Computer and Information Science; Vol. 7, No. 3; 2014 ISSN 1913-8989 E-ISSN 1913-8997 Published by Canadian Center of Science and Education
67
Autonomous Monitoring of River Level with Real Time Event Prediction
Zamshed I. Chowdhury1, Md. Istiaque Rahaman2 & Shahriar I. Chowdhury3
1 Institute of Information Technology, Jahangirnagar University, Savar, Dhaka, Bangladesh
2 Department of Electrical & Electronic Engineering, Ahsanullah University of Science & Technology, Dhaka, Bangladesh
3 Desme Bangladesh, Dhaka, Bangladesh
Correspondence: Zamshed I. Chowdhury, Institute of Information Technology, Jahangirnagar University, Savar, Dhaka-1342, Bangladesh. Tel: 880-277-910-4551. E-mail: [email protected]
Received: April 9, 2014 Accepted: April 22, 2014 Online Published: July 29, 2014 doi:10.5539/cis.v7n3p67 URL: http://dx.doi.org/10.5539/cis.v7n3p67
Abstract
Observation of water level at various river sites could provide valuable insight about probable disaster in advance to initiate disaster management protocol as early as possible. We have developed an autonomous remote river water level monitoring network with event prediction algorithm at the server while maintaining a substantially low manufacturing cost. The WSN is comprised of several chosen sites based on their statistics with intelligent sensors for water level measurement. The sensors are autonomous in nature to account for any disturbance in node environment and also within the network. The real time data are transmitted to a remote server through GPRS for further processing. Server extracts information and simulates various real time parameters such as water level rise rate, time remaining to exceed the critical level for a particular site etc. A prediction algorithm running on the server side predicts the measured level values for each node over a period of time. A prototype system is implemented with six nodes at different points and has yielded satisfactory results.
Keywords: data acquisition module, flood, GSM network, PIC18f4550, prediction analysis, real time, river level, WSN
1. Introduction
In coastal countries where rivers are flowing everywhere through lands, flood is considered to be one of the most common natural disasters. About 100,000 persons are killed and over 1.4 billion people are affected during the last decade of the 20th century as described by S. N. Jonkman (Jonkman, 2005). Continuous monitoring of river water level at alarming points could provide with valuable information that can be utilized to warn authorities and inhabitants well ahead in time. A review on conventional river water level measuring showed that river level monitoring by constructing piezometer on the location under investigation is somewhat costly and requires human presence for data collection (Winter, LaBaugh, & Rosenberry, 1988). Therefore, information cannot be gathered ahead in time. The modern sensor network for this purpose has pressure sensors installed at different points to acquire data with high precision. But the use of this kind of networks is limited by the cost involved with installation and maintenance. For developing and under developed countries where such costly method is not a viable option, the threat of mass destruction by sudden flood is still at large. The only solution is to find a low cost system that could provide a near precision measurement and timely warning so that if water level is about to exceed the critical level, concerning authorities might be able to be prepared to minimize the figure of flood caused deaths and immense harms. This paper illustrates such an autonomous wireless sensor network (WSN) for twenty four hour river water level surveillance and alarming tasks.
This article is a revised and expanded version of a paper entitled Design and Deployment of a Robust Remote River Level Sensor Network, presented at IEEE Sensors Applications Symposium, TX, USA, February 22-24, 2011(Chowdhury et al., 2011). The sensor network architecture is illustrated using a block diagram shown in figure 1. At the heart of this network, a server collects data periodically sent from installed nodes (Node 1, Node 2… Nodes N) placed at different sites along the rivers. The nodes are responsible for raw data collection, storage and transmission to the server after initial processing. Existing GSM (Global System for Mobiles) network is
used to con widely use for distanc method (G as well as that are in operate wi Radio Serv a database natural crit updated as predicts va suddenly, t priority sta
This articl are presen presented.
concluding 2. System 2.1 Basis o The conce horizontal and K. T.
vertical di calculating compute a converts e signal hav signal of t cannot pas an equival US signal reception o after the tr should be, twice the l The resulti periodicall 2.2 The M The data a
nnect the node ed technique to ce measureme Gueuning et al.
centralized n nstalled near b
ithout direct h vice) to a remo e software wh
teria and estim s soon as new alues of critica then sensors w ate or to detect
e is organized ted in section
The result a g section discu Architecture of Distance Me ept behind the
distance meas Park (1998).
istance. The m g the node ent and store yield lectrical signa ving a frequenc
the same frequ ss through and lent electronic s and uses a of echo. The s ransmission ha , l = (340 × t)/
length to get to ing data is sto ly to a remote Mounting Struct acquisition sys
es to a server.
o build wireles nt. One such w , 1996). Senso etwork using bank of the ri human supervi ote server. The hich receives d mates critical p w raw data are al parameters f would sample t t any sudden c
Figu as follows: th 2. In the next and analysis s usses the future easurement an
vertical distan surement. This The data acqu measurement m tities (in this c ded water level al to US signal cy equal to the
uency and tra d is received by
signal of the s timer and cap sound wave tr as occurred and
/2 meters. The o the receiver f red into the EE server and for ture
stem and the G
Connecting wi ss sensor netwo
work is the m or networks are
WLAN (Posto iver where dep sion. The data e server which data, extract in
arameters. The e received from for each node.
the water level hange.
ure 1. Network he whole system
section, the re section discus e improvemen nd Vertical Dis nce measureme
s is a well-kno uisition system
module conta case, a microc l data. The mo and vice vers e characteristic ansmits it. The y another cryst
source frequen pture module avels through d till the recep e factor of 2 c from the transm EPROM by th r data recovery GSM network
ireless sensors orks. Some no measurement o e designed as d olache et al., 2 pth is modera a are uploaded is designated nformation an ese newly obta m the sensing If the levels ap l information w
architecture b m architecture mote server de sses the resul nts.
tance ent is very sim own concept w m used here ad ains an ultraso
controller), rea odule block dia a. The transmi c frequency of e US signal re tal termed as re ncy. The proce
to get the tim air at a veloci ption, the horiz comes from the
mitter. Each m he processor fo y, in case of a d k interfacing ci
s together throu otable works ha f distance usin distributed net 2005). The no ate. These emb d by the nodes
as Central Mo nd further proc
ained paramete nodes. Based pproach critica with a shorter
block diagram from both har esign process a lts from the i
milar to the one which is first b
dopts the simi onic transmitte al-time clock a agram is show itter is a piezo f the crystal an eflects back fr eceiver. This c essor governs t me elapsed be ity of 340 ms- zontal distance e fact that the measured level or later process delivery failure ircuit is integr
ugh the use of ave been done ng time of flig twork (Cardell odes are data a bedded system s through GPR onitoring Unit
cesses the info ers are refresh d on the prior al values at a c
acquisition pe
rdware and sof and the associa implemented
e using ultraso rought into att ilar technique er-receiver pa and non-volati wn in figure 2.
o-electric crysta d converts it to rom any obsta crystal convert
the transmissio etween the tra
-1. If the timer e of the obstacl
transmitted si is marked usin sing which inc e.
rated into a sin
f GSM network e on sensor net
ght and phase l-Oliver et al., acquisition sys ms are designe RS (General Pa (CMU) is basi formation base ed periodically information, C constant rate o eriod to be at a
ftware perspec ated algorithm prototype and
onic (US) signa tention by M.
but for measu ir, a processo ile data memo The US transd al which recei o an equivalen acle that the s s the signal ba on and receptio ansmission and r count is t sec
le from the mo ignal must trav ng real time st cludes sending
ngle module a k is a
work shift 2005) stems
ed to acket ically ed on y and CMU r rise high
ctives ms are d the
al for Toda uring r for ry to ducer ves a nt US ignal ack to on of d the conds
odule verse tamp.
g data
nd is
www.ccsen mounted o transmitter near the bo experience result, the module is the water e inside the which cou of the tube data acqui through th and outsid work (Cho
Once the v which are measures t thickness o height of t value of h that stays sites. The with differ and includ
net.org/cis on top of a tube
r and receiver ottom end and e the air pressu
water level re placed at diffe enters the tube tube. The thi uld reflect back
e almost witho isition system.
he window nea de the tube bec owdhury et al.,
vertical height in fact the ph the distance fr of the box is 0 the slab above h is calculated above the wat height should rent types of sa ded in the firmw
e like structure r both facing d d the other nea
ure building in emains at the s erent chosen s e and floats a h
ckness is chos k the incident U out any friction
. When the bo ar the river bo come equal. T , 2011).
Figure 2. D t is being mea hysical parame om the module 0.05 m and the
e the water is at node site a ter surface vari be more than alt. This partic ware of the eac
Computer an e made of PVC downward (tow ar the top end.
nside the tube same level as sites. When the hollow box m sen to be enou US signal. A r n as the tube h ottom end of t ottom. The wat The whole arra
Data Acquisiti asured, the wat
ters associated e to the floatin height of the
d meter, the e and transferred ies from site to what it is at a cular dimensio ch of the node
Figure 3. M
nd Information S
69 C. The module
wards the grou The purpose o as water ente outside of the e tube is place ade of plastic ugh to provid radius of 10 cm has a diameter
the tube is sun ter continues t angement is sh
ion System (DA ter column hei d with the tube ng box on the w
tube from the effective water d to the CMU o site due to th site which is n on is determine e processors to
Mounting struc Science e is oriented ve
und surface).
of two separat ers and compre e tube. Each m ed at a node fro
having a thick e an opaque m m makes the b
of 12 cm. Thi nk for the first to rise in the t hown in figur
AS) block diag ight is calcula e. It is evident water surface w
module to the r column heig for further pro he variation in nearer to sea w ed manually at calculate effec
cture
ertically from t The tube has e windows is t esses the air co mounting struct
om where data kness of 5 cm medium for th ox move up an is box acts as t t time, water f tube until the e 3 as suggest
gram
ated by exclud t that the data which is denot e bottom surfac ght is h = (L –
ocessing. The n density of th water as sea wa
t each site prio ctive height of
Vol. 7, No. 3;
the surface wit two openings that water doe olumn inside.
ture and assoc a is to be colle which was alr he ultrasonic s
nd down the le the obstacle fo flows into the water levels in ted in the prev
ding the bias v acquisition sy ted by x meter ce is L meter. I – x – d) meter.
portion of the e water at diff ater is more co or to the install f water column
2014 th the : one es not As a ciated ected, ready ignal ength or the tube nside vious
alues ystem . The If the . The e box ferent ntent ation n.
2.3 Node H The hardw finally ass mentioned 1) Data A DAM inco i) US ii) US iii) D US signal circuit elem TX. Two h current wi frequency and proces which was with a low signal is th reception s The data EEPROM USBv2.0 troublesho EEPROM node mod employed due to cha long enoug sense the t 2) Networ network, R modem an module en AT (Atten (Subscribe sent as pa identifies a nodes.
Hardware Arch ware is designe
sembled into d below:
Acquisition Mo orporates sever S signal genera S signal recept Data processing generator and ments. NE555N hex inverters a
ith capacitor, of the signal a ssing block in s received by R w noise amplif hen fed to a co should have am
processing an for the storag compliance w ooting or readi
, external mod dules. As the v
which provide ange in temper
gh which sets temperature de rk interface: T RS232 interfac nd RS-232 lev nsures conserv ntion) comman
er Identificatio ackets using G a particular no
hitecture ed following a
a single mod odule (DAM):
ral hardware b ator and transm
tion and proce g and storage b
d transmitter N provides the are connected
about twice and the whole ncludes a US r
RX is amplified fier and a gain omparator whi
mplitude less nd storage blo ge. PIC 18f455 which makes ing the data s de of storage is velocity of so es the convers rature. The con
a limit on the epending on w To connect any
ce is used. Th el converter. T vation of powe
nds from the on Module) fro GPRS. Before ode in the netw
Figure 4. A
modular appro dule. Each of
The distance locks inside a mitter block
ssing block an block
block consists e oscillation ne in parallel to i the voltage fr generation pr receiver (RX), d by 60 dB wit n of 20 dB in ich compares t than the refere ock incorporat
50 is chosen d it possible t tored in modu s chosen due to ound is depen
ion value that nversion error e measurable d hich the calcul y module to th his section inc This interface er while in off- MCU to conn om a network sending, data work. This is a
Architectural b
oach. Each no the sub-modu measuring se DAM as illust nd
s of a US tran eeded for the t increase the st rom the inver rocess is contro
an amplifier a th a two stage n the next stag the amplified e
ence value and tes a microco due to its 13 1 to use in HID ule. Even thou o large amount ndent on temp corrects the er becomes critic distance by the
lation of water he network an cludes Microco is controlled b -line mode. Th nect to the G
service provid a are encapsula two digit deci
lock diagram o
ode module com ules is design
ction and the trated in figure
nsmitter (TX), transmitter. Th trength of the rter output is olled by a mic and a threshol amplifier. The ge is prvided b echoed signal d should not b
ntroller for pr 10-bit A/D cha D (Human In ugh this partic
t of data to be perature, a tem
rror generated cal when the d e module. A te
r column heigh nd to make it ontroller Unit by the MCU. T he modem is c
SM/GPRS net der with GPR ated in pre-de imal ID that se
of system hard
mprised of fou ned for specif
associated cir e 4:
, a multi-vibra he inverter is u
transmitted pu applied to th crocontroller. U
ld detector. Th e first stage giv by general pur to a reference be considered
rocessing task annels and the nteraction De cular chip has
accumulated d mperature com d by the change
distance to be emperature sen ht is optimized able to send (MCU), a GP The control ov connected to th twork. The m S enabled. Th efined format eparates amon
dware
ur sub-module fic functionalit rcuitry form D
ator and addit used for the dri ulse. By cuttin he transmitter.
US signal rece he ultrasonic s ves a gain of 4 rpose op-amp.
voltage. Any to be a valid e ks and an ext e special featu evice) mode w
256 bytes bu during operatio mpensator circu e of sound vel
measured beco nsor is employ d.
data over wir PRS enabled G ver communic he MCU to rec modem uses a he measured da with a header ng different sen
s and ty as DAM.
tional ve of ng dc The ption ignal 40 dB The false echo.
ternal ure of when uilt-in on by uit is ocity omes ed to reless GSM ation ceive SIM ata is r that nding
www.ccsenet.org/cis Computer and Information Science Vol. 7, No. 3; 2014
71
3) Real Time Clock (RTC) generation: To stamp the events in real time, there is a real time generation section.
This is actually a RTC integrated circuit (DS1307) employed to operate in I2C slave mode over a two-wire serial bus with MCU acting as the master. The design is done according to the application note provided by Maxim Integrated Products (2006). Timing information is stored in 56 bytes of NV (Non-Volatile) SRAM from where it is fetched by master when necessary.
Along with these sections, a few voltage divider networks are used to feed voltage signals to MCU. This section actually monitors different parameters from other sections, e.g. node battery voltage. The voltage is converted by MCU and sent to the server along with other data.
2.4 Explanation of Acquisition Process
When the system is switched on, MCU initializes the modules necessary for system operation. After the initializations are being done, GLOBAL interrupt is enabled to allow the external interrupts. The contents of count register of TIMER1 and capture register are cleared. MCU starts sending 40 KHz pulse whose transmission length can be varied from the firmware. After a transmission pulse is sent, there is a chance that the transmitted signal could directly fall on the receiver. This could trigger a wrong detection. To avoid this scenario, the signal from the echo detector is disregarded until a certain period of time is elapsed which ensures the avoidance of reception of the transmitted pulse directly instead of the echoed signal. MCU waits 1 ms and then enables the detector circuit. Just prior to this, the capture module and timer modules are turned on. TIMER1 starts to count the number of pulses between the turn on and another pulse to indicate the end. When an echo is received, the capture module turns off the timer module and the count value is stored. From the timer count, the time taken by the echoed signal to traverse the length of the tube above water is calculated. If the time taken by the sound wave to return to the module is t seconds, the one way distance traversed by the transmitted wave before being captured by RX will be: l = (v × t )/2, where v = velocity of sound at a specific temperature. If the clock of TIMER1 is configured as to count 1µs/count, the value of the TIMER1 count register will indicate the number of instances elapsed since US wave is being transmitted. It takes k microseconds to execute one instruction for a PIC MCU with an F MHz clock where, k=1/F. This count value is used as the value of t. Hence, the TIMER1 module must be configured very carefully depending on the system clock. Before transmitting the signal, MCU samples the temperature sensor data to get the instantaneous temperature value and sets the value of v to be used in that calculation. The dependency of velocity of sound on temperature is governed by this formula:
c= 331 ms-1+0.6 ms-1 C-1 ×T °C (1) Where T is temperature in degree Celsius. A lookup table (LUT) containing values of velocity for every 1 °C increase in temperature between the range 5 °C-50 °C is used by firmware after acquiring temperature revision data for appropriate value of v. The actual water column height is determined afterwards which is stored and sent to remote server. The whole system process is illustrated in Figure 5.
2.5 Node firmware
The software works in two successive steps: one is for acquiring column height and another for storing and sending the data over internet. MCU is loaded with software to carry on the signal generation, reception, data acquisition, data storage and data transmission tasks. The coding steps can be visualized with the help of a flow chart which is shown in the Figure 6. After the system is switched on, built-in timer module TIMER0 is initialized with a calculated pre-scalar so that the timer counts up to 65535 which indicates a timeout of 65 ms at 4 MHz clock speed used for controlling purpose of transmission period. TIMER1 is initialized to capture the echoed signal from the slab surface. Capture module CCP1 is turned off to prevent the module from capturing noise. A temperature revision data is input to A/D channel which is used to get the current velocity of sound from a LUT. As the river surface should have some disturbances in terms of current and waves, there must be situations when the successive readings will yield data which are different from each other significantly. To mitigate this effect, multiple reading are taken during a single measurement and then averaged to get a value that lies in between the maximum and minimum values with minimum aberration from the true value. A variable r is declared to have a null value that indicates the number of readings taken. Enabling the interrupts, MCU declares another variable k which is also assigned a null value at the beginning of a process. The limiting value of k defines the number of readings per measurement cycle to be made. Signal transmission is initiated by MCU and continues for a period of 0.5 ms. Reception unit is turned on 1 ms later to eliminate the wrong reception of the transmitted pulse instead of echoed one. Capture mode is initiated by enabling CCP1 module. Interruption can be caused by either TIMER0 overflow or echo reception. If TIMER0 is overflowed, MCU restart the TIMER and reinitiate pulse sending sequence. A distance of (65535×v)/2 should be covered by the transmitted pulse train
before the interrupt f actual heig transmitter to the time After the c from the i averaging and data in Prior to go node pow Transmissi handles th measurem is set. MC timer expi whole set Sequence of next me less than 1 value grea operates in change be value, the in consequ the value o
timer is overf from capturing ght of water c r-receiver pair e being waited calculation, k i initiation of p all the reading n successive m oing in to the tr wer source in t ion MACRO he protocol for ment sequence i CU then waits f ires, MCU dis of data is sent variable r is th easurement cy 12. This limitin ater than 12 ma
n two basic m etween two suc firmware swit uence ensures of r.
flown, so a bia g the echo ma olumn. The te and the slab s d before enablin
is incremented pulse transmiss gs. The curren memory locatio ransmission M terms of perc
prepares the doing so. The id byte. Once t for the ACK (A sables the com t again and the hen increment ycle. Before co
ng value ensur akes MCU to g modes: low pr
ccessive meas tches to the hig
a more freque
as of (65535×v akes MCU to emperature dat surface using th
ng pulse recep d by unity and sion repeats. A nt time is fetch ons in EEPROM MACRO, anoth
centage of pow modem for se e data set cons
the send comm Acknowledgm mmunication m e communicati ted by unity an ommencing an
res the number go to the top of iority and hig urements. If th gh priority lev ent sequence of
Figure 5
v)/2 is added to get the captur ta and capture he formula dis ption to accoun if it holds a v After 2000 re hed from RTC
M.
her A/D conve wer left. Then
ending the co sists of 2 extra mand is issued ment) from the module and ste ion module is nd MCU steps nother measure r of measurem f the execution gh priority mo
he rate of rise el and re-initia f measuremen
. Process time
o the final valu re value and i
value yields t sused in section nt for the true t value less than adings, MCU and data is tim rsion takes pla n MCU goes ollected inform
bytes which i , each data byt server to arriv eps out of the disabled regar s into a wait st ement, r is che ments within a n list and to cle
des. The mod e (RR) is posit alizes the value nts. The final w
line
ue of calculate nitiate the cal the vertical di n 2.1. The cap time gap betw 2000, the who
calculates the me stamped b ace to get the c into the trans mation to the include one no te is sent and a ve. If the ACK MACRO. If t rdless of ACK tate which last ecked to have period of 24 h ear the variabl des are depend
tive and greate e of r to be gre wait state is adj
ed distance. A lculation to ge stance betwee pture value is a een the two ev ole process sta e final distanc by writing that current status o smission MAC
remote server ode id byte and
a timer of one K arrives befor timer runs out K is received or ts till the begin a value equal hours is to be 1 le. Sensor hard dent on the ra er than a thres eater than 12.
justed accordi valid et the n the added vents.
arting ce by time of the CRO.
r and d one hour re the t, the r not.
nning to or 12. A dware ate of
shold This, ng to
www.ccsen
3. Central CMU con written usi follows th channels a Server wa informatio sequence.
is maintain number. T successful
net.org/cis
l Monitoring U nsists of a hard
ing Visual Stu he stop-n-wait and increases t aits for a pac on and deliver Rn is the sequ ned. This helps This kind of sce extraction of
Unit (CMU) a dware section udio. A graphi t ARQ (Autom the reliability o cket to arrive.
rs to the datab uence number a s to detect any enario generat information c
Computer an
Figure 6. Nod and Server Al
along with a ical description matic Repeat of the commun
When a pac base. The dat assigned for d y missing data tes an error me
auses Rn to inc
nd Information S
73
de software flo lgorithms an algorithm ru
n of CMU is Request) pro nication as sug cket arrives an tabase enters t
ata arrived at t when the rece essage to indic crease by unity
Science
owchart
unning with a given in Figur tocol. This pr ggested by Fo and the data i the data using the site. For ea ived data do n cate any distur y. Prior to sen
a database. Th re 7. The rece rotocol is wel orouzzan (B. A is not corrupt g the node id
ach node, a se not contain the rbance during nding an ackno
Vol. 7, No. 3;
he database GU eiver site algor
ll suited for n A. Forouzan, 2
ted, it extracts and measure eparate Rn sequ
expected sequ transmission.
owledgement t 2014
UI is rithm noisy 009).
s the ment uence uence Each o the
correspond is greater t per day to 3.1 Event P There are the one tha is chosen 2007).
1) Linear r the mathem residual be and the y approach f relationshi 2) Second by straight appropriat level states 3) Markov model to p generating technique Markov m labeled sta any pair o
ding node, it c than the thresh
be changed to Prediction Alg some useful m at would produ
after careful e
regression: If t matical expres etween the mo calculated wi for prediction ip. The observ
order regressi t line. For this te polynomial t s, a set of data v model: Hidd
predict faults g a sequence o proposed by F model is an ext
ates S = {Si} ( of states, and a
checks to see if hold defined fo o a value greate gorithm models for deve
uce a result th evaluation of t
F
there is a straig ssion for the st odel and the o ith the linear m
of future even ations may or ion: As the ob s kind of cases to the data usin a is obtained an
den Markov m and failures i of most likely F. Salfner, P.
tension of a fin 1≤ i ≤N), a a vector π = [π
f the sequence or a particular er than 12. The eloping a succ at matches clo the following
Figure 7. CMU
Figure 8. C ght line contain traight line is:
observations. T model is mini nts but with a may not have bservations cou s, a curve wou ng polynomial nd compared w model (HMM) n systems. In events where Troሷger, and S nite discrete ti a square stocha πi] specifying
e count exceed node, CMU e e algorithm is cessful event pr
osely with the three differen
U general block
CMU process c ning a set of p : y = a0+a1x+e The sum of the imized to get constraint; wh a linear behav uld exhibit ma uld be best suit
l regression. U with the actual is adopted as our system, w all the observa . Tschirpke (S ime Markov ch astic matrix A
an initial prob
ds twelve. The expects the nod illustrated in F rediction algor actual data. Th nt techniques (
k diagram
cycle
paired observat e; where a0 an e squares of re
the best fitted hich is that the vior; hence ther arked patterns t ted to fit data.
Using a quadrat data.
s the third sol we used it to ations are trea Salfner, Troሷger
hain (DTMC).
= [aij] defining bability for eac
rate of rise is de to change th Figure 8.
rithm. The cha he model of pr (Chapara & C
tions- (x1, y1), nd a1 are coeffi
esiduals betwe d data set. Th e data set mus re is always a this would be . An approach tic equation to lution. HMM model a pred ated as a failur r, & Tschirpke . DTMCs are g transition pro ch state. Hidd
monitored and he maximum c
allenge is to ch rediction algor Canale, 2006; R
(x2, y2)…… (x ficients and e i een the measu his is a fairly st maintain a l margin of erro poorly represe h would be to f o describe the w
is a well-prac dictive behavio re, according t e, 2009). A hi defined by a s obabilities betw en Markov mo
d if it count
hoose rithm Ross,
xn, yn);
is the red y good inear or.
ented fit an water cticed or for o the idden set of ween odels
www.ccsen extend DT be predicte Each of th each day f against the diagram, it
Due to le developing 4. Results 4.1 Prototy Prior to th verified at measurabl previous w surface. Th firmware t The densit system. Th Buriganga 25-35 met time wind choosing t due to hea For these The install can interfe at the node due to pow data is lost 4.2 Measu The princi the perform actual valu
net.org/cis TMCs in that a
ed by identifyi hese technique for 30 days. Th e average of ac
t is evident tha
Figure 9. B ast amount of g the predictio
& Discussion ype Developm he installation t the laborator e distance of work (Chowdh his limit impo to use the time ty factor in d he whole syste a, Dholessori a ters and have r ow as this is t this window is avy rain. Near t reasons, place lation sites we ere with the U
e allows for th wer failure, it c
t.
urement Accura ipal concern w mance of the s ue. To determ
an output is pro ing special pat es is used to im
he adequacies cquired data fo at HMM has yi
Bar diagram sh f aberration fr n algorithm.
ns
ent and Install of the module ry test environ data acquisiti hury et al., 20 osed a drawbac er overflow me determining the
em is assemble and Turag, wit recent records the time of the s that during th
the bank of the es near the ban ere chosen not
S signal and th he timely repla could be usefu acy
with any measr system will be mine the accur
Computer an oduced each ti tterns of errors
mplement thre of models we or 5 days. The ielded predicti
howing compa from the actua
lation es at node site nment for hori
on systems is 011), node mod ck on the tube echanism to inc
e water level ed and six nod th two nodes s of causing flo
e year when th his time the de ese rivers, dep nks are qualifi t to be too crow
hereby degrad acement of the ul by reading a
rement system solely depene racy yielded b
nd Information S
75
ime the DTMC the system is ee different mo ere tested by p
comparative p ion those expe
arison among p al data which
es, the distance zontal and ver between 0.3 dules cannot b e length used a
crease the mea brought more des are placed along each riv ood. The mon he rivers are no
epth of rivers pth is fairly les ed as the insta wded and nois de the performa e power source all the data for
m is the accurac edent on the de by the prototy
Science C enters a stat
experiencing.
odels based on lotting the pre predictions are rienced a mini
predictions usin are acquired
e measuring c rtical paths wi
m to 3.0 m in be mounted m at various node asurable distan e precision in at six differen ver. All these nths of March- ot too deep or are subjected s and GSM ne allation sites w sy such as harb ance. The feat e in nodes. Ho r that particula
cy of the meas egree of closen ype system, in
e. The occurre n the average edictions calcu e shown in Figu
imum deviatio
ng different m manually, HM
capability of pr ith a minimum n the prototyp more than 3.0 m
e points. The m nce solved this
measurement nt points of thr rivers have an -August are ch r too shallow. A
to continuous etwork is acces which provided
bors as there m ture of monitor owever, if any ar day from EE
surement beca ness of the me nitially we set
Vol. 7, No. 3;
ence of failure of observation ulated by the m ure 9. From th on from true va
models
MM is chosen
rototype modu m of errors. A pe designed du
m above the w modification i
particular pro t than the prev
ree different ri n average dep hosen to be the
Another reaso and rapid cha ssible by the n d sufficient de might be noise ring power sou
node is shut d EPROM so th
ause of the fact easured value t
up a single n 2014
s can ns of model he bar alues.
n for
ule is As the uring water n the blem.
vious ivers:
th of e test n for anges odes.
epths.
e that urces down at no
t that to the node,
collected w were made days time error over 4.3 Measu The screen
Figure 10.
water level da e concurrently and compared these 100 sam urement Results
nshot of entry w
. Illustration of
ata manually a y so as to ensur d to determine mples and the a
s
window of the
f accuracy of m th
and compared re the clarity o the degree of average error c e database softw
measurement b he average val
with the data of evaluation. A
closeness. Fig alculated.
ware in Figure
by percentage lue of error exp
a sampled by A total of 100 gure 10 illustar
e 11 (a) shows
error over 100 perienced
our system. B samples were rtes the distribu
all the nodes w
0 samples. The
Both measurem e collected dur
ution of percen
with their nam
red line repre ments
ing 3 ntage
mes,
sents
www.ccsen
Figu
last measu in order of than ninety delayed da the details pipe used backgroun than thirty twelve me Figure 11 and node 2 The real ti instead of session an
net.org/cis
ure 11. (a) GU
ured levels, ser f reading sequ y five percent ata were sent b of a particula and the measu nd which turns y percent. This easurement rec (c). Figure 12 2 from Burigan ime monitorin f the time whe nd the use of a
(b) I window show
rver time of rec uence. The nod t of data are re by the nodes a ar node with in ured levels of t red upon rece s window also ceived. The g
(a-d) illustrate nga river.
g produced be en the data are arithmetic mea
Computer an
wing database (c) Measure ception of data de power statu eceived within at busy hour w nformation asso
that node along eiving a status o contains the raphical repre es the details f etter result than e received by t an reduced the
nd Information S
77 (a)
entries with n ement graph w a from the nod us represents th n 20 seconds fr which explains
ociated with th g with the nod
of power leve option for sho esentation of t
from two more n previous sys the server. Als effects of dist
Science
node parameter window
des, node time he percentage from the begin the delay. The hat node e.g. r de times etc. Ea
el available in owing a graph the measureme e nodes, name stem as it give so, the multipl turbances on t
(c) rs (b) Node det
(real time) for of power left nning of measu e window of F river name, no ach reading is
that particular hical interpreta ents is shown ly node 1 from es the estimate
le measuremen the water surfa
Vol. 7, No. 3;
tails window
r data collectio on the node. M urement cycle.
Figure 11 (b) sh de id, length o
marked with g r node which is ation based on in the windo m Dholesshori ed time in real nts during a s ace. The perio
2014
on etc.
More . The hows of the green s less n last ow of river
time ingle odical
ups and do hour wind measurem 95.83%. H window is that period sampled d critical lev Figure 13 offers the against the deviation t we used th samples ar an arithme node durin illustartes precision g acquired d day oscilla
Figure
4.4 Probab During the
owns in the wa dow, maximum ment data were However, this i s considered, th d. The lost me data. The critica
vel is calculate shows the ev user with the a e manually me to determine th he average of a re deviated. As etic average of
ng a 15 days that the perf graphically. T data. Also the r
ates between 0
12. Database e
ble Drawbacks e deployment
ater levels wer m 12 and minim
during peak h is the worst ca he success rat easurement dat al time (time t d based on the ent prediction average levels easured values he statistical ac
actual measure s each day is s f these standar period. The formance of th
he average sta red line repres 0.2 and 0.7 whi
(a)
(c) entries (a and c
s of Prototype process of th
e due to the pe mum 11 data hour (i.e. 11 am ase scenario as te rises up to 9 ta were later re to exceed the c e average of th bar diagram s for the next f s to determine
ccuracy of the ement from th susceptible to d rd deviations i
changes in ev he devised ne andard deviati senting the actu
ich indicates th
c) from two mo entr System and So his prototype s
eriodic high tid were received m – 2 pm) wh s this represen 98.89% with o etrieved from critical level) i he all the sampl
for an individ five days. The the measurem e measurement hat particular n different stand s used to deter veryday standa etwork in term ion is below 0 ual standard d hat the precisio
ore nodes with ries (b and d) olutions system, the ins
de and low tide d from the nod hich explains t nts the minimu only 2 addition the node and s updated once le data receive dual node. The measured valu ment acuracy. W
t samples foun node as the ave dard deviation
rmine the stan ard deviation ms of measur 0.6 which sug eviation calcu on of measurem
h respective GU
stallation sites
e which is natu des. We analyz the data recept um success rat nal measureme used in statist e per day. The ed on any day.
e algorithm run ues from the n We have used nd from the nod
erage value fro values of sam ndard deviation
is depicted in rement accurac gests very goo lated from me ment of data is
(b)
(d) UI windows sh
s were chosen
ural. In twenty zed that the lo tion success ra te. When a 15 ent data loss du tical analysis o e time to excee nning at the s nodes are comp d modified stan des. In our ana om which the mple from any n
n for that parti n Figure 14 w
cy and as we od accuracy o easurements at
s satisfactory.
howing databa
n to be at the y four oss of ate of days uring of the d the erver pared ndard alysis,
node node, cular which ell as of the each
ase
close
www.ccsen vicinity of challengin end packet to the netw network w network ot from each
Also, the dependent change dra in less acc be useful i 4. Conclus The result helps to es For furthe wireless c signal hop such as po
net.org/cis f the banks o ng task featurin
t delivery rate work is of utm where each no
ther than GSM node to CMU
F dependence o t on atmospher
astically over a curate measure
in determinatio sion ts show that th stimate a time er research, pl
ommunication ping network a orts, marine ha
of the rivers s ng wide geogr requirements most importanc de will be con M. The land ba U.
igure 14. Stan of velocity of ric pressure an a short period
ments. Sensin on of true valu he sensor netw frame for taki laces where G n protocols sta architecture. F rbor etc. digita
Computer an so that GSM raphic network and hostile sy ce here. One s
nnected to a l ased node wou
Figure 13. P
ndard Deviation sound on only nd humidity. I
of time and h g air pressure ue of height of work can prov ing necessary GSM network
andardized by For DAMs to
al polarity corr
nd Information S
79 network can k coverage, hig ystem deploym
olution would land based con uld serve the pu
Prediction grap
n of measurem y temperature In rough cond hence affect the
and relative hu water column.
vide predictab measures to m
is not existen IEEE such a measure dista relator suggest
Science
be utilized fo ghly dynamic ment environme d be to arrange
nnecting node urpose of trans
ph window
ments in 15 day is considered ditions of weat e velocity of s umidity along .
bility about th mitigate damag
nt can be cho as ZIGBEE (IE
ances more ac ted by a group
or sensor netw radio transmi ents. The meth e the nodes in e by another c sferring the ac
ys window d. Actually the
ther, these two sound. This, in
with ambient
e possible flo ges of human osen. For thos
EEE 802.15.4) ccurately in a p of scientists (
Vol. 7, No. 3;
working. This ssions, high en hod of connect
a distributed m connecting wir cquired inform
e parameter is o parameters c n effect could r temperature w
od occurrence life and prope se cases, adva ) could be use noisy environ (Nakahira, Kod
2014 is a nd to tivity mesh reless mation
also could result would
e and erties.
anced ed in ment dama,
Morita, & Okuma, 2001) could be proved to be a solution. USB modem can be a great replacement in this system for utilizing 3G (Third Generation) connectivity for faster and more reliable networking. Firmware for USB connected modem could also reduce the power and spatial concerns for RS232 supported modems. The data analyzed by the server can find additional uses in other applications and researches also. The nature of water level of any particular river in each season can be taken into account to plan for the agricultural aspects of the nearby localities. Architects can decide whether the place under consideration is reachable to flood and design structures accordingly. Geographically important information such as time of high tide, low tide and any behavior other than normal could be extracted from the analyzed data. Using multi sensor-multi node technique in single installation site, change in depth of river bed can be understood. Advanced statistical analysis of the information could provide pattern of river level rise and fall over a long period of time which could be useful in understanding the effects on sea level by phenomenon like global warming. With sufficient data, a database could be built to observe the current scenario of the water level rise over a period of time which could give us a fair idea about the effect of global warming over a particular geographical area.
References
Cardell-Oliver, R., Kranz, M., Smettem, K., & Mayer, K. (2005). A reactive soil moisture sensor network:
Design and field evaluation. International Journal of Distributed Sensor Networks, 1(2), 149-162.
http://dx.doi.org/10.1080/15501320590966422
Chapara, S. C., & Canale, R. P. (2006). Numerical Methods for Engineers (5th ed.). New York, USA, McGraw-Hill.
Chowdhury, Z. I., Imtiaz, M. H., Azam, M. M., Sumi, M. R. A., Rahman, M. R., Alam, F., … Hassan, N. (2011).
Design and deployment of a robust remote river level sensor network (pp. 244-249), Proceeding of IEEE Sensors Applications Symposium. Texas, USA. http://dx.doi.org/10.1109/SAS.2011.5739776
Forouzan, B. A. (2009). Data Communication and Networking (4th ed.) New Delhi, India, McGraw-Hill.
Gueuning, F. E., Varlan, M., Eugene, C., & Dupuis, P. (1997). Accurate distance measurement by an autonomous ultrasonic system combining time-of-flight and phase-shift methods. IEEE Transactions on Instrumentation and Measurement, 46(6), 1236-1240. http://dx.doi.org/10.1109/IMTC.1996.507414
Jonkman, S. N. (2005). Global perspectives of loss of human life caused by floods. Natural Hazards, 34, 151-175. http://dx.doi.org/10.1007/s11069-004-8891-3
Maxim Integrated Products. (2006). Using a DS1307 with a PIC Microcontroller. Retrieved from http://www.maxim-ic.com/app-notes/index.mvp/id/3921
Nakahira, K., Kodama, T., Morita, S., & Okuma, S. (2001). Distance measurement by an ultrasonic system based on a digital polarity correlator. IEEE Transactions on Instrumentation and Measurement, 50(6), 1748-1752.
http://dx.doi.org/10.1109/19.982975
Park, K. T., & Toda, M. (1996). Short distance ultrasonic distance meter. U.S. Patent No. 5, 483, 501. Retrieved from http://www.google.com/patents/US5483501
Postolache, O. A., Girao, P. M. B. S., Pereira, J. D., & Ramos, H. M. G. (2005). Self-organizing maps application in a remote water quality monitoring system. IEEE transactions on instrumentation and measurement, 54(1), 322-329. http://dx.doi.org/10.1109/TIM.2004.834583
Ross, S. M. (2007). Introduction to Probability Models (10th ed.), San Diego, USA, Academic Press.
Salfner, F., Troger, P., & Tschirpke, S. (2009). Cross-core event monitoring for processor failure prediction. In High Performance Computing & Simulation, 2009. HPCS'09. International Conference on (pp. 67-73).
IEEE. http://dx.doi.org/10.1109/HPCSIM.2009.5191988
Winter, T. C., LaBaugh, J. W., & Rosenberry, D. O. (1988). The design and use of a hydraulic potentiomanometer for direct measurement of differences in hydraulic head between groundwater and surface water. Limnology and Oceanography, 33(5), 1209-1214.
Copyrights
Copyright for this article is retained by the author(s), with first publication rights granted to the journal.
This is an open-access article distributed under the terms and conditions of the Creative Commons Attribution license (http://creativecommons.org/licenses/by/3.0/).
