Technical ReportPDF Available

Artificial Neural Network Based Nuclear Facility Inspection System and Computerized Exploration System for Radiological Ground Surveys

Authors:

Abstract and Figures

1. Using machine learning to efficiently carry out visual inspection of nuclear facilities by detecting and verifying defects in steel surfaces, crack in stainless steel surfaces and concrete surfaces. 2. To establish a nuclear area observation system. 3. To establish a radiological ground survey system. 4. To detect radioactive materials. 5. To clear and verify rumors about radioactive sources.
Content may be subject to copyright.
MILITARY INSTITUTE OF SCIENCE AND TECHNOLOGY
Artificial Neural Network Based Nuclear Facility Inspection System
and Computerized Exploration System for Radiological Ground
Surveys
Project Paper
Course Code: NSE 384
Course Name: Project Design
SUPERVISOR
Lt Col Faisal Kader, PhD, EME
AUTHORS
Name
ID Number
Fahim Sabab Siddique
202028010
Kristy Gourab Sinha
202028011
Safkat Islam Safin
202028032
Submission Date: 07.12.20
SUPERVISOR’S APPROVAL
The report title “Artificial Neural Network Based Nuclear Facility Inspection System and
Computerized Exploration System for Radiological Ground Surveys” submitted by Fahim Sabab
Siddique, ID No- 202028010, Kristy Gourab Sinha, ID No-202028011 and Safkat Islam Safin, ID
No- 202028032 has been accepted as satisfactory in partial fulfillment of the requirement of the
degree of Bachelor of Science in Nuclear Science & Engineering.
_______________________________________________
Lt Col Faisal Kader, PhD, EME
Professor
Department of Nuclear Science and Engineering
Military Institute of Science and Technology (MIST), Bangladesh.
DECLARATION
We hereby declare that this is our original work and it has been written by us in its entirety. We
have duly acknowledged all the sources of information which have been used in the report.
_______________________________
Fahim Sabab Siddique
Department of Nuclear Science and Engineering
Military Institute of Science and Technology
13th December, 2022
_______________________________
Kristy Gourab Sinha
Department of Nuclear Science and Engineering
Military Institute of Science and Technology
13th December, 2022
_______________________________
Safkat Islam Safin
Department of Nuclear Science and Engineering
Military Institute of Science and Technology
13th December, 2022
ACKNOWLEDGEMENT
We genuinely thank the Almighty for all of his blessings in our life and in the accomplishment of
this project. The authors are grateful to everyone who contributed to this research. The writers will
always be indebted to them.
Lt Col Faisal Kader, PhD, EME served as the authors' mentor throughout their tenure at the Nuclear
Science and Engineering Department at MIST. When needed, he has offered every kind of help.
An example would be his morality in both his work and personal lives life. The authors like to
thank their supervisor sincerely for exposing them to the field of nuclear science and engineering.
Working with him has been a privilege.
Additionally, the authors would like to express their gratitude to A S M Nasim, Lecturer in the
Department of Nuclear Science and Engineering at MIST, for his insightful and practical guidance
throughout the research. He was helpful when required and respected the writers' effort.
Finally, the authors wish to thank their parents for raising them and supporting them every step of
the way. They helped them get to where they are now. In order to thank their families for their
assistance, inspiration, and consideration over the semester, the writers would want to convey their
thanks.
i
Table of Contents
1 Introduction .......................................................................................................................................... 3
1.1 Background ................................................................................................................................... 3
1.2 Objectives...................................................................................................................................... 4
2 Literature Review ................................................................................................................................. 6
3 Components ......................................................................................................................................... 9
3.1 ESP32-CAM Wi-Fi + Bluetooth Camera Module ........................................................................... 9
3.1.1 Features .............................................................................................................................. 10
3.2 ESP32-CAM External Antenna ..................................................................................................... 11
3.3 ESP-32S ESP-WROOM-32 Development Board 38P NodeMCU .................................................. 11
3.3.1 Features .............................................................................................................................. 12
3.4 ESP8266 CH340 NodeMCU Wifi Module Lua V3 ........................................................................ 13
3.4.1 Features .............................................................................................................................. 14
3.5 Mini servo motor sg90 ................................................................................................................ 15
3.5.1 Features .............................................................................................................................. 15
3.6 U-Blox NEO-6M GPS .................................................................................................................... 16
3.6.1 Features .............................................................................................................................. 16
3.7 Geiger Counter Kit Nuclear Radiation Detector Gamma Ray Build Radiation Monitoring Station
17
3.7.1 Features .............................................................................................................................. 18
3.7.2 Specification ........................................................................................................................ 19
3.8 ThingSpeak .................................................................................................................................. 19
3.9 Pipes and Joints ........................................................................................................................... 20
3.9.1 Specification of pipes and joints ......................................................................................... 21
3.10 Aluminum plate ........................................................................................................................... 22
3.11 DC motor (12V) ........................................................................................................................... 23
3.11.1 Specifications ...................................................................................................................... 23
3.12 Robot Smart Car Wheel .............................................................................................................. 24
3.12.1 Features .............................................................................................................................. 24
3.12.2 Specifications ...................................................................................................................... 25
3.13 PVC Pipe Clamp and Screws ........................................................................................................ 25
ii
3.14 BTS7960 Motor Driver Module ................................................................................................... 27
3.14.1 Features .............................................................................................................................. 27
3.15 L298N Motor Driver Module ....................................................................................................... 28
3.15.1 Features .............................................................................................................................. 28
3.16 2200mah LiPo Battery ................................................................................................................. 29
3.16.1 Specifications ...................................................................................................................... 29
................................................................................................................................................................ 29
4 Methodology ...................................................................................................................................... 30
4.1 Sensor to Server Data Transfer Flow Chart ................................................................................. 30
4.2 Control Unit ................................................................................................................................. 31
4.3 Circuit Diagram for GPS Tracker .................................................................................................. 32
4.4 Training Convolutional Neural Network ..................................................................................... 33
4.5 Creating User Interface ............................................................................................................... 34
5 Result .................................................................................................................................................. 35
5.1 Concrete Surface Crack Detection Model Using Convolutional Neural Network ....................... 35
5.2 Stainless Steel Surface Crack Detection Model Using Convolutional Neural Network .............. 35
5.3 Steel Surface Defect Detection Model Using Convolutional Neural Network ............................ 35
5.4 User Interface ............................................................................................................................. 35
5.5 GPS Channel Stats ....................................................................................................................... 37
5.6 Geiger Counter Channel Stats ..................................................................................................... 38
5.7 Radiation Data Mapping ............................................................................................................. 38
5.7.1 ThingSpeak Channel Feed ................................................................................................... 38
5.7.2 Radiation Mapping Mist Plaza ............................................................................................ 39
5.8 Esp32 Cam Feed .......................................................................................................................... 40
5.9 Validation .................................................................................................................................... 41
5.9.1 For Geiger Counter Kit Nuclear Radiation Detector Gamma Ray Build Radiation
Monitoring Station .............................................................................................................................. 41
5.9.2 For Convolutional Neural Network ..................................................................................... 41
6 Conclusion .......................................................................................................................................... 42
7 Future scope ....................................................................................................................................... 44
iii
Table of Figures
Figure 1: ESP32-CAM WIFI+ BLUETOOTH CAMERA MODULE ....................................................... 10
Figure 2: ESP32-CAM EXTERNAL ANTENNA ..................................................................................... 11
Figure 3: ESP-32S ESP-WROOM-32 DEVELOPMENT BOARD 38P NODEMCU ............................... 12
Figure 4: ESP8266 CH340 NodeMCU Wifi Module Lua V3 .................................................................... 13
Figure 5: Mini servo motor SG90 ............................................................................................................... 15
Figure 6: U-Blox NEO-6M GPS ................................................................................................................. 16
Figure 7: GEIGER COUNTER KIT NUCLEAR RADIATION DETECTOR GAMMA RAY BUILD
RADIATION MONITORING STATION .................................................................................................. 18
Figure 8: Lira UPVC pipe .......................................................................................................................... 21
Figure 9: 45-degree elbow joint (left), 90-degree elbow joint (middle), PVC cap (right) .......................... 21
Figure 10: Size description of the Aluminum plate .................................................................................... 22
Figure 11: DC motor (12V) ........................................................................................................................ 23
Figure 12: Robot Smart Cart Wheel ........................................................................................................... 24
Figure 13: PVS pipe clamp for connecting pipes........................................................................................ 26
Figure 14: M4 size screw ............................................................................................................................ 26
Figure 15: BTS7960 Motor Driver Module ................................................................................................ 27
Figure 16: L2980N Motor Driver Module .................................................................................................. 28
Figure 17: 2200mah LIPO Battery.............................................................................................................. 29
Figure 18: Sensor to Server data transfer flow chart................................................................................... 30
Figure 19: Flow chart of control system ..................................................................................................... 31
Figure 20: Flow chart of overall system ..................................................................................................... 31
iv
Figure 21: Circuit diagram of GPS tracker ................................................................................................. 32
Figure 22: Stainless Steel surface data set .................................................................................................. 33
Figure 23: Concrete surface data set [13] .................................................................................................. 33
Figure 24: Steel surface defect data set [14] ............................................................................................... 34
Figure 25: Concrete Surface Crack Detection Model Accuracy ................................................................. 35
Figure 26: Stainless Steel Surface Crack Detection Model Accuracy ........................................................ 35
Figure 27: Steel surface Defect Detection Model Accuracy ....................................................................... 35
Figure 28: Image Input to Website Interface .............................................................................................. 36
Figure 29: Classification Result .................................................................................................................. 36
Figure 30: Longitude vs Date...................................................................................................................... 37
Figure 31: Latitude Vs Date ........................................................................................................................ 37
Figure 32: CPM Vs Date ............................................................................................................................. 38
Figure 33: Radiation Mapping Mist Plaza .................................................................................................. 39
Figure 34: ESP32 cam feed ......................................................................................................................... 40
Figure 35: Steel Surface Defects Epoch Vs Loss/Accuracy. ...................................................................... 41
Figure 36: Concrete Surface Defects Epoch Vs Loss/Accuracy. ................................................................ 42
1
ABSTRACT
In order to improve production efficiency, decrease operating costs, improve operational safety
management, and accomplish risk avoidance, artificial intelligence technology is being introduced
into the nuclear industrial chain. Vital and extensively utilized techniques to guarantee the safe
operation of the nuclear industrial chain include regular equipment inspection and remaining
useful life estimation. Manual inspections of the nuclear facilities parts are challenging due to the
high heat and radiation concerns. Some of which are submerged in water to maintain cooling. As
a result, in the traditional routine inspection of facilities cracks, workers frequently do a remote
review and manual detection on the video records of underwater reactors, which is time-consuming
and difficult. Delays cost money mostly because fewer plants are available. The cost of acquiring
replacement power to match a 1000-MWe reactor's output is $500000[1]. Therefore, we suggest
convolutional neural network architecture for automated crack identification in both stainless steel
and concrete surface and defects in steel surface. By focusing on both the channel and the spatial
axis, our proposed architecture makes the best use of the concrete and steel crack surfaces,
achieving a mean accuracy of 99.39% in detecting concrete crack surface, 98.61% in detecting
defects in steel surface and 99.97% in detecting crack in stainless steel surface. The proposed
method has the advantage of being quicker, more reliable, and reproducible than the manual
method now in use. Then there is a strong impetus to utilize remotely controlled robots to collect
radiological measurements throughout due to the need to minimize any risks associated with the
manual collection of radiological measurements in reactor and to adhere to the principles of
ALARP. The robot is also used in carrying visual inspection in valves measuring contamination
levels in components and using various sampling methods to identify contamination utilizing
camera modules and sensors linked to a microcontroller. In our design, the vehicle is controlled
2
by a joystick, and a receiver within the car receives data from the NRF24L01 module. Accordingly,
the motors' speed is controlled. To calculate µSv/h and count cpm, we use the GM Counter
Module. We send these data to the ThingSpeak channel using a microcontroller, then we develop
a radiation map using python folium. If the cpm and dosage exceed the allowable background
radiation level, we may also alert our authorities by utilizing the IFTTT protocol. The primary
objectives of the work are to conduct surveys in emergency situations, to carry out remote
environmental monitoring, to reduce operating delays due to inspections, to identify radioactive
substances in order to determine substance theft, to clear rumors by collection and verification of
data, and to monitor radioactive substances during normal operations and accidental scenarios.
KEYWORDS: ALARP, ThingSpeak, cpm, IFTTT, convolutional neural network.
3
1 INTRODUCTION
1.1 BACKGROUND
Cracking is a common damage caused by stress in a material that can easily be exaggerated by
other factors, including corrosion, fatigue, high pressure, and material of construction. Crack
detection is the process of detecting the crack in the structures using any of the processing
techniques. The crack detection issues with nuclear equipment are extremely crucial because of
the unique working environment of the nuclear reactor. Due to severe accidental condition, cracks
in the containment building occurs. Accidents such as LOCA (Loss of coolant accident) that
releases large amount of radioactivity in the containment building. Which causes cracking to occur
in a nuclear reactor’s containment building. Another major reason of cracking in concrete building
is earthquake. Due to the shockwave produced during earthquake cracks are produced in concrete
structure of containment building. To prevent these accidental scenarios nuclear facility is
inspected on regular periods. Because of the unique characteristics of nuclear power plants, the
nuclear power inspection robot must be operated remotely. A robot can be used for this task. The
inspection robot's primary function is to perform inspection, prevention, reporting, and safety
emergency operations on the instruments in order to ensure the safe operation of the nuclear power
plant. The inspection tasks involve scanning concrete surfaces of the facilities, inspecting steel
surfaces of nuclear equipment, calculating “Remaining estimation lifeof equipment and so on.
Ground radiological survey is the assessment of radiation risks associated with the creation, usage,
or existence of radioactive materials under a certain set of circumstances. Surveys give a precise
measurement of radiation levels in an area and look for radioactive material that has accidentally
been spilled on a person, a surface, or a piece of equipment. When conducting radiological surveys,
4
the ALARA principles of time, distance, and shielding are employed to minimize radiation
exposure. Radiological surveys are done frequently following each day of using or experimenting
with radioactive material, after the radioactive material has been transferred from stock solutions,
if there is a chance that the radiation levels or contamination will change after each experimental
run, cleaning up after a minor radioactive spill or an emergency. Area monitoring is inspecting a
specific area. Areas where nuclear technologies are regularly used employ radiation area
monitoring systems. They continuously scan the atmosphere for any rise in radiation levels, and
when they do, they sound an alarm. This task is particularly important during nuclear
decommissioning. As the area decommissioned needs to be monitored continuously. Nuclear
decommissioning is the process of dismantling a nuclear facility to the point where it no longer
requires radiation protection measures. Source detection is an important phenomenon in the field
of nuclear engineering. Source detection refers to finding radioactive sources. This source
detection plays an important role in clearing rumors about radioactive source. It prevents
unnecessary anxiety caused by radioactive source. As people have a misunderstanding about
radioactive sources and harm caused by them so there is a chance that any rumor can cause public
anxiety. Source handling refers to Source handling refers to transportation and management of
radioactive sources. A robotic system can be used to effectively manage and safely store
radioactive materials by smart handing.
1.2 OBJECTIVES
1. Using machine learning to efficiently carry out visual inspection of nuclear facilities by
detecting and verifying defects in steel surfaces, crack in stainless steel surfaces and
concrete surfaces.
5
2. To establish a nuclear area observation system.
3. To establish a radiological ground survey system.
4. To detect radioactive materials.
5. To clear and verify rumors about radioactive sources.
6
2 LITERATURE REVIEW
Fu-Chen Chen and Mohammad R. Jahanshahi established a deep learning-based crack detection
using convolutional neural network and naïve bayes data fusion in NB-CNN: Deep Learning
Based Crack Detection Using Convolutional Neural Network and Naïve Bayes Data Fusion[2].
The system represents real time video analysis of steel surface cracks by using naïve bayes data
fusion. This system can detect cracks on underwater metallic surfaces from nuclear inspection
videos making inspection task easy and efficient. The advantage of this work is it has a higher hit
rate than any method previously used. It can identify microscopic cracks with low contrast and
variable brightness that are otherwise undetectable. One problem is that the CNN requires a large
amount of training data (e.g., more than 100 000 samples) to converge and prevent overfitting.
Another issue is that the CNN's computations are significantly reliant on a GPU. CNN
computations could be five times slower if a GPU is not used. Furthermore, the suggested NB-
CNN merely detects crack locations without quantifying crack features.
Detection of cracks in nuclear power plant using spatial-temporal grouping of local patches was
developed by Stephen J. Schmugge, Lance Rice, N. Rich Nguyen, John Lindbergy, Robert Grizziy,
Chris Joffey and Min C. Shin.’’[3]. By fine-tuning a deep neural network to find local patches that
contain cracks, which are subsequently clustered in spatial-temporal space for group-level
classification, they suggested a crack detection approach for nuclear power plant inspection videos.
We compare the proposed method to existing ones and test it using data from 17 films totaling
around 150,000 inspection video frames.
7
Devereux, M.G., Murray, P. and West, G.M have presented a new, automatic method for
analyzing video footage captured during routine inspection of the AGR cores in the UK’s fleet of
nuclear power plants in their work “A New Approach for Crack Detection and Sizing in Nuclear
Reactor Cores[4]. A set of techniques were described for detecting the presence in crack in AGR
fuel channel inspection images. The author also presented a method for detecting circular channel
features knows as trepanned holes whose dimensions are known and can be used for scaling. The
proposed system has the advantage that it operates in a fast, robust and repeatable way which
removes the subjectivity associated with manually analyzing image data. The accuracy of the
process was 92.0%.
Keir Groves, Emili Hernandez, Andrew West, Thomas Wright and Barry Lennox in their Robotic
Exploration of an Unknown Nuclear Environment Using Radiation Informed Autonomous
Navigationestablished an autonomous ground vehicle that is designed for exploring unknown
environments which contain sources of ionizing radiation, such as might be found in a nuclear
disaster site or a legacy nuclear facility’’[5]. A gamma radiation detector and an autonomous
navigation package with a radiation avoidance software were installed on an unmanned ground
vehicle. The entire system was tested experimentally in a complex semi-structured environment
with two radiation sources. During the experiment, the robot correctly detected both sources and
avoided areas with high radiation levels while travelling between user-defined waypoints. The
robot utilized a CSIRO Navigation Pack, a Thermo Fischer RadEye G-10 personal gamma
dosimeter with a laptop for communicating with the dosimeter, and a Rajant BreadCrumb ES1
communication node. The CSIRO Navigation Pack (also known as "CatPack") is a 2.5D
localisation and navigation solution for ground vehicles. Its sensor suite includes a Velodyne Puck
(VLP-16) LiDAR, a LORD Microstrain CV5-25 Attitude and Heading Reference System (AHRS),
8
four RGB cameras with wide angle lenses to cover the entire plane of the robot, and LED lights to
function in underground conditions. The advantage of this founding is that an autonomous system
can be developed that can create radiation costmap and alert itself from going into high dose areas
which are harmful for the system. Disadvantage of the work is it cannot create a 3D mapping due
to not using a 3D LIDAR sensor.
Benjamin Bird, Matthew Nancekievill, Andrew West, Jim Hayman, Chris Ballard, Will Jones,
Shaun Ross, Toby Wild, Tom Scott, and Barry Lennox in their VegaA small, low cost, ground
robot for nuclear decommissioning introduced Vega, a compact, low-cost, potentially disposable
ground robot designed for nuclear decommissioning which has been created specifically to aid in
characterization and inspection procedures such as 2D and 3D mapping, radiation scanning, and
sample retrieval’’[6]. Vega was built to do teleoperated exploration through difficult terrains, using
an X86 onboard computer, a huge array of sensors, and differential drive continuous tracks.
Gamma spectrometers, alpha/beta radiation sensors, LiDARs, lighting, and a small robotic arm are
among the sensors and actuators installed in Vega. A 5 DOF arm is fitted for material handling. A
major disadvantage is that mechanical and electrical failure can occur in robot in extreme radiation
environment. The manipulator can only handle small amount (<500g) substance.
Kui Qian, Aiguo Song, Jiatong Bao, and Huatao Zhang in their Small Teleoperated Robot for
Nuclear Radiation and Chemical Leak Detection established a small robot that can detect
radiation and chemical leak. The robot is particularly useful during nuclear and chemical attack.
‘’[7]The robot can climb stairs, dash over rubble, and navigate tiny passageways with ease. The
remote terminal replicates real-time video, audio, detection data, and other sensor readings while
the operator remains at a safe distance. The robot monitors radiation levels and recycles radioactive
9
sources, as well as responding to toxic, hazardous, flammable, explosive, or dangerous things
using a small manipulator. It also uses remote control terminal and a chassis to house the
components. The robot can detect radiation and chemical leak during accidental scenarios. It can
also handle radioactive and chemical substances using a 5-DoF manipulator and variable gripper.
It can smartly handle radioactive substances and dangerous chemical products and transport them
which is very dangerous task for human. The robot doesn’t have extensive ground clearance. It
cannot do radiation mapping. It only determine radioactive substances by seeing in the camera.
3 COMPONENTS
3.1 ESP32-CAM WI-FI + BLUETOOTH CAMERA MODULE
Based on ESP32 chips, the ESP32-CAM is a WIFI+ Bluetooth dual-mode development board with
antennas built into the PCB. It can function as a bare minimum system independently. ESP
integrates WiFi, traditional bluetooth and BLE Beacon, with 2 high-performance 32-bit LX6
CPUs, 7-stage pipeline architecture, main frequency adjustment range 80MHz to 240MHz, on-
chip sensor, Hall sensor, temperature sensor, etc. 3. Fully compliant with WiFi 802.11b/g/n/e/i and
bluetooth 4.2 standards, it can be used as a master mode to build an independent network
controller, or as a slave to other host MCUs to add networking capabilities to existing devices. The
ESP32-CAM is extensively applicable in several IoT applications[8]. It is appropriate for wireless
10
monitoring, wireless identification with QR codes, wireless industrial control, and smart home
devices.
Figure 1: ESP32-CAM WIFI+ BLUETOOTH CAMERA MODULE
3.1.1 Features
1. Ultra-small 802.11b/g/n Wi-Fi + BT/BLE SoC module
2. Low-power dual-core 32-bit CPU for application processors
3. Up to 240MHz, up to 600 DMIPS
4. Built-in 520 KB SRAM, external 4M PSRAM
5. Supports interfaces such as UART/SPI/I2C/PWM/ADC/DAC
6. Support OV2640 and OV7670 cameras with built-in flash
7. Support for images WiFI upload
8. Support TF card
9. Support multiple sleep modes
10. Embedded Lwip and FreeRTOS
11. Support STA/AP/STA+AP working mode
11
12. Support Smart Config/AirKiss One-click distribution network
13. Support for serial local upgrade and remote firmware upgrade (FOTA)
14. Support secondary development [8]
3.2 ESP32-CAM EXTERNAL ANTENNA
When we don't utilize an external antenna, projects that include streaming video typically collapse
because of unstable connectivity. Therefore, in order for our projects to operate properly, we just
need one antenna[9]
Figure 2: ESP32-CAM EXTERNAL ANTENNA
3.3 ESP-32S ESP-WROOM-32 DEVELOPMENT BOARD 38P NODEMCU
It is the 38 Pin version. 2.4GHz Dual-Mode WiFi + Bluetooth Development Board, 38 pin. Ultra-
Low power consumption works perfectly with the IDE. ESP-WROOM-32 ESP32 ESP-32S
Development Board 2.4GHz Dual-Mode WiFi+ Bluetooth Dual Cores Microcontroller. The
12
ESP32 is integrated with Antenna switches, RF Balun, power amplifiers, low-noise amplifiers,
filters, and management modules, and the entire solution occupies the least area of PCB. 2.4 GHz
Wi-Fi plus Bluetooth dual-mode chip, with TSMC Ultra-low power consumption 40nm
technology, power dissipation performance, and RF performance is the best, safe and reliable, easy
to extend to a variety of applications[10].
Figure 3: ESP-32S ESP-WROOM-32 DEVELOPMENT BOARD 38P NODEMCU
3.3.1 Features
1. High performance-price ratio.
2. Small volume, easily embedded to other products.
3. Strong function with support LWIP protocol, Freertos.
4. Supporting three modes: AP, STA, and AP+STA.
13
5. Supporting Lua program, easily to develop.[10]
3.4 ESP8266 CH340 NODEMCU WIFI MODULE LUA V3
The NodeMcu is an open-source firmware and development kit that helps you to prototype your
IoT product with few Lua script lines. The Development Kit based on ESP8266, integrated GPIO,
PWM, IIC, 1- Wire and ADC all in one board.
Figure 4: ESP8266 CH340 NodeMCU Wifi Module Lua V3
14
3.4.1 Features
1. Uses CH340G instead of CP2102.
2. Wireless 802.11 b/g/n standard
3. WiFi at 2.4GHz, support WPA / WPA2 security mode
4. Support STA/AP/STA + AP three operating modes
5. Built-in TCP/IP protocol stack to support multiple TCP Client connections (5 MAX)
6. Support UART / GPIO data communication interface
7. Remote firmware upgrade (OTA)
8. Support Smart Link Smart Networking
9. ESP8266 has IO Pin
10. Don’t need to download resetting
11. For Arduino like hardware IO
12. Open-source, Interactive, Programmable, Low cost, Simple, Smart, WI-FI enabled
13. WI-FI MCU ESP8266 integrated and easy to prototyping development kit.
14. NodeMCU has built-in USB-TTL serial with super reliable industrial strength CH340G for
superior stability on all supported platforms.
15. Advanced API for hardware IO, which can dramatically reduce the redundant work for
configuring and manipulating hardware.
15
3.5 MINI SERVO MOTOR SG90
A servo motor is a type of motor that can rotate with great precision. Normally this type of motor
consists of a control circuit that provides feedback on the current position of the motor shaft, this
feedback allows the servo motors to rotate with great precision. If you want to rotate an object at
some specific angles or distance, then you use a servo motor. It is just made up of a simple motor
which runs through a servo mechanism. If motor is powered by a DC power supply then it is called
DC servo motor, and if it is AC-powered motor then it is called AC servo motor. This SG90 Micro
servo with 1.5kg/cm stall torque is a good choice for our project. It comes with servo arms and
screws for easy installation.
Figure 5: Mini servo motor SG90
3.5.1 Features
1. High resolution
16
2. Accurate positioning
3. Fast control response
4. Constant torque throughout the servo travel range
5. Excellent holding power
3.6 U-BLOX NEO-6M GPS
To navigate the distance with a wider range, then U-Blox Neo-6M GPS Module is needed. Our
project is equipped with a GPS module that allows them to know their location relative to a
network of orbiting satellites. Connecting to signals from these satellites allows the bot to perform
functions such as position hold, exploring areas by navigation etc. This is a complete GPS module
that is based on the NEO 6M GPS.
Figure 6: U-Blox NEO-6M GPS
3.6.1 Features
1. 5Hz position update rate
2. Operating temperature range: -40 TO 85°C UART TTL socket
17
3. EEPROM to save configuration settings
4. Rechargeable battery for Backup
5. The cold start time of 38s and Hot start time of 1s
6. Supply voltage: 3.3 V
7. Configurable from 4800 Baud to 115200 Baud rates. (default 9600)
8. SuperSense ® Indoor GPS: -162 dBm tracking sensitivity
9. Support SBAS (WAAS, EGNOS, MSAS, GAGAN)
10. Separated 18×18mm GPS antenna
.
3.7 GEIGER COUNTER KIT NUCLEAR RADIATION DETECTOR GAMMA RAY BUILD RADIATION
MONITORING STATION
The Geiger Counter produces an audible click and blinks a LED each time it detects a radioactive
particle. It has a Data output jack, that outputs a +5V pulse overtime a radioactive particle is
detected. It also has a headphone jack for private listening. Typically the Geiger counter clicks 10-
20 times a minute due to normal background radiation[11]
18
Figure 7: GEIGER COUNTER KIT NUCLEAR RADIATION DETECTOR GAMMA RAY BUILD RADIATION MONITORING STATION
3.7.1 Features
1. 5V power supply, or 1.5V 3x battery; 1.2V 4x battery, current: 30mA - 12mA for the
detection of 20mR/h ~ 120mR/h of gamma rays and 100 ~ 1800 off variables / points / cm
2 of the soft beta ray.
2. Sound and Light alarm
3. Interrupt the output interface, through this interface can be connected to the microcontroller
and then displayed on the LCD.
4. Supports most of the Geiger tube: M4011, STS-5, SBM20, J305, etc. (the 330~600V
operating voltage of the Geiger tube can be supported).
5. Support the computer (PC) data acquisition, MATLAB analysis and processing.
6. Compatible with Arduino: (recommended UNO R3 Arduino, or any other arbitrary with
5V and external interrupt INT)[11]
19
3.7.2 Specification
1. Name: Radiation Detector system
2. Model:508295
3. Accuracy ±2ppm from 0°C to +40°C
4. I2C Address 0x64
5. Geiger tube Technical parameters diameter: Φ10±0.5mm
6. Geiger tube Total length: 90±2mm
7. Geiger tube Starting voltage: < 350V
8. Geiger tube Recommended operating voltage: 380V
9. Geiger tube Minimum plateau length: 80V
10. Geiger tube Maximum plateau slope: 10%/80V
11. Geiger tube Extreme operating voltage: 550V
12. Geiger tube The maximum count rate: 25 times / min
13. Geiger tube Life: > 1 x 10^9 pulse
14. Geiger tube Medium temperature: -40 ~ 55
15. Geiger tube Size:108x63x20mm [11]
3.8 THINGSPEAK
ThingSpeak is an open-source Internet of Things application and API that uses the HTTP Protocol
to store and retrieve data from sensors. The mathematical computing program MATLAB from
MathWorks is integrated with ThingSpeak, enabling users to analyze and visualize uploaded data
using MATLAB without having to buy a MATLAB license from MathWorks. The ThingSpeak
20
API is used to connect an object to the Internet of Things. The interface enables easy
communication capabilities to IoT gadgets as well as fascinating extra applications (such as
ThingTweet, which will be covered further in a later section)[12]. ThingSpeak, which has APIs
for gathering data generated by sensors and APIs for reading that data from applications, plays the
primary role of continuously updating data. There are two sections to the paper. In one section of
the paper, we have to program something to send data. And in the second section, the other party
must view the data. ThingSpeak sits in the middle and makes it handy to do both. The paper uses
easily accessible hardware to build a proof-of-concept IoT system to monitor radiation, location
etc. Further, this can be modified with different sensors or actuators for building something for
individual purposes. Thus a direct access to all the environmental parameters is given to the user
after the above stated procedure is completed.
3.9 PIPES AND JOINTS
For our proposed model, Lira UPVC pipes were used. These pipes were cut into several pieces and
then connected with different types of joints. The diameter of the pipes used in the model is 1 inch.
Lira UPVC pipe was used in our model because of the strength and durability because they have
high tensile and impact strength. PVC caps were also used to close the open side of the pipe.
21
Figure 8: Lira UPVC pipe
Figure 9: 45-degree elbow joint (left), 90-degree elbow joint (middle), PVC cap (right)
3.9.1 Specification of pipes and joints
Item no
Part name
Size Description
Quantity
1
Lira UPVC Pipe
22 inch
2
2
Lira UPVC Pipe
18 inch
2
3
Lira UPVC Pipe
10.5 inch
2
4
Lira UPVC Pipe
7.5 inch
2
5
Lira UPVC Pipe
12 inch
2
6
90-degree PVC elbow
90-degree angle
6
22
7
45-degree PVC elbow
45-degree angle
4
8
PVC cap
2
3.10 ALUMINUM PLATE
Aluminum plate has many applications due to its strength and durability, and is ideal for use
outdoors due to aluminum’s excellent corrosion resistance. For our proposed model aluminum
plate is used with a length of 20 cm, width of 8 cm and a thickness of 3 mm. It was used to connect
the two pipe bodies with screw for making the joint stronger. 3 holes were made in the aluminum
plate to connect the pipes at three different parts.
Figure 10: Size description of the Aluminum plate
23
3.11 DC MOTOR (12V)
12-volt DC motor is a rotary motor which can convert the direct current into mechanical energy or
convert mechanical energy into DC power. It means that the 12-volt DC motor can interconvert
electric energy and mechanical energy. When it is operated as a DC motor, electric energy is
converted into mechanical energy. The motor was used to run the wheel. Six motors were
connected to the wheel with a hex.
Figure 11: DC motor (12V)
3.11.1 Specifications
Rated power: 20-25 (W)
Product Type: Micro DC Gear Motor
Rated voltage: 12/24(V)
Rated current : 1-2 (A)
Rated speed: 300(rpm)
Rated torque: 13kgcm , Dimensions: 37(mm)
24
3.12 ROBOT SMART CAR WHEEL
The wheel is made up of high-quality rubber which gives maximum traction while operating. The
wheel is strong, shock-proof and sturdy as it features a nylon reinforced plastic rim. It is 85mm
large and 38 mm width.
Figure 12: Robot Smart Cart Wheel
3.12.1 Features
1. Car hub is reinforced nylon, very sturdy.
2. Tire with sponge liner for more strength
3. With upgraded tire tread for greater friction.
4. New design wheel for better combination with the motor
25
3.12.2 Specifications
1. Loading Capacity -3 kg
2. Weight -60 gm
3. Wheel Diameter- 85 mm
4. Color- Black (Tire), Blue (Rim)
5. Wheel Width -38 mm
6. Body Material- Plastic
7. Grip Material- Rubber
8. Weight -0.064 kg
9. Dimensions- 10 × 10 × 5 cm
3.13 PVC PIPE CLAMP AND SCREWS
A pipe fitting is a common term used to describe any seal or grommet that holds two things
together. You can make gaskets from rubber, plastic, metal, and other materials. The material
choice depends on why you need to use a gasket. But as you'll learn, there's more to gaskets than
a seal preventing a leak. In our proposed model metal gaskets were used to tight the pipes.
26
Figure 13: PVS pipe clamp for connecting pipes
For our proposed model M4 screws were used. The M4 machine screw has a thread diameter of 4
millimeters. It is used in giving structural stability.
Figure 14: M4 size screw
27
3.14 BTS7960 MOTOR DRIVER MODULE
Motor drivers acts as an interface between the motors and the control circuits. Motor require high
amount of current whereas the controller circuit works on low current signals. So the function of
motor drivers is to take a low-current control signal and then turn it into a higher-current signal
that can drive a motor
3.14.1 Features
1. Input voltage (V): 6 27.
2. Maximum current (A): 43.
3. Input level (V): 3.3 5.
4. Duty cycle: 0 100%.
5. Path resistance: 16 mΩ at 25°C.
Figure 15: BTS7960 Motor Driver Module
28
3.15 L298N MOTOR DRIVER MODULE
Motor drivers acts as an interface between the motors and the control circuits. Motor require high
amount of current whereas the controller circuit works on low current signals. So the function of
motor drivers is to take a low-current control signal and then turn it into a higher-current signal
that can drive a motor
3.15.1 Features
1. Maximum motor supply current: 2A per motor.
2. Current Sense for each motor.
3. Heatsink for better performance.
4. Power-On LED indicator.
5. Double H bridge Drive Chip: L298N
Figure 16: L2980N Motor Driver Module
29
3.16 2200MAH LIPO BATTERY
3.16.1 Specifications
1. Voltage: 11.1V
2. Capacity: 2200mah
3. Cell Type: Li-polymer
4. Configurations: 3S
5. Continuous Discharge Rate: 35C
6. Max Burst Rate: 70C
7. Discharge(Output) Lead: T-Plug
8. Wire Length: 100mm
9. Weight: 196g
10. Dimensions:106*35*25
Figure 17: 2200mah LIPO Battery
30
4 METHODOLOGY
4.1 SENSOR TO SERVER DATA TRANSFER FLOW CHART
Figure 18: Sensor to Server data transfer flow chart
31
4.2 CONTROL UNIT
Figure 19: Flow chart of control system
Figure 20: Flow chart of overall system
32
4.3 CIRCUIT DIAGRAM FOR GPS TRACKER
The connections between the node MCU and the NEO 6M module is simple. The VCC and GND
pin of the NEO 6M module is connected to the positive and negative power rail of the breadboard,
and the 3v3 and GND pin of the node MCU is connected to the positive and negative power rail
of the breadboard as shown in the below fig. The RX and TX pin of the NEO 6M is connected to
the D1 and D2 pins of the node MCU. The Node MCU is powered by using USB.
Figure 21: Circuit diagram of GPS tracker
33
4.4 TRAINING CONVOLUTIONAL NEURAL NETWORK
Figure 22: Stainless Steel surface data set
Figure 23: Concrete surface data set [13]
i. The training test consisted of 20000 concrete images having no crack on surface and 20000
images having crack on its surface.
ii. The training test consisted of 2000 steel surface images having no crack and 2000 images
having crack on its surface
34
iii. The training test consisted of 6 classes of steel defect images (crazing, inclusion, patches,
pitted, rolled, scratches) having 240 images of each defect.
Figure 24: Steel surface defect data set [14]
4.5 CREATING USER INTERFACE
After training the neural network a user interface was created using python flask in google colab
platform. In the interface the frames captured by the ESP32 cam are inserted. Then images are
35
inserted into the interface and with the click of a button we can know about the characteristics of
the surface (e.g. concrete crack surface, concrete non crack surface, defects in steel surface etc.)
5 RESULT
5.1 CONCRETE SURFACE CRACK DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORK
Figure 25: Concrete Surface Crack Detection Model Accuracy
5.2 STAINLESS STEEL SURFACE CRACK DETECTION MODEL USING CONVOLUTIONAL NEURAL
NETWORK
Figure 26: Stainless Steel Surface Crack Detection Model Accuracy
5.3 STEEL SURFACE DEFECT DETECTION MODEL USING CONVOLUTIONAL NEURAL NETWORK
Figure 27: Steel surface Defect Detection Model Accuracy
5.4 USER INTERFACE
36
Figure 28: Image Input to Website Interface
Figure 29: Classification Result
37
5.5 GPS CHANNEL STATS
Figure 30: Longitude vs Date
Figure 31: Latitude Vs Date
38
5.6 GEIGER COUNTER CHANNEL STATS
Figure 32: CPM Vs Date
5.7 RADIATION DATA MAPPING
5.7.1 ThingSpeak Channel Feed
Observation
Number
Latitude
Longitude
Radiation(μSv/h)
1
23.83847
90.35836
0.139
2
23.83843
90.35839
0.13
3
23.8384
90.35839
0.146
4
23.83826
90.35829
0.136
5
23.83822
90.35837
0.137
6
23.83814
90.35837
0.144
7
23.83711
90.35903
0.149
39
8
23.83808
90.35839
0.147
9
23.83806
90.3584
0.149
10
23.83783
90.35829
0.154
11
23.83789
90.35829
0.158
12
23.83797
90.35822
0.16
13
23.83801
90.35824
0.15
14
23.83822
90.35823
0.152
15
23.83832
90.35824
0.14
16
23.83851
90.3582
0.146
17
23.83855
90.35831
0.135
5.7.2 Radiation Mapping Mist Plaza
Figure 33: Radiation Mapping Mist Plaza
40
5.8 ESP32 CAM FEED
Figure 34: ESP32 cam feed
41
5.9 VALIDATION
5.9.1 For Geiger Counter Kit Nuclear Radiation Detector Gamma Ray Build Radiation
Monitoring Station
Count from the Geiger counter = 0.147 μSv/h
Count from Geiger Counter Kit Nuclear Radiation Detector Gamma Ray Build Radiation
Monitoring Station = 22 cpm

 μSv/h [The GM tube M4011 conversion index is 151 and 151 cpm = 1 μSv/h]
= 0.146 μSv/h
Percentage of error 
 
= 0.68%
5.9.2 For Convolutional Neural Network
Figure 35: Steel Surface Defects Epoch Vs Loss/Accuracy.
42
Figure 36: Concrete Surface Defects Epoch Vs Loss/Accuracy.
6 CONCLUSION
The proposed system is an effective and handy solution for exploration for radiological ground
surveys, nuclear area observation. It can successfully detect cracks in concrete surface and stainless
steel surface and defects in steel surface using convolutional neural network. Binary detection
model is an effective method of analyzing surface of materials. By analyzing the surfaces, we can
easily determine the current situation of materials and the necessity of any action. This proposed
method requires considerable amount of time but can detect surface crack in concrete with 99.39%
accuracy and steel with 98.61% accuracy for defects and 99.97% accuracy for stainless steel crack
surface. Established prototype can perform the task of radiation monitoring, visual inspection of
concrete and steel surfaces. It can also represent the outputs from Geiger Counter Kit Nuclear
Radiation Detector Gamma Ray Build Radiation Monitoring Station in terms of cpm (count per
minute) data, cpm Vs Date graph, radioactivity (μSv/h) Vs Date graphs on ThingSpeak. It can also
utilize the U-Blox NEO-6M GPS Module output to create Latitude Vs Date graph and Longitude
43
Vs Date graph, location on ThingSpeak. Our proposed system can also perform successfully the
task of radiation mapping by using python folium which is a significant perspective to verify
rumors about nuclear threat. The proposed architecture shows efficient result making the system
more attractive.
44
7 FUTURE SCOPE
1. Using steel body and box instead of pipe for carrying more loads, protection from radiation
of all robot components and for better service.
2. Advanced robotic arm using linear actuators for carrying more loads.
3. Improving the architecture of Convolutional Neural Network for better accuracy in training
and testing.
4. Improving the efficiency of the detection system by real time video analysis.
5. Improving the dataset for Convolutional Neural Network by collecting steel surface image
from a nuclear reactor.
6. Developing customized website for better representation of result data.
7. Improving human robot interaction interface.
8. Addition of better detector and scintillation detector for radioactive material detection
REFERENCES
[1] “1] T. Moore, ‘Robots for nuclear power plants,’ IAEA BULLETIN , 1985. - Google Search.”
https://www.google.com/search?q=1%5D+T.+Moore%2C+%22Robots+for+nuclear+power+plant
s%2C%22+IAEA+BULLETIN+%2C+1985.&oq=1%5D++T.+Moore%2C+%22Robots+for+nuclear+pow
er+plants%2C%22+IAEA+BULLETIN+%2C+1985.&aqs=chrome..69i57.827699j0j7&sourceid=chro
me&ie=UTF-8 (accessed Dec. 14, 2022).
[2] F. C. Chen and M. R. Jahanshahi, “NB-CNN: Deep Learning-Based Crack Detection Using
Convolutional Neural Network and Naïve Bayes Data Fusion,” IEEE Transactions on Industrial
Electronics, vol. 65, no. 5, pp. 43924400, May 2018, doi: 10.1109/TIE.2017.2764844.
[3] S. J. Schmugge et al., “Detection of cracks in nuclear power plant using spatial-temporal grouping
of local patches,” 2016 IEEE Winter Conference on Applications of Computer Vision (WACV), May
2016, doi: 10.1109/WACV.2016.7477601.
45
[4] M. G. Devereux, P. Murray, and G. M. West, “A new approach for crack detection and sizing in
nuclear reactor cores,” Nuclear Engineering and Design, vol. 359, p. 110464, Apr. 2020, doi:
10.1016/J.NUCENGDES.2019.110464.
[5] K. Groves, E. Hernandez, A. West, T. Wright, and B. Lennox, “Robotic exploration of an unknown
nuclear environment using radiation informed autonomous navigation,” Robotics, vol. 10, no. 2,
Jun. 2021, doi: 10.3390/ROBOTICS10020078.
[6] B. Bird et al., “Vega—A small, low cost, ground robot for nuclear decommissioning,” J Field Robot,
vol. 39, no. 3, pp. 232245, May 2022, doi: 10.1002/ROB.22048.
[7] K. Qian, A. Song, J. Bao, and H. Zhang, “Small teleoperated robot for nuclear radiation and
chemical leak detection,” Int J Adv Robot Syst, vol. 9, Sep. 2012, doi: 10.5772/50720.
[8] “ESP32-CAM WiFi + Bluetooth Camera Module.”
https://bdspeedytech.com/index.php?route=product/product&product_id=3503 (accessed Dec.
07, 2022).
[9] “ESP32-CAM External Antenna Robotics Bangladesh.” https://store.roboticsbd.com/internet-of-
things-iot/1852-esp32-cam-external-antenna-robotics-bangladesh.html (accessed Dec. 07, 2022).
[10] “ESP-32S ESP-WROOM-32 Development Board 38P NodeMCU Robotics Bangladesh.”
https://store.roboticsbd.com/arduino-bangladesh/1820-esp-32s-esp-wroom-32-development-
board-38p-nodemcu-robotics-bangladesh.html (accessed Dec. 07, 2022).
[11] “Geiger Counter Kit Nuclear Radiation Detector Gamma Ray Build Radiation Monitoring Station
Robotics Bangladesh.” https://store.roboticsbd.com/sensors/1111-geiger-counter-kit-nuclear-
radiation-detector-radiation-monitoring-station-robotics-bangladesh.html (accessed Dec. 07,
2022).
[12] “*1 Maureira, Marcello A. Gómez, Daan Oldenhof, and Livia Teernstra. ‘ThingSpeak–an API and
Web Service for the Internet of Things.’ World Wide Web (2011). - Google Search.”
https://www.google.com/search?q=*1+Maureira%2C+Marcello+A.+G%C3%B3mez%2C+Daan+Ol
denhof%2C+and+Livia+Teernstra.+%22ThingSpeak%E2%80%93an+API+and+Web+Service+for+th
e+Internet+of+Things.%22+World+Wide+Web+(2011).&oq=*1+Maureira%2C+Marcello+A.+G%C
3%B3mez%2C+Daan+Oldenhof%2C+and+Livia+Teernstra.+%22ThingSpeak%E2%80%93an+API+a
nd+Web+Service+for+the+Internet+of+Things.%22+World+Wide+Web+(2011).&aqs=chrome..69i
57.1412j0j7&sourceid=chrome&ie=UTF-8 (accessed Dec. 14, 2022).
[13] “Surface Crack Detection | Kaggle.” https://www.kaggle.com/datasets/arunrk7/surface-crack-
detection (accessed Dec. 09, 2022).
[14] “Metal Surface Defects Dataset | Kaggle.” https://www.kaggle.com/datasets/fantacher/neu-
metal-surface-defects-data (accessed Dec. 09, 2022).
46
ResearchGate has not been able to resolve any citations for this publication.
ResearchGate has not been able to resolve any references for this publication.