Research Article System-Aware Smart Network Management for

Research ArticleSystem-Aware Smart Network Management forNano-Enriched Water Quality Monitoring

B Mokhtar12 M Azab234 N Shehata256 and M Rizk12

1Department of Electrical Engineering Faculty of Engineering Alexandria University Alexandria Egypt2Center of Smart Nanotechnology and Photonics (CSNP) SmartCI Research Center Alexandria University Alexandria Egypt3Informatics Research Institute City of Scientific Research and Technological Applications Alexandria Egypt4ACIS ECE Department University of Florida Gainesville FL USA5Department of Engineering Mathematics and Physics Faculty of Engineering Alexandria University Alexandria Egypt6USTAR Bioinnovations Center Utah State University Logan UT USA

Correspondence should be addressed to B Mokhtar bassemmokhtargmailcom

Received 18 April 2016 Revised 19 July 2016 Accepted 20 July 2016

Academic Editor Constantin Apetrei

Copyright copy 2016 B Mokhtar et al This is an open access article distributed under the Creative Commons Attribution Licensewhich permits unrestricted use distribution and reproduction in any medium provided the original work is properly cited

This paper presents a comprehensive water quality monitoring system that employs a smart network management nano-enrichedsensing framework and intelligent and efficient data analysis and forwarding protocols for smart and system-aware decisionmaking The presented system comprises two main subsystems a data sensing and forwarding subsystem (DSFS) and OperationManagement Subsystem (OMS) The OMS operates based on real-time learned patterns and rules of system operations projectedfrom the DSFS to manage the entire network of sensors The main tasks of OMS are to enable real-time data visualizationmanaged system control and secure system operation The DSFS employs a Hybrid Intelligence (HI) scheme which is proposedthrough integrating an association rule learning algorithm with fuzzy logic and weighted decision trees The DSFS operation isbased on profiling and registering raw data readings generated from a set of optical nanosensors as profiles of attribute-valuepairs As a case study we evaluate our implemented test bed via simulation scenarios in a water quality monitoring frameworkThe monitoring processes are simulated based on measuring the percentage of dissolved oxygen and potential hydrogen (PH)in fresh water Simulation results show the efficiency of the proposed HI-based methodology at learning different water qualityclasses

1 Introduction

The detection of water quality parameters such as dissolvedoxygen (DO) and potential hydrogen (PH) in aqueousmedia is important for wide variety of applications includingenvironmental monitoring biomedical research and pro-cess control [1ndash3] Compared to currently used techniquesfluorescence-based sensing technique has significant advan-tages over other procedures due to fouling avoidance thefact that there is no need for a reference electrode and theresistance to exterior electromagnetic field interferences [4ndash9] One of the most promising optical nanostructures is ceriananoparticles due to its oxygen capability storage low-costsynthesis and adequate sensitivity [10 11] This paper offersa comprehensive monitoring framework as an integration

between nanotechnology and trustworthy wireless sensornetworks

The main objective of our work is to develop a completesensing platform for real-time monitoring of two waterquality parameters DO concentration and PH value inaqueous media [12 13] Our system as shown in Figure 1goes behind local single location monitoring to a networkedsensing of DO and PH concentrations at multiple locationsacross streams water treatment facilities hydroponic farmsand aquafarms The system comprises two main subsys-tems which are the data sensing and forwarding subsystem(DSFS) and Operation Management Subsystem (OMS)The DSFS employs our proposed hybrid intelligence (HI)technique which is embedded at powerful on-site compu-tation nodes for efficient data analysis classification and

Hindawi Publishing CorporationJournal of SensorsVolume 2016 Article ID 3023018 13 pageshttpdxdoiorg10115520163023018

2 Journal of Sensors

Zone 1

Zone 2

Zone N

Managementplatform

Administrationand decision

support thoughreal-time visual image

of the network

Powerful centralized nodefor intermediate processing

Sink

Sensor node

Control messagesPartially analyzed feedback

2

N

Management

Administraand decisi

support thoreal-time visua

of the netw

Figure 1 Comprehensive system architecture of data management and decision making

forwarding processes Powerful nodes receive the digitalizedsignal from one or more running nanosensors merge thecollected data with the geographical location of the sensingelement(s) and run the HI techniques for learning datapatterns and taking on-site decisions Based on the weightof reached decisions data might be forwarded to sinks andremote off-site management servers in a real-time fashionfor further analysis with security extensions The real-timemanagement of the overall system with security policiesrsquoenforcement is performed by the OMS The OMS operatesbased on learned patterns and rules of system operationsprojected from the DSFS to manage the entire network ofsensors The main tasks of OMS are to enable real-time datavisualization managed system control and secure systemoperation

As shown in Section 3 the main advantage of havingsuch level of autonomic control and management is theability to predict some of the sensor feedback based onthe analysis of the overall network feedback [14ndash16] Suchability facilitates optimizing the sensor power usage bycontrolling the activation periods of sensors leading to muchbetter sensor utilization expanding the lifetime of the entirenetwork and reducing the system cost Further such abilityfacilitates detecting and fixingexcluding any misbehavingor problematic sensing nodes that can massively reduce thesystem accuracy Additionally the automated system coulddefinitely overcome the problems of manual data collectionthat requires extensive effort and timeThis capability enablesalmost continuous real-time monitoring of DO with meansto identify and address sensor drift helping researchers toconstruct test models based on the flow of the monitoredwater supply Both DO and PH are sensed using fluorescence

quenching technique depending on the reduction of visibleemitted fluorescent optical intensity based on the changeof the quencher concentration of dissolved oxygen or thechemical impact of PH in ceria nanoparticles itself

Additionally we present in this paper a HI-based schemefor data classification and forwarding in wireless sensornetworks (WSNs) enabling energy-efficient better utilizedresources and optimizedQoS operationsThemain objectivesof our proposed scheme by adopting concepts in the infor-mation theory and Artificial Intelligence (AI) [17ndash19] are asfollows

(i) Building efficient reconfigurable information-drivendata classification and forwarding methodology forWSNs

(ii) Learning interesting routing-based attributes andgenerating forwarding models

(iii) Enabling the capability of learningexpectingdetect-ing abnormal data flows and making classificationand generating rules

The organizing of the remaining sections of this paperis as follows Section 2 Materials and Methods discussesthree subsections including themethodology of data sensingforwarding subsystem with the proposed HI-based dataclassificationforwarding scheme the experimental setup forsensing of DO and PHoptical nanosensors and the trustwor-thy Operation Management Subsystem The obtained resultsare analyzed and discussed in Section 3 Section 4 concludesthe paper and highlights our future work

Journal of Sensors 3

2 Materials and Methods

21 Data Sensing and Forwarding Subsystem (DSFS) Thenet-working infrastructure of sensorsrsquo communication and dataforwarding processes within the DSFS depends on havinga network of wirelessly communicated sensors organized inclusters The continuously growing interest and developedresearch inWireless SensorNetworks (WSNs) lead tomakingsuch networks widely used in applications related to vari-ous fields such as environmental monitoring [20] WSNsprovide the communication framework for such applicationsenabling data sensing collection and forwarding until reach-ing the final data destination center for further data analysisand decision making As known WSNs comprise limitedenergy and computation communicating nodes thus thisleads to great challenges on the lifetime of such networks andconsequently their reliability So there is a need to proposean efficient data forwarding scheme that considers the energyconstraints inWSN and the importance of interesting data inorder to route data that have large information gain In theliterature many trials provided solutions for having efficientdata forwarding and routing targeting energy saving at sensornodes using opportunistic routing theory [21 22] Otherresearchers consider data routing and forwarding dependingon data aggregation and clustering approaches [23] Anotherwork provides an effective data collection using a set ofpowerful mobile nodes which increases the probability ofhaving available and reliable communication channels [24]The inter distance and noise level between sensor nodes issometimes considered for designing data forwarding strate-gies [25] However to the best of our knowledge prior relatedwork did not provide a data classification and forwardingscheme forWSNs leveraging the capabilities of simple hybridintelligence techniques and information theory concepts toextract data semantics and learn interesting informationAccordingly there will be an ability to control classificationlevels and amounts of data forwarded from areas of interestto remote data analysis and decision making locations Thiswill consequently aid in optimizing power consumptionand memory utilization at sensor nodes Leveraging con-cepts form information theory this paper offers a novelhybrid intelligence-based scheme for data classification andforwarding inwireless sensing communication environmentsenabling energy-efficient better utilized resources and opti-mized QoS operations

As a case study we run this scheme for a WSN directedto monitor water quality levels The proposed methodologydepends on profiling raw data readings generated froma set of optical nanosensors as profiles of attribute-valuepairsThen data patterns are learnt adopting association rulelearning algorithm clarifying the most frequent attributesand their related values According to the discovered sets ofattributes a set of fuzzy membership functions are directedto produce a discrete sample space and a specific mem-bership class for each attribute based on its value Basedon calculated attribute-dependent entropies and informationgains weighted probabilistic decision trees are built to helptake decisions of data forwarding and to generate long-termfuzzy rules One of the main objectives of our proposed

scheme by adopting concepts in the information theoryand Artificial Intelligence (AI) is building efficient reconfig-urable information-driven data classification and forwardingmethodology for WSNs [17ndash19]

211 Hybrid Intelligence-Based WSN-Based Data Classifica-tion and Forwarding Scheme Figure 2 shows the buildingblocks of the proposed scheme for data classification andforwarding for WSNs Those blocks illustrate the main usedalgorithms and operations We will discuss the comprisedblocks shown in Figure 2 as follows

(i) The shared memory an accessible memory located atsensor nodes with powerful storage and processingresources Data readings are stored as profiles ofattribute-value pairs Each data profile refers to aspecific reading from a certain sensor node located ata location where that sensor is directed to execute adefined function such as reading lead-related qualitylevel at fresh water

(ii) Association rule learning algorithm amachine learn-ing algorithm which is adopted to learn data pat-tern at the shared memory and to extract the mostfrequent attributes located in the memory accordingto defined minimum support and confidence thresh-oldsThose attributes will be forwarded to the fusiliersin order to be classified based on attributesrsquo values

(iii) Fuzzifiers operating engines which adopt fuzzy logicvia defining a set of defined fuzzy membership func-tions to generate specific finite discrete classes for theattributes based on their values [26]

(iv) Statistical analysis this defines a set of statisticaloperations that are performed on data profiles storedin the shared memory in order to aid in extractingsome information leading to computing the infor-mation gains For instance a specific attribute willbe analyzed to know how many times it is foundin readings related to pollutants based on measuredlevels of dissolved oxygen and lead in fresh water atusing specific types of sensors

(v) Entropy and information gain calculation it runsmathematical operations based on information con-cepts to calculate the information entropy due to clas-sified attributes and the possible outcome classes afterdecision making Then it computes the informationgains for each attribute with respect to the calculatedclass entropy Accordingly the attribute which has thegreatest weight on making decisions will be knownThen other attributes with weights ordered in adescending order will be used to form the decisiontree The next section will discuss an illustrativeexample to show how calculations are performed

(vi) Decision trees and fuzzy rules generation basedon the above calculations weighted entropy-baseddecision trees will be formed and used to makedecision According to these designed decision treesand possible attribute classes a set of fuzzy rules can


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------

Association rule learning

algorithm

Sets of learned attributes

Fuzzifiers(sets of fuzzymembership

functions)

Statistical analysis(attribute-value

probability)

Entropy and information

gain calculation

Generation of weighted decision trees and sets of

fuzzy rules

Stored data patterns

Shared memory at advanced powerful

sensor nodes

Attribute 1

Attribute 2

Attribute 3

Set of attributes with finite sample space

Set of relevant attributes with infinite sample space

Figure 2 The proposed data classification and forwarding scheme

be generated which shows the different cases thatmight be faced when getting a set of attributes withincertain range of values

212 Overview of the Technical Insights of the Proposed Schemefor EfficientWater Quality Monitoring In this subsection wewill discuss the main concepts on which the proposed hybridintelligence scheme for data classification and forwardingis developed We assume a small-scale WSN comprising aset of communicating sensor nodes where some of thosenodes called advanced nodes are with powerful computingand storage capabilities This WSN is directed to collect datarelated to various classes of water pollutants and forwarddata to a final destination unit in order to make further dataanalysis and management Collected data and readings fromsensor nodes are stored in a shared on-demand accessiblememory where data entries are indexed according to sensorsrsquoidentifications and their locations Advanced nodes can playthe role of cluster heads in case of forming a cluster hierarchyfor the designed WSN Also advanced nodes analyze thepatterns of stored data where these data refer to readingsabout a certain pollutant The following subsection will dis-cuss building weighted classifiers using fuzzy logic entropyand decision trees which are adopted by sensor nodes [27]

The following points illustrate how the weighted clas-sifiers are built and work in advanced sensor nodes Asdiscussed previously readings from sensors are stored asattribute-value pairs in a shared memory

(i) The fuzzy logic divides the registered attributes withan infinite real space into finite sets with discrete finitespace

(ii) According to obtained attributesrsquo levels and expectedrelated classes a statistical analysis can be done whichwill show the percentageprobability of each attributeto a possible class

(iii) Based on data pattern and using the entropy of allpossible classes and the conditional entropy withrespect to their related attributes a fuzzy weighteddecision trees can be built Equation (1) calculates theclass entropy

119867(119862) = minus

119899

sum

119862=1

119901119862log2119901119862 (1)

where 119901119862shows the probability of a certain class found in the

analyzed patterns It can be calculated based on the frequencyof a certain class in data patterns For example if there are10 data profiles where 5 of them refer to a water quality classof high level type then the probability of high level qualityclass is 50 Then to know the weight of each attribute thatwill affect getting the main class we will use the conditionalentropy as described in

119867(119862

119860) = sum

119886isin119860

(119901119886)119867(119862

119860 = 119886)

= minussum

119886isin119860

(119901119886)

119899

sum

119862=1

119901(119862(119860=119886))

log2119901(119862(119860=119886))

(2)

where 119886 is a possible value for attribute 119860 and 119867(119862(119860 =119886)) is the conditional entropy of having a certain class whenattribute 119860 is of value 119886

Assume that we have three fuzzy attributes (ie 119860 =3) that can affect the process of classification We needto know the more affecting attribute that will be the firststage the classifier This can be obtained from calculatingthe conditional entropy for all attributes and their possiblevalues which can be defined from the fuzzifiers by which theattributes are with a finite sample space For instance thereis an attribute 119860 type such as sensor type which can take two


Water quality classifier

07

03

06

04

Sensor type

attribute

1

0

Reading level

attribute

Sensor status

attribute

08

02

07

03

1

0

075

025

Dangerous (low quality)

(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF

O2

Figure 3 Example of a weighted decision tree

different string values (ie 119886 = 2) which are119883 and 119884 This isa binary attribute From patterns we found that the value of20 of patterns which contain 119860 type is119883 Then (119901

119886=119883) = 02

and (119901119886=119884) = 08

Then we will use (1) and (2) to calculate the informationgain as stated in

Information Gain = 119868 = 119868 (119862 119860) = 119867 (119862) minus 119867(119862119860) (3)

Based on the calculated gain we can know the attributes withhigher priorities and those attributes will be at the first stagesor nodes in the decision tree-based classifier The more theinformation gain we will have for an attribute the higher theprobability that such attribute will be in the first decision treenodes

For example if 119867(119862) = 15 bits and we have threeattributes which are 119860 type 119860 level and 119860 status and eachattribute has two possible string values where119867(119862119860 type) =11 bits 119867(119862119860 level) = 07 bits and 119867(119862119860 status) = 05 bitsthis means that information gain according to 119860 type is thelargest one So the first node in the decision tree will be thesensor type attribute Figure 3 shows an example of a decisiontree

(i) Some fuzzy rules from the built decision trees can beextracted [28] Relying on those rules data forward-ing and routing decisions can be taken For instancea rule of a low level of quality can be extracted based

on the designed decision tree as shown in Figure 3As an example if we have a set of ON dissolved O

2

sensors with high level readings then this refers to alow quality level

As a case study with an illustrative example we discussa numerical example which discusses the designed for-warding scheme We assume that WSN is employed toforward quality-related data readings from a set of sensorsto a final destination in a fresh water context like a riverThe implemented WSN will comprise two main types ofsensors which are normal sensors and advanced sensorsThe proposed scheme will be implemented in advancedsensor nodes which possess powerful computing resourcesand communication capabilities Advanced nodes will beable to register readings from normal nodes in a sharedmemory at them allowing other nodes to access and get moreinformation (eg learning data patters concerning a certainpollutant during a specific period in the year) Readingswill be represented by advanced nodes in their memoriesas data profiles of attribute-value pairs The data profile willshow the identification (ID) of the data source sensor nodetype of sensor node status of the sensor (ON OFF andmaintenance) and level of readings (eg high or low)

Normal nodes will be equipped with two types of sensorsthat can provide readings about the percentage of dissolvedoxygen and PH as indication to the quality of the fresh waterin the studied river The practical measurements of dissolvedoxygen in fresh water show the following concentrations


Dangerous(low quality)

Moderate(moderate quality)

Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8

Figure 4 FMFs for dissolved oxygen concentration

where the last two ones can refer to a high degradation in thewater quality due to pollution [29]

(a) Safe (sim8 parts per million or ppm)(b) Moderate (5ndash7 ppm)(c) Stressful (3ndash5 ppm)(d) Dangerous (lt3 ppm)

Concerning the concentration level of PH in a fresh water weadopt the following three levels as indication to having acidiclevels [29]

(a) Low acidic (6ndash8)(b) Moderate acidic (4ndash6)(c) Highly acidic (lt4)

The designed scheme in advanced nodes will adopt fuzzylogic that employs fuzzy membership functions (FMFs) toclassify the readings levels of each sensor For readings con-cerning the dissolved oxygen and the discussed concentrationlevels previously FMFs are designed as shown in Figure 4 toconsider the following the first concentration as best qualitythe second concentration as moderate quality and the lasttwo concentrations as low quality For the concentration levelof absorbed lead other FMFs are used to define the firstconcentration level as safe level or optimum quality of waterand the second concentration level as moderate quality andthe last concentration level as the lowest quality of water

In this example we assume that we have one finaldestination four advanced nodes and five normal nodesAlso we have two types of sensors in each sensor node formeasuring levels of DO concentration and PH The dataprofile of a reading stored in amemory of an advanced sensornode will show the ID of the related sensor node type ofsensor reading level and status of the node

Considering that each region in a river is covered bya set of sensor nodes with specific IDs for monitoringwater quality so each area will have data patterns (basedon registered data profiles) that can be learned to knowand detect the main pollutants that might exist and theexpected consequences and recommend countermeasure fortreatment

Table 1 Data profiles of sensors

ID Sensor type Reading level Sensor status Decision1 DO High (lt3 ppm) ON Dangerous1 PH Low (5) ON Moderate2 DO No (6ndash8 ppm) ON Optimum2 PH Low (4) ON Moderate3 DO High (lt3 ppm) ON Dangerous3 PH NA OFF Maintenance4 DO High (lt3 ppm) ON Dangerous4 PH Low (6) ON Moderate5 DO High (4 ppm) ON Dangerous5 PH No (8) ON Optimum

In normal operation we get two readings from eachnormal sensor node through the day So we expect 10readings per day and 300 readings per month As shown inTable 1 assume thatwe have for a region of interest in the riverthe following daily data profiles collected from five operatingsensors through a certain month

The advanced sensor nodes will learn the patterns ofdata profiles stored in their memories Advanced nodes willlearn Apriori-based association rule learning algorithm [30]for learning the main frequent attributes in the stored dataprofiles which are sensor ID sensor type reading leveland sensor status Additionally advanced nodes will employstatistical analysis algorithm for generating some statisticsabout stored data profiles such as how many profiles containO2and high pollution level [31]From Table 1 advanced nodes can build a weighted

decision tree-based on entropy and information gain in orderto classify data flows and know the more important flowsHence they can allocate more resources and offer morebandwidth

For building the weighted decision tree we will have thefollowing procedures There are 10 data profiles with threemain attributes which are sensor type reading level andsensor status Also there are main four decisions in the tablewhich aremaintenance no pollution low pollution and highpollution From (1) we can calculate the decision entropy asfollows

119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863

= minus119901Dangerouslog2119901Dangerous

minus 119901Moderatelog2119901Moderate

minus 119901Optimumlog2119901Optimum minus 119901Maintlog2119901Maint

= minus04log204 minus 03log

203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)


Then we will use (2) and (3) to compute the information gainfor each interesting attribute of the main three we have

For the first attribute sensor type

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)

= 18466 minus 05 (minus08log208 minus 02log

202)

minus 05 (minus06log206 minus 2 times 02log

202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)

Similarly for the other two attributes

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816

10Entropy (High) minus |Low|

10Entropy (Low)

minus|No|10

Entropy (No) minus |NA|10

Entropy (NA)

= 18466 minus 04 (0) minus 03 (0) minus 02 (0) minus 01 (0)

= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10

Entropy (ON) minus |OFF|10

Entropy (OFF)


20333

minus 0222log20222) minus 01 (0) = 18466 minus 09 (052

+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)

From the above results we find that the higher informationgain is obtained for the reading level attribute (18466 bits)and then for the sensor status (1337 bits) and then for thesensor type (08 bits)This means that the advanced powerfulsensor nodes will forward data to the final destinationmanagement units once they detect high reading level fromoperating sensor nodes without giving more weight to thesensor type or the main cause for the low water quality levelsAccordingly a weighted decision tree can be formed andadopted by advanced nodes to take decision and forwarddata to the final destination A set of rules can be generatedaccording to the built decision tree such as the following if

Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2

Figure 5 Experimental setup of optical sensing for dissolvedoxygen

Laser module

Optical fibers

Optical NPs

Pump

Washing system

Commercial sensor node

Photodetector

Microcontroller

Figure 6 Suggested prototype design of the compact sensor

there is a high level reading from an operating DO or PHsensor then forward the data to the destination

22 Optical Materialsrsquo Synthesis and Sensing Setup Ceriananoparticles are prepared using a chemical precipitationtechnique for simplicity and relative low-cost precursors[32ndash34] 05 g of cerium (III) chloride (heptahydrate 999Aldrich Chemicals) is added to 40mL deionized water asa solvent The solution is stirred at rate of 500 rpm for2 hours at 50∘C with added 16mL of ammonia Thenthe solution is stirred for 22 hours at room temperatureThe synthesized ceria nanoparticles are characterized usingUV-Vis spectroscopy (dual beam PG 90+) to detect theabsorbance dispersion and consequent bandgap calculation

The experimental test bed used to detect the change ofthe fluorescence intensity peak due to the change of DOconcentration and PH value in the aqueous solution is shownin Figures 5 and 6 The fluorescence spectroscopy systemconsists of a near-UVLEDof central wavelength of 430 nm asan excitation source exposed to a three-neck flask containingDI water solution of the synthesized ceria nanoparticlesOxygen and nitrogen gases are fed through individual linesthrough double-holes cork placed into one of the necks on


Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm


Fuzzifiers(sets of Fuzzymembership

functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level

attributeSensor status

attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF


Figure 7 The sensor operation management framework

the flask and controlled by a mass flow rate controller Theprobe of a commercial DO meter (Thermo Scientific A500with a measurement range up to 50mgL) is inserted in thesecond neck of the flask to measure DO concentration Thefluorescence signal is collected from the colloidal solutionscanned by a Newport Cornerstone 1300 monochromatorpositioned at a 90∘ angle to the excitation signal forminimumscattering effects Then a photomultiplier tube (NewportPMT 77340) is connected to a power meter (Newport Powermeter 1915-R) for fluorescence intensity monitoring Samesetup is used for PH detection at normal DO concentrationby removing both oxygen and nitrogen inlets with varyingadded concentrations of acid (HCL) The value of PH ismeasured using same meter (Thermo Scientific A500) butwith the optimum probe for PH detection

23 Autonomous Sensor OperationManagement System Fig-ure 7 is a block diagram of the sensor framework where themain sensing element is interfaced with an autonomouslymanaged wireless sensor network for water quality monitor-ingThe sensing function is interconnected to the cyber layerfor control management and monitoring

The nanosensor is interfaced with an A2D chip apowerful microcontroller chip a GPS chip and wirelesstransmission module The microcontroller is programed to

control the activation and deactivation of the sensing elementand control the measurement configuration if necessaryThe microcontroller receives its guideline from a remotemanagement server Each sensor has a unique identifier thatis used in all transmissions along with the geographicallocation of the sensor in case of mobility The system is builtto be as generic as possible allowing more sensing elementsto be attached to the same sensing elements if the applicationrequires that

The system is built to scale where sensors are groupedinto different zones The sensor feedback and location aredispatched frequently based upon either event change andquery or a predetermined schedule to a central data collectionnode at each zoneThis node applies partial analysis and datagrouping and dispatches a comprehensive zone status-reportto the management server for further analysis and guidanceThe algorithms used to establish such analysis are applicationdependent We devised a simple case study with a simplemodel for the excremental study just to reflect the effect ofsuch automation on the quality of the system output

3 Results and Discussion

31 Evaluation of HI-Based Data Classification and Forward-ing Scheme In this subsection we conduct a simulation


Normal sensorAdvanced sensor

PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank

More information about PH level in this

area

More information

concentration from this area

about dissolved O2

Dissolved O

-2

Dissolved O2

Figure 8 Simulation scenario of the proposed HI-based scheme for data classification

scenario for evaluating the proposedHI-based schemeof dataclassification and forwarded data adopted by the powerfuladvanced nodes in the DSFS subsystem

311 Simulation Setup In this section we provide a pre-liminary simulation study for the proposed classificationand forwarding scheme Using the discussed example inSection 212 we conducted a simple scenario of a WSN-based pollution monitoring and data forwarding networkusing Java based discrete time event-driven WSN simulatorcalled J-Sim [35]We target a river as a network context wheresensors will be distributed along a river bank Figure 8 showsa layout of the scenario which depicts the network topologyand communication amongst sensor nodes We assume alinear WSN where sensors forward data to the next availablesensors We have two different types of sensors which arethe normal node and the advanced node The last typecan access registered data patterns and decide whether theforwarded data are of a certain importance or not Accordingto the calculated information entropy and gain besides theconstructed weighted binary decision trees the advancednodes will forward data to the final network destination Weassume that we have two types of pollutants that will resultin various levels of pollution indicators which depend onmeasuring the concentrations of dissolved oxygen (DO) andPH levels in regions of interest in the river In case of havinglow measurements of DO and low levels of PH this meansthat we have indication of a combined low water quality

level On the other hand if there are measurementsrsquo levelsfrom one sensor type refering to a certain low water qualityindication this means that we have an explicit low waterquality level Table 2 shows the simulation parameters Table 3shows a set of key metrics which are used to study the offeredQoS and network performance

Figure 9 shows that advanced nodes allocate moreresources and bandwidth for data flows concerning high levelreadings from dissolved O

2sensors The figure shows that

at early time low water quality levels are detected and thenhuge amount of data are allowed to flow from normal sensornodes to the destination Also some advanced nodes findhigh water quality levels based on measured dissolved O

2

concentrations Once the water quality-related data patternsreach a certain information gain advanced nodes allocatemore resources and forward the related data to the finaldestination So in case of optimum water quality levels lesstraffic is found since small information gain is calculated (iewe have in this case normal operations with low informationentropy)

Figure 10 shows that there are higher data throughputsmeasured at the final destination in case of detecting lowwater quality levels in case of having one source (onlydissolved O

2concentrations) or with two sources (dissolved

O2concentrations and PH levels) This is because more data

packets will be forwarded amongst sensor nodes till reachingthe destination

Figure 11 depicts the response of the WSN to a lowwater quality level that exists widely in the studied context


High water quality throughputLow water quality throughputOptimum throughput

50 100 150 200 250 3000Simulation time (seconds)

0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)

Dissolved O2-related water quality data throughput

Figure 9 Data throughput based on detected dissolved oxygen-related water quality

Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105

Figure 10 Average maximum throughput measured at runningvarious simulation scenarios

Water quality level class

High quality levelLow quality level

020406080

100120140160

Det

ectio

n tim

e (se

cond

s)

Figure 11 Detection time of dissolved oxygen-related water qualitylevel

Table 2 Simulation parameters

Simulation parameter ValueNumber of sensor nodes 11Number of normal sensor nodes 6Number of advanced nodes 4Number of final destinations 1Initial node power 1000 wattsPacket size 512 bytesPacket time to live (TTL) 255 seconds

Communication bandwidth Variable 200 1000 104 106bytessec

Types of sensors Two dissolved O2 PH

Dissolved O2concentrations Low high

PH levels Low highNumber of defined low waterquality classes Three low high no

Status of operating sensors ON OFFExpected number of datapatterns 8

Pollution threshold gt50 of stored patternsPatterns analysis and reasoningfrequency Every 15 seconds

Number of generated rules 3Simulation time 300 seconds

15 45 75 105 135 165 195 225 255 285Simulation time (seconds)

Optimum qualityCombined high quality levelCombined low quality levelExplicit O2-related low quality level

minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)

Figure 12 Average residual power per sensor node versus thesimulation time

Many advanced nodes are able to detect low dissolved O2

concentrations at the studied context and hence they allowdata flows to pass to the final destination

We studied the average residual power at the first deadsensor node As shown in Figure 12 the degradation in powerincreases as there are readings related to low water qualitylevels Hence more data packets will be forwarded to thefinal destination Also the figure shows that combined lowwater quality levels which comprise readings related to lowquality indicators based on the measured concentrations ofdissolved O

2and PH levels result in more degradation in


Table 3 QoS and network performance metrics

Metric Unit DescriptionReceived data throughput Bytesseconds The amount of data received by the final destination

Average maximum throughput Bytesseconds The average of maximum throughput captured at the final destination whenrunning the same scenario for more than one time

Water quality level detection time Seconds Delay introduced by the time needed to measure amounts of data readings fordetermining a specific water quality level

Average residual power per sensor Watt The average amount of measured residual power at sensors after running thesimulation scenario and forwarding data

Absorbance versus wavelength plot for undoped ceria

Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200

250versus E plot(120572E)2

(120572E)2

(b)

Figure 13 (a) Absorbance dispersion for the synthesized ceria nanoparticles and (b) direct allowed bandgap calculation

the residual power level compared with the case of explicitlow water quality level in case of having many measurementsof low O

2levels

32 Optical Nanoparticles Characterization Figure 13(a)shows the resulting absorbance dispersion of the synthesizednanoparticles Based on the resulting absorbance measure-ments the allowed direct bandgap semiconductor of thesynthesized nanoparticles can be found through the followingequation [36]

120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)

where 119860lowast is a constant for the given material depending onits refractive index and effective masses of both electronsand holes 119864 is the absorbed optical energy and 119864

119892is the

direct allowed bandgap energy Then (120572119864)2 is plotted withthe absorbed optical energy and the intersection with 119909-axisgives the value of bandgap energy as shown in Figure 13(b)

Figure 14 shows the change of the visible fluorescenceemission intensity at 520 nm from the ceria nanoparticleswith increasing DO concentration at normal PH sim7 under

Decrease DO in ceria in normal water

CBA

Ceria = 870 ppmDO = 607 ppmDO = 426 ppm

DO = 317 ppmDO = 209 ppm

0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)

Figure 14 Visible fluorescence spectra at different DO concentra-tion within neutral PH (sim7)


Decrease PH in ceria

Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283

Figure 15 Visible fluorescence spectra at different PH levels due toincreasing acid concentration within neutral DO (sim8 ppm)

near-UV excitation We speculate that this is due to a releaseof oxygen stored in the ceria lattice when the nanoparticlesare introduced into the solution Also the emitted fluo-rescence emission is found to be reduced with increasingthe added acidic concentration which consequently reducesthe value of PH as shown in Figure 15 at normal DOconcentration level (sim8 ppm)

4 Conclusions

This work presented a smart network management systemfor efficient data classification and forwarding and decisionmaking The system comprised two main subsystems a datasensing and forwarding subsystem (DSFS) and OperationManagement Subsystem (OMS) The DSFS adopted a novelhybrid intelligence (HI) scheme for data classification andforwarding in wireless sensor networks integrating infor-mation theory concepts machine learning fuzzy logic andweighted decision tress Such adoption led to better energyconsumption improved resource utilization and optimizedQoS operation Simulation scenarios discussed the per-formance of a WSN employing the proposed scheme formonitoring water quality indicators Simulation results ofthe proposed HI-based data classification and forwardingscheme showed that the DSFS works efficiently with lowtime overhead and success in achieving small levels of powerconsumption for operating sensors Additionally we imple-mented the presented approach constructing a small test bedemploying nano-enhanced sensing elements Analyzing suchelements showed a clear change in the visible fluorescenceintensity with the variation of DO ratio The reason is thedeveloped oxygen vacancies concentration formed insideceria nanoparticles which act as DO receptor The entire

system operation is managed by an automated manage-ment system Results showed the clear effect of the smarttrustworthy and automated data collection and analysis onthe quality and accuracy of sensing Having such systemis essential for easy data consolidation and informationextraction Different experiments were conducted to showthe effect of having such automation in enhancing the sensorand the network lifetime even in presence ofmalicious nodes

The proposed system is an adequate solution for acomprehensive automated management of DO sensors inthe aqueous media The automation platform within theOMS is built to be generic and can be easily modified tobe used with wide variety of other applications and sensingelements achieving wide scope of applications Our futurework includes testing the proposed scheme experimentallyfor other applications and in large-scale scenarios

Competing Interests

The authors declare that they have no competing interests

Acknowledgments

The presented work is part of the awarded Grant ldquoARP2013R132rdquo funded by Information Technology Industry Devel-opment Agency (ITIDA Egypt)

References

[1] P R Warburton R S Sawtelle A Watson and A Q WangldquoFailure prediction for a galvanic oxygen sensorrdquo Sensors andActuators B Chemical vol 72 no 3 pp 197ndash203 2001

[2] M A Acosta P Ymele-Leki Y V Kostov and J B LeachldquoFluorescent microparticles for sensing cell microenvironmentoxygen levels within 3D scaffoldsrdquo Biomaterials vol 30 no 17pp 3068ndash3074 2009

[3] A Mohyeldin T Garzon-Muvdi and A Quinones-HinojosaldquoOxygen in stem cell biology a critical component of the stemcell nicherdquo Cell Stem Cell vol 7 no 2 pp 150ndash161 2010

[4] C-S Chu and Y-L Lo ldquoOptical fiber dissolved oxygen sensorbased on Pt(II) complex and core-shell silica nanoparticlesincorporated with sol-gel matrixrdquo Sensors and Actuators BChemical vol 151 no 1 pp 83ndash89 2010

[5] W C Maskell ldquoInorganic solid state chemically sensitivedevices electrochemical oxygen gas sensorsrdquo Journal of PhysicsE Scientific Instruments vol 20 no 10 pp 1156ndash1168 1987

[6] R Sanghavi M Nandasiri S Kuchibhatla et al ldquoThicknessdependency of thin-film samaria-doped ceria for oxygen sens-ingrdquo IEEE Sensors Journal vol 11 no 1 pp 217ndash224 2011

[7] X-D Wang and O S Wolfbeis ldquoOptical methods for sensingand imaging oxygen materials spectroscopies and applica-tionsrdquo Chemical Society Reviews vol 43 no 10 pp 3666ndash37612014

[8] L Chen S Xu and J Li ldquoRecent advances in molecularimprinting technology current status challenges and high-lighted applicationsrdquo Chemical Society Reviews vol 40 no 5pp 2922ndash2942 2011

[9] G Mistlberger and I Klimant ldquoLuminescent magnetic parti-cles structures syntheses multimodal imaging and analytical


applicationsrdquo Bioanalytical Reviews vol 2 no 1 pp 61ndash1012010

[10] N Shehata K Meehan and D E Leber ldquoFluorescence quench-ing in ceria nanoparticles dissolved oxygen molecular probewith relatively temperature insensitive Stern-Volmer constantup to 50∘Crdquo Journal of Nanophotonics vol 6 no 1 Article ID063529 2012

[11] N Shehata K Meehan M Hudait N Jain and S GaballahldquoStudy of optical and structural characteristics of ceria nanopar-ticles doped with negative and positive association lanthanideelementsrdquo Journal of Nanomaterials vol 2014 Article ID401498 7 pages 2014

[12] R Ramamoorthy P K Dutta and S A Akbar ldquoOxygensensors materials methods designs and applicationsrdquo Journalof Materials Science vol 38 no 21 pp 4271ndash4282 2003

[13] A Oczkowski and S Nixon ldquoIncreasing nutrient concentra-tions and the rise and fall of a coastal fishery a review of datafrom theNile Delta Egyptrdquo Estuarine Coastal and Shelf Sciencevol 77 no 3 pp 309ndash319 2008

[14] M Azab andM Eltoweissy ldquoBio-inspired evolutionary sensorysystem for cyber-physical system defenserdquo in Proceedings ofthe 2012 12th IEEE International Conference on Technologies forHomeland Security (HST rsquo12) pp 79ndash86 IEEE Waltham MassUSA November 2012

[15] CHill andK Sippel ldquoModern deformationmonitoring amultisensor approachrdquo in Proceedings of the 22nd FIG InternationalCongress Washington DC USA April 2002

[16] E A Garich Wireless automated monitoring for potentiallandslide hazards [MS thesis] Texas AampM University 2007

[17] B Mokhtar and M Eltoweissy ldquoHybrid intelligence forsemantics-enhanced networking operationsrdquo in Proceedings ofthe 27th International Florida Artificial Intelligence ResearchSociety Conference (FLAIRS rsquo14) pp 449ndash454 Pensacola BeachFla USA May 2014

[18] B Mokhtar and M Eltwoeissy ldquoHybrid intelligence forsmarter networking operationsrdquo in Handbook of Research onAdvanced Hybrid Intelligent Techniques and Applications SBhattacharyya P Banerjee D Majumdar and P Dutta EdsIGI Global 2015

[19] B Mokhtar and M Eltoweissy ldquoTowards a data semanticsmanagement system for internet trafficrdquo in Proceedings of the6th International Conference on New Technologies Mobility andSecurity (NTMS rsquo14) pp 1ndash5 IEEE Dubai UAE April 2014

[20] J Yick B Mukherjee and D Ghosal ldquoWireless sensor networksurveyrdquoComputerNetworks vol 52 no 12 pp 2292ndash2330 2008

[21] J Luo J HuDWu andR Li ldquoOpportunistic routing algorithmfor relay node selection in wireless sensor networksrdquo IEEETransactions on Industrial Informatics vol 11 no 1 pp 112ndash1212015

[22] L Pelusi A Passarella and M Conti ldquoOpportunistic network-ing data forwarding in disconnected mobile ad hoc networksrdquoIEEE Communications Magazine vol 44 no 11 pp 134ndash1412006

[23] R Kumar andN Singh ldquoA survey on data aggregation and clus-tering schemes in underwater sensor networksrdquo InternationalJournal of Grid and Distributed Computing vol 7 no 6 pp 29ndash52 2014

[24] M Di Francesco S K Das and G Anastasi ldquoData collection inwireless sensor networks with mobile elements a surveyrdquo ACMTransactions on Sensor Networks (TOSN) vol 8 no 1 article 72011

[25] K Seada M Zuniga A Helmy and B KrishnamacharildquoEnergy-efficient forwarding strategies for geographic routingin lossy wireless sensor networksrdquo in Proceedings of the 2ndInternational Conference on Embedded Networked Sensor Sys-tems (SenSys rsquo04) pp 108ndash121 Baltimore Md USA November2004

[26] J J Buckley and E Eslami An Introduction to Fuzzy Logic andFuzzy Sets Springer Science amp Business Media 2002

[27] S K Debray S Kannan and M Paithane ldquoWeighted decisiontreesrdquo in Proceedings of the Joint International Conference andSymposium on Logic Programming (JICSLP rsquo92) pp 654ndash668Washington DC USA November 1992

[28] K Matiasko J Bohacik V Levashenko and S Kovalık ldquoLearn-ing fuzzy rules from fuzzy decision treesrdquo Journal of Informa-tion Control andManagement Systems vol 4 no 2 pp 143ndash1542006

[29] C A Atkinson D F Jolley and S L Simpson ldquoEffect ofoverlying water pH dissolved oxygen salinity and sedimentdisturbances on metal release and sequestration from metalcontaminated marine sedimentsrdquo Chemosphere vol 69 no 9pp 1428ndash1437 2007

[30] R Agrawal T Imielinski and A Swami ldquoMining associationrules between sets of items in large databasesrdquo ACM SIGMODRecord vol 22 no 2 pp 207ndash216 1993

[31] N Vasconcelos and A Lippman ldquoStatistical models of videostructure for content analysis and characterizationrdquo IEEETrans-actions on Image Processing vol 9 no 1 pp 3ndash19 2000

[32] H-I Chen and H-Y Chang ldquoHomogeneous precipitationof cerium dioxide nanoparticles in alcoholwater mixed sol-ventsrdquoColloids and Surfaces A Physicochemical and EngineeringAspects vol 242 no 1ndash3 pp 61ndash69 2004

[33] N Shehata K Meehan I Hassounah et al ldquoReduced erbium-doped ceria nanoparticles one nano-host applicable for simul-taneous optical down- and up-conversionsrdquoNanoscale ResearchLetters vol 9 no 1 article 231 pp 1ndash6 2014

[34] N Shehata K Meehan and D Leber ldquoStudy of fluorescencequenching in aluminum-doped ceria nanoparticles potentialmolecular probe for dissolved oxygenrdquo Journal of Fluorescencevol 23 no 3 pp 527ndash532 2013

[35] A Sobeih J C Hou L-C Kung et al ldquoJ-Sim a simulationand emulation environment for wireless sensor networksrdquo IEEEWireless Communications vol 13 no 4 pp 104ndash119 2006

[36] J Pankove Optical Processes in Semiconductors Dover Publica-tions New York NY USA 1971

International Journal of

AerospaceEngineeringHindawi Publishing Corporationhttpwwwhindawicom Volume 2014

RoboticsJournal of

Hindawi Publishing Corporationhttpwwwhindawicom Volume 2014


Active and Passive Electronic Components

Control Scienceand Engineering

Journal of



RotatingMachinery


Hindawi Publishing Corporation httpwwwhindawicom

Journal ofEngineeringVolume 2014

Submit your manuscripts athttpwwwhindawicom

VLSI Design



Shock and Vibration


Civil EngineeringAdvances in

Acoustics and VibrationAdvances in



Electrical and Computer Engineering

Journal of

Advances inOptoElectronics


Volume 2014

The Scientific World JournalHindawi Publishing Corporation httpwwwhindawicom Volume 2014

SensorsJournal of


Modelling amp Simulation in EngineeringHindawi Publishing Corporation httpwwwhindawicom Volume 2014


Chemical EngineeringInternational Journal of Antennas and

Propagation




Navigation and Observation



DistributedSensor Networks



Zone 1

Zone 2

Zone N

Managementplatform



of the network


Sink

Sensor node


2

N

Management

Administraand decisi

support thoreal-time visua

of the netw

Figure 1 Comprehensive system architecture of data management and decision making

forwarding processes Powerful nodes receive the digitalizedsignal from one or more running nanosensors merge thecollected data with the geographical location of the sensingelement(s) and run the HI techniques for learning datapatterns and taking on-site decisions Based on the weightof reached decisions data might be forwarded to sinks andremote off-site management servers in a real-time fashionfor further analysis with security extensions The real-timemanagement of the overall system with security policiesrsquoenforcement is performed by the OMS The OMS operatesbased on learned patterns and rules of system operationsprojected from the DSFS to manage the entire network ofsensors The main tasks of OMS are to enable real-time datavisualization managed system control and secure systemoperation

As shown in Section 3 the main advantage of havingsuch level of autonomic control and management is theability to predict some of the sensor feedback based onthe analysis of the overall network feedback [14ndash16] Suchability facilitates optimizing the sensor power usage bycontrolling the activation periods of sensors leading to muchbetter sensor utilization expanding the lifetime of the entirenetwork and reducing the system cost Further such abilityfacilitates detecting and fixingexcluding any misbehavingor problematic sensing nodes that can massively reduce thesystem accuracy Additionally the automated system coulddefinitely overcome the problems of manual data collectionthat requires extensive effort and timeThis capability enablesalmost continuous real-time monitoring of DO with meansto identify and address sensor drift helping researchers toconstruct test models based on the flow of the monitoredwater supply Both DO and PH are sensed using fluorescence

quenching technique depending on the reduction of visibleemitted fluorescent optical intensity based on the changeof the quencher concentration of dissolved oxygen or thechemical impact of PH in ceria nanoparticles itself

Additionally we present in this paper a HI-based schemefor data classification and forwarding in wireless sensornetworks (WSNs) enabling energy-efficient better utilizedresources and optimizedQoS operationsThemain objectivesof our proposed scheme by adopting concepts in the infor-mation theory and Artificial Intelligence (AI) [17ndash19] are asfollows

(i) Building efficient reconfigurable information-drivendata classification and forwarding methodology forWSNs

(ii) Learning interesting routing-based attributes andgenerating forwarding models

(iii) Enabling the capability of learningexpectingdetect-ing abnormal data flows and making classificationand generating rules

The organizing of the remaining sections of this paperis as follows Section 2 Materials and Methods discussesthree subsections including themethodology of data sensingforwarding subsystem with the proposed HI-based dataclassificationforwarding scheme the experimental setup forsensing of DO and PHoptical nanosensors and the trustwor-thy Operation Management Subsystem The obtained resultsare analyzed and discussed in Section 3 Section 4 concludesthe paper and highlights our future work














Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3










119867(119862) = minus

119899

sum

119862=1

119901119862log2119901119862 (1)



119867(119862

119860) = sum

119886isin119860

(119901119886)119867(119862

119860 = 119886)

= minussum

119886isin119860

(119901119886)

119899

sum

119862=1

119901(119862(119860=119886))

log2119901(119862(119860=119886))

(2)





07

03

06

04

Sensor type

attribute

1

0

Reading level

attribute

Sensor status

attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF

O2



119886=119883) = 02

and (119901119886=119884) = 08







2







Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8















119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863





203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)




119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation






















Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3










119867(119862) = minus

119899

sum

119862=1

119901119862log2119901119862 (1)



119867(119862

119860) = sum

119886isin119860

(119901119886)119867(119862

119860 = 119886)

= minussum

119886isin119860

(119901119886)

119899

sum

119862=1

119901(119862(119860=119886))

log2119901(119862(119860=119886))

(2)





07

03

06

04

Sensor type

attribute

1

0

Reading level

attribute

Sensor status

attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF

O2



119886=119883) = 02

and (119901119886=119884) = 08







2







Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8















119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863





203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)




119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3










119867(119862) = minus

119899

sum

119862=1

119901119862log2119901119862 (1)



119867(119862

119860) = sum

119886isin119860

(119901119886)119867(119862

119860 = 119886)

= minussum

119886isin119860

(119901119886)

119899

sum

119862=1

119901(119862(119860=119886))

log2119901(119862(119860=119886))

(2)





07

03

06

04

Sensor type

attribute

1

0

Reading level

attribute

Sensor status

attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF

O2



119886=119883) = 02

and (119901119886=119884) = 08







2







Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8















119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863





203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)




119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











07

03

06

04

Sensor type

attribute

1

0

Reading level

attribute

Sensor status

attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF

O2



119886=119883) = 02

and (119901119886=119884) = 08







2







Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8















119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863





203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)




119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












Safe(high quality)

5 (ppm)

Mem

bers

hip

class

3

1

7 8















119867(119863) = minus

4

sum

119863=1

119901119863log2119901119863





203 minus 02log

202

minus 01log201

= 0529 + 0521 + 04644 + 03322

= 18466 bits

(4)




119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












119868 (119863 119860) = 119867 (119863) minus 119867(119863

Type)

= 18466 minus

1003816100381610038161003816O21003816100381610038161003816

10Entropy (O

2)

minus|PH|10

Entropy (PH)


202)


202)

= 18466 minus 05 (02575 + 04644)

minus 05 (04422 + 2 times 04644)

= 18466 minus 0361 minus 06855 = 08 bits

(5)


119868 (119863 119860) = 119867 (119863) minus 119867(119863

Level) = 18466

minus

1003816100381610038161003816High1003816100381610038161003816


10Entropy (Low)

minus|No|10


Entropy (NA)


= 18466 bits

119868 (119863 119860) = 119867 (119863) minus 119867(119863

Status) = 18466

minus|ON|10


Entropy (OFF)


20333


+ 05283 + 0482) = 18466 minus 09 (0566256)

= 1337 bits

(6)


Monochromator

PMT

Powermeter

DOPH meter

UV LED

Visible light

Inlet N2O2


Laser module

Optical fibers

Optical NPs

Pump

Washing system


Photodetector

Microcontroller






Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation










Managementplatform



of the network


Sink

Sink Sensor node


Attributes----------------------------------------------

----------------------------------------------------

---------------------

----------------------------------------------------


algorithm



functions)


probability)


gain calculation


Fuzzy rules



sensor nodes

Attribute 1

Attribute 2

Attribute 3



07

03

06

04

Sensor type

attribute

1

0

Reading level


attribute

08

02

07

03

1

0

075

025


(low quality)

Moderate

Safe(high quality)

(high quality)

(high quality)

(high quality)

Safe

Dangerous

Safe

Quality classes

Moderate

Safe

Pb

High

Low

ON

OFF

ON

OFF

High

Low

ON

OFF

ON

OFF












PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











PH levelDestination

PH-relate

d data

flow

related data flow

Read

ings

Read

ings

River bank


area

More information


about dissolved O2

Dissolved O

-2

Dissolved O2








2










0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation












0

200000

400000

600000

800000

1e + 006

Thro

ughp

ut (b

ytes

sec

)



Opt

imum

qua

lity

Aver

age m

axim

um

Com

bine

d hi

ghqu

ality

leve

l

Com

bine

d lo

wqu

ality

leve

l

low

qua

lity

leve

lEx

plic

itO

2-r

elat

ed

0123456789

thro

ughp

ut(b

ytes

sec

onds

)

times105




020406080

100120140160

Det

ectio

n tim

e (se

cond

s)













minus2000

200400600800

10001200

Aver

age r

esid

ual p

ower

per n

ode (

wat

t)













Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation
















Undoped ceria

400 500 600 700 800 9003000

05

1

15

2

25

3

35

4

120582 (nm)

120572(A

U)

(a)

1 15 2 25 3 35 4 45

E (eV)Undoped ceria

Eg = 33 eV0

50

100

150

200


(120572E)2

(b)



2levels


120572119864 = 119860lowast

(119864 minus 119864119892)12

(7)


119892is the




CBA



0

02

04

06

08

1

12

14

Inte

nsity

550 600 650 700 750500120582 (nm)




Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











Ceria PH = 816

550 600 650 700 7505000

01

02

03

04

05

06

07

08

09

1

Inte

nsity

(au

)

120582 (nm)

PH = 666PH = 410

PH = 347PH = 283



4 Conclusions




Competing Interests


Acknowledgments


References









































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation








































RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation











RoboticsJournal of





Journal of



RotatingMachinery





VLSI Design



Shock and Vibration







Journal of



Volume 2014


SensorsJournal of





Propagation









Documents

Research Article System-Aware Smart Network Management for