You are on page 1of 6

Proceedings of 14th International Conference on Computer and Infonnation Technology (ICCIT 2011) 22-24 December, 2011, Dhaka, Bangladesh

A Reliable Structural Health Monitoring Protocol


Using Wireless Sensor Networks
t ·
+

Md. Abdul Hamid, M. Abdullah-AI-Wadud , Muhammad Mahbub Alam


Dept. of Information & Communications Engineering, Hankuk University of Foreign Studies, Yongin, Korea.
t
Dept. of Industrial and Management Engineering, Hankuk University of Foreign Studies, Yongin, Korea.
+
.

Dept. of CIT, Islamic University of Technology, Board Bazar, Gazipur, Bangladesh.


hamid@hufs.ac.kr, wadud@hufs.ac.kr, mma@iut-dhaka.edu

damage or deterioration, and determining the structure


Abstract condition or health. SHM is an active area of research
In this paper, we present a distributed reliable devoted to systems that can autonomously and
structural health monitoring (SHM) protocol using proactively assess the structural integrity of bridges,
hierarchical wireless sensor networks. We assume that buildings, and aerospace vehicles. Recent technological
the sensor network consists of (i) sensors capable of advances promise the eventual ability to cover a large
sensing the structural health conditions (e. g. , strain, civil structure with low-cost wireless sensors that can
vibration, pressure, temperature, etc. ), (ii) cluster-head continuously monitor a building'S structural health, but
nodes that collect and process the sensory researchers face several obstacles to reaching this goal,
measurements from the sensors and prepares a local including high data-rate, data-fidelity, and reliability
report and (iii) a sink that collects the local reports. requirements. A combination of factors (e.g., recent
Sensors are organized into single hop clusters, each catastrophic natural disasters, terrorist activities, rapidly
managed by a cluster-head node. A cluster head should deteriorating infrastructure, and increasing costs for new
take the sensor-faults into account when preparing the construction, etc.) points to the need for a robust, low­
local report based on the measurements of its cost, low-power, and high-performance autonomous
subordinate sensors. In majority decision�rule for wireless monitoring system for the "health " and safety
combining measurements, the judgement about health of buildings, bridges, and other civil structures. Existing
status will follow the majority where a correct majority building monitoring systems are typically hard-wired,
decision requires that a majority of observing nodes sparse, expensive, and provide only limited data. The
provide accurate measurements. Using Baysian use of wireless system would potentially be inexpensive
approach to form a judgment is problamatic without and extremely versatile, enabling the widespread use of
additional information or assumptions (for example, the sensors with the capability of sensing various health
difficulty of knowing conditional probabilities). conditions such as strain, temperature, vibration
Dempster-Shafer theory of evidence based approach measurements.
overcomes these limitations. Unlike the simple binary The design of a robust structural health assessment
decision, it produces a judgment value between 0 and 1 system would be applicable not only to bridges, dams
that reflects the degree of belief in that judgment. It and major buildings but also industrial facilities such as
discounts the unreliable observer's measurements. Our chemical plants, oil refineries, biotech and
results through extensive simulations show the efficacy pharmaceutical plants, high precision manufacturing
of the proposed scheme. We believe that our design plants and other facilities requiring continuous, real­
might lead to the development of commercial, cost­ time monitoring of large areas. However, there are two
effective, yet efficient and reliable distributed SHM fundamental challenges in such monitoring problem
systems to effectively monitor and intelligently detect using wireless sensor networks. First, the detection
structural health under various loads. accuracy is limited by the amount of noise associated
with the measurement and the reliability of sensor
Keywords: Structural Health Monitoring, Wireless nodes. The sensor nodes are usually low-end
Sensor Networks, Dempster-Shafer, Performance. inexpensive devices and sometimes exhibit unreliable
behavior. For example, a faulty sensor node may send a
I. INTRODUCTION physical measurement that reflects the severe condition
A wireless sensor network (WSN) consists of spatially though, really it is not such in an event region and vice
distributed autonomous sensors to monitor physical or versa. This might add an additional layer of complexity
environmental conditions, such as temperature, sound, for the damage assessment problem. Hence, it is
vibration, pressure, motion or pollutants, and to essential to consider these factors during health
cooperatively pass their data through the network to a monitoring to achieve better overall assessment
main location. In particular, one of the important sensor accuracy and to limit the effects of inaccurate
network applications is monitoring the structural health measurement or faulty behavior of individual
[1], [2], [3], [4]. Structural Health Monitoring (SHM) components to a minimum. Second, a WSN consists of
targets monitoring the in-situ behavior of a structure thousands of sensors that are in charge of both
such as a bridge, assessing its performance, detecting monitoring and data transmission tasks. The data
collected by each sensor is transmitted via a network
987·161284·908·91111$26.00 © 201 I IEEE
consisting of other sensors towards a well identified works have demonstrated the feasibility of continuous
destination node called sink. In the basic setting of a structural data collection using a wireless network [5],
WSN, each individual piece of data is thus [3], [4]. However, existing approaches do not consider
independently transmitted over several hops towards the the measurement uncertainty posed by the sensor nodes
sink and each sensor node is involved in the forwarding and the problem of centralized data coIIection in which
of a large number of data pieces originated from other sink coIIects aII the raw data, and so on.
sensors. In the resource constrained WSN environment, There are several formalisms able to mathematicaIIy
forwarding of large amounts of data becomes the major model someone's degrees of belief. A very popular one
focus of energy and bandwidth optimization efforts. The is the Bayesian Theory, which is based on a prior
source of energy for a sensor node is most often an knowledge of a probability distribution. Krishnamachari
attached battery cell. Hence, centralized health and Iyengar introduced a fault-tolerant event detection
monitoring algorithms, where all sensor nodes need to method [6] for wireless sensor networks using Bayesian
transmit their individual measurements directly to a approach [7]. Based on the observation that the sensor
central monitoring node (sink), are not suitable for a faults are likely to be stochastically uncorrelated, while
wireless sensor network due to energy constraints. event measurements are likely to be spatially correlated,
Moreover, since high volume of data has to traverse a they propose letting an individual sensor node
large number of radio hops, network wiJI suffer from communicate with its n neighbors and using their binary
high packet loss, buffer drops and increased delays and decisions to correct its own decision. A majority voting
will tend to be grossly unfair towards the nodes far scheme is shown to be the optimal decision scheme for
away from the sink. A localized, distributed and reliable fault correction in their work. The scheme has the
monitoring system is highly preferred for wireless advantage of being completely distributed and localized,
sensor networks. where each node only needs to obtain information from
Thus, in this paper, a distributed monitoring approach its neighbor sensors in making its decisions. Authors in
has been put forward as an essential technique to [8] proposed a distributed fault-tolerant event detection,
achieve fault-tolerant and reliable structural data which is an improvement of [6] considering two
acquisition. The idea of distributed monitoring is to additional important questions: (i) how to address both
have each of the independent sensors sends observation the noise-related measurement error and sensor fault
to its corresponding cluster-head sensor that generates a simultaneously in fault-tolerant detection, and (ii) how
local decision from the small area surrounded by the to choose a proper neighborhood size n for a sensor
cluster-head node. Then the cluster-head sensor sends node in fault correction such that the energy could be
the decision through other cluster-head sensors en-route conserved. Clearly, the Bayesian approach used in [6]
to the sink in order to reduce redundancy and minimize and [8] requires complete knowledge of both prior and
bandwidth usage. Since, the structural-response data conditional probabilities, which might be difficult to
acquisition is distributed (i.e., cluster-head sensors determine in practice. We often estimate prior
continuously collect structural response data from their probabilities from empirical data, or, in the absence of
local regions), each of the cluster-head nodes holds the empirical data, we assume them to be uniform or some
final report regarding the health condition. All the other distribution. The outcome reflects these
reports (decisions) from aII the fusion sensors (cluster­ assumptions, so the Bayesian approach's critics often
heads) have to be reached to the sink, because loss of a point out that the method is not weII equipped to handle
particular packet will provide completely zero states of ignorance. Another model is the Theory of
knowledge about a region. Hence, such a distributed Evidence or Dempster-Shafer theory [9], which
monitoring requires almost 100% reliability. Moreover, provides a method for combining evidences from
we explicitly introduce the sensor fault probability into different sources without prior knowledge of their
the reliable data collection process using the Dempster­ distributions. In this latter method, it is possible to
Shafer theory of evidence based approach, which helps assign probability values to sets of possibilities rather
the cluster heads to make reliable reports based on the than to single events only, and it is not needed to divide
sensor measurements. all the probability values among the events, thus
The rest of this paper is organized as follows. Section II modeling more naturally certain classes of problems.
describes an overview of existing works. Section III We introduce Dempster-Shafer theory of evidence in
provides network model and assumptions. Section IV our health assessment technique in which no conditional
presents the proposed fault-tolerant data coIIection or prior probabilities need to be known. Knowing only
scheme in details. Section V presents performance the individual sensor's reliability, Dempster's rules for
evaluation through simulations. FinaIIy, Section VI combination gives a numerical procedure for fusing
concludes this paper. together multiple pieces of measurements from
unreliable sensors.
II. RELATED WORKS
III. NETWORK MODEL AND ASSUMPTIONS
Structural health monitoring is a vast, interdisciplinary
area of research whose literature spans several decades We consider a sensor network, where sensors are
[1]. Wireless sensor networks promise cheap and dense organized into single hop clusters, each managed by a
instrumentation for structural monitoring [2]. Recent cluster-head node. There are N cluster-head sensors
denoted as Ci where i 1,2,... , N, and a data collection
= fault using the concept of Dempster-Shafer theory of
center, called sink that form the infrastructure network. evidence in the measurement of structural health.
All the cluster-head sensors are connected to the sink in
A. Distributed Structural Health Assessment
a multi-hop fashion and the sink connects the network
to an external network (i.e., Internet). There are ni A. 1 Local Report Generation without Sensor Fault:
sensors denoted as Si,j' where j 1,2, ... , ni, associated
= Let us consider a distributed structural health
with C in one hop. For simplicity, we use Sj to denote assessment system where each cluster consists of n
sensing node Si,j throughout this paper. The network is a sensors as shown in Fig. 1. The measurements from n
tree like topology where non-leaf nodes are the cluster­ sensors are then combined at the cluster-head to make a
head nodes, and leaf nodes are the sensor nodes as final decision regarding the structural health condition.
depicted in Fig. 1. Each sensor node, Sj, (j 1,2,... , n) senses to measure
the structural health condition for Z different
=

We assume that a sensor node is in charge of


performing different types of application tasks (e.g., applications (e.g., vibration, pressure, strain,
sensing the strain, vibration, pressure temperature, etc. temperature, etc.). The sensed reading of Sj for the klh (k
in context of structural health assessment) and sending = 1,2,... , Z) application is denoted as R ; . Then, C can
its measurements to the cluster-head node. A cluster­
head node is responsible to collect sensory prepare the local report Rep 0 rt : for klh application
measurements from its leaf nodes to prepare a local using
report. Sensing nodes receive commands from and send
data to their cluster-head nodes, which pre-process the k k k k
Report, = (avg(Rj),max(Rj),min(Rj)), (1)
data as a report and forward them to other cluster-head
sensors in a multi-hop path (multi-hop whenever a
cluster-head cannot reach the sink in a single hop). Each where avg(R ; ), max(R ; ) andmin(R ; ) are the
report contains the overall health condition surrounding
average, maximum and minimum values, respectively,
the cluster-head node. Then, cluster-head nodes send
of the n sensor readings for the f(h application. Finally,
their local reports to the sink. Sensing nodes and
C can send this report to the sink.
cluster-head sensors are assumed to be stationary.
The approach presented above uses the raw data to be
sent to the sink by the fusion nodes (cluster-heads).
Indeed, fusion node prepares the report according to (1),
instead of preparing the final decision regarding the
health condition surrounded by the fusion node. Let us
consider one step further, in which individual cluster­
head node processes the raw data from the sensing
nodes and sends a local decision to the sink.
Suppose that each fusion node is given a predefined
reference threshold regarding the health condition of a
particular application. Let R :'1 denote the predefined
reference threshold of the klh application. The fusion
node Ci compares the individual readings, R ; , of the n

sensors against the R :'1 . The sensor reading reflects the


healthy condition if it follows the reference value, else it
reflects the unhealthy condition. Let a ; = 1 if sensor
S/s reading reflects healthy condition for the kth
application, and f3; = 1 otherwise. Based on that, C

computes the value H ,k ,which reflects whether the


space under observation of C maintains a certain level
of health condition for the kth application, according to
Fig. 1 A hierarchical sensor network model for structural health
monitoring (2).

IV. FAULT-TOLERANT DATA COLLECTION (2)


PROTOCOL

In this section, we first describe the distributed health


assessment with/without considering the sensor fault. Thus, C sends H ,k to the sink as a local decision. If H :
Then, we formally introduce the inclusion of sensor
is too low (i.e., majority of the sensing nodes' reading
reflect the worse condition around CD, the health Another differential is that there is no need to distribute
condition must be too bad around the cluster-head Ci. the whole probability among the events. The remaining
Hence, if H / is less than a reference health value, unavailable evidence is kept unassigned, not assigned to
the complement event as done in the classical
H :'1 ' monitoring center (sink) will consider this event probability theories. These two differentials allow this
to be alarming. Besides, Ci may also send Report,k to theory to model more precisely the natural reasoning
process on evidence accumulation, making it
the sink as described earlier.
progressively more popular.
A.2 Local report generation with sensor fault: In the As an example of how to apply Dempster-Shafer to
absence of noise related measurement error and sensor reasoning, suppose that a sensing node Sj is either
fault, the above strategy works well since the sensors reliable with probability 0.9 or unreliable with
within a small region in the deployment area are likely probability 0.1. Also suppose that node Sj provides the
to have similar measurements, and the structural health measurement (to its corresponding cluster-head) which
H ,'
condition provided by C is calculated from the states that the structural condition is healthy. If node Sj
itself is reliable, then its measurement is accurate, but if
majority consensus. However, though this distributed Sj is not reliable, then its measurement is not necessarily
health assessment system can combine (at the fusion inaccurate. Node S/s measurement provides evidence
sensor) data from multiple nodes to estimate the health for 0.9 degrees of belief that structural condition is
status, the observing sensor nodes might not be reliable healthy, but a zero degree of belief that structural
due to the sensor measurement error, noise and some condition is not healthy (not 0.1). The zero does not
other reasons. Hence, this might be infeasible to necessarily imply in impossibility, it means that node
guarantee that the health status provided by H ,' is S/s measurement (evidence) provides no support to the
essentially genuine. So, when the fusion sensor belief that structural condition is unhealthy. An interval
combines the offered measurements from sensor nodes bounded by 0.9 and 0 might constitute a type of belief
about a particular health status (e.g., vibration), reading function. In the presence of uncertainty, Dempster­
of every sensor mat not be fully reliable. Consequently, Shafer is concerned with bounds for probabilities of
a sensor may contribute contrary evidence provability rather than computing probabilities of truth.
(measurement) to the [mal judgment at the fusion sensor This formalism provides methods for combining the
(For example, when health status is judged to be healthy evidences provided by different sources (i.e., sensors),
at the fusion sensor, a sensor node might have offered being the Dempster's Rule the de-facto method.
its measurement reflecting the status to be unhealthy). A Frame of Discernment or Environment in Dempster­
In majority decision-rule for combining measurements, Shafer is a set of primitive hypothesis. Let X denote the
the judgment about health status follows the majority. frame of discernment. It must (i) be exhaustive, in the
However, a correct majority decision requires that a sense of being complete, containing all possible
majority of observing nodes provide accurate primitive (atomic) solution, and (ii) have mutually
measurements. Moreover, all the sensors are given the exclusive primitive elements. Any hypothesis A will
same importance in (2) regardless of their reliability. refer to a subset of X for which observers can present
Using Bayesian approach to form a judgment is evidence. The set of all possible subsets of X, including
problematic without additional information or itself and the null set, is called a power set. There are
assumptions (for example, the difficulty of knowing the three types of values that can be assigned to hypotheses:
prior and conditional probabilities). (i) the basic probability assignment, or mass function,
Dempster-Shafer theory of evidence based approach (ii) belief function, and (iii) plausibility function, as
overcomes these limitations. It produces a judgment described in the following.
value between 0 and 1 that reflects the degree of belief The basic probability assignment, or mass function,
in that judgment. It discounts the unreliable observer's assigns some quantity of belief (evidence) to the
measurements. In our distributed structural health elements of the power set. The mass function, m,
assessment approach, the Dempster-Shafer theory of assigns a number in the interval [0, 1] to each subset of
evidence is applied that considers measurement error of X, where 0 means no belief, and I means certainty. The
sensor fault to accurately assess the structural health sum of all assignments is equal to I, meaning that the
status. We now first briefly explain the basics of the right hypothesis is in the frame of discernment. The
Dempster-Shafer theory of evidence, and then present probability not assigned to any element, named "non­
our approach that applies this concept. assigned belief' is assigned to X The mass function
follows the following two conditions:
B. Basics of Dempster-Shafer Theory of
Evidence
1
m(¢)= ° I

Unlike the Bayesian Theory, the Theory of Evidence I
m(A):?:O, VA (3)
X
[8] [9] does not need prior knowledge of the probability I.
distribution, and it is able to assign probability values to I m(A)=1 I
sets of possibilities rather than to single events only. Ac;;;;X J
Thus, meA) is the measure of the belief assigned by a C. Application of Dempster-Shafer Theory of
given evidence to A, where A is any element of the Evidence in Fault-tolerant Data Collection in
power set. As meA) deals with the belief assigned to A SHM
only, and not to subsets, none belief is forced by the In the context of SHM, there are two possibilities
lack of knowledge. concerning the health condition of a civil structure:
The belief function, Bel, measures how much the healthy or unhealthy. Therefore, according to the
information given by a source support the belief for a Dempster-Shafer theory, the frame of discernment
specified element as the right answer. Thus, belief consists of two possibilities regarding the value
function maps each hypothesis A to a value as calculated by the fusion node: X = {H ,H} , where H

1
means structural condition is healthy and H means
I unhealthy. So, the power set of X has four elements:
Be/(¢)= 0,

L m(B) · (4)
X
Be/(A)=
J
- -

. B
'e null set, hI = {H}, hz = {H} , and h3 = = {H ,H} .
Suppose that the basic belief (reliability) of sensor Sj is
The Bel(A) represents the weight of evidence supporting PI. If node Sj provides the measurement that is healthy,
A's provability, On the other hand, the upper probability then its mass value assignment (according Dempster­
function, or plausibility function, PI, measures how Shafer) will be the following: m;Chl) Pj' m;Chz) 0, and
= =

much the information given by source does not m;Ch3) I - Pj' If node Sk provides the measurement that
=

contradict a specified element as the right answer. In is unhealthy, its basic probability assignment will be the
other words, the plausibility is the weight of following: mihl) =0, mk(hz) P k, and mk(h3) I - Pk. In
= =

evidence that do not refute B. The plausibility function this way, a cluster-head can have the mass values based
to the element A is: on the decisions of its n subordinate sensors. Then,
according to Dempster's rule of combination in (6), the
1
I
PI(¢)=O,
combined mass values can be calculated using
PI(A)= . L m(B) · (5)
�nH*¢ J 1
I
m(¢)=O,
"
There are several ways of taking the final decision. For L TI m (S )
, , i, (8)
.
instance, decision can be made by choosing the meA)= v,,,,,,��,�
hypothesis with the maximum mass, belief or J
plausibility.
where K represents the degree of conflict given by
B. 1 Rules of combination of evidences from
different observers: Evidences obtained from two
independent sources and expressed by the two mass (9)
functions ml and mz on the same power set can be
properly combined to obtain a joint mass function ml,z,
The Dempster's rule of combination is expressed by Thus the combined mass values, from the evidences
orthogonal sum and normalization as given by provided by the n sensors, can be calculated for hI and
hz. The fusion sensor finally prepares the report based
m(¢)=O,
on the higher value between m(hl) and m (hz).

L m,(B)m,(C) (6) C. 1 Computing the Basic Belief p: A significant


�n
B �c_��A ________ difficulty lays in the problem of determining the
m(A)=
1- K
reliability a node. Due to lack of proved probability
distribution model or due to unavailability of accurate
where K represents the degree of conflict given by mathematical analysis, in many syatems using
Dempster-Shafer theory, the probability numbers are in
K 2: m,(B)m,(C). fact simply assigned by expert opinion [10]. But the
=
(7)
BnC=¢
problem with this is that over time, the sensor device's
capability of sensing may deteriorate or change. So,
According to Dempster's rule of combination, the dynamism is not considered in this approach.
evidences ml(B) and mz(C) the two sources are Another approach might be to use historical data to get
combined by considering the product ml(B)mzCc), the ratio of correctness. However, if entire history is
which focuses on the intersection Bn c. Obviously, as considered, this impact will be very less. For example,
there could be more than one pair of sets that have the if a sensor node provides 1000 correct readings, after
same intersection, a sum of products is present in (6). that suddenly it begins to provide wrong readings. So,
Moreover, some intersections may be empty. Therefore, the ratio or percentage of the correct readings will still
in order to satisty the conditions in (3) for the basic be high, and it will take long time to consider its
probability assignments, the products ml(B)mz(C) where reliability to be low. Therefore, realiability of a sensor
Bn c ¢ must not be included in ml,Z, and the mass
= node might not be judged properly.
values should be normalized using I-K.
We propose to use a reliability calculation mechanism much less when the proposed method is used. It shows
by each cluster-head node based on the recent history the stability of the proposed scheme.
(i.e., immediate past) of each sensor nodes. Let pk-l
J

p
VI. CONCLUSIONS
and �"" be the immediate past and current,
In this paper, we have presented a protocol for structural
respectively, reliabilities of sensor Sj at the kth decision health monitoring using wireless sensor networks. Here
interval. The updated reliability p � of Sj' to be used every cluster-head uses the Dempster-Shafer theory of
during the (k+ l)th decision interval, is calculated using evidence to take the final decision while aggregating
different measurements for the subordinate nodes in that
cluster. Results through simulations show that the
(10) proposed method achieves higher reliabilities than the
existing event detection mechanisms.
where t5 is the tuning parameter that controls the rate of
update, and p �u" is calculated by REFERENCES

ClIrr
[1) F.-K. Chang, Structural Health Monitoring: The
L
PJ (11)
:::: _

W
1 ' Demands and Challenges. CRC Press, 2001.
[2] J. P. Lynch, A. Sundararajan, K. H. Law, A. S.
Kiremidjian, and E. Carryer, "Power-efficient data
where W is the window size representing the number of
past readings to consider and f is the number of management for a wireless structural monitoring
incorrect readings in this window. system, " Proc. of the 4th International Workshop
on Structural Health Monitoring, 2003, pp. 15-17.
V. PERFORMANCE EVALUATION [3] N. Xu et al., "A wireless sensor network for
structural monitoring, " Proc. of SENSYS. ACM
In this section, we evaluate the performance of the Press, 2004, pp. 13-24.
proposed fault-tolerant data collection protocol through
[4] K. Mechitov, W. Kim, G. Agha, and T. Nagayama,
simulations using C++. We examine the effectiveness of
"High-frequency distributed sensing for structure
the proposed protocol in terms of the percentage of
monitoring, " Proc. of First IntI. Workshop on
detection errors in the reports generated by the cluster­
Networked Sensing Systems (INSS), 2004.
heads.
To compare the detection performance of the proposed [5] J. P. Lynch, A. Sundararajan, K. H. Law, A. S.
Dempster-Shafer theory-based detection method, we Kiremidjian, E. Carryer, H. Sohn, and C. R. Farrar,
implement two well-known methods such as Bayesian "Field validation of a wireless structural monitoring
detection and Neyman-Pearson detection method [11]. system on the alamosa canyon bridge, " Proc. of
We also applied the method based on k-out-of-n rule SPIE's 10th Annual International Symposium on
presented in [8], which attempts to minimize the Smart Structures and Materials, San Diego, CA,
detection error in the Bayesian and Neyman-Pearson USA, March 2-6, 2003.
detections. [6] B. Krishnamachari, and S. Iyengar, "Distributed

TABLE!
bayesian algorithms for fault-tolerant event region
AVERAGE DETECTION ERRORS (%) MADE BY CLUSTER-HEADS WHEN detection in wireless sensor networks, " IEEE
APPLIED DIFFERENT METHODS Transactions on Computers, vol. 53, 2004, pp.
Ba esian Neyman-Pearson Proposed 241-250.
n Original k-out-of-n Original k-out-of-n method
[7] D. Koller, and N. Friedman, Probabilistic
3 20.49 1 2.43 20. 1 0 1 3. 1 1 9.02
5
Graphical Models - Principles and Techniques.
21.10 8.30 20.33 8.86 7.76
7 2 1 .45 6.0 1 20.35 6.43 5.23 MIT Press, 2009.
9 2 1 .60 4.46 20. 1 5 5.65 3.8 1 [8] X. Luo, S. Member, M. Dong, and Y. Huang, "On
n number of sensors III each cluster. distributed faulttolerant detection in wireless sensor
networks, " IEEE Transactions on Neural Networks,
=

We test the proposed detection scheme in several vol. 55, 2006, pp. 58-70.
simulated sensor networks. We use 100 clusters in our [9] S. Salicone, Measurement Uncertainty: An
simulations. We vary number of the subordinate sensor Approach via the Mathematical Theory of
nodes under the supervision of each cluster-head. Evidence. Springer, 2006.
Therefore, the total number of sensors is different in
[10]1. R. Boston, "A signal detection system based on
different simulations. Every result presented in this
dempster-shafer theory and comparison to fuzzy
detection, " IEEE Transactions on Systems, Man,
section is the calculated average of 1000 runs of a
simulation. The summary of results of our simulations
and Cybernetics, vol. 30, pp. 45-51.
is presented in Table 1. The probability of sensor faults
is kept 10% for all the simulations. [11] P.K. Varshney, Distributed Detection and Data
Table 1 clearly demonstrates that the error rates in the Fusion. New York: John Wiley & Sons, 1991.
structural health report done by the cluster-heads are

You might also like