SensorsandActuatorsA228(2015)159–169 ContentslistsavailableatScienceDirect Sensors and Actuators A: Physical journal homepage: www.elsevier.com/locate/sna System modelling and device development for passive acoustic monitoring of a particulate-liquid process ManuelTramontanaa,AnthonyGachagana,∗,AlisonNordonb,DavidLittlejohnb, RichardO’Learya,AnthonyJ.Mulhollandc aDepartmentofElectronicandElectricalEngineering,UniversityofStrathclyde,204GeorgeStreet,Glasgow,G11XW,UK bWestCHEM, De partment ofPu reandAp pliedChemis tryandCen tr eforProcess Ana lyticsan dCont rolTechn olo gy(CP ACT),UniversityofStrathclyde,295 CathedralStreet,Glasgow,G11XL,UK cDepartm entofM athemat ics and Statistics,UniversityofStrathclyde,26RichmondStreet,Glasgow,G11XH,UK a r t i c l e i n f o a b s t r a c t Articlehistory: Thispaperpresentsthedevelopmentofapassiveultrasonicmonitoringsystemforthedetectionof Receiv ed1December2014 acou sticem ission(A E)cr eatedbychem ica l particles strikingth einnerwall ofareac tor vess el.Thefin ite RAeccceepivteedd i1n 3 rMevaisrechd 2fo0 r1m5 13 March 2015 element (FE) code PZFl ex was u se d to analy ze the co mplex in ter action s bet we e n chem ical par ticle s and thevesselwall.A4-layer2Dmodelwasdevelopedcomprisingaliquidloadmediumandaglass-oil-glass Availableonline22March2015 combinationcorrespondingtothejacketedvesselreactor.Themodelhasbeenexperimentallyvalidated withexcellentcorrelationachieved.TheexcitationfunctionwasderivedfromHertz’stheoryandusedas Keywords: themodelstimuluscorrespondingtoparticlesstrikingtheinnerglasswall.AnalysisoftheFEsimulations Passiveacoustics providedthetransducerspecificationsforapassiveultrasonicmonitoringsystem.Thesystemcomprises Heterogeneousreactionmonitoring PFianrittiecleel esimzee nantda n caolnycsiesnt ration stwenos titriavnitsyd.uIcmerpso wrtiathn tcloym,tphleemseennstiatirvyi tcyhaorfatchteerrisetsiocsn:a nnatrrtroawn sbdauncdewripdrtohv/hidigehs sdeinscsriitmiviintya;t wioindeobfapnadr/tliocwle Stacke dtransd ucer concentrati on.Moreover ,the broaderb and wid thofthe off-resonan tdeviced emonstratespo te ntialfor Off-resonanttransducer insituestimationofparticlesize.Theperformanceaffordedbythisapproachhasconsiderablepotential forreal-timeprocessmonitoringinthechemicalsandpharmaceuticalindustries. ©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense (http://creativecommons.org/licenses/by/4.0/). 1. Introduction of vessel or pipe walls [2], although collisions with any internal structuresorbetweenparticlesmayalsocontribute,todetermine Acoustic monitoring techniques offer a significant advantage informationrelatedtothestatusoftheprocess.Measurementof over optical techniques such as near infrared and Raman spec- AEhasbeenusedsuccessfullytomonitor,forexample,high-shear trometriesforprocessmonitoringinthattheycanbeappliedto granulation[3–6],powderblending[7],drying[8]andvariousflu- samplesthatareopticallyopaquewithouttheneedforanysample idizedbedprocesses[9–15],heterogeneousreactions[16–20],and preparation. In particular, the ability of acoustic waves to pene- thetransportofpowders[21,22],tablets[23]andslurries[24]. trateopticallyopaquemediasuchasstainlesssteelenablesacoustic A number of experimental and theoretical studies have been techniquestobeconfiguredtooperateinanon-invasivemodeof publishedtounderstandhowdifferentfactorsaffecttheAEsignal operation,e.g.bysensorattachmentorlocationofamicrophone generated. In a series of papers, Leach et al. [25–30] investi- closetotheoutervesselwall,withouttheneedforincorporationof gatedAEgeneratedbycollisionsbetweenspherical,cylindricaland awindowinthevessel.Whilethereareonlyafewreportsoftheuse irregular-shapedparticlesinarotatingvessel.Acondensermicro- ofnon-invasiveactiveacoustictechniquesforinsituprocessmoni- phonewasusedtodetecttheAEgeneratedbythecollisions,with toring[1],passiveacousticshasbeenusedmorewidelyparticularly the frequency of emission inversely related to particle size. The forthenon-invasivemonitoringofparticulateprocesses.Passive samerelationshipwasalsoobservedforthecollisionoftwosteel acousticmonitoringtechniquesusetheacousticemission(AE)gen- balls[31],andforcollisionsbetweenglassspheres[32,33]andsed- erated by collisions of particles primarily with the inner surface imentgravelinwater[34].Inaddition,itwasdeterminedthatthe amplitudeofAEincreasedwiththenumberofcollidingspheres, butthatthefrequencyofAEwasunaffected[32].AEgeneratedby ∗ theimpactofobjectswithsurfaceshasbeeninvestigatedusinga CEo-mrraeislpaodnddreinssg: aau.gtahcohra. gTaenl.:@ +s4tr4a 0th1.4a1c. u5k48(A 2.5G3a5c.hagan). mic rophon eo ratrans ducer attached tot hesu rfaceoftheim pacte d http://dx.doi.org/10.1016/j.sna.2015.03.022 0924-4247/©2015TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense(http://creativecommons.org/licenses/by/4.0/). 160 M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 Table1 Particleinformationusedinboththeexperimentalandsimulationinvestigations. Mnemonic Experimental Simulated((cid:2)m) distribution((cid:2)m) Size(cid:3) 0<x<251 200 Size (cid:4) 2 51 < x <5 00 400 Size (cid:5) 500 < x < 853 600 structure.TheamplitudeofAEarisingfromcollisionsbetweena steelplateandaballbearingincreasedasthenumberofimpacting objectswasincreased,butasforcollisionsbetweenspheres,the frequencyofAEdidnotchange[35].Duringthepneumaticcon- veyanceof co al, them ax imumfr equen cyofth evi brationalm odes Fig.1. MeasuredAEpowerspectraof40gdm−3ofitaconicacidintoluenewith ofthest ruc ture exc itedonimp actwasfo un dto decreasew ithan part icl e sizes (cid:3), (cid:4) and (cid:5). increaseinparticlesize[36,37].Instudiesofthemixingofdrypow- ders[7,38]andparticlesinaliquid[17],theAEsignalincreasedwith anincreaseinthemassandsizeofparticles,whiletheportionof onthebroadbandAEsignals,eachparticlesizerangeofitaconicacid AEatlowerfrequenciesincreasedwithparticlesize. wasaddedstepwise,upto200g,totoluene. In previous experimental work [16,17], AE generated by the E xampl esofAEs pec tra ofd if fer entparticlesizes,(cid:3),(cid:4)and(cid:5), impact of particles in a fluid with the internal wall of a 1 L jacketed of 40gdm−3 of ita conic ac id in toluen e are ill ustrate d in Fig. 1. glass reactor was studied. More recently, mathematical models The principal frequencies of interest lay in the 0–350kHz range, werederivedtodescribetheAEgeneratedbyparticleimpactswith withcomponentsabove200kHzdemonstratinglesssensitivityto the vessel wall [39–41]. The vessel wall was modelled as a sin- changesinparticlesizestothoseatlowerfrequencies. glelayercircularplateandanalyticalexpressionswerederivedto Avarietyofdataanalysismethodswereemployedtodetermine describetheimpactofparticleswiththeplate.Whileitwaspos- both particle size and concentration information. Analyzing the sibletoobtainparticlesize[39,40]andparticleconcentration[41] signalareaundereachAEspectrum,asillustratedinFig.1,demon- information,itbecomesincreasinglydifficulttoderivemathemat- stratedthebestopportunityforparticleconcentrationandsizing icalexpressions,whichcanbesolvedanalytically,describingthe information.Particlesizeinformationwasextractedbyconsider- propagationofacousticwavesformorecomplexgeometriesand ingtheratioofsignalareabetween55and200kHz,wherethereis materialproperties.Insuchcases,ithasbeenshownthatanumer- cleardiscriminationbetweenthedifferentparticlesizeranges,to icalmethodlikefiniteelement(FE)modellingcanbeused[42,43]. theoverallsignalareabetween55and500kHz.Fulldetailsofthe Therefore,inthiswork,thegenerationofAEfromtheimpactof resultsobtainedaregiveninreference[17]. particles with a reactor vessel wall for the experimental set-up Tocharacterizethefrequencyresponseofdifferentcomponents described in references [16,17] was investigated using FE mod- of the experimental equipment (the jacketed reactor vessel, the elling.AFEmodeloftheexperimentalset-upcomprisingaliquid Nano30(PhysicalAcoustics)transducerandthepreamplifier),the loadmediumandtheglass-oil-glasscombinationcorrespondingto inner face of the vessel was excited using a ‘pencil lead break’ thereactorvesselwallstructurewasdevelopedusingPZFlex(Wei- (Hsu-Nielson source) [44] to simulate impulse excitation of the dlingerAssociatesInc.,NewYork,USA).TheFEmodelwasusedto system. Fig. 2 shows the measured frequency spectrum of the investigateparticlesizeandconcentrationcharacteristicsthrough systemimpulseresponse.OncomparingtheexperimentalAEspec- analysisofthefrequencyspectraassociatedwiththegeneratedAE tra(Fig.1)withthismeasuredimpulseresponse,thesimilarityis fromtheparticle–wallcollisions.Theresultsoftheinvestigation clearlyevident.Therefore,itcanbeconcludedthattheacousticAE havebeenusedtodefinetheultrasonictransducersystemspeci- signalsobservedaremodifiedbythefrequencytransferfunction ficationforanewpassiveacousticmonitoringapproach,through ofthetransducer,andbythefilteringeffectsofthereactorvessel whichapairofultrasonictransducershavebeendesignedandfab- materialandelectronicdevices. ricated.Interestingly,thecomplementarycharacteristicsofthese transducers can provide additional particulate information from theheterogeneoussystemunderinvestigation. 2. Experimentalpilotstudy Theexperimentalset-upthatwasmodelledinthepresentstudy isdescribedindetailinreferences[16,17],withthemainfindings summarizedhere.Theexperimentalapparatusemployedconsisted of:a1Lglassreactor(VWRInternational,Dorset,UK)withanoil jacket,whichwasconnectedtoaheater–chillerunitfortemper- ature regulation; a glass stirrer rod and paddle connected to an overhead stirrer motor; a Nano30 AE sensor (Physical Acoustics Limited,Cambridge,UK)attachedtotheouterwallofthereactor vessel;andaPCfordataacquisitionandprocessing. BroadbandAEsignalswerecollectedofitaconicacidparticles (SigmaAldrich,Dorset,UK)mixingin500mLoftoluene(Bamford Laborat ories,Ro chdale, UK) at20◦C .T oge ner ate differen tparticle sizeranges,theitaconicacidwassievedintothreefractions(see Tabl e1).Toi nve stigateth eeffe ctso fpartic lesiz eand concentra tion rFeiga.c t2o.r ,Etxrapnesridmuceenrtaalnlyd mpreeaasmurpelidfi esrp.ectral profile of the impulse response of the M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 161 totheparticlesizedistributionsutilizedintheearlierexperimen- tal work, as described in Table 1. Fig. 4 illustrates the temporal andspectralprofilescorrespondingtoparticleimpactonthewall ofthereactorforthethreesimulatedparticlessizesdescribedin Table1.Itisimportanttonotethatasparticlesizeincreases,the impactforceamplitudeincreasesandthedurationofthecontact islonger.Thus,theAEassociatedwithsmallparticleshavewider bandwidthwhencomparedtolargerparticleimpacts.Itisacknowl- edgedthatparticle–wallcollisionsdooccuratanglesotherthan thenormaldirectionandmoreover,therewillbeatangentialcom- ponentinvolvingtheparticlesliding,orrolling,againstthewall. Interestingly,iftheparticleandwallaresufficientlyhard,asinthe Fig.3. Finiteelementmodelofthevesselreactor. caseconsideredhere,thenthesetangentialcomponentswillbesec- ondorder.Hence,tosimplifytheexcitationschemeforthiswork Table2 onlyelasticnormalimpactsareconsidered. Materialproperties. Material Density (kg m−3) (Lmonsg−i1tu)dinal velocity (Smhesa−r1 v)elocity 3.3. FE model validation Model validation and verification are critical in the develop- OToilluene 982690 1926225 00 mentofa simulatedm ode l.Themathe mat icalmo del ofth eimpact Glass 2484 6192 3143 of the particle striking a surface has been validated in the liter- ature[46],henceonlythevalidationofthereactorvesselmodel was performed for this work. It was necessary to reproduce an 3. Simulationmethodology identicalinputfortheexperimentaldataandpredictedmodel,and thencomparethefunctionaloutputofthetwosystems.Forthis 3.1. Modellingofthevesselreactor purpose,acustomizedtankwasbuilttakingintoaccountthe2D nature of the computerized model. The tank was manufactured Theapproachundertakentoanalyzethecomplexinteractions usingPlexiglass.Onewallwascomposedoftwoparallelsheetsof betweenthechemicalparticlesandthevesselwallusedFEmod- glass of 3mm thickness and the 4.5mm cavity in between con- elling techniques based on PZFlex software. Fig. 3(a) shows the tained 500mL of the same oil that was employed in the reactor threedimensional(3D)4-layermodeldeveloped,comprisingaliq- heater–chiller.Duringthevalidationexperiments,a250kHzwide- uidloadmediumandaglass-oil-glasscombinationcorresponding band immersion transducer (Alba Ultrasound Ltd., Glasgow, UK) to the jacketed vessel reactor. Due to computational demand of waspositionedinthemiddleofthevesselandemployedasatrans- such a 3D FE model, a 2D longitudinal section of the cylindrical mitter.APVDFhydrophoneimmersedinsidethetankandaPVDF model was utilized throughout this work as illustrated in Fig. 3(b). strip(s ize 30m m×80mm) attachedo ntheo uter wall oft he con- Samplesofglassandoilfromthereactorwerecollectedandtheir taineractedasbroadbandreceivers.Analuminiumcasewasalso acousticandphysicalpropertieswerecharacterized.Thesevalues usedtoshieldthePVDFstripfromtheenvironmentalnoiseinthe wereobtainedbymeasuringthetimeofflightinaglasssampleof lab.Moreover,thePVDFhydrophoneimmersedinthetankcanbe thickness10mmandasampleof50mLofoil.Table2reportsthe consideredacousticallytransparenttothepropagatingultrasonic measuredacousticmaterialproperties(density,longitudinalveloc- energyattheoperatingfrequencyof250kHz.Thereceiverswere ity,shearvelocity)utilizedthroughouttheFEmodellingdiscussed directlyconnectedviaacoaxialcabletotheAgilentInfiniiumoscil- inthispaper. loscopewhichwasinterfacedtoaPCthroughaGPIBconnectionto collectreceivedwaveformsfromthePVDFsensors. 3.2. Simulationoftheexcitationfunction AsrepresentedinFig.5,thevalidationofthevesselreactorwas conductedintwodifferentstages.Inthefirststage,a10Vpp20- ExcitationoftheFEmodelisillustratedinFig.3(b)andcorre- cycletoneburst,withacentrefrequencyof250kHzwasusedto spondstoapressureloadingattheinnerglasswallusingsignals drive the ultrasonic transducer and the propagation waveforms derivedfromthemathematicalmodelofimpactofparticlesofdif- werecollectedbythehydrophoneandthePVDFstripreceivers. ferentsizes.TheexcitationfunctionusedasthesourceofAEwas Thefirstarrivalwaveformonthehydrophonewasselectedman- derivedfromHertz’stheory[40,45]andthemodeloutputisthe uallyandstoredintoafile.Inthesecondstageofthevalidation,the out-of-plane displacement on the outer glass wall in the centre signal output acquired by the hydrophone was employed as the of the model. Assuming an elastic normal impact, each particle signalinputintheFEmodelasillustratedinFig.5(b).Inthissim- colliding against the wall of the vessel generates a force normal pletechnique,theultrasonictransducerresponsecanbesimulated to the surface. The determination of the impact source function withoutknowingitsinternalcharacteristicasthesignalreproduced as a function of time yields the impact time and the magnitude isdependentuponthefrequencyresponseoftheacoustictransmit- oftheforce,bothofwhichdependuponparticlesizeandveloc- ter.Thus,thepredictedoutputpressurevariationwascollectedat ity of collision. Table 3 shows the model predictions for contact thecentreofthesimulatedmodel,wherethePVDFsensorwastheo- timeandtheimpactforceforparticlesizeschosentocorrespond reticallysituatedandcomparedwiththeexperimentallymeasured waveform. Table3 TheexperimentaldataandFEmodelpredictionarecompared Parametersofthepredictedparticleimpact. inFig.6todetermineiftheFEmodelhassufficientaccuracyforuse Sizeofparticle Experimental Contacttime((cid:2)s) Force(N) inthissimulationprogramme.Theplotsshowastandingwavepat- ((cid:2)m ) distribution ter nes tablishedb ythelayerst ruct ure.T heses ta ndingwa vemo des 200 Size(cid:3) 7.08×10−6 4.25×10−7 arisefrominterferencebetweenthereflectedwavesandtheinci- 400 Size (cid:4) 1.20 × 10−5 9.60 × 10−7 dent wave s.Themaxim umofth ew aveisloc atedat 265 kHz and 600 Size (cid:5) 1.65 × 10−5 1.50 × 10−6 thisc orrespo nds tothethick ne ssm odeof th eultraso ni ctra nsm itter 162 M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 Fig.4. Predictedimpactprofiles(temporalandspectral)for200,400and600(cid:2)mparticlesemployedasinputexcitationfunction. used during the validation process. In general, the model shows studyandthepublishedtheoryintheliteratureregardingparticle excellentcorrelationwiththeexperimentalresponse;theerrorin concentrationandparticlesizeeffects.Theparticlesizesusedinthe thepredi ctedmaxim apos itio nsislessthan 9%forth em ainfo ur simulationstu dyw ere200 ,40 0and6 00(cid:2) m,whic hcor respo nd to peaks. This good agreement is in part due to the careful charac- theexperimentalparticledistributionrangesusedinthefeasibility terizationofalloftherelevantmaterialpropertiesutilizedinthe study(seeTable1). modelforeachlayer. 4.1. Concentrationeffects 4. FEsimulationresults The effects of different numbers of particles hitting the inner OncetheFEmodelwasvalidated,anumberofparameterswere wallofthereactorvesselweresimulated.Anexampleofthepre- selectedtoprovideaninsightintotheAEgeneratedbytheparticle dictedAEspectracausedby1,3and7particlesofafixedparticle impacts on theinne rw allofav esse l.O neo bjectiveof th emo delling diamet er of200(cid:2) m,imp act ing i nth e middleof th e inner wallof workwastoascertainiftheconceptualmodelpredictionofthesys- thevesselisshowninFig.7.Here,themultipleexcitationscheme temoutputwouldbeconsistentwiththeexperimentalfeasibility ispurelytheoretical,withthepressureloadingfunctionassociated Fig.5. Illustrationofexperimentalandsimulationconfigurationsusedinthevalidationprocedure. M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 163 1 F E M p red ic tio n 0 .9 E x p e rim e n ta l res pon s e 0 .8 0 .7 a.u. 0 .6 e d plitu 0 .5 Am 0 .4 0 .3 0.2 0 .1 0 0 1 2 3 4 5 F r equ e ncy [H z] x 1 05 Fig. 8. Histogram of predicted AE spectra arising from the impact of one particle of differentsizes. Fig.6. ComparisonofmodelvalidationresultsforexperimentalresponseandFEM prediction. 4.2. Particlesizeeffects Toevaluatetheeffectsofparticlesize,theexcitationfunction with each particle size applied to the same node in the model correspondingtooneparticleofeachsizedetailedinTable3was tosimulatedifferentconcentrationeffects.Thepressurevariation appliedastheinputtotheFEmodel.Theresultantnormalizedsig- of the external wall was considered as the output of the sys- nalprofilesareplottedinFig.8andextendbetween1and170kHz. tem; hence, the shape of the predicted frequency spectra were Moreover,onethousandparticlesoffixedsizeimpingingagainst notinfluencedbythetransferfunctionofthetransducer,aswas thefirstglasslayerofthevesselwerealsosimulatedtoevaluate thecasefortheexperimentalresultsobtainedduringthefeasibil- the effect of multiple particle impacts. In this work, the number ity study. Here, only the filtering effects of the reactor structure of events was limited to 1000 for computational efficiency. To itselfmoderatethesimulateddata.Inparticular,thereactorves- simulateascenarioclosertoapracticalsituation,thesimulation selbehavesasamechanicallowpassfilter,withthegeneratedAE softwarerandomizedboththespatialimpactpositionandrelative restrictedtothefrequenciesbelow180kHz.Interestingly,theorig- timesforeachAEeventassociatedwiththe1000particlesample. inalexperimentalworkdemonstratesfrequencycomponentsupto Fig.9illustratesthepredictedoutputwhenthesystemissubjected 380kHz,seeFig.1andthesehigherfrequencycomponentscorre- tomultipleparticleimpacts.Theplotisnormalizedtohighlightthe late directly with resonant modes in the transducer itself. From spectra content. A frequency shifting effect is evident, where an Fig.7,theFEmodelhaspredictedtworesonantmodesat74.4and increaseinthediameteroftheobjectimpactingthevesselresults 89kHzcorrespondingtotheresonancesassociatedwiththeglass- inadecreaseinAEfrequency.Again,thiseffectisconsistentwith oil-glassconfiguration,whichwerecalculatedtobe75and90kHz, theexperimentalworkundertakenbytheauthorsanddiscussedin respectively.Anincreaseinthenumberofimpactingparticlescor- Section2[17]. respondstoanincreaseinparticleconcentrationandresultedinan Thechangesinsignalareabetween0and150kHzforthethree overallincreaseofAEspectralenergy.Importantly,thefeasibility differentparticlesizerangesasafunctionofnumberofparticles studyofthisheterogeneousreactionproducedasimilarrelation- (N)isillustratedinFig.10.Anincreaseinthenumberofimpact- ship[17]. ingparticlescorrespondstoanincreaseinparticleconcentration andresultedinanoverallincreaseofAEspectralenergy.Impor- tantly,theseeffectsareconsistentwithpublishedtheoreticaland experimentalstudiesofcollisionsbetweenglassspheres[32],aball Fig.7. PredictedAEspectraarisingfromtheimpactof1,3and7particleswiththe Fig.9. PredictedAEspectraarisingfromtherandomizedimpactof1000particles vesselwall. ofdifferentsizes. 164 M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 Fig.10. RelationshipbetweentheAEspectralenergyandthenumberofimpactingparticlesandtheparticlesize. bearingandaplate[35],andotherinvestigationsofthemixingof monitoringsystemthatprovidesenhancedsensitivityandoperates particulatesystems[7,16,17,38]. overthebroadfrequencyrangeof0–180kHz. Itwasdecidedthatthemostappropriatewaytomatchtheultra- 4.3. DiscussionofFEsimulationresults sonicsystemperformancewiththesignificantfindingsoftheFE simul ationpr ogrammewas tode velo padual-tra nsducera pp roa ch. The FE simulation programme has been used to enhance To enhance the sensitivity of the monitoring system, a resonant the un ders tanding of the underlyin g dy namic s of t his complex 40 kHzultra son ictransduc er was proposed.W hereas, th ebroad- ultr asonic wave pr op agat ion system. Important ly, the following ban dsy stemrequ irementsco uldb ematched usingano ff-re sonant statement shave tobeconside redinth econtextoft hisw ork:only mode ofope ration.Itwas consid er edthatth eset wo transducers a 2D FE model has been utilized; a well-mixed system has been assumed;andtheentiresystemoperatesatacommontempera- ture. The validated model has provided the following properties associatedwithparticleimpactsonaglassreactorvessel: • Thereactorvesselisahighlyattenuatingsystem,withonly1.1% oftheenergyinthehorizontalplanetransferredthroughtothe receptionsensorposition. • Thepredic tedsys temsensitivitytoparticleimpactatthefurthest excitationpointinthemodelisonly3%weakerthananimpact in the centre of the model, thus the AE sensor position is not themostcriticalparameterassociatedwithsystemSNR.Inter- estingly,thissupportstheexperimentalresultsacquiredduring thepilotstudyandpresentedinreference[17]. • Part icles izeha sad irectcorre lat iontothe energyinthesystem. • TheAEs pect ral pr ofiles arestrongly de pen denton p arti clesize. • The AE spectralr esponse sar ewithina 0–180kHz fre quency band, withthemaincontributionsbetween10and60kHzandapeak around40kHz. 4.4. Ultrasonictransducerspecification Analysis of the simulation results provided the basis of Fig. 11. Magnitude of electrical impedance for (a) stacked device and (b) off- ultrasonic transducer design specifications for use in a passive reso nan cedevice. M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 165 Fig.12. AEspectraofitaconicacidintoluene,andtheimpulseresponsesofthetransducerandsystemacquiredusing(a)thestackedtransducerand(b)theoff-resonance transducer. offeredcomplementarityintheirmonitoringcapabilitiesandcould propertiestomatchthedesiredspecification.Ineachdevice,the be used to provide both particle concentration and particle size loadmediumwastheglasswallofthereactor. informationonheterogeneoussystemsfoundinthechemicalsand The resonant 40kHz transducer design (stacked device) uti- pharmaceuticalindustries. lizedastacked3–1connectivitycompositetransducercomprising fivelayersofPZT4Dceramicconnectedmechanicallyinseriesand electricallyinparallel,withasingleslotineachverticalsideincor- 5. Experimentalresults poratingasoftpolymerCIBA/GEIGYCY208/HY956[48].Thedevice had an a c tive lateral d imension o f 9mm×9mm , en caps ulated 5.1. Transducerfabrication withinapolymercylinderofdiameter45mmandthetransducer hadan o verallthic knessof 37 mm. The two ultrasonic transducer designs utilized piezoelectric To accommodate the broadband ultrasonic receiver design cerami c com posite con figurations [47] to t ailor the piezoelectric specifi cation,a1–3c onne ctivitypiez oelectricce ramiccom posite 166 M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 Fig.13. Powerspectraofdifferentmassesofitaconicacidin500mLoftoluenemeasuredusingthestackedtransducer. transducer [47,49–51] was designed to operate below its fun- response,i.e.forthetransducer,vesselandpre-amplifier,isdomi- damental resonant mode (off-resonance device). Again, PZT4D natedbythevesselresonancesasobservedintheoriginalfeasibility ceramicwastheactivephaseandthesamesoftpolymerwasused study[17]. asthepassivephaseina30%ceramicvolumefractiondevicewith dim en sionsof 18mm × 1 8m m×7.8m m.The fundam entalt hick- 5.2. Concentrationeffects ness mechanical resonance frequency was designed to occur at 240kHz,withnoothersignificantresonancesbelowthethickness Fig.13showsthespectracorrespondingtodifferentconcentra- mod edu etoth eh ighla teralattenu ationofthe softp olym erincor- tionso fita conic acid ofafixe dparticlesize (cid:3) in500mL oftoluene poratedintothepiezocompositedesign. andemployingthestackedtransducerastheAEsensor.Asexpected Fig. 11 illustrates the magnitude of the electrical impedance, the peak sensitivity is at the resonance frequency of the device, both experimental and simulated, for each transducer config- with discrimination between each particle concentration clearly uration. There is excellent correlation between the theory and observed. experimental data for the stacked configuration, as shown in The difference in signal energy for the particle size ranges Fig.11(a).Theagreementisnotasgoodfortheoff-resonancedevice, described in Table 1, as a function of concentration, employing Fig. 11(b), although it is clear that there is only one significant thestackedtransducerisillustratedinFig.14.AssumingAEaris- resonancemodebelow200kHz. ingfromNparticlesimpactingwithconstantvelocityontheinner Moreover,apre-amplifierwasintegratedintoeachtransducer layeroftheglassreactor,theintensityorenergywillbepropor- housingtoimprovesystemSNR.Thepre-amplifierhadavoltage tionaltothesquarerootofN[35].However,powerspectrawere gainof40dB,withlownoisecomponentschosenandpass-band calculatedinthiswork,sotheintensityofpowerspectrashould filters to complement the transducer operating characteristics, varylinearlywiththemassconcentrationofparticlesandhenceN. includedinthedesign. Thevariationofenergyisapproximatelylinearwithconcentration To evaluate the transducer performance, three experimental measurementswereacquiredforeachconfiguration:theimpulse 1100 responseoftheunloadedtransducer;theimpulseresponseofthe <251um transduce r atta ched to th e vessel an d th e AE de tected by e ach 1000 >251um <500um >500um <853um devicefromparticulateinteractioninthevessel.TheHsu-Neilson 900 pencilleadtestwasusedasawidebandexcitationsourceforboth oftheimpulsemeasurements.Thetwomeasurementswerethen 800 (csoimzep(cid:4)ar)esdu swpietnhd tehde AinE5 s0p0ecmtrLao afctqouluireende fraonmd a1g0i0ta gt eodf iatatcaosntiicr aractide a.u. 700 y of250rpm.Fig.12(a)and(b)illustratesthespectralcharacteristics g 600 er forthestackedandoff-resonanttransducerconfigurations,respec- n E 500 tively. It is clearly evident that the stacked transducer provides superiorsensitivityandtheoff-resonantcompositeoperatesovera 400 widerfrequencyrange.Itshouldbenotedthatthefrequencyspec- 300 trumoftheoff-resonantdeviceisnotflatbetween0and180kHz. First, the resonant mode of the manufactured device is slightly 200 lowerat170kHzandtherearesomelateralmodeactivitypresent 100 at 20 and 35kHz. Nevertheless, this device is considered appro- 0 20 40 60 80 100 pr iate for this app licationasind icate dbyth ew idebandre sponse Mass [g] associatedwiththeAEfromtheparticulatesuspensionshownin Fig.14. EffectsofbothconcentrationandparticlesizeontheAEsignalenergy Fig.12(b).Itisalsointerestingtonotethatthe‘system’impulse mea sure dusing th estac kedtransducer . M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 167 shiftingeffectsarenotimmediatelyobviousfromFig.15andhence, thespectraweredissectedintoanumberoffrequencyrangesto examinetheirpotentialtodiscriminatebetweendifferentparticle sizes. The energy of each spectrum in the range between 60 and 200kHz was computed as a percentage of energy over the total bandwidth (0–200kHz). Fig. 16 illustrates this approach for dif- ferent particle sizes; the percentage of area was invariant with concentration,butincreasedwiththeparticlesizerangeandcould beusefulforestimatingthemeansizeofparticlesinliquid.Most commercialAEmonitoringsystemswouldmissthistypeofinfor- mationastheyoftenconvertthesignaltoaDClevel.Importantly, theshifttowardslowerfrequencies,correspondingtoanincrease inparticlesize,couldbeanimportantindexforinsituparticlesize characterization. 6. Conclusions A model-based approach has been described through which tFhige. o1f5f-. reAsEo pnoawntetrr aspnesdcturcae orf. different sizes of itaconic acid particle measured using thea cousticemissi oncharact erist icsass ociatedwi thparticl e–wall impacts in a reactor vessel have been analyzed to provide a upto20gdm−3 (i.e.10gin500mL)Abovethisconcentrationthe transducerdesignspecificationforanon-invasivepassiveacous- sen sit ivit y changes due to the satu ration of pa rticles in tolu ene ticmonitor ingsys tem.Thetran sdu ce rspecification lendsi tselfto or the pre sence of othe r s ourc es of AE ( par ticle–imp ell er colli- im plementatio n throug h tw o separate devices: a l ow-fre quen cy sio ns, particle–p art icles c ollisions ) [1 7]. In fact, at values above (40kHz)sensitiv econfigu ratio nandabr oadband (1 0–180kHz)off- 20gd m−3thedegreeofn on-lineari tydep en dson th epartic lesize, reso nant structure .Thecomplem en ta rityofthes etwode vices has wi th thelarges tpartic les givingamore linearre spo nse .FromFi g.14 been sho wn to pro vide discrimination o f p articl e co ncentrat ion alone the reisno improve ment o nthe inform ationthat isach iev ed comb inedwi th particles izequalification . using RMSs ig na lsbecauseitw ou ldn otbepossibl eto di stinguish Theabi lityt omonit orch angesinparticlesizeandconcentra- betwe en,fo rexam ple,thes ig nalene rgy of ahighco nce ntrationof tioninf ormati on isofgre atinteres tt omany phar mac euticaland smallpar ticle sfromal ow concen tration of l argep articles. food industrypro ce sse s.The reistyp ica llyac ompromisebetw een the s ensing c apabilities of su ch a monit or ing system, with the following key priorities desired: non-invasive; low power; high 5.3. Particlesizeeffects sensitivity; and particle size discrimination. The dual transducer approach described here offers a route to in situ process moni- Interestingspectralfeatureswereapparentwhenemployingthe toringbycombiningallthesedesirableattributesusingultrasonic off-resonanttransducer.Fig.15illustratesthespectrafor100gof transducers. Importantly, it would be possible to integrate both differentparticlesizesofitaconicacidin500mLoftoluenecollected transducer configurations described in this paper into a single using the off-resonant transducer. Note that the peak at 90kHz package,withadequatelateraldampingbetweenthetwophases correspondstotheresonanceoftheoillayer.Theincreaseinthe ofthetransducervitaltominimizecrosstalk. relativeintensityofthesignalsinthelowerfrequencyregionsas Thispaperhasinvestigatedparticleconcentrationandsizeand particle size increased is in agreement with previously reported notaddressedthepracticalissuesofmixtureswithdifferentparti- studies of the impact of objects with surfaces [7,17,35–38]. An clematerialproperties.Forexample,theAEspectralcharacteristics increaseinparticlesizecausesanincreaseinAEmagnitudeand fortwoparticlesofthesamesize,butdifferentdensity,illustrate adecreaseinAEfrequencyowingtoadependenceontheparti- that the frequency content is similar, with some of the higher clesurfaceareaandimpacttime,respectively[33].Thefrequency frequencycomponentsreducedinamplitudeforthedenserpar- ticlecase.Thiswouldbeanadditionallevelofcomplexityofusing 1 the acous tic t echniq ue de scribed in this pa per. Howev er , it is <251um 0.9 >251um <500um the authors’ proposition that if the mixtures contained different >500um <853um materials,thisiswhereadditionaltechniquessuchasRamanspec- 0.8 trometry would be needed to provide chemical information and ge 0.7 hence,co mplem ent theinfo rm ationpr ovidedby AE.Ifapro cess a cent 0.6 ccuonltttaoindeedd muciexttuhreeps aorft idcilfefesrizeento fmeaatcehrimalast, ewrihaillefr oitm wtohueldA Ebes idginfafil-, er 0.5 AEc ou ldbeus edt oprovid ea sig natur eofthe proce ssa nd hence, a p 0.4 cou ldbe use dtod et ectanyp r ocesschan ge s. e r A 0.3 0.2 Acknowledgements 0.1 The authors would like to acknowledge EPSRC (GR/S8599/01, 00 20 40 60 80 100 GR/N27 644/01) and a CPA CT industrial cas e stude ntship for the Mass [g] fundingthrough wh ich thisco llaborative proj ectwasesta blish ed. Inaddition,theRoyalSocietyisthankedfortheawardofaUniver- Fig.16. PercentareaofAEspectra,measuredusingtheoff-resonanttransducer,for sityResearchFellowshiptoAlisonNordon. differentconcentrationsandparticlesizesofitaconicacidintoluene. 168 M.Tramontanaetal./SensorsandActuatorsA228(2015)159–169 References [32] P.D.Thorne,Themeasurementofacousticnoisegeneratedbymovingartificial sedi ments,J .Aco ust.Soc.Am.7 8 (1985)10 13–1 023. [1] J.A.Bamberger,M.S.Greenwood,Measuringfluidandslurrydensityandsolids [33] P.D.Thorne , D.J.Fode n,G ener atio nofun derwatersoundbycollidingspheres, con centrationn on-i nvasively,Ul trasonics42 (200 4)5 63–56 7. J.Ac oust.Soc .Am .84(1 988)2144– 21 52. [2] R.Hou,A.Hun t,R.A.Williams, Acousticm oni toringo fhydrocyclones,Powder [34] P .D.Thor ne,L abor ato ryand marinemeasurementsontheacousticdetection Te chno l.1 24(20 02) 176–187. ofse diment transport,J. Aco ust.Soc. Am.80(1986) 899 –91 0. [3] M.Whita ker, G.R.Ba ker,J.Westrup,P.A.Goulding,D.R.Rudd,R.M.Belchamber, [35] J.H idaka,A. Shimosaka ,H .Ito,S. Miw a,Ins tan taneou smeasurementofparticle M. P.Collins,A ppl ication o facoustic emi ssiontoth em onitor inga ndendpoint si zeandfl ow ratebyth ep ara me terso fimpactsound betweenpart icle sanda dete rminatio nofahigh sh eargran ulationp roc ess, Int.J.Pharm .2 05( 2000) circu larp late ,KON A 10( 1992)175–1 83 . 79–91. [36] F.R.Bloc k,H.U .Obst, No ninvasi vesize-controlofpneumaticallyconveyedpar- [4] L.Briens,D.Daniher,A.Tallevi,Monitoringhigh-sheargranulationusingsound ticle s,Part .Par t.Syst .Char.10(19 93)353–356 . an dvibra tio nmeasu rem ents,I nt.J.Pharm .331(2007 )54–60. [37] G.P.H ancke ,R.M alan ,Amo da lanalys istechniquefortheon-lineparticlesize [5] D.D aniher,L .Briens,A.Talle vi, En d-point det ection inhigh-sheargranula- mea suremen to fpneum a tically conveye dpulverize dc oal, IEEETra ns.Instr um. tio nusings ou ndand vi bration signalanal ysis,Powd er Technol.18 1(2008) Meas.47(199 8) 114–122. 130– 136. [38] P.J.Til y,S .Porad a,C.B.Scruby,S.Lidington,Monitoringofmixingprocesses [6] J.F.Gamble,A.B.Dennis,M.Tobyn,Monitoringandend-pointpredictionof usi ngac ou sticemi ssion ,in:N. Ha rnby,H.B enkreira,K.J .C arpenter ,R.Mann as mallscal ewe tgranul atio nproce ssusingaco usti cemission ,Pharm.De v. (Eds.) ,FluidM ixingIII,I nst itu tionofC he micalEngin eer s,Rugby,1 98 8,pp. T echnol .14(2 009 )299–304. 75–94 . [7] P.Allan, L.J. Bellam y,A.Nordon,D.Littlejohn,Non-invasivemonitoringofthe [39] G.Carson,A.J.Mulholland,A.Nordon,A.Gachagan,G.Hayward,Theoretical m ixingo fp harmaceu ti calpowd ers bybroadb andacoustic emission,A na lyst an alysisof ult rasonicvibra tio nspectra fr ommultip lep article-pla teimpacts, 135(20 10 )518–524. IEEETra ns. Ultrason.F erroelectr .Freq.C ontro l56(200 9)1034–1041. [8] L.Br iens,R. Smith,C.Briens,Monitoringofarotarydryerusingacousticemis- [40] G.Ca rson,A .J.Mulhol land,A.Nord on,M .Tramo nta na,A.G achagan,G.Hayward, si ons,Pow d erTech no l.181( 2008)115– 12 0. Pa rticlesi zing usingpassi ve ultrason ic measuremen to fparticle– wa llimpact [9] J.D.W ang,C.J. Ren,Y.R .Yan g,Char acterizationofflowregimetransitionand vibratio ns,J.So undV ibr.317 (2008)14 2–157. par ticlemo tio nusin gac oustic emissionmeasur em enti nagas– solidfluid ized [41] G.Carson, A .J.Mulh ollan d,A .Nordo n,M.Tramontana,A.Gachagan,G.Hay- bed,AIC hEJ.56 (2010 )1173–1 183. wa rd,Estim ati ngparticleco nc entration us ingpassiveult ras onicmeasu re ment [10] S.Ma tero,S . Pou tiainen ,J.Leskinen,K.Jarvinen,J.Ketolainen,S.P.Reinikainen, ofimp actvibratio ns,IEEE Trans.Ultraso n.Fer roelectr .Freq.Con trol56(2009) M .Hakulin en ,R.Lappala in en,A.Poso ,T hefeasibi lit yofusingac ous ticemissions 34 5–352. for monitoring o ffluidizedbed g ranul ation ,Chemom .I ntell.L ab.Syst. 97(2009) [42] F.Moser,L.J.Jacobs,J.M.Qu,Modelingelasticwavepropagationinwaveguides 75– 81. w iththe fini teelem entm et hod,NDTE Int.32 (199 9)225–234. [11] Y.F.Zhou,K.Z.Dong,H.Zhengliang,J.D.Wang,Y.R.Yang,Faultdetectionbased [43] G. H ayw ard, A. Gach agan, R. Ham il ton , D .A. Hu tchins, W.M.D. Wright, ona coust icem ission -e arlyagglome rati onreco gni tionsy stem influidize dbed Ce ramic–epox yc ompositetr ans ducersfor nonco ntactingu ltrasonic applica- rea ctor,Ind .Eng.Chem.Res .50(2011)847 6–8484. tions,in:F.L.Li zzi(Ed.),N ewDevelop men tsinUltrason icTransdu cersand [12] G.R.Flat en,R .Bel chamb er,M .C ollins,A .D.Walmsley,Caterpillar—anadaptive Trans duc erSy stem s,SPIE –Int SocOpticalEng ine ering,Belli ngham,1992 ,pp. algo rithmf or detectingpro ces schang esfr omacousti cemissionsign als,Anal. 49–56. Chim.Acta 54 4(2005)2 80–291 . [44] T.Boczar,M.Lorenc,Determiningtherepeatabilityofacousticemissiongen- [13] M.Ha lsten sen, P.deB akker,K.H.Esbensen,Acousticchemometricmonitor- er atedby the Hsu-Ni elsencalibrati ngs ource,Mol.Q ua ntumAco ust.25(2 004) ing ofanindus tri alg ranulat ionp roduction process— aPATfeasibil itystudy, 177–19 2. Che mo m. Intell.Lab .Syst.84(20 06)88–97. [45] A.Akay,M.Latcha,Soundradiationfromanimpact-excitedclampedcircular [14] K.Naelap ää,P.V eski ,J.G. Ped ersen, D.Anov,P.Jørgensen,H.G.Kristensen,P. pl atein ani nfiniteb affle,J .Acoust.S oc.Am .7 4(1983)640–6 48. Be rtelsen,Ac ou sticmo nit oringofafl ui disedb ed coatingpr ocess ,Int.J.Pharm . [46] D.J. B ut tle, C.B. Sc ruby, C h aracteri zatio n o f p article impact by quantitative 332(2007 )90–97. aco ustic-em issi on,Wear 137(1990)63–9 0. [15] H.T sujimot o,T.Yokoyama,C.C.Huang,I.Sekiguchi,Monitoringparticleflu- [47] W.A. Smith, B.A. A uld, Mod eling 1 –3 composite piezoelectrics—thickness- idi zationinafl ui dizedbedgr anul atorwit h anacoustic emissionsen sor,Pow der mode oscilla tions ,IEEE Trans.Ult rason .Ferroelect r.Freq.Control38(1991) Technol. 11 3 (2000)8 8–96 . 40–47 . [16] A.Nordo n,R. J.H.Wa ddell,L.J.Bellamy,A.Gachagan,D.McNab,D.Littlejohn,G. [48] R.L.O’Leary,M.Kijowski,G.Hayward,T.McCunnie,3–3connectivitymul- Ha yward,M onito ringofah ete rogeneou s reactionby ac ousticem is sion,Analy st tilay ered pie zoe lectric co mp osites, in: M .P. Yuhas (Ed.) , IEEE Ultras onics 129(2004 )463–467. Symposiu m,IEEE,New York,2004,p p.1 686–1 689. [17] A.N ordon, Y.Carella,A.Gachagan,D.Littlejohn,G.Hayward,Factorsaffect- [49] G.Hayward, J.A.H ossa ck,Un idime nsio nalmodelingof1–3compositetrans- in gbroadb and acoust ice missionm eas urementso fa heterogen eousre action, du cers,J.Aco ust. Soc.Am. 88(1990)599–6 08. Ana lyst131(20 06)323– 330. [50] J.A.Hos sa ck,G.Ha ywa rd,F init e-elem entanalysisof1–3compositetransducers, [18] P.D.We ntze ll,A.P.W ade,Chemicalacousticemissionanalysisinthefrequency IEEE Trans.U lt rason.Ferr oelectr.Freq.C ontrol38 ( 1991 )618–629 . dom ain,Anal. Che m.61( 1989)263 8–2642. [51] A. Co chran , P. Reyn olds, G. Hay ward , Progre ss in stac ked piezocomposite [19] Z.Cao,B .-F.W ang,K .-M .Wang ,H.-G.Lin,R.-Q.Yu,Chemicalacousticemis- ul trasonictr an sducersfor lo wfrequenc yapplica tio ns,Ultra sonics36(1998) sio nsfr omg asevolu tionp rocesse sreco rded byap iezo electrictr ansducer ,Sens. 969–977. Actua torsB 50 (1998)27 –37. [20] D.Betterid g e,M .T.Josl in,T.Lilley,Acousticemissionsfromchemicalreactions, An al.Chem.5 3(19 81)10 64 –1073 . Biographies [21] K.H.E sbense n, M.Hals tensen,T.T.Lied,A.Saudland,J.Svalestuen,S.deSilva, B.H ope,Acous tic chemometri cs— from no isetoinfor m ation,Chem o m. Intell. La b.Syst .44(1998 )61–76. ManuelTramontanaobtainedaM.Sc.degreeintelecom/electronicengineeringat [22] J.Hu ang, S.O se,S.d eSilva,K.H.Esbensen,Non-invasivemonitoringofpow- thePolit ecnicodiMila no.Here c eived hisPhD ,fr omStrathclydeUn iversityfort he d erbreak ag edu rin gp neum atic transporta tionusingaco usticchemo me trics, dev elopmento fpa ssive/ac tiv eacoustic sys tems forp rocessmonit oringandc ont rol. Pow derTechn ol.129 (2003)130 –138. Hehasworke df orOxfordTech nicalSol utionas ele ctricald esignengine erd evelop- [23] K. Albio n, L. Br iens , C. Br iens, F. Berruti, Detection of the breakage of ing ani nertials yst em.In2 008,hejo inedFlex tro nicsasas ystema rchitect. Hismain ph armaceu tic al table ts in pneu ma tic trans port, Int. J . Ph arm . 322 (20 06) are aof researc hinteres tc overs the design ofvariousm e di calprod uctsandt ech nolo- 119–129. gies suc hasgluc osemete rs,we arab leECG pa tches,e nergyha rvesting, Blue toothlow [24] K.Albion,L.Briens,C.Briens,F.Berruti,S.McDougall,Detectionofoversized ener gyan d wireless chargin g. m aterialin a hydrot ran sports lur rypipeu si nganon-inv asiveacou sti cmethod, [[[[22225678]]]] aPeafHMMMrccmoo....WoowFFFmiuu...s d.LLsssL aetteeiKeoiicraaaccwon ccc -T eushheahem,smn,, ,c P tGGh,Gii oi sc.sn.M.AAwsAseoi.i..o.md loF RR.Rne n.u1u isurs,L9bb ,sbU Te0Piiiinenaolno t(cc,,nrw , 2hhJ Jas.S.0Cn,Cd,s i0oP.oz.eC 9WWenolrh.) w iTaa 1cii3lerlns6dll6ac iia e2aa1c(hlr13mtmy–ne9T s3(orss ie717isl,,c z 7.9 P1P hoa2)8a.a tfn115rr i o optt5)(ii lna1c3c2.rl 9l–13eeto71i93--fc8s5s– l(ii)ce821zz ys2.e39e l6 8o7iddn3f.8ie ds–)ittrr2e1rri6rcie5bma7g7ul.u– itnpl1iaoaa6rtnr 7 its oc.ihcnhal aefprrseoa cmffrtroe o amrmcizo attuthhsieeotiiincrr spUedipuAnoneurlnnt onvcbtritfhoaveahleiremlesocso,or asnfipspbuotiymeaiintroo lys ddGeanis n n coasauo o ftcnatlf uthh dcSorusaret aflttor gtsihaDasrcoaetaissennrh s a-api caocc pialbnuso ytprr dnurtdtolam hidpeacinen l.dae fes dHotn dddaiurtoeiu p r sp nocopetrive rfcleeoi iretczcEafsor eaoll c ra eit2eo vpnicol0oderetdf noredc y iatcd t cerhrrhe aaarieesicnlarsx s sygC a ct Pcseneriath,oa nnd iaanDnnt dtrrEstci,r oe rdlliafuoes nuryf cldo otc .tti himremerHn eorcUg i a hnSsa tlgnn iettucisrrocint paaqhnhEgtseun-noh ocpdeorgcnirse filliooi.oynscccftdge reEoryeeusen.vr,s c sigeHiitnennirignaev g1r 1 ee,(ch 0 9ehEeha0r9iEv sig 6nEair hn wel)guf t sp oaeoaaeortrrtna weikttrdoshhece nthdeeasr, [29] o1Mf6 .9Fr.– i g1Li7ed6a .cpha, rtG ic.Ale.s Rfurobmin , aJc.Co.u sWti cil lieammiss, si Aonns a,l yPsiosw o dfe rp oTleycdhisnpoel.r se1 9 sy(s1t9e7m8s) ASsptlrieascotthnrco lNsycodorepdywo fnhre oormbet stahhineee Uhden alidv BerSrecssi et(yHa roocnfh sDf)eu ilrnlho cawhmea.m nSdhisest retynh ieaonnr dmr eaos vePaehrdDc htionf teshloleol iwUd-npsitvoaesttress iNwtyMi toRhf [30] o1Mf8 .9Fr.–i gL1ie9da5 cp.ha,r Gtic.Ale. sR ufrboimn, Jt.hCe. iWr ialclioaumssti,c A-enmaliysssiiso no,f Pao Gwaduesrs iaTnec shinzeo ld. i1st9r i(b1u9t7io8n) iwCnPh AtehCreeT D.s Bheepetawi srtecmeuner rn2et0n 0ot4fl y Pa aunrdsee 2 na0 ino1dr2 ,Al esphcpteul iwre edars .C Hah eRerom myiasalt irS nyo rcaeites ttehya erU cUnhinviinevtreesrriste iyst ytRs eoasfre Sea trtrchaheth Fdce ellyvlodewel-, [31] eJ.t Herids aokfa i,m Ap. Sahctim soousnakda b, Se.t Mweiwena ,t Twhoe pefafretcictsle osf, pKaOrtNicAl e7 p (r1o9p8e9r)t i4e–s 1o4n. the param- aonpdm seing tn oafl ipnr soictues sspinegct fr oor s cporpoicce msse masounrietmo reinngt sa (nodp tc iocnalt raonld. a coustic), chem om etrics