. And look a peak magnitude
values on the graph of the FFT. from the resulting data
shown graph FFT on each axis x, y, and z.

(a)

(a)

(b)

Fig. 7. Three-axis sensor node data response when mounting

(b) bolts tight (a) data respons when the bolt loose (b)

Fig. 6. Three-axis sensor node data response node 1 (a)

graph time sensor node with FFT (b)

Test on the model of the bridge is done by taking the

sensor data when given a load moving and not given a load
moving. When the constituent components of the bridge like
a bolt that is still attached to the full and is not complete. then
carried out observations on the outcome of the reading of the
sensor. Shown in figure 7 and 8.

(c)

(d)

Fig. 8. FFT graph when bolts tight (c), FFT graph when bolt
loose (d).

The accelerometer sensor node is set with a 10 ms delay

and an analysis of 1000 data inputs is performed. The higher
peak value difference is generated on the “x” and “y” node
axes when there is a change in the bridge structure at the
maximum value with the FFT computation of the node 4
sensor data which detects the state of the bridge at the time
of the bolt nut condition removed, and with a complete bolt
nut condition. When there is a moving load across the bridge it appears
Node 4 is the sensor node that is placed closest to the that the accelerometer sensor readings are changing. And in
location of the nut and bolt release on the bridge model. The the result seen at the maximum sensor reading value has a
maximum output value of sensor node 4 can be seen in higher value when the bolt release.
Table II. From some tests that have been done known that FFT can be
Table II Output maximum value of data sensor node 4 done on accelerometer sensor data. FFT is performed on
each axis. And the faster of sensor readings are used the
Condition LSB X (g) Y (g) Z (g) higher the sampling frequency. Accelerometer sensor can be
Complete used as vibration detector and also displacement at bridge
bolt 31,2 0,5616 1,3416 2,496 monitoring system. The FFT process is performed on the
accelerometer data to know the value in the real of
Lost bolt 31,2 3,1824 1,0296 4,056 frequency.

ACKNOWLEDGMENT
Thanks to Mr Sudarmono for support at Politeknik Negeri
Semarang, for the bridge prototype.
REFERENCES
[1] G. N. Kawchuk, J. Hartvigsen, T. Edgecombe, N. Prasad, and J.
H. Van Dieen, “Structural health monitoring ( vibration ) as a
tool for identifying structural alterations of the lumbar spine : a
twin control study,” Nat. Publ. Gr., no. March, pp. 6–11, 2016.
[2] K. Bassett and K. Bassett, “Vibration Based Structural Health
Fig. 9. Graph of amplitude maximum FFT axis x,y,z data
sensor when bolts tight and loose Monitoring for Utility Scale Wind Turbines By,” 2010.

When the model of the bridge is passed through the [3] Menteri Pekerjaan Umum dan Perumahan Rakyat, “Perencanaan
moving load, the accelerometer sensor installed produces a Sistem Monitoring Kesehatan Struktur Jembatan,” p. 23, 2015.
different data change output similar to that of the bridge [4] J. Yang, F. Chang, and M. M. Derriso, “Structural Health
model body. The change is also obtained by reading the Monitoring,” Struct. Heal. Monit., 2013.
different sensor output values. The peak value of the [5] E. Setijadi, S. Bp, and P. Suprobo, “Design of Large Scale
resulting FFT is in the range of 0-1 Hz. And when the Structural Health Monitoring System for Long-Span Bridges
bridge model changes by removing the bolts, there is an Based on Wireless Sensor Network,” pp. 169–174, 2013.
increase in data readings as can be seen in the average FFT
[6] S. K. Shome, S. Sen, K. Mondal, and U. Datta, “Development
peak value graph in Figure 9.
and Performance Analysis of Wireless Sensor Node for Structural
V. CONCLUSION Health Monitoring using Fast Fourier and Wavelet Transform,”
pp. 491–495, 2016.
From the research that has been done, it can be seen that
the output data from the accelerometer sensor in the form of [7] A. Skraba, A. Kolozvari, D. Kofjac, V. Stanovov, and E. Semen,
raw data, the analysis needs to be done first, and calibrate “Streaming Pulse Data to the Cloud with Bluetooth LE or
the sensor and match the datasheet of the sensor used. The NODEMCU ESP8266,” pp. 428–431, 2016.
raw data generated by the sensor is analyzed by converting [8] I. Setiawan, B. Setiyono, and T. B. Susilo, “Hasil Uji Kalibrasi
it according to the sensor datasheet into units of "g" or "mg" Sensor Accelerometer,” pp. 118–122.
so that it can be used further. And after observation, knows [9] G. Description, A. Ideas, and K. Specifications, “MMA7455 3-
that the raw output value of the sensor data obtained still Axis Accelerometer Module (# 28526 ) Package and Pin
meets the value of g as found in the datasheet of the Descriptions + Z + X,” vol. 8333, no. 916, pp. 1–7, 2009.
accelerometer sensor ADXL 345 when the sensor is still and
the bridge is not passed through the load.