L f e / s o u / n /oh 
NEURAL NETWORKS BASED TYRE IDENTIFICATION 
FOR A TYRE INFLATOR OPERATIONS 
A Thesis submitted to the 
Department of Electrical Engineering, University of Moratuwa 
On partial fulfilment of the requirement for the 
Degree of Master of Science in Industrial Automation 
LIBRARY 
UNIVE RSITY OF MORATUWA, SRI LANKA 
MORATUWA 
by 
KAHANDAWA APPUHAMILLAGE 
GAYAN CHANAKA KAHANDAWA <3 \ 
G 2 \ - 3 
Supervised by 
Dr. Lanka Udawatta 
University of Moratuwa ' ' V 5 
91251 
Department of Electrical Engineering 
University of Moratuwa 
Sri Lanka 
March 2008 
91251 
DECLARATION 
The work submitted in this dissertation is the result of my own investigation, 
except where otherwise stated. 
It has not already been accepted for any degree, and is also not being 
concurrently submitted for any other degree. 
Kahandawa K. A. G. C. 
26.03.2008 
endorse the declaration by the candidate. 
Dr. Lanka Udawatta 
Acknowledgement 
In completion of this project, first and foremost gratitude is due to Dr. Lanka 
Udawatta for supervising, guiding and encouraging throughout the project giving 
invaluable advice. Also I thank Dr. D.C Bandara for giving valuable advice to make 
this project a complete success. 
I would thank all the lecturers of the postgraduate study course in the Department of 
Electrical Engineering of the University of Moratuwa, Sri Lanka who imparted the 
theoretical knowledge and the encouragement in bringing up this academic work with 
excellent cooperation and guidance 
Further I would like to thank the staff members of the Department of Electrical 
Engineering for spending their valuable time to provide relevant instructions and 
service for the progress of the research experiment. 
Also I should like to thank many individuals, friends and colleagues, who have not 
been mentioned here personally, for making this product a success. 
Finally I thank to my family members for helping and encouraging me throughout the 
fieldwork of the research and being patient with me. I may not be able to accomplish 
this task without their support. 
Abstract 
Tire industry has become one of the huge industries at present since the number of 
vehicles are rapidly increasing. Since the tire directly related to the vehicle safety, 
durability, running cost and comfort level of the passengers tire selection and 
maintenance addressed widely. Maintaining a tire is a duty of the vehicle user and he 
may has to check the tire rapidly for wear, cuts and other irregularities. The most 
important activity of tire maintenance is to maintain the pressure of the tire. 
The tire pressure is an important issue since it directly relates to the safety of the 
vehicle, durability, running cost and comfort level of the passengers as mentioned 
above. A rapid check for tire pressure is essential since tire pressure may reduce 
ii 
normally with the time apart from losing pressure when the tire drives over a pothole 
or hump. Checking the tire is done using a machine called tire inflator. This machine 
must be accurate and user friendly since the operator has to trust on it. Tire inflators 
are available mostly in tire shops and in gas stations. Tire inflators are mainly two 
types, Analog tire inflators and Digital tire inflators. Digital tire inflators arrived the 
market recently and analog tire inflators are getting replaced with digital tire inflators 
as the convenience of operation. 
Digital tire inflators read the pressure with a pressure sensor and with this sensor the 
dynamic pressure readings are not possible. Hence to get the static pressure the 
inflation process has to be stopped. In other words, while in the inflation, the tire 
inflator has no idea about the tire pressure until the inflation stops. In this case, to 
have an idea about the rise of pressure, there must be a method to identify the type of 
the tire. 
The tire identification mechanism must be fast, accurate and reliable. The other 
requirement is that the tire has to be identified online and this process must not delay 
the inflation process. 
The main tusk of this exercise is to develop an artificial neural network based tire 
identification method. A developed tire inflator model was used to collect information 
and to test the tire identification process. 
To develop the network, three inputs were considered. By expanding the number of 
layers in the network experiment was carried out. The results were successful and that 
the network with two hidden layers zero percent error achieved. 
Content 
Declaration 
Acknowledgement 
Abstract 
List of Figures 
List of Tables 
Chapter 1 
1. Introduction to the working environment 
1.1 Background 
1.2 Pressure inside tire 
1.3 Goal 
Chapter 2 
2. Problem Identification 
2.1 Properties of the Tire 
2.1.1 Temperature Resistance 
2.1.2 Tread Wear Number 
2.1.3 Traction of Tire 
2.1.4 Maximum load capacity and tire speed 
2.1.5 Manufacturing date 
2.1.6 Tire Pressure 
2.2 Tire Inflator 
2.2.1 Operation of Tire Inflator 
2.2.1.1 The Analog Tire Inflator 
2.2.1.2 The Digital Tire Inflator 
2.2.2 Controller of the Inflator 
2.2.3 The Tire Inflator Model 
Chapter 3 
3. Theoretical Background 
3.1 Introduction to Neural Networks 
3.2 Engineering of Brain 
3.3 Neuron Physiology 
3.3.1 Neuron 
3.4 Artificial Neural Networks 
3.4.1 Basic Model of Neuron 
3.4.2 Learning in Artificial Neural Networks 
3.4.3 Characteristics of ANN 
3.4.4 Neural Network Topology 
3.4.5 Single Layer Networks 
3.4.6 Multilayer Neural Networks 
3.4.7 The Backward Propagation Algorithm 
Chapter 4 32 
4. Development of Tire Inflator Model 32 
4.1 Tire Inflator Model 32 
4.1.1 The Microcontroller 3 3 
4.1.2 The Pressure Sensor 33 
4.1.3 Pneumatic Valve 33 
4.1.4 LCD Screen 35 
4.2 Working of the Tire Inflator 36 
4.3 Tire Identification 38 
4.4 Simple On-Off Algorithm 39 
4.5 The Tire Identification System 40 
Chapter 5 41 
5. Neural Network Model 41 
5.1 Data Analysis 41 
5.2.1 Experiments With Artificial Neural Networks 43 
5.2.2 Three Input MPL 43 
5.3 Network with a Single Hidden Layer 44 
5.4 Network with Two Hidden Layers 47 
Chapter 6 52 
6. Discussion, Conclusion and Future Developments 52 
6.1 Discussion and Conclusion 52 
6.2 Future Developments 53 
7. References 54 
Appendices 
Appendix A A1 
Appendix B B1 
Appendix C CI 
Appendix D D1 
Appendix E El 
Appendix F • F1 
v 
List of Figures 
Figure 1.1 Correctly Inflated and Underinflated Tire 2 
Figure 1.2 Tire Connector 3 
Figure 1.3 Dynamic Pressure Rise inside a Tire with Time 4 
Figure 2.1 Tire Details Available in the Tire Side Wall 6 
Figure 2.2 Temperature Resistance Mentioned in Tire 7 
Figure 2.3 Tread Wear Number 8 
Figure 2.4 Traction of a Tire 8 
Figure 2.5 Load Index and Speed Symbol of the Tire 9 
Figure 2.6 Load Capacity Vs Inflation Pressure for Three Different Tires 9 
Figure 2.7 Tire Manufactured Year 11 
Figure 2.8 Maximum Permissible Pressure that Tire Could Withstand 12 
Figure 2.9 Tire Carrying Capacity with the Tire Pressure 13 
Figure 2.10 Infrared Photograph of a Tire 13 
Figure 2.11 Overinflated Tire 14 
Figure 2.12 Under Inflated Tire and Correctly Inflated Tire 15 
Figure 2.13 Analog Tire Inflators 16 
Figure 2.14 Digital Tire Inflator 17 
Figure 2.15 Pressure Signal Analysis 18 
Figure 2.16 Working flow chart of a Tire Inflator 20 
Figure 3.1 The Biological Neuron 23 
Figure 3.2 The Parts of the Neuron 23 
Figure 3.3 The Synapse 24 
Figure 3.4 Basic Neuron Model 26 
Figure 3.5 Common Neural Network Topologies 29 
Figure 4.1 The Tire Inflator Hardware 32 
Figure 4.2 SX. 150 Pressure Sensor 33 
Figure 4.3 5/3 Valve: 5 Port, 3 Position, Center Closed 34 
Figure 4.4 Port Allocation of the Valve 34 
Figure 4.5 The LCD Display 35 
Figure 4.6 Fiberglass Enclosure 35 
Figure 4.7 Final Look of the Tier Inflator 36 
Figure 4.8 Valve Positions ' 36 
Figure 4.9 Variation of Pressure Rise of Several Tire Types 39 
Figure 4.10 IR Tire Identifier 40 
Figure 4.11 IR Scanning 40 
Figure 5.1 Variation of Inputs 41 
Figure 5.2 The Plot of Three Different Tire Types 42 
Figure 5.3 The Plot of four Different Tire Types 42 
Figure 5.4 Architectural Design of the System 44 
Figure 5.5 Network Model with one hidden Layer 44 
Figure 5.6 Variation of Percentage Misclassified error for Training Set 46 
with a Number of Iterations for Network with one hidden layer 
Figure 5.7 Variation of Percentage Misclassified error for Validation Set 46 
with a Number of Iterations for Network with one hidden layer 
Figure 5.8 Network Model with two hidden Layer 47 
vi 
Figure 5.9 Variation of Percentage Misclassified error for Training Set 49 
with a Number of Iterations for Network with two hidden layer 
Figure 5.10 Variation of Percentage Misclassified error for Validation Set 49 
with a Number of Iterations for Network with two hidden layer 
vi i 
List of Tables 
Table 2.1 Temperature Resistance of a Tire 7 
Table 2.2 Maximum Load Carrying Capacity of a Tire with Load Index 10 
Table 2.3 Maximum Speed Limit of a Tire 11 
Table 2.4 Tire Classification 19 
Table 4.1 Pressure Rise of Several Tires for an Inflation Time of 0.25s 38 
Table 5.1 Misclassified Percentage for Network with Single Hidden Layer 45 
Table 5.2 The Final Network Weights 47 
Table 5.3 Misclassified Percentage for Network with Single Hidden Layer 48 
Table 5.4 The Weights of the Network with Two hidden layers 50 
Table 5.5 Predicted Values with Actual Values 51 
vii i