Project topics and materials
Get Complete Project Material File(s) Now! »
Wireless Sensor Networks(WSN)
Wireless Sensor Networks (WSN) is considered as a significant technology ever since its birth, A wireless sensor network (WSN) consisting of autonomous sensor nodes can provide a rich stream of sensor data representing physical measurements. It can be described as a network of distributed self-powered nodes that could sense or exchange with environment. The main advantage of WSN is that it could be easily and rapidly installed and gather information for a long period of time, pro-viding an enormous quantity of sensor data. WSN based applications have shown a rapid growth in a variety of fields, including target tracking and surveillance, natural disaster relief, health monitoring, environment exploration and geological sensing, etc [5].
So far, most of the sensor networks deployed involve a relatively limited number of sensor nodes. They are usually connected to a central processing unit where all signal processing is performed [6]. On the contrary, the WSN is a wireless distribu-ted network, in which the signal processing is often done with acquisitions. To better understand the necessity of deploying WSN in real applications, some description and statement should be made :
– Wireless
Cabled sensors networks work perfectly when nodes can be wired to stable energy sources and reliable infrastructure of communications. However, in many practical applications, the monitored target-area does not equipped any of these, for example, when the monitoring target is a group of wild ani-mals in the nature. Therefore sensor nodes should rely on local, finite, and relatively small energy sources as well as wireless communication channels, this would open a new door for wider applications with mobility and Auto-nomy.
– Distrubuted sensing
When a precise location of the interest-area is unknown in a surveillance zone, WSN allows a distribution of more autonomous sensors in the place closer to the wanted monitoring area. Instead of using only one or few sensors, this gives more signal to noise ratio (SNR) and better opportunities for the line of sight. SNR can be addressed in some cases by the deployment of a high sensitivity sensor, however, the line of sight of and more generally disturbance of noise cannot be processed by the deployment of a sensor with high sen-sitivity. Thus, distributed sensing provides more robustness under different environmental conditions.
– Distributed processing
It may be considered reasonable that in the cabled sensor networks, data can be communicated back to a central processing unit. However, for the sensor nodes of WSN, there are two main barriers : first, the finite energy budget is the first primary constraint. RF (Radio Frequency) Communication makes the main energy consumer. Secondly, most wireless sensor network defined limited data transmission rate. Thus, we need to process data as much as possible inside the network to reduce the energy consumption as well as the number of bits transmitted, particularly over longer distances.
Artificial Neural Network(ANN)
Computer has become an necessary tool in engineering. Engineers have used various computer applications to improve their efficiency and performance. Ever since early 70s, Artificial Intelligence (AI) have been implemented by engineers to perform specialized tasks design. Although computers are involved in a variety of engineering activities, currently, the main software applicable areas are with well-defined rules, such as the sophis-ticated analysis, graphic and CAD applications, etc. However, where there are no defined rules or heuristics, the use of computer is very limited. Artificial Neural Networks (ANN) are another AI application that has recently been widely used to model nonlinear system, or system with unknown dynamics in many different domains of science and engineering [7]. ANN has been found to be extremely useful in situations where the rules are either unknown or are very difficult to explicit. Some of the main attributes of ANN can be listed as follows :
– ANN can learn and generalize from examples to produce practical solutions to problems.
– They can perfectly cope with situations where the input data of the network is unclear or incomplete.
– ANN are able to adapt solutions in time and to compensate from changing circumstances.
– The data for training an ANN can be theoretical data, experimental data or empirical evidence based on reliable experiences.
ANN can be considered as a good generalized approximator based on the experience of a set of training data, it contains no explicit rules. Although it may not have the exact formality as the traditional parametric approaches, it is still A powerful tool that can produce perfect approximations when formal traditional solutions has difficulties with insufficient knowledge of the problem. It is chosen for applications where precision of traditional techniques can not meet the requirement that the problem is too complex to model with rules [4].
Combination of WSN and ANN in modeling
So far, WSN and ANN together have hardly been used in modeling. However, we think there are two main advantages of applying WSN and ANN in modeling and system identifications. First, the nature of WSN and ANN make them a combi-nation : WSN could be easily and rapidly implemented, providing a huge quantity of sensor data. These data sources can be essential for the ANN to identify a fine grained model. Second, they have high practical values : traditional mathematical modeling, Take the thermal modeling of building rooms as an example, approaches like [8] are usually established with well defined equations, they are usually used in general simulations instead of practical applications. It is mainly because these models are based on elements such as room thermal capacitances/resistance, air-flow rate, heat transfer coefficient, heat gain coefficient, etc. These parameters are difficult to measure precisely in buildings. Also, the dynamic behavior of some phenomenon is very complex 1, It is nearly impossible to obtain an accurate ma-thematical model with limited number of system parameters. The WSN system, on the contrary, is highly transplantable as it could be quickly equipped under any environmental surroundings to gather real-time thermal data. Additionally, with its self-adaptive learning and mapping ability, ANN can directly simulate the rela-tions between the modeling object’s inputs and outputs. Based on the two reasons, it seems that the combination of WSN and ANN can be a valuable and reasonable solution for modeling.
Physical phenomenon involved in buildings
The three main physical phenomena involved in the thermal response of buil-dings are :
– Radiation : transferring heat from one object to another by electromagnetic waves (without contact).
– Conduction : the heat spreads within the material (solid, liquid, gas).
– Transport : The heat transfer between the air and the solid material resulting from the displacement of the particles (from the air) on the interface.
For a homogeneous material, we have the following equations to describe the be-havior of heat transfer through it. If the thickness of the material is T , the thermal resistance R considering the conductivity of the material λ is expressed by : R = T (1.1) λ.
The heat transfer coefficient U 2 is an indicator of energy efficiency. The value of U is expressed as the inverse of R : U = 1 = λ (1.2) R T.
Table of contents :
Chapitre 1 The phenomenon involved in building
1.1 Building Science : Thermal Phenomena
1.2 Physical phenomenon involved in buildings
Chapitre 2 Mathematical model of a room
2.1 Thermal modeling of building
2.1.1 Effective Indoor Thermal Time Constant(EITTC)
2.1.2 Mathematical model of room E106 in UTLN
2.1.3 Limitation of the established mathematical model
2.2 Proposed new modeling method
Part II Artificial Neural Networks and Wireless Sensor Network
Chapitre 3 Neural Network Basics
3.1 Neural network basics
3.2 Model of McCulloch-Pitts
3.3 Perceptron
3.4 The Delta Rule
3.5 Multilayered Neural Network and Back-propagation
3.6 Learning mode
3.6.1 Activation functions
3.7 Initializing the network’s weights
3.8 Momentum
Chapitre 4 The Back-Propagation algorithm
4.1 Back-propagation (BP) algorithm
4.2 Practical aspects of the Back-propagation algorithm and Momentum
4.2.1 Procedure of network learning with Back-propagation
Chapitre 5 Overview of the WSN technology
5.1 Wireless Sensor Networks (WSN)
5.2 A short history
5.3 Applications of sensor network
5.3.1 Military applications
5.3.2 Environmental applications
5.3.3 Health applications
5.3.4 Home applications
5.3.5 Other commercialized applications
5.4 WSN topologies
5.5 The Low power wireless technologies : A comparison
Chapitre 6 Specification and Protocol
6.1 IEEE 802.15.4 and ZigBee
6.2 IEEE 802.15.4 specification
6.2.1 Network addresses of IEEE 802.15.4
6.2.2 The physical layer of IEEE 802.15.4
6.2.3 The MAC layer of IEEE 802.15.4
6.2.4 The frame format of IEEE 802.15.4
6.2.5 Functionality of units in IEEE 802.15.4
6.3 ZigBee
6.3.1 Types of devices and the network topology of ZigBee
6.3.2 The stack ZigBee
6.3.3 Configuration of a network ZigBee
Chapitre 7 Development of a practical WSN system
7.1 WSN system development
7.1.1 MRF24J40MA with PIC18LF4520 microcontroller
7.1.2 Platform based on Freescale MC13224
7.1.3 WSN solution with cc2530 microcontroller from TI
7.1.4 Z-stack of Texas Instrument
7.1.5 The embedded system on CC2530
Chapitre 8 Feasibility of our WSN system : Link quality test
8.1 WSN in buildings : signal quality analysis
8.1.1 Packet Error Rate (PER) test
8.1.2 Signal strength indication and link quality indication
8.1.3 Experiments and analysis
