[PDF] The Online Journal of Science and Technology

Download The Online Journal of Science and Technology...

ISSN2146‐7390

TheOnlineJournalof ScienceandTechnology

Prof.Dr.Aytekinİşman Editor‐in‐Chief Prof.Dr.MustafaŞahinDündar Editor HüseyinEski TechnicalEditor

www.tojsat.net April2016

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Copyright©2016‐THEONLINEJOURNALOFSCIENCEANDTECHNOLOGY Allrightsreserved.NopartofTOJSAT’sarticlesmaybereproducedorutilizedinanyformorbyanymeans, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system,withoutpermissioninwritingfromthepublisher. PublishedinTURKEY ContactAddress: Prof.Dr.MustafaŞahinDündar‐TOJSAT,EditorSakarya‐Turkey

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Message from the Editor-in-Chief Dear Colleagues, TOJSAT welcomes you. TOJSAT would like to thank you for your online journal interest. We are delighted that almost 50,000 academicians, teachers, and students from around the world have visited for two years. It means that TOJSAT has continued to diffuse new trends in science and technology. We hope that the volume 6, issue 2 will also successfully accomplish our science and technology goal. TOJSAT is confident that readers will learn and get different aspects on science and technology. Any views expressed in this publication are the views of the authors and are not the views of the Editor and TOJSAT. TOJSAT thanks and appreciate the editorial board who have acted as reviewers for one or more submissions of this issue for their valuable contributions. TOJSAT will organize ISTEC-2016 International Science and Technology Conference (www.iste-c.net) in Austria. This conference is now a well-known science and technology event. It promotes the development and dissemination of theoretical knowledge, conceptual research, and professional knowledge through conference activities. Its focus is to create and disseminate knowledge about science and technology. ISTEC - 2015 conference book has been published at http://www.iste-c.net Call for Papers TOJSAT invites you article contributions. Submitted articles should be about all aspects of science and technology. The articles should be original, unpublished, and not in consideration for publication elsewhere at the time of submission to TOJSAT. Manuscripts must be submitted in English. TOJSAT is guided by it’s editors, guest editors and advisory boards. If you are interested in contributing to TOJSAT as an author, guest editor or reviewer, please send your cv to [emailprotected].

April 01, 2016 Prof. Dr. Aytekin ISMAN Sakarya University

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Letter from the Editor Dear Tojsat Readers,

Now, we are reading 6th volume of The Online Journal of Science and Technology. We finished the istec 2015 conference hold in Saint Petersburg, Russia. These papers receives interests from all over the areas of science and technology. The selected, peer-reviewed papers were accepted for publication apart from contributions from all over the World. The accepted and published papers are from information technology to mechanical engineering.

April 01, 2016 Prof. Dr. Mustafa S. DUNDAR Sakarya University

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Editor‐in‐Chief Prof.Dr.AytekinİŞMAN‐SakaryaUniversity,Turkey

Editor Prof.Dr.MustafaŞahinDÜNDAR‐SakaryaUniversity,Turkey

TechnicalEditor HüseyinEski,SakaryaUniversity,Turkey

EditorialBoard Prof.Dr.AhmetAPAY,SakaryaUniversity,Turkey Prof.Dr.GilbertMbothoMASITSA,UniversirtyofThe FreeState,SouthAfrica Prof.Dr.AntoinetteJ.MUNTJEWERFF,Universityof Amsterdam,Netherlands Prof.Dr.GregoryALEXANDER,UniversityofTheFree State,SouthAfrica Prof.Dr.ArvindSINGHAL,UniversityofTexas,United States Prof.Dr.Gwo‐DongCHEN,NationalCentral UniversityChung‐Li,Taiwan Prof.Dr.AytekinİŞMAN,SakaryaUniversity,Turkey Prof.Dr.Gwo‐JenHWANG,NationalTaiwan Prof.Dr.BilalGÜNEŞ,GaziUniversity,Turkey UniversityodScienceandTechnology,Taiwan Prof.Dr.BrentG.WILSON,UniversityofColoradoat Prof.Dr.HellmuthSTACHEL,ViennaUniversityof Denver,UnitedStates Technology,Austria Prof.Dr.CaferÇELİK,AtaturkUniversity,Turkey Prof.Dr.J.AnaDONALDSON,AECTFormer Prof.Dr.Chih‐KaiCHANG,NationalUniversityof President,UnitedStates Taiwan,Taiwan Prof.Dr.MehmetAliYALÇIN,Sakarya Prof.Dr.Chin‐MinHSIUNG,NationalPingtung University,Turkey University,Taiwan Prof.Dr.MustafaS.DUNDAR,Sakarya Prof.Dr.ColinLATCHEM,OpenLearning University,Turkey Consultant,Australia Prof.Dr.NabiBuxJUMANI,InternationalIslamic Prof.Dr.DeborahE.BORDELON,GovernorsState University,Pakistan University,UnitedStates Prof.Dr.OrhanTORKUL,SakaryaUniversity,Turkey Prof.Dr.DonM.FLOURNOY,OhioUniversity,United Prof.Dr.PaoloDiSia,UniversityofVerona,Italy States Prof.Dr.ÜmitKOCABIÇAK,SakaryaUniversity,Turkey Prof.Dr.Feng‐ChiaoCHUNG,NationalPingtung University,Taiwan Assist.Prof.Dr.EnginCAN,SakaryaUniversity,Turkey Prof.Dr.FinlandCHENG,NationalPingtung Assist.Prof.Dr.HüseyinOzanTekin,Üsküdar University,Taiwan University,Turkey Prof.Dr.FrancineShuchatSHAW,NewYork Assist.Prof.Dr.TuncerKORUVATAN,TurkishMilitary University,UnitedStates Academy,Turkey Prof.Dr.FrankS.C.TSENG,NationalKaohsiungFirst Dr.AbdulMutalibLEMAN,UniversitiTunHussein UniversityosScienceandTechnology,Taiwan OnnMalaysia,Malaysia Prof.Dr.GianniViardoVERCELLI,Universityof Dr.AbdülkadirMASKAN,DicleUniversity,Turkey Genova,Italy Dr.AlperTolgaKUMTEPE,AnadoluUniversity,Turkey

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Dr.AtillaYILMAZ,HacettepeUniversity,Turkey Dr.MarthaPILARMéNDEZBAUTISTA,EAN University,Bogotá,Colombia Dr.BekirSALIH,HacettepeUniversity,Turkey Dr.MdNorNoorsuhada,UniversitiTeknologiMARA Dr.BerrinÖZCELİK,GaziUniversity,Turkey PulauPinang,Malaysia Dr.BurhanTURKSEN,TOBBUniversityofEconomics Dr.MohamadBINBILALALI,UniversitiTeknologi andTechnology,Turkey Malaysia,Malaysia Dr.ChuaYanPIAW,UniversityofMalaya,Malaysia Dr.MohamedBOUOUDINA,Universityof Dr.ConstantinoMendesREI,InstitutoPolitecnicoda Bahrain,Bahrain Guarda,Portugal Dr.MohammadRezaNAGHAVI,Universityof Dr.DanielKIM,TheStateUniversityofNew Tehran,Iran York,SouthKorea Dr.MohdRoslanMODHNOR,Universityof Dr.Dong‐HoonOH,UniversiyofSeoul,SouthKorea Malaya,Malaysia Dr.EvrimGENÇKUMTEPE,AnadoluUniversity,Turkey Dr.MuhammedJAVED,IslamiaUniversityof Bahawalpur,Pakistan Dr.FabricioM.DEALMEIDA Dr.MuratDİKER,HacettepeUniversity,Turkey Dr.FahadN.ALFAHAD,KingSaudUniversity,Saudi Arabia Dr.MusaDOĞAN,MiddleEastTechnical University,Turkey Dr.FatimahHASHIM,UniversitiMalaya,Malaysia Dr.MustafaKALKAN,DokuzEylülUniversiy,Turkey Dr.FatmaAYAZ,GaziUniversity,Turkey Dr.NihatAYCAN,MuğlaUniversity,Turkey Dr.FonkSOONf*ck,UniversitiSains Malaysia,Malaysia Dr.NilgünTOSUN,TrakyaUniversity,Turkey Dr.GalipAKAYDIN,HacettepeUniversity,Turkey Dr.NursenSUCSUZ,TrakyaUniversity,Turkey Dr.HasanMUJAJ,UniversityofPrishtina,Kosovo Dr.OsmanANKET,GülhaneAskeriTıp Akademisi,Turkey Dr.HasanKIRMIZIBEKMEZ,Yeditepe University,Turkey Dr.PiotrTOMSKI,CzestochowaUniversityof Technology,Poland Dr.HasanOKUYUCU,GaziUniversity,Turkey Dr.RajaRizwanHUSSAIN,KingSaudUniversity,Saudi Dr.HoSooonMIN,INTIInternational Arabia University,Malaysia Dr.RamdaneYOUNSI,PolytechnicUniversity,Canada Dr.Ho‐JoonCHOI,KyonggiUniversity,SouthKorea Dr.RıdvanKARAPINAR,YuzuncuYılUniversity,Turkey Dr.HyoJinKOO,WoosukUniversity,SouthKorea Dr.RıfatEFE,DicleUniversity,Turkey Dr.Jae‐EunLEE,KyonggiUniversity,SouthKorea Dr.RuzmanMd.NOOR,UniversitiMalaya,Malaysia Dr.JaroslavVesely,BRNOUNIVERSITYOF TECHNOLOGY,CzechRepublic Dr.SandeepKUMAR,SunyDownstateMedical Center,UnitedStates Dr.JonChaoHONG,NationalTaiwanNormal University,Taiwan Dr.SanjeevKumarSRIVASTAVA,MitchellCancer Institute,UnitedStates Dr.JosephS.LEE,NationalCentralUniversity,Taiwan Dr.SelahattinGÖNEN,DicleUniversity,Turkey Dr.KendraA.WEBER,UniversityofMinnesota,United States Dr.SenayCETINUS,CumhuriyetUniversity,Turkey Dr.KimSunHEE,WoosukUniversity,SouthKorea Dr.SharifahNorulAKMAR,Universityof Malaya,,Malaysia Dr.LatifKURT,AnkaraUniversity,Turkey Dr.ShengQUENYU,BeijingNormalUniversity,China Dr.LiYING,ChinaCentralRadioandTV University,China Dr.SunYoungPARK,KonkukUniversity,SouthKorea Dr.Man‐KiMOON,Chung‐AngUniversity,South Dr.TeryL.ALLISON,GovernorsState Korea University,UnitedStates

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Dr.TürkayDERELİ,GaziantepUniversity,Turkey Dr.YueahMiaoCHEN,NationalChungCheng University,Taiwan Dr.UnerKAYABAS,InonuUniversity,Turkey Dr.YusupHASHIM,AsiaUniversity,Malaysia Dr.WanMohdHirwaniWANHUSSAIN,Universiti KebangsaanMalaysia,Malaysia Dr.ZawawiISMAIL,UniversityofMalaya,Malaysia Dr.WanZahWANALI,UniversitiPutra Dr.ZekaiSEN,IstanbulTechnicalUniversity,Turkey Malaysia,Malaysia

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

TableofContents APLANARROBOTDESIGNANDCONSTRUCTIONWITHMAPLE

EnginCAN,ÖzkanCANAY

ASTOCHASTIC‐OPTIMIZATIONMODELFORDETERMININGTHEOPTIMALMICRO‐SITINGOFWIND TURBINES

AkinerTuzuner,IssaAlmassri,SelcukGoren

ANINQUARYSTUDYOFNiOFILMSDEPOSITEDWITHSOL‐GELSPINCOATING

GuvenTurgut

COMPARATIVEANALYSISOFCURRENTMETHODSINSEARCHINGOPENEDUCATIONCONTENT REPOSITORIES

BenneaserJohn,V.Thavavel,JayakumarJayaraj,A.Muthukumar,PoornaselvanKittuJeevanandam

CREATINGONTOLOGYBASEDCONCEPTMAPSWHICHCANBEQUERIEDINCOMPUTERENVIROMENT

MehmetMilli,EmreÜnsal,ÖzlemAktaş

ELASTICFOUNDATIONEFFECTSONTHREEDIMENSIONALARCHDAMS

MuhammetKARABULUT,MuratEmreKARTAL,MuratCAVUSLI,DeryaTANRIVERMIS,BayramKAYALAR

EXPERIMENTALINVESTIGATIONOFDUCTILITYOFREINFORCEDCONCRETEBEAMSSTRENGTHENEDWITH POLYPROPYLENEFIBERS

RıfatSEZER,AbdulhamidARYAN

FRICTIONANDWEARPERFORMANCEOFHIGHDENSITYPOLYETHYLENE/STYRENE‐BUTADIENERUBBER POLYMERBLENDS

SezginErsoy

LOWCOSTWIRELESSSENSORNETWORKSFORENVIRONMENTMONITORING

EmreÜnsal,MehmetMilli,YalçınÇEBİ

PANELCOINTEGRATIONANALYSISOFINTERNATIONALTOURISMDEMAND:SAMPLEOFANTALYA

SabriyeGÜVEN,MehmetMERT

POSITIONOPTIMISATIONOFGEDETECTORSINNUCLEARRESONANCEFLUORESCENCE(NRF) EXPERIMENTBYUSINGMONTECARLOMETHOD

HüseyinOzanTekin,IskenderAkkurt

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

THEHEALTHINFORMATIONPROFESSIONALINEHEALTH:ETHICALCONSIDERATIONSFORAN INTERJURISTICALSETTING

Eike‐HennerW.Kluge

THEIMPROVEOFCOMBUSTIONPROPERTIESONWOODENMATERIALBYUSINGLIQUEDNITROGENAND BORICACID

C.Özcan,Ş.Kurt,R.Esen,C.Özcan

THEMISSINGPERSONSFINDINGSENSITIVETOMOVEMENTSWITHIMAGERECOGNITIONSYSTEMAND NUMERICDATAPEOPLEOFLOSSINTURKEY

EmreAVUÇLU,FatihBAŞÇİFTÇİ

USINGVIKORMETHODFORANALYZINGOFQUALIFICATIONLEVELSANDTRANSITIONTOEMPLOYMENT OFEUROPEANUNIONANDCANDIDATECOUNTRIES

EmreIPEKCICETIN,H.HandeCETIN

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

A PLANAR ROBOT DESIGN AND CONSTRUCTION WITH MAPLE Engin CAN1, Özkan CANAY2 1

Kaynarca School of Applied Sciences, Sakarya University, Turkey, [emailprotected] 2 Vocational School of Adapazarı, Sakarya University, Turkey Abstract: Maple is used to do numerical computation, plot graphs and do exact symbolic manipulations and word processing. In this study we demonstrate how Maple can be used for the simulation of a planar robot. This offers the possibility to become familiar of mathematical modelling. The mechanism under consideration is a so-called F-mechanisms (Can & Stachel, 2014), i.e., a planar parallel 3-RRR robot with three synchronously driven cranks. It turns out that at this example it is not possible to find the poses of the moving triangle exactly by graphical methods with traditional instruments only. Hence, numerical methods are essential for the analysis of motions which can be performed by a planar robot. Keywords: Maple, scientific computing, mathematical modelling, planar mechanism, planar parallel 3-RRR-robot.

Introduction F-mechanisms were first introduced and analyzed by Can (2012) and Can & Stachel (2014). These are high-speed planar mechanisms with modifiable compulsory courses based on parallel robots simultaneously driven cranks. Staicu (2008) has the kinematics of such robot treated; the website of Ahamed provides a controllable interactive simulation of parallel planar manipulators. In the following, we review the characteristics of F-mechanisms: Definition A Fehrer-mechanism (F-mechanism in short), is a kinematic chain with 8-links Σ 0 ,…, Σ 7 and 9 revolute joints A 0 , B 0 , C 0 , A 1 ,…, C 2 (see Figure 1) with the following properties: R

1) There are three driving cranks A 0 A 1 ⊂ Σ 1 , B 0 B 1 ⊂ Σ 2 , and C 0 C 1 ⊂ Σ 3 . They rotate with the same angular velocity ω about the respective anchor points A 0 , B 0 and C 0 , all fixed in the frame Σ 0 . The links Σ 1 and Σ 3 rotate counter-clockwise; Σ 2 rotates either counter-clockwise (direct F-mechanism) or clockwise (indirect Fmechanism). 2) The bars A 1 A 2 ⊂ Σ 4 , B 1 B 2 ⊂ Σ 5 , and C 1 C 2 ⊂ Σ 6 connect the active cranks with the moving frame Σ 7 . The points A 2 , B 2 , C 2 are attached to Σ 7 . 3) Variable phase shiftings between the cranks enable to modify the constrained motion Σ 7 /Σ 0 . �� The lengths of the cranks are denoted by a 1 = 𝐴𝐴 0 𝐴𝐴1 , b 1 = 𝐵𝐵0 𝐵𝐵1 and c 1 = 𝐶𝐶0 𝐶𝐶1 . The bars' lengths are a 2 = 𝐴𝐴1 𝐴𝐴2 , �� b 2 = 𝐵𝐵1 𝐵𝐵2 and c 2 = 𝐶𝐶1 𝐶𝐶2 .

Figure 1: A planar parallel 3-RRR robot (An indirect F-mechanism)

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Maple Procedure and Construction of the Motion After all mathematical representation was given by Can & Stachel (2014), we will only give design and construction of a planar parallel robot with Maple which is displayed in Figure 2 also addressed in Fig. 3 and Fig 4. Its dimensions are as follows: fixed triangle: A 0 = (0.0, 0.0), B 0 = (52.5, 8.0), C 0 = (40.0, 99.0), lengths of cranks: a 1 = 19.0, b 1 = 14.0, c 1 = 16.0, phase shifts: ä𝑏𝑏 = 2430, 𝛿𝛿𝑐𝑐 = −150, lengths of bars: a 2 = 35.0, b 2 = 34.0, c 2 = 54.0, moving triangle: A 2 = (0.0, 0.0), B 2 = (40.0, 18.0), C 2 = (−7.0, 28.0).

Figure 2: An indirect F-mechanism displayed using Maple From here we start to write the program with using Maple: > > > > > > > > > > > > > > >

restart: with(linalg):with(plots): A0:=[0,0]: B0:=[52.5,8]: C0:=[40,99]: A2g:=[0,0]: B2g:=[40,18]: C2g:=[-7,28]: la1:=19: lb1:=14: lc1:=16: la2:=35: lb2:=34: lc2:=54: omegab:=-1: phasgradb:=243: phasgradc:=-15: phasb:=evalf(phasgradb*Pi/180): phasc:=evalf(phasgradc*Pi/180): pi:=evalf(Pi): x:=eva200lf(326*Pi/180): epsilon:=pi: mitte:=x: start:=mitte-epsilon: ende:=mitte+epsilon: anz:=200: dif:=ende-start: w:=dif/anz: idx0:=1:

Now we set the points A 1 , B 1 and C 1 to depending on tangent f1, f2 and f3 half drive angle. In addition, the angle of rotation is t := 𝜑𝜑70 introduced. Furthermore, we introduce the shifting vector trans = (u, v):= u 0 and an orthogonal matrix dreh := A. > > > > > > > > > >

Arm:=[((1-f^2)/(1+f^2)),(2*f/(1+f^2))]: A1:=A0+la1*(subs(f=f1,Arm)): B1:=B0+lb1*(subs(f=f2,Arm)): C1:=C0+lc1*(subs(f=f3,Arm)): trans:=[u,v]: co:=(1-t^2)/(1+t^2): si:=2*t/(1+t^2): dreh:=matrix(2,2,[co,-si,si,co]): A2:=simplify(matadd(evalm(dreh &* A2g),trans)): B2:=simplify(matadd(evalm(dreh &* B2g),trans)): C2:=simplify(matadd(evalm(dreh &* C2g),trans)):

Equivalent to the system of equations (4) in Can & Stachel (2014) are now the equations AG1 = 0, AG2 = 0 and AG3 = 0 :

www.tojsat.net

The Online Journal of Science and Technology - April 2016

> > > > > >

Volume 6, Issue 2

AG1:=numer(simplify(((A2[1]-A1[1])^2+(A2[2]-A1[2])^2)-la2^2)): a1:=coeff(AG1,u^2): AG2:=numer(simplify(((B2[1]-B1[1])^2+(B2[2]-B1[2])^2)-lb2^2)): a2:=coeff(AG2,u^2): AG3:=numer(simplify(((C2[1]-C1[1])^2+(C2[2]-C1[2])^2)-lc2^2)): a3:=coeff(AG3,u^2):

It is followed by the elimination of u 0 2 = u2 +v2 by subtraction. We take into account only the coefficients a i of u2. There remain two linear equations eq [1] and eq [2], we solve for u and v: > > > > > > > > > > > > >

eq[1]:=eval(numer(combine(a1*AG2-a2*AG1))): eq[2]:=eval(numer(combine(a1*AG3-a3*AG1))): G1:=collect(expand(eq[1]),[u,v]): G2:=collect(expand(eq[2]),[u,v]): P:=coeff(G1,u): Q:=coeff(G1,v): S:=coeff(G2,u): T:=coeff(G2,v): R:=-subs(u=0,v=0,G1): U:=-subs(u=0,v=0,G2): mat:=matrix(2,2,[P,Q,S,T]): vec:=[R,U]: loesung:=linsolve(mat,vec): uu:=loesung[1]: vv:=loesung[2]: tt:=numer(simplify(subs(u=uu,v=vv,AG1))):

We denote the individual values of the angle drive for the cranks A 0 A 1 , B 0 B 1 and C 0 C 1 respectively with ff1[i], ff1[i] and ff3[i], for i = 1, ..., anz (anz:=200:) and start the main loop of the program: > > > > >

for i from 0 to anz do ff1[i]:=start+i*w: ff2[i]:=omegab*ff1[i]+phasb: ff3[i]:=ff1[i]+phasc: tt_werte[i]:=[fsolve(evalf(subs(f1=tan(ff1[i]/2),f2=tan(ff2[i]/2), f3=tan(ff3[i]/2),tt)))];

Case of i = 0 we choose one of the zeros arbitrary (see Figure 2). > t_neu[0]:=tt_werte[0][idx0]: Otherwise, we find the number of zeros from using > nbr[i]:= nops(tt_werte[i]);

and search for fixed i the solution tt werte[i][j], 1 ≤ j ≤ nbr[i] that the calculated approximation value from the previous position naeh[i] := tneu[i − 1] + dtneu[i − 1] comes closest. This is done as follows: > naeh[i]:= t_neu[i-1] + dt_neu[i-1]; > idx[i]:= 1; > for j from 2 to nbr[i] do if abs(tt_werte[i][j]-naeh[i]) t_neu[i]:=tt_werte[i][idx[i]]:

The here proposed selection of the 'right' zero point must be observed four conditions: > idx[I]:= 1; > if abs(naeh[I]) t_neu[i]:=tt_werte[i][idx[i]];

Velocity Analysis of the Motion In the following we calculate an approximate value of the next correct position. > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > >

A1:=A0+la1*(subs(f=tan(ff1[i]/2),Arm)); B1:=B0+lb1*(subs(f=tan(ff2[i]/2),Arm)); C1:=C0+lc1*(subs(f=tan(ff3[i]/2),Arm)); d1:=A1-A0: d2:=B1-B0: d3:=C1-C0: vA1:=w*[-d1[2],d1[1]]: vB1:=w*omegab*[-d2[2],d2[1]]: vC1:=w*[-d3[2],d3[1]]: X[i]:=[x1[i],x2[i]]: Y[i]:=matrix(2,2,[0,-x3[i],x3[i],0]): u_werte:=evalf(subs(f1=tan(ff1[i]/2),f2=tan(ff2[i]/2), f3=tan(ff3[i]/2),t=t_neu[i],uu)); v_werte:=evalf(subs(f1=tan(ff1[i]/2),f2=tan(ff2[i]/2), f3=tan(ff3[i]/2),t=t_neu[i],vv)); transm:=[uu_werte,vv_werte]: c2:=evalf(subs(t=t_neu[i],co)): s2:=evalf(subs(t=t_neu[i],si)): drehm:=matrix(2,2,[c2,-s2,s2,c2]): A2:=convert(evalf((matadd(evalm(drehm &* A2g),transm))),list): B2:=convert(evalf((matadd(evalm(drehm &* B2g),transm))),list): C2:=convert(evalf((matadd(evalm(drehm &* C2g),transm))),list): vA2:=simplify(matadd(evalm(Y[i] &* A2),X[i])): vB2:=simplify(matadd(evalm(Y[i] &* B2),X[i])): vC2:=simplify(matadd(evalm(Y[i] &* C2),X[i])): V11:=matadd(vA2,-vA1): V12:=matadd(A2,-A1): V21:=matadd(vB2,-vB1): V22:=matadd(B2,-B1): V31:=matadd(vC2,-vC1): V32:=matadd(C2,-C1): GL1:=simplify(innerprod(V11,V12)); GL2:=simplify(innerprod(V21,V22)); GL3:=simplify(innerprod(V31,V32)); P1:=coeff(GL1,x1[i]): P2:=coeff(GL2,x1[i]): P3:=coeff(GL3,x1[i]): Q1:=coeff(GL1,x2[i]): Q2:=coeff(GL2,x2[i]): Q3:=coeff(GL3,x2[i]): R1:=coeff(GL1,x3[i]): R2:=coeff(GL2,x3[i]): R3:=coeff(GL3,x3[i]): T1:=-subs(x1[i]=0,x2[i]=0,x3[i]=0,GL1): T2:=-subs(x1[i]=0,x2[i]=0,x3[i]=0,GL2): T3:=-subs(x1[i]=0,x2[i]=0,x3[i]=0,GL3): mat:=matrix(3,3,[P1,Q1,R1,P2,Q2,R2,P3,Q3,R3]): vec:=[T1,T2,T3]: losung:=linsolve(mat,vec): x1[i]:= losung[1]: x2[i]:= losung[2]: x3[i]:= losung[3]: dt_neu[i]:= eval((1+t_neu[i]*t_neu[i])*x3[i]/2): od:

Displaying of the Motion And finally should animation of the motion with Maple generated, one must proceed according to the calculation of the positions within the loop as follows: > > > > >

for i from 0 to anz do AA1[i]:=A0+la1*(subs(f=tan(ff1[i]/2),Arm)); BB1[i]:=B0+lb1*(subs(f=tan(ff2[i]/2),Arm)); CC1[i]:=C0+lc1*(subs(f=tan(ff3[i]/2),Arm)); uu_werte:=evalf(subs(f1=tan(ff1[i]/2),f2=tan(ff2[i]/2), f3=tan(ff3[i]/2),t=t_neu[i],uu)); > vv_werte:=evalf(subs(f1=tan(ff1[i]/2),f2=tan(ff2[i]/2), f3=tan(ff3[i]/2),t=t_neu[i],vv));

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

> trans2:=[uu_werte,vv_werte]; > c2:=evalf(subs(t=t_neu[i],co)): > s2:=evalf(subs(t=t_neu[i],si)): > dreh2:=matrix(2,2,[c2,-s2,s2,c2]): > AA2[i]:=convert(evalf((matadd(evalm(dreh2&*A2g),trans2))),list): > BB2[i]:=convert(evalf((matadd(evalm(dreh2&*B2g),trans2))),list): > CC2[i]:=convert(evalf((matadd(evalm(dreh2&*C2g),trans2))),list): > Dreieck[i]:=polygonplot([AA2[i],BB2[i],CC2[i]],color=red): > Gelenk_a01[i]:=polygonplot([A0,AA1[i]],thickness=2): > Gelenk_a12[i]:=polygonplot([AA1[i],AA2[i]],thickness=2): > Gelenk_b01[i]:=polygonplot([B0,BB1[i]],thickness=2): > Gelenk_b12[i]:=polygonplot([BB1[i],BB2[i]],thickness=2): > Gelenk_c01[i]:=polygonplot([C0,CC1[i]],thickness=2): > Gelenk_c12[i]:=polygonplot([CC1[i],CC2[i]],thickness=2): > od: > Konfiguration:=seq(display(Dreieck[i],Gelenk_a01[i],Gelenk_b01[i], Gelenk_c01[i],Gelenk_a12[i],Gelenk_b12[i],Gelenk_c12[i]),i=1..anz): > display(Konfiguration,scaling=constrained,insequence=true);

Figure 3: Different poses of given example

Figure 4: Diagram showing 𝜑𝜑70 in given example

If you want to have an overview how many rotational angle 𝜑𝜑70 are possible, for how many positions the moving plane Σ 7 for which drive angles (here with ff1[i] denotes), so it can be drawing in a graph (see Figure 4) discrete ff1- values the corresponding - each in degrees- 𝜑𝜑70 value (in this case tt_werte[i][j]) . R

Acknowledgements

This work has been supported by the Research Fund of the Sakarya University with the project number 2014-1702-001.

References Can, E. & Stachel H. (2014). A planar parallel 3-RRR robot with synchronously driven cranks. Mechanism and Machine Theory 79 (pp. 29-45). Can, E. (2012). Analyse and Synthese eines schnelllaufenden ebenen Mechanismus mit modifizierbaren Zwangläufen. PhD Thesis, Vienna University of Technology. Can, E. (2015). Über die Verwendung von MAPLE für die Simulation von Mechanismen. Teaching Mathematics and Computer Sciences 13/1 (pp. 21-39). Ahamed, S. Simulation of a 3-DOF 3-RRR Planar Parallel Robot, www.parallemic.org/Java/3RRR.html Gander, W. & Hrebicek, J. (2004). Solving Problems in Scientific Computing Using Maple and Matlab. Heidelberg Springer, Berlin. Staicu, S. (2008). Kinematics of the 3-RRR planar parallel robot. U.P.B. Sci. Bull. Ser. D. 70/2 (pp. 3–14). Wunderlich, W. (1970). Ebene Kinematik. BI–Hochschultaschenbücher, Bd. 447. Bibliographisches Institut, Mannheim.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

A STOCHASTIC-OPTIMIZATION MODEL FOR DETERMINING THE OPTIMAL MICRO-SITING OF WIND TURBINES Akiner Tuzuner1, Issa Almassri2, Selcuk Goren3 1

Istanbul Kemerburgaz University, Department of Industrial Engineering, Istanbul/Turkey Tel.: +90-5452821975 Email: [emailprotected] 2

Istanbul Kemerburgaz University, Graduate School of Science, Istanbul/Turkey Tel.: +90-5349260612 Email: [emailprotected]

Abdullah Gul University, Department of Industrial Engineering, Kayseri/Turkey Tel.: +90-5534810001 Email: [emailprotected]

Abstract: We propose a general model for the placement of wind turbines in a rectangular grid formation over a flat area. For better realism, we consider stochastic wind speeds and directions, in conjunction with the wake effects that upstream turbines impose on downstream ones. The objective is to pack as many turbines as economically optimal in a given area, i.e. to maximize the expected MW output per dollar of capital investment and O&M costs per meter square. Due to the complex structure of the mathematical model, we apply a hybrid approach of MonteCarlo sampling of wind speeds and directions together with the Nelder-Mead heuristic method to search for the optimal horizontal and vertical spacing of the turbines. Results of a case study based on a real dataset of wind speeds and directions, a selected commercial turbine’s approximated power curve, and industry estimates of costs is discussed. Keywords: wind energy, wind power, wind investment, micro-siting, wake effect

Introduction Wind energy has gained great attention because it represents an important option for reducing the reliance on hydrocarbons for energy production, especially for electricity. With the current technology, one challenge faced in wind farm design is the appropriate placement of the individual wind turbines in order to optimally harvest the energy from the wind. Grouping the turbines leads to a reduction of the power produced by the downstream wind turbines due the so-called “wake” effect. That is, if a turbine is within the area of the turbulence caused by another turbine, or the area behind another turbine, the wind speed suffers a reduction, and therefore, there is a decrease in its electricity production. On the other hand, spacing the turbines too far apart requires a larger land area for the wind farm, which may not only bring added costs but also be just infeasible. Because of the considerable amount of power loss due to the wake effects in a large wind farm, it is considered one of the main issues that should be focused on when optimizing the wind turbine position. Several studies have been conducted in recent years in order to maximize energy production and the efficiency of the turbines. These studies have focused primarily on obtaining the optimal placement of turbines within a wind farm, based on their output efficiency. To state this problem, we define two objectives: (i) maximization of energy production, and (ii) minimization of the total cost and land area requirements. In the present research we use an objective function that represent the expected MW output per dollar of capital investment and O&M costs per meter square of land (i.e. MW/$m2). By use of the Nelder-Mead heuristic method adapted to stochastic wind speeds and directions, we search for the optimal vertical and horizontal spacing (respectively in the North-South and East-West coordinate axes) of a rectangular grid of wind turbines. That is, given a fixed number of turbines in a rectangular formation, the goal is maximize the area-density of the expected power return on the investment. The remainder of the paper is organized as follows. Section two formulates the model of the wake model used in the wind farm. Section three presents describes the wind farm layout optimization model, including such components and parameters as the power curves, the wind direction and speed distribution, and cost estimation. Section four presents the two case scenarios and discusses the results of the optimization process. Finally, in section five we present our conclusions and discuss futures avenues of research on the problem.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

The Wake Model Wind turbine wakes are generally divided into three different regions, as described in Figure 1. These are the near wake region, the transition region, and the far wake region. This model assumes the wake region is an empirical linear expansion region, Jensen Model, starting from behind the wind turbine, as shown in Figure 2. Here the model is characterized by a uniform velocity profile at any distance 𝑥𝑥 in the downstream behind the turbine.

Figure 1: Wake Regions Behind a Wind Turbine To start analyzing the model we first write the wind speed at a distance 𝑥𝑥 behind the turbine. Let us consider Figure 2 again, where we have a turbine 𝑇𝑇0 generating a wake region. We can write the equation of the wake radius at distance 𝑥𝑥 from turbine 𝑖𝑖 as: (1) 𝑟𝑟𝑥𝑥 = 𝑟𝑟0 + 𝛼𝛼𝛼𝛼

Where 𝑟𝑟0 is the rotor radius of turbine 𝑇𝑇0 , and 𝛼𝛼 is a decay factor expressing how fast the wake breaks down. An analytical equation is given for 𝛼𝛼 concerning the height 𝑧𝑧 of the turbine generating the wake and the constant surface roughness 𝑧𝑧0 , which depends on the terrain characteristics in the form of 𝛼𝛼 =

0.5

ln�

(2)

𝑧𝑧 � 𝑧𝑧0

If we consider 𝑣𝑣0 as the ambient wind speed and 𝑣𝑣1 as the wind speed at distance 𝑥𝑥, then we can write a balance of mass equation as (3) 𝜋𝜋𝑟𝑟02 𝑢𝑢 + 𝜋𝜋(𝑟𝑟𝑥𝑥2 − 𝑟𝑟02 )𝑣𝑣0 = 𝜋𝜋𝑟𝑟𝑥𝑥2 𝑣𝑣1 , where 𝑢𝑢 is the wind speed just behind turbine 𝑇𝑇0 .

Figure 2: Single Wind Turbine Wake Model A study of the aerodynamics of wind turbines provides us with an analytical equation connecting the ambient speed with the wind speed right behind the turbine. From this we can write 𝑢𝑢 = �1 − 𝐶𝐶𝑇𝑇 𝑣𝑣0 ,

Where the term 𝐶𝐶𝑇𝑇 is the thrust coefficient of the turbine. Solving the previous equation for 𝑣𝑣1 , we obtain 𝑟𝑟 2

𝑣𝑣1 = 𝑣𝑣0 �1 − �−1�1 − 𝐶𝐶𝑇𝑇 � 02� .

(4)

𝑟𝑟𝑥𝑥

The previous equation expresses the wind speed at distance 𝑥𝑥 behind a turbine 𝑇𝑇0 when the radius of the wake at

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

that distance is 𝑟𝑟𝑥𝑥 . In case of a wind farm, where we have two turbines, 𝑖𝑖 and 𝑗𝑗, and Turbine 𝑗𝑗 is in the wake region of Turbine 𝑖𝑖, then the wind speed affecting Turbine 𝑗𝑗 is given by 𝑟𝑟 2

(5)

𝑣𝑣𝑗𝑗 = 𝑣𝑣0 �1 − �1 − �1 − 𝐶𝐶𝑇𝑇 � 𝑖𝑖2�. 𝑟𝑟𝑗𝑗

From Equation 5 we can see how the speed received by Turbine 𝑗𝑗 is reduced due to the wake effect of Turbine 𝑖𝑖. Therefore, the power output of Turbine j will be reduced too. Equation 5 gives the resulting wind speed at a given turbine when we considered that the turbine is totally covered and affected by the wake effects. Which means that the wind speed is the same over all the turbine surface. But in wind farm we can find partial shadowing which means the turbine surface is partially in the wake area and not totally. Therefore, the turbine will be affected by different wind speed at its rotor’s sweep area and the calculation of the power output needs more attention. To this end, the speed from the wake effects affecting the turbine can be rewritten to take the partial shadowing into account: 𝑣𝑣𝑥𝑥 = 𝑣𝑣0 �1 + ��1 − 𝐶𝐶𝑇𝑇 − 1�

𝑟𝑟𝑖𝑖2 𝐴𝐴𝑗𝑗𝑗𝑗ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑟𝑟𝑗𝑗2

𝐴𝐴𝑗𝑗

(6)

�,

where 𝐴𝐴𝑗𝑗 is the total area of the turbine and 𝐴𝐴𝑠𝑠ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 is the shadowed area by the affecting turbine. Here we must note that there are different shadowing possibilities between the two turbines: complete shadowing, quasicomplete, partial shadowing and no shadowing. Therefore before applying the formula we must determine the type of shadowing depending on the horizontal and vertical distances between the two turbines and the direction of the wind. This study considers these detailed geometric calculations, whose details are not presented here.

Figure 3: Bird’s Eye View Illustration of Multiple Wakes within a Wind Farm In a wind farm each turbine will produce a wake area, as shown in Figure 3. This will cause a turbine to possibly be under the effect of multiple wakes caused by different turbines. Therefore, a method to combine the different single wakes effects is required. One of the used methods is the sum of squares of velocity deficit, and this method is useful here since it provides a formula for the deficit wind speed as 𝛿𝛿𝑣𝑣𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥 = 1 −

𝑣𝑣𝑥𝑥 𝑣𝑣0

= �1 − �1 − 𝐶𝐶𝑇𝑇 �

𝑟𝑟𝑖𝑖2 𝐴𝐴𝑗𝑗𝑗𝑗ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑟𝑟𝑗𝑗2

𝐴𝐴𝑗𝑗

(7)

And combining the multiple wakes effects, we obtain

𝛿𝛿𝑣𝑣𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥 = �∑𝑗𝑗∈𝑈𝑈𝑖𝑖�𝛿𝛿𝑣𝑣𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥 � ,

(8)

Where 𝑈𝑈𝑖𝑖 is the collection of turbines affecting turbine 𝑖𝑖 by its wake. Then the total speed at Turbine 𝑖𝑖 will be given by (9) 𝑣𝑣𝑖𝑖 = 𝑣𝑣0 �1 − 𝛿𝛿𝑣𝑣𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥𝑥 �.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Combining our calculations, we conclude that the wind speed affecting Turbine 𝑗𝑗 is given by 𝑣𝑣𝑖𝑖 = 𝑣𝑣0 �1 − �∑𝑗𝑗∈𝑈𝑈𝑖𝑖 ��1 − �1 − 𝐶𝐶𝑇𝑇 �

𝑟𝑟𝑖𝑖2 𝐴𝐴𝑗𝑗𝑗𝑗ℎ𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎𝑎 𝑟𝑟𝑗𝑗2

𝐴𝐴𝑗𝑗

� �.

(10)

Wind Farm Layout Optimization Model

We consider a rectangular grid of a certain number of turbines where they are separated by a distance of 𝑥𝑥 on the horizontal dimension and a distance of 𝑦𝑦 on the vertical dimension (see Figure 4). For this study, we simply take the horizontal axis to be the exact East-West direction and the vertical axis to the North-South direction. While it is possible to choose the axes in a arguably more efficient fashion (e.g., setting the vertical axis to be the speedweighted mean of directions), or even defining an additional optimization variable for the “tilt” angle of the rectangular layout, these generalizations are left as a topic for further study. The objective is to find the most optimal values of 𝑥𝑥 and 𝑦𝑦 with respect to the objective of maximizing the economic return per grid area as explained previously. As also shown in Figure 4, the stochastic wind blows from the direction 𝜃𝜃 and with a speed of 𝑣𝑣(𝜃𝜃) (i..e whose distribution depends on 𝜃𝜃) at a prespecified height equal to the turbines’ hubs. We also assume here that the wind’s direction and magnitude applied uniformly all over the wind farm in the vertical and horizontal planes. In other words, each turbine is assumed to receive a uniform wind direction and speed over its entire rotor speed area and all turbines receive the same direction and speed regardless of their position on the grid. In reality wind’s speed usually increases with height (i.e. upper sections of the turbines sweep area will receive higher speeds) and, due to the considerably high volatile nature of wind, distant locations in a given wind farm can be facing significantly different speeds (and perhaps directions) as wind travels throughout the farm’s considerable large area. Nevertheless, the uniformity assumption that we make are not just for mathematical simplicity, but, rather, because of the limited time- and height- resolution of typical wind data and also since the errors should partly cancel each other out. In our model, the turbines still receive different speeds due to the complex wake effects occurring within the wind farm. Given the “adjusted” speed for each turbine, we use the turbine’s specific “power curve” to estimate its power output of the turbine.

Figure 4: Wind Farm Grid Layout The power curve of a wind turbine describes the turbine’s generated power versus the wind speed, p = f(v), when the wind is perpendicular to the rotational plane. When the speed of wind reaches a threshold of so-called “cutin” value, the turbine starts generating power, and as the speed increases the power generated increases nonlinearly to its maximum value at the so-called “saturation” speed (or, rated speed). Due to structural stability and other concerns, the turbine is regulated to generate a steady maximum (or, rated) output between the saturation speed and the so-called “cut-out” speed, beyond which the blades are stopped as operating the turbine at such “storm” winds poses mechanical and safety hazards. Based on the theory laid out in Betz’s Law and also empirical evidence, the increasing portion of the power curve between the cut-in and saturation speeds is often modeled as a cubic function of the form 𝑐𝑐𝑣𝑣 3 , where c is a constant depending on the turbine’s mechanical properties and of the air ﬂow (density, temperature, etc.) and v is the wind speed. In reality, however, the increasing portion begins deviating from the cubic behavior and curls rather smoothly the maximum plateau level. The shape of the increasing section of the power curve is suitably described by a (approximately cubic) convex lower section and a concave and flattening upper-section (see Figure 5).

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

For this study, we considered several different models of Vestas turbines and we approximated their power curves by ﬁtting suitable polynomials to their discrete speed-power data that was available from the manufacturer. Twopiece polynomials were fitted (using Mathematica 9) to the increasing portion of the data –one to the convex lower half and a another to the concave upper half. The polynomials vary between 4 and 8 in degrees in order to yield an exact ﬁt to the data points and some degree of subjectivism was involved in identifying the two separate portions in order to obtain the best ﬁt. The power curve fitting results for 8 Vestas models of various rated powers and other parameters are shown in Figure 5. For our case study, chose to use the Vestas V90 with 3MW maximum power output.

Figure 5: Estimated Power Curves for Several Vestas Wind Turbine Models For a better realism of the optimization process, we consider stochastic wind speeds and directions. Wind speed and direction measurements collected by the YEGM (Turkish Renewable Energy General Directorate) at various sites Turkey during 2003-2011 were obtained. The wind data is comprised of 10-minute apart measurements of speed and direction at 30 meters height. The total length of the measurement period (i.e. number of years) vary between sites, and the data had to be cleaned rigorously for missing or obviously erroneous values. To assure continuity of the used data, two suitable sites were chosen to be used in the case study. One is the Biga site, which is near the same-named town of the Çanakkale province in the Marmara region of Turkey. The second is the Tavas site, which is a town of the Denizli Province in the inner Aegean region of Turkey. The Biga site was selected partly because it is one of the strongest-wind sites in Turkey. The Tavas site, on the other hand, can be classified as a moderate-to-low wind site. This choice of this particular site pair is motivated by comparison purposes. We use one year’s length of data for each site, yielding a total of 52704 (10 minutes data over 366 days) speeddirection data points for each site. Moreover, since the data were taken at a 30 meters height, they were converted to a more meaningful hub height of 100 meters using the standard Equation 13 given below. In Equation 13, 𝛼𝛼 is an empirically derived coefficient that varies dependent upon the stability of the atmosphere. For this study, we adopt the standard value of 0.143 for this parameter. v100 = 𝑣𝑣30 �

100 𝛼𝛼 � 30

(11)

Figure 6 below presents a more detailed picture of the two sites’ wind properties in the form of a modified “wind rose”, i.e. a joint circular histogram of the distribution of the wind direction and average wind speed at that site. The histograms consist of 5° bins for the direction, where the color of the bin corresponds to the average speed of the wind in that direction according to the given color scale.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Figure 6: Annual Wind Roses for the Biga and Tavas Sites It is once apparent that the Biga site is considerably windier than the Tavas site and that the wind at both sites blow predominantly in the Northeast direction. Other observations from Figure include that for both locations wind almost never blows in the South-East quadrant of the direction coordinates (and for Tavas, also almost never in the North-West quadrant) and that there’s very moderate wind occurrence in the opposite direction of the predominant North-East. These other details of the presented Wind roses could be explained better in a moreinformed context of the geography and topography of the sites. To verify the commonly encountered Weibull behavior of the wind speeds, a fitting process of the Weibull distribution over the whole data was performed. Based on Figure 7, we can justify that the wind speed data agrees strongly with a Weibull distribution, and, therefore, we can safely perform random variate generation using the estimated Weibull parameters in order to use for our simulated samples.

Figure 7: Weibull Distribution Fit for the Biga Site Wind Speed Data Based on the analyses of Wiser and Bolinger (2013). for the wind farm development project carried out in the USA in 2012, the average capital cost for projects with more than 5 MW output is around 1,900 $ /kW. In our case, we take the cost to be around 2,000 $ /KW. In addition to this initial capital cost, we also take the O&M costs into account. Because O&M costs are realized over time, they are customarily reported in dollars per energy produced, i.e. $/MWh. To get an estimate for this rather elusive cost type, we again refer the issues of the D.O.E. reports (Wiser and Bolinger, 2012, 2013), which give the example of two wind farm operators reporting that the total operating costs are around 20 $/MWh. Assuming an economic life span of the turbines of 20 years and using a discount rate of 6%, the present worth of the total O&M costs can be calculated as 𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐 = ($2,000,000/𝑀𝑀𝑀𝑀) ∗ 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 (𝑀𝑀𝑀𝑀) + (11.469𝑦𝑦𝑦𝑦) ∗ ($20/𝑀𝑀𝑀𝑀ℎ) ∗ (8,784ℎ𝑟𝑟/𝑦𝑦𝑦𝑦 ∗ 𝑚𝑚(𝑀𝑀𝑀𝑀)),

(14)

where 𝑃𝑃𝑚𝑚𝑚𝑚𝑚𝑚 is the total power output and 𝑚𝑚 is the mean power production.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Optimization Scenarios and Results Performing the optimization using the whole data set (52704 point) for every x-y pair of spacing of the turbines require a huge amount of calculation time. Therefore, make use of a hybrid approach of Monte-Carlo sampling of wind directions combined with a Weibull random-variate generation of the speed from the corresponding Weibull fit. We provide an itemized description of our approach for simulating one direction-speed pair as follows: • We divide the wind direction into 360 discrete categories (i.e. slices) in increments of 1° Celsius and calculate the frequency of the direction data in each category. • For each category we fit a Weibull distribution to the wind speed data corresponding to the direction data in that slice. • For each sample to be simulated, we first generate a random discrete direction via Monte-Carlo simulation using the categories’ relative frequencies (we take the simulated direction to be the middle of that slice, i.e. 0.5°, 1.5°,…., 359.5°). • Next we generate a wind-speed random variate using the estimated Weibull distribution for that slice. • Based on the simulated direction and speed, we calculate the wake effects and the effective speed facing each individual turbine on the grid. • We next calculate each turbine’s output using the power curve and average the results over all the objective function. • We finally obtain the value of the objective function by dividing the average output by the total cost of a turbine and by the total area of the grid. The histogram in Figure 8, which is actually a “flattened-out” wind rose, shows the distribution of the wind direction and the average wind speed for the 360 1-degree slices for the Biga site. Each slice has its individual estimated Weibull speed distribution, which are not presented here.

Figure 8: distribution of wind speed over direction Due to the complex nature of the model, mostly because of the incorporation of the wake effects, a differential calculus formulation of the objective function is not trivial. Due to the low variable-dimensionality of the problem and the considerable computational effort required to perform the calculations for each single simulated sample, the Nelder-Mead heuristic method presents itself as a suitable algorithm. With two decision variables (x and y) the method requires only 3 points in the x-y plane to implement. For our stochastic model, however, the objective function value for each Nelder-Mead point needs to be defined as the average of many simulations. Deciding that 1000 simulations for each point is a reasonable compromise between representative accuracy and computational power, we implemented the algorithm with the stopping criterion that the objective function value has not improved by more than 0.01% for the past 100 iterations. In this study, we present the results for two square cases of the grid, a 10x10 grid and a 30x30 grid. The optimization results for the Biga and Tavas sites are presented in, respectively, Table 1 and Table 2 below.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Table 1: Case Study Results for the Biga Site (Vestas V90 – 90 m Rotor Diameter). Number of turbines in x-axis 10 30

Number of turbines in y-axis 10 30

Spacing distance in the x-axis 584.3 m 590.6 m

Spacing distance in the y-axis 211.4 m 221.9 m

From Table 1, we see that the x-spacing is considerable larger than the y-spacing for both grid sizes (the ratio x/y is 2.77 for the 10x10 grid and 2.67 for the 30x30 grid). This result is consistent with the predominant wind directions at the Biga site as evident from Figures 6 and 7 –winds are stronger and more frequent in the EastNorth-East direction (first octant of the compass). A larger x-spacing would thus help reduce wake effects more than a larger y-spacing. Given 90-meter rotor diameter of the Vestas V90 – 3MW turbine, this indicates a spacing of 6.49 diameters along the x-axis and a spacing of 2.34 diameters along the y-axis for the 10x10 grid and very close values for the 30x30 grid. These numbers also compare reasonably with the industry rule of thumb of spacing the turbines by 2 diameters in the perpendicular line to the prevailing wind direction and by 7 diameters along this direction (note that our grids is not oriented to face the “prevalent” wind direction perpendicularly, but are in the fixed East-West and North-South axes). The result that turbines are slightly more packed in a smaller grid can be explained by the “boundary” effect as follows: Turbines at certain borders of the grid will be impacted much less (the East and South borders in the case of Biga) from the wake effects as compared to turbines in the interior and on the opposite borders. The fact that the “border” turbines constitute a greater fraction of the total number of turbines in a smaller grid as compared to a larger grid will compensate better for internal wake losses, and, thus, it may allow tighter optimal packing of the whole turbine set. Results for the Tavas site in Table 2 are quite close to the values for the Biga site and bear largely similar interpretations. However, comparing the two sites is somewhat confusing. Both the x- and y-spacing for the Tavas 10x10 grid is slightly larger than their counterparts for the Biga site, possibly owing to the fact that the prevailing winds at the Tavas site are slightly more in the North direction as compared to Biga. However, both of spacing values are for the 30x30 grid are slightly less at the Tavas site than at Biga, thus, running counterintuitive to the previous explanation. It is also noteworthy that while the y-spacing at Tavas increases slightly from the 10x10 to the 30x30 grid, as consistent with the “border turbines” effect, the x-spacing decreases slightly. Given the complex nature of the model and the non-deterministic behavior of a heuristic algorithm, we caution from over-comparing the two sites and postulating unsupported conclusions. Table 2: Case Study Results for the Tavas Site (Vestas V90 – 90 m Rotor Diameter) Number of turbines in x-axis 10 30

Number of turbines in y-axis 10 30

Spacing distance in the x-axis 591.5 m 583.4 m

Spacing distance in the y-axis 215.6 m 218.8 m

Conclusion and Future Research The general problem of the optimal micro-siting of wind turbines a complicated one due to several reasons. This paper aims to develop a mathematical for a special case of the model where turbines of the same type are placed at uniformly spaced grid points over a flat wind farm area in the form of a rectangle whose edges sit on the EastWest and North-South directions. The model’s main focus is to incorporate the complex wake effects that may occur within a wind farm that critically impact the total output. This paper also differs from other studies that consider only the total output in that we consider an economic objective, the return on investment per area required. The wake effects are modeled in an idealized –yet mathematically quite complicated– formulation a la Jensen (Katic et al., 1986). The consideration of stochastic wind speeds and directions, and the use in the case study of real measurement data from 2 sites in Turkey, is yet another aspect of the study that increases its realism. The numerical and analytical complexities introduced by the incorporation of wake effects and the use of real data are dealt with by adopting a hybrid approach of using the heuristic Nelder-Mead optimization method based on Monte-Carlo simulation of random samples of wind direction-speed pairs. Despite the fixed orientation of the grid, the model’s result seem to validate the industry’ simplistic rule of thumb practice of spacing the turbines by 2 and 7 rotor diameters along, respectively, the perpendicular and parallel direction of the “prevalent” winds. Introducing a third decision variable corresponding to the “tilt” angle of the rectangular grid, which would be optimized together with the horizontal and vertical turbine spacings, could yield even-more confirming results, but this is the subject of future generalizations of the model. The case study results

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

also indicate –albeit with an exception– that a larger wind farm (i.e. having more turbines) requires larger spacing between the turbines in order to operate optimally, possibly owing to the fact that a greater portion of the turbines in a larger wind farm are in the farm’s interior region where the wake effects are stronger than on the boundaries. Nevertheless, we offer these interpretations with some degree of caution as the aforementioned observations and possible other ones ask for a more generalized model as well as many more site cases to be analyzed, which again, are among planned extensions of the present study.

Acknowledgements

This study is part of a larger research project funded by TÜBİTAK (Turkish Scientific and Technological Research Council) under Grant No. 113M503. The authors would also like to thank Istanbul Kemerburgaz University for its logistical support including the provision of the facilities where the project is carried out.

References Attias, K. (2011). Optimal Layout for Wind Turbine Farms, Ben-Gurion University. Donovan, S. (2005). 40th Annual Conference: Wind Farm Optimization, Operational Research Society of New Zealand, Wellington, New Zealand. Katic, I., Hojstrup, J., & Jensen N. O. (1986). European wind energy conference and exhibition: A simple model for cluster efficiency. Rome, p. 407-10. Kulunk E. (2011). Fundamental and Advanced Topics in Wind Power: Aerodynamics of Wind Turbines, ISBN: 978-953-307-508-2. Kusiak A., (2010). Design of wind farm layout for maximum wind energy capture, The University of Iowa. Liu, F. and Wang, Z. (2014). Offshore Wind Farm Layout Optimization Using Adapted Genetic Algorithm: A Different Perspective, Virginia Commonwealth university. Renkema, D. J. (2007). Validation of wind turbine wake models using wind farm data and wind tunnel measurements, Delft University of Technology. Samorani, M. (2013) The Wind Farm Layout Optimization Problem, University of Alberta. Schlez, W. New Developments in Precision Wind Farm Modelling, Garrad Hassan and Partners Ltd. Tao, H. (2011). The Assessment of Dynamic Wake Effects on Loading, Delft University of Technology, The Sailhawt Website, https://sites.google.com/site/sailhawt The Windpower Program Website, http://www.wind-power-program.com/turbine_characteristics.htm Zhang, X. & Wang, W. (2009). Wind Farm and Wake Effect Modeling for Simulation of a Studied Power System, Xi’an Jiaotong University.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

AN INQUARY STUDY OF NiO FILMS DEPOSITED WITH SOL-GEL SPIN COATING Guven Turgut Erzurum Technical University, Erzurum/Turkey [emailprotected] Abstract: Y incorporated nickel oxide thin films have been deposited via sol-gel route by using spin coater. The structural, morphological and optical properties of films have been inquired. The films had nano-sized polycrystalline cubic structure. The optical characterizations indicated that the band gap and Urbach energy continuously went up with Y contribution. The present investigation reveals the properties of nickel oxide are healed and controlled with Y contribution and these films are probably good candidates for various applications. Keywords: NiO, Sol-gel, Y doping

Introduction Nickel oxide (NiO), which is one of transition metal oxides, is an intriguing material for magnetic and electrochromic applications, organic light emitting diodes, catalysts, smart windows and solar cells (Reguig et al., 2007; Turgut et al., 2015) antiferromagnetic feature, good optical transmittance with a wide band gap of 3.6-4.0 eV, thermal and chemical stableness (Patil et al., 2002; Gowthami et al., 2014). A stoichiometric NiO is a good insulator with resistivity of about 1013 ohm.cm, but non-stoichiometry, which is resulted from nickel vacancies and oxygen interstitials, makes it a p-type semiconductor material (Joshi et. al., 2006). The structural, magnetic, optical, electrical, morphological and electrochromic properties of NiO material can be improved by doping with foreign elements such as Li, Cu, K, P, Y and other elements. To the our the best knowledge, there is no report about investigation of Y doping effect on characteristic features of NiO. Therefore, it is necessary to make various studies in order to understand detailed influence of Y doping on properties of NiO. Hence, in present study, pure and Y doped NiO films have been fabricated via a sol-gel spin coating technique which is simple, safe, cheap and capable of fabricating hom*ogenous and good quality films (Turgut et al., 2015) and Y doping effect on crystallographic, morphological and optical properties of NiO films have been investigated.

Material and Method Pure (YNO-0), 1 at. % (YNO-1), 2 at. % (YNO-2), 3 at. % (YNO-3), 5 at. % (YNO-5) and 7 at. % (YNO-7) yttrium doped NiO thin films were prepared with spin coating by using nickel(II) acetate tetrahydrate, yttrium(III) chloride, methanol and monoethanolamine. After the solutions were prepared, they were aged for three months and finally solutions were dribbled onto micro slide glasses that were spun at velocity of 3500 rpm for 25 seconds by spin coater. The pre-grown layers were sintered at 200 oC for 10 min. This procedure was reiterated for ten times. Lastly, the whole samples were annealed at 550 oC for 1 hour in air. The crystallographic, surface and optical features of pure and Y doped films were inquired by a Rigaku/Smart Lab x-ray diffractometer (XRD), Nanomagnetic Instruments AFM (atomic force microscope), Perkin Elmer UV/VIS Lambda 35 spectrophotometer.

Results and Discussions The XRD patterns in Fig.1 reveal all films have polycrystalline cubic bunsenite NiO structure (JCPDS card no: 47–1049) with (200) preferential orientation. Sta et al. (2014) observed the similar crystallographic structure for Li doped NiO films prepared by sol-gel spin coating. As seen from Fig.1, the peaks’ intensities gradually decrease with Y contribution, which suggests Y doping causes a deteriorating in the crystallinity of NiO. The textured coefficient (TC) is calculated by Eq. (1) TC(hkl) =

1 N

I(hkl) /I0

(1)

( ) ∑ I(hkl) /I0

where I hkl , I 0(hkl) and N are the intensities of relative and standard peak, the number of peaks.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Figure 1. The XRD patterns of YNO thin films

Figure 2. The variation of TC values of NiO with Y doping The alteration of TC values with Y doping (Fig. 2) indicates that the TC values of some peaks are greater than one and this remarks the more crystallites than one are directed at those orientations (Duman et al., 2014). The mean crystallite size (D) and micro-strain values for samples are identified by Eqs. (2) and (3) 𝐷𝐷 =

0.9𝜆𝜆

(𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽)

𝜀𝜀 = �

𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠

(2) 𝜆𝜆

� �� − (𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽𝛽)� 𝐷𝐷

(3)

where 𝛽𝛽 is full width at half of the peak maximum (FWHM) and it is computed by β2 = β2obs − β2inst (βobs is measured from XRD, βinst is instrumental broadening). The mean crystallite size value for undoped sample continuously decreases from 21.32 nm to 20.80 nm, 17.77 nm, 17.06 nm, 15.76 nm and 13.84 nm for YNO-1, NYO-2, NYO-3, NYO-5 and NYO-7 samples. However, the micro-strain value gradually increases from 1.96x103 , to the values of 2.01x10-3, 2.35x10-3, 2.64x10-3, 2.98x10-3 and 3.04x10-3 with Y doping. These results clearly indicate that the crystallinity of films decreases with Y doping ratio. The decreasing crystallinity of NiO with Y doping can be resulted from higher ionic radius of Y3+ (0.90 Å) than Ni2+ (0.69 Å) (Greenwood and Earnshaw, 1997). The difference between ionic radii can introduce a lattice distortion and an increase in amount of disorders. The 2D and 3D AFM micrographs (Fig. 3) show there are nano-sized particles on the surfaces of films and their distribution is nearly hom*ogenous. These images also exhibit the particle size incessantly decreases with raising Y doping ratio. This alteration is very good consistent with variation of crystallite size determined from XRD analysis. The surface roughness value of NiO film initially decreases from 1.23 nm to 0.98 nm, 0.94 nm and 0.62 nm with Y doping up to 3 at. % content and then it starts to increase to the values of 0.96 nm and 1.05 nm with more Y doping.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

YNO-0

YNO-1

YNO-2

YNO-3

YNO-5

YNO-7

Figure 3. The AFM images of YNO thin films

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

The transmittance and (αhν)2 vs. hν graphs are given in Fig. 4a-b. The transmittance values are about 80 %-96 % over wavelength range of 450-1100 nm. The value of optical band gap (Eg ) is calculated by Eqs. (4), (5) α = ln (1/T)/d

(4)

αhν = K(hν − Eg )1/2

(5)

where α, T, d and K are absorption coefficient, transmittance, film thickness and constant. As seen from (αhν)2 vs. hν graphs (Fig. 4b), the E g value for pure NiO film continuously increases from 3.274 eV to 3.809 eV with Y doping. An increasing in E g with Y doping is probably resulted from decreasing crystallite size, which is clearly seen from XRD and AFM analysis

(a)

(b)

Figure 4. (a) The optical transmittance (b) (αhν)2 vs. (hν) curves of YNO films

This is known to be quantum confinement effect and band gap of a semiconductor is reversely proportional to crystallite size (Lin et al., 2005). The similar result was also observed for P doped NiO films (Das et al., 2010). The absorption tail is also inquired for deposited films. The absorption edge is called as the Urbach tail (Urbach, 1953) and it is given by Eq. (6) α(E, T) = α0 exp �

E−E0

Eu (T,X)

�

(6)

The Urbach energy (E u ) is related to temperature (T) and crystal disorders (X). However, Cody et al. (1981) indicated non-thermal component depending on the structural disorders. Hereby, E u values are determined by Eq. (7) ln α = E

∙ �ln(α0 ) +

�

(7)

E u =Δ(hν)/Δ(lnα) and it is based only on the degree of structural disorders. From Fig.5, the Urbach energy value of pure NiO gradually increases with increasing Y doping ratio. This suggests Y incorporation into NiO causes a significant raise in the E u values as a result of increasing structural disorders, which is seen from XRD analysis. An interesting point for this study is that E g and E u have a similar variation with Y doping ratio. The transitions from tail and band to tail because of expanding of Urbach tail would cause a decrease in E g . But as indicated before, decreasing crystallite size is dominant factor for increasing optical band gap. A similar tendency between E g and E u was found for ZnO semiconductor material (Demirselcuk and Bilgin, 2013).

Conclusion In this investigation, Y doped NiO samples were synthesized for the first time by sol-gel spin coating. The crystallographic and topographic investigations showed the samples were made of nano-sized particles with bunsenite NiO structure. The crystallinity and crystallite size decreased with Y content, however the values of micro-strain, optical band gap and Urbach energy increased with Y contribution level. It can be concluded the features of NiO films may be altered with Y doping.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Figure 5. The curves of lnα vs. photon energy

References Cody, G.D., Tiedje, T., Abeles, B., Brooks, B., Goldstein, Y. (1981). Disorder and the Optical-Absorption Edge of Hydrogenated Amorphous Silicon. Physics Review Letters vol. 47 pp. 1480-1483. Das, N.S., Saha, B., Thapa, R., Das, G.C., Chattopadhyay, K.K. (2010). Band gap widening of nanocrystalline nickel oxide thin films via phosphorus doping. Physica E vol. 42 pp. 1377-1382. Demirselcuk, B. and Bilgin, V. (2013). Ultrasonically sprayed ZnO:Co thin films: Growth and characterization. Applied Surface Science vol. 273 pp.478-483. Duman, S., Turgut, G., Özçelik, F.Ş., Gürbulak, B. (2014). The synthesis and characterization of sol–gel spin coated CdO thin films: As a function of solution molarity. Materials Letters vol. 126 pp. 232-235. Gowthami, V., Perumal, P., Sivakumar, R., Sanjeeviraja, C. (2014). Structural and optical studies on nickel oxide thin film prepared by nebulizer spray technique. Physica B vol. 452 pp.1-6. Greenwood, N.N., Earnshaw, A. (1997). Chemistry of the Elements. 2nd ed. Oxford: Reed Edu Prof Publishing. Joshi, U.S., Matsumoto, I., Itaka, K., Sumiya, M., Koinuma, H. (2006). Combinatorial synthesis of Li-doped NiO thin films and their transparent conducting properties. Applied Surface Science vol. 252 pp. 2524-2528. Lin, K.-F., Cheng, H.-M., Hsu, H.-C., Lin, L.-J., Hsien, W.-F. (2005). Band gap variation of size-controlled ZnO quantum dots synthesized by sol–gel method. Chemical Physics Letters vol. 409 pp. 208-211. Patil, P.S., Kadam, L.D. (2002). Preparation and characterization of spray pyrolyzed nickel oxide (NiO) thin films. Applied Surface Science vol. 199 pp. 211-221. Reguig, B.A., Khelil, A., Cattin, L., Morsli, M., Bernède, J.C. (2007). Properties of NiO thin films deposited by intermittent spray pyrolysis process. Applied Surface Science vol. 253 pp. 4330-4334.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Sta, I., Jlassi, M., Hajji, M., Ezzaouia, H. (2014). Structural, optical and electrical properties of undoped and Lidoped NiO thin films prepared by sol–gel spin coating method. Thin Solid Films vol. 555 pp.131-137. Turgut, G., Sonmez, E., Duman, S. (2005) Determination of certain sol-gel growth parameters of nickel oxide films. Ceramics International vol. 41 pp. 2976-2989. Urbach, F. (1953). The Long-Wavelength Edge of Photographic Sensitivity and of the Electronic Absorption of Solids. Physics Review vol. 92 pp.1324.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

COMPARATIVE ANALYSIS OF CURRENT METHODS IN SEARCHING OPEN EDUCATION CONTENT REPOSITORIES Benneaser John SVP, Engineering – AppNexus Inc., New York, USA Research Scholar, Karunya University, Coimbatore, India [emailprotected] V. Thavavel College of Computer and Information Sciences, Prince Sultan University, Saudi Arabia [emailprotected] Jayakumar Jayaraj Department of Electrical and Electronics Karunya University, Coimbatore, India [emailprotected] A. Muthukumar Department of Computer Science & Engineering PSG Institute of Technology and Applied Research Neelambur, Coimbatore, India [emailprotected] Poornaselvan Kittu Jeevanandam Program Manager, Tata Consultancy Services Edison, NJ, USA [emailprotected] Abstract: The current generation of learners are living in a eConnected society where the technology and content are open. Open learning enables learners to be self-determined and interest-guided. To make online learning successful, it is critical that learners need effective ways of finding the appropriate learning resources. However, due to the generally unstructured nature and overwhelming quantity of learning content, effective learning remains challenging. This study compares different features offered by the Open learning content search platforms, and analyzes the past one-year website usage metrics data to gather insights about the usage. This paper also discusses the gaps in the current Open Content search dilemma and proposes potential solutions. Keywords: Open Educational Resources(OER), Content search, Open Learning, Massive Open Online Courses(MOOC), Distance Education, Learner Centered Learning

Introduction Computer-aided instruction (CAI) has evolved from its humble origins, to the level of Massive Open Online Course (MOOC), which was introduced in 2008 as open online courses aimed at unlimited participation. The Internet and the entire World Wide Web (WWW) constitute the largest and most comprehensive knowledge base in the history of the world. Learners are living through this information explosion (Chiou & Shih, 2015). Currently, e-learning is not simply providing the course materials, while the trend of Massive Open Online Courses (MOOCs) and the concept of flipped classrooms (Goodwin & Miller, 2013) is well applied everywhere. Rai & Chunrao (2015) states that “In recent years, Massive Open Online Courses (MOOCs) have attracted millions of learners around the world, through various MOOC providers, such as edX, Coursera, and Udacity. MOOCs allow millions of learners to enroll in courses form reputed universities such as Harvard University, Stanford University, Massachusetts Institute of Technology (MIT), and University of California at Berkeley (UCB). Outside of MOOCs, professors are creating and releasing their own content using tools such as Slideshare and YouTube. Every day, millions of learners make use of free, open online tools, and resources (MacDonald, 2015). Open content learning resources such as MIT’s OpenCourseWare project (OCW), TED videos, Khan Academy, YouTube videos, and the MERLOT (Malloy & Hanley 2001; Hanley 2015) project are a few examples of systems through which millions of learners learn on the web every day. Open learning enables learners to be self-determined and interest-guided. Stacey (2013) educators to “Go beyond open enrollments and use open pedagogies that leverage the entire web not just the specific content in the MOOC platform”. Learners are often unable identify which material is needed, useful, and required at their level. Hence, open content learning design must assimilate the material from various sources and provide a new pedagogy that is appropriate to the needs of today’s learners (Smyth, Bossu & Stagg, 2015). This paper explains the design for an Open Content Social Learning (OCSL) system that leverages Open Content to deliver an adaptive and personalized experience accounting for the pedagogical needs of the learners and similar learners

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

and the need to recommend learning activities in a pedagogically effective order. A great majority of these Open Educational Resources (OER) initiatives are based on established web based technology platforms and have accumulated large volumes of quality resources which are shared openly. However, one limitation inhibiting the wider adoption of OER is the current inability to effectively search for academically useful OER from a diversity of sources (Yergler, 2010). While the open content grows in popularity, and the proliferation of repositories and portals for open content, it becomes more difficult for potential users to find the content they need (Dichev & Dicheva, 2012). Learners are often unable to identify which material is needed, useful, and required at their level. Hence, open content learning design must assimilate the material from various sources and provide a new pedagogy that is appropriate to the needs of today’s learners (Smyth, Bossu & Stagg, 2015). In this paper, we present our analysis of the current Open learning content platforms. We reviewed the research papers relevant to OER search platforms, studied the survey results from learners and teachers, collected & analyzed the metrics from Alexa.com for the top eight OER search platforms, and then compared the features of the top eight OER search platforms. After the study and analysis, we proposed some potential solutions to overcome the OER adoption due to the massive and diversified volume of content by providing effective personalized search.

History of OER, OCW and MOOC The term Learning Object, was first popularized by Wayne Hodgins in 1994 when he named the CedMA working group "Learning Architectures, APIs and Learning Objects", which has become the Holy Grail of content creation and aggregation in the computer-mediated learning field (Polsani, 2006). In 1998, David Wiley coined the term “open content”, to which the principles of the open source free software can be applied to content. (Caswell, Henson, Jensen, & Wiley, 2008). This movement helped the creation of the first widely adopted open license for content called “Open Publication License” (Wiley & Gurrell, 2009). In 2001 Larry Lessig and others founded the Creative Commons (Commons, 2009) and released a flexible set of licenses which improved Open Publication License’s confusing license option structure. One role of Creative Commons, in the history of OER, is to increase the credibility and confidence in their legally superior, much easier to use licenses brought to the open content community. Also in 2001, MIT announced its OpenCourseWare(OCW) initiative to publish nearly every university course for free public access for non-commercial use (West & Victor, 2011). MIT’s OpenCourseWare played a critical role in the history of OER with its brand and commitment (Yuan, MacNeill & Kraan, 2008). Since first being coined by UNESCO in 2002, the term Open Educational Resources has evolved to meet the fast pace of the changing and diverse contexts in which it has now been used (Bossu, Bull, & Brown, 2012). Open Educational Resources (OERs) are teaching and learning materials that anyone can use and share freely, without charge. The worldwide OER movement is rooted in the idea of high quality education at no cost. The Cape Town Declaration (2007) states that “Educators worldwide are developing a vast pool of educational resources on the Internet, open and free for all to use. These educators are creating a world where each and every person on earth can access and contribute to the sum of all human knowledge. They are also planting the seeds of a new pedagogy where educators and learners create, shape and evolve knowledge together, deepening their skills and understanding as they go.” The term MOOC was developed in 2008 by Dave Cormier and Bryan Alexander to describe a course experiment utilizing connectivism (Moe, 2015). In summary, OER initiatives have resulted in the development of open content and open courses in higher education. OCW is a free and open digital publication of university-level educational materials. MOOCs are free online courses without formal entry requirement nor participation limit. OERs, OCW and MOOC are closely related to the Openness movement in education promoting the ideas of how people should produce, share, and build education.

Related Research One of the challenges facing open learning is that while the open content grows in popularity and we witness the proliferation of repositories and portals for OER content, it becomes more difficult for potential users to find the content they need. The power in OER is not in their production; it is in their reuse by other educators and learners. If OER discovery is improved and simplified, many OER aspirations such as widespread remix, repurposing, and redistribution could become part of the educational practice (Dichev & Dicheva, 2012). The Paris OER Declaration (2012) states the need for more research in this area as “encourage the development of user-friendly tools to locate and retrieve OER that are specific and relevant to particular needs”. Unwin (2005) argues that the problem with open content is not the lack of available resources on the Internet, but the inability to effectively locate suitable resources for academic use. Research shows (Mercer, Koenig, McGeachin & Tucker, 2011) that learners frequently arrive at open content items from outside search engines rather than by browsing through the repository's organizational structure. Jamali & Asadi (2010) state that scientists are

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

increasingly relying on Google to find scholarly literature and college students overwhelmingly use search engines as a starting point of their information searches. Open content on the web can be found with some basic meta-data, such as the title, document type, and location, but additional metadata are required for the content to enable effective searching. Indexing, categorization and tagging methods are critical to search the content to offer a personalized learning experience (Barros, Costa, Magalhães,& Paiva, 2015). OER efforts led to a fragmented landscape of concurrent metadata schemas or interface mechanisms that were not designed to offer mechanisms to enable the exchange of resources between these repositories. Recent studies show that this scenario makes it harder to reuse the resources located in OER Repositories. The motivation to learn and engage with the e-Learning solution is key to its effectiveness, especially when the effectiveness is defined as the time spent on the learning platform instead of spending too much time finding for the right content. 1. How do the learners’ find OER content online? 2. Is there a significant difference in the time spent on the site based on the traffic from direct visits to search engine driven traffic? 3. Is the learner engagement with the OER platform varies based on the features offered by the platform?

Comparison of Current OER Search Platforms Most of the search platforms currently use standard search techniques by combining conventional information retrieval techniques that are based on page content, such as word vector space (Salton, & McGill, 1983), with link analysis techniques based on the hypertext structure of the Web, such as PageRank (Brin & Page, 1998) and HITS (Devi, Gupta, & Dixit, 2014). Web search engines built on standard search techniques, parse text into tokens to be indexed into an inverted index for any relevant information about documents (such as categories, subject or other attributes). The results are then ranked to obtain an ordered list of results. The PageRank (Page, Brin, Motwani & Winograd, 1999) value for page u is dependent on the PageRank values for each page v that is contained in the set Bu (the set that contains all of the pages that link to page u), divided by the number L(v) of links from page v. The PageRank value for any page u can be expressed as

The PageRank algorithm (Brin & Page, 1998) attempts to provide an objective estimate of the Web page’s importance. However, the importance of the Web page is subjective for different users. Learners loose confidence in open content if their search results produce random irrelevant content. The true relevancy of a page depends on the interests, goals and existing knowledge of the individual users; a global ranking of a web page might not necessarily capture the importance of a page for a given individual user.

Data Collection and Analysis The purpose of this study was to examine learners’ OER interaction patterns, effectiveness of OER content search and features offered in the OER platforms for effective learning. Currently, there are no research reports with the detailed metrics of the OER resources used, usage patterns of the OER repository and how satisfied the users are. Survey based results may not be able to provide the overall effectiveness because of the volume and diversified nature of the open content and users. Jansen & Molina states (2006) that the Alexa.com ranking is an indicator of the popularity of an engine. In this study, we use Alexa.com (Xun, 2015) to examine the performance of the most commonly used eight OER search platforms. We selected curriki.org, oeconsortium.org, oercommons.org, mooc-list.org, merlot.org, class-central.org, cnx.org and coursetalk.org for further analysis. Alexa.com’s bounce rating (Arslan & Seker, 2014) is an indicator of the number of bounces, which means that the users just visit a single page and then leaves the web site. The higher bounce rates the lower the web reputation index, while the lower values indicate a higher reputation value. Lower percentage generally indicates that the user spends more time on the web site. Page Views per User (Chu, Chen, Jia, Pouwelse, & Epema 2014) is another indicator to calculate the number of pages visited by a single user. The higher number means that the user is spending more time to visit more pages and we consider these views as an indicator to a more attractive web site. The following table produces the result of the top 8 OER search platforms and the performance metrics (average page views per user, bounce rate and time spent on site) of global internet users who visited the site over the past 1 year.

www.tojsat.net

The Online Journal of Science and Technology - April 2016

Volume 6, Issue 2

Table 1: Top eight OER search platforms and performance metrics over the past 1 year _____________________________________________________________________________________ Site

Number of page views/user

Bounce Rate

Time on site (minutes)

______________________________________________________________________________________ curriki.org

4.4

38.40%