Thermal behavior of an engineered fuel and its constituents for a largerange of heating rates with emphasis on heat transfer limitationsOdile Vekemans, Jean-Philippe Laviolette, Jamal Chaouki *Chemical Engineering Department, cole Polytechnique de Montral, C.P. 6079, succ. CentreVille, Montral, QC H3C 3A7, CanadaARTI CLE I NFOArticle history:Received 23 September 2014Received in revised form 8 December 2014Accepted 9 December 2014Available online 15 December 2014Keywords:Engineered fuelRefuse derived fuelWastesTGADevolatilizationHeat transferABSTRACTEngineered fuels (EF) manufactured fromwaste can be an advantageous substitute for coal or other fossilfuel in (co-)combustion, gasication or pyrolysis processes. Unfortunately, because of theirheterogeneity, the thermal behavior of such fuels can often be complex, limiting their application. Inthe present study, the pyrolysis of a heterogeneous commercial EF composed ofbers and plastics wasinvestigated using a TGA apparatus over a large range of heating rates (from5 to 400C/min). At a heatingrateof5C/min, theEFdevolatilizationcurve wassimply aproportionalsumofthedevolatilizationcurves of its individual components. When the heating rate was increased up to 100C/min, however, ashift in the devolatilizationTG curve of the bers, plastics and EF to higher temperatures was observed asaconsequenceof heat transfer limitations withinsamples. Furthermore, differences betweentheproportional sumof the devolatilizationcurves of the individual components of the EFanditsexperimentalcurvewereobserved, andincreasedwithincreasingheatingratesupto400C/min. Amodel was developed to correct for heat transfer limitations by considering thermal phenomena such asheat transfer limitations between the TGA and the sample, change in sample heat capacity and effect ofendothermicreactionsonsampletemperature. ThismodelpredictedtheshiftofEFdevolatilizationtowards higher temperature with increasing heating rates, which suggests that no signicant chemicaleffects occurred between the EF components. 2014 Elsevier B.V. All rights reserved.1. IntroductionThe global energy market is expected to depend on fossil fuelsfor the foreseeable future, even though the combustion of fossilfuels is the main source of air pollutant emissions such as NOx, SO2,mercury, CO2, etc. [3]. New and increasingly stringent regulationshavethusemerged, whichincentivizethedevelopmentofnewtechnologiestoreducepollutantemissionsof fossil fuel powerplants. One such technology is waste-derived fuel, which has thepotential to become a cleaner complement to fossil fuels throughco-combustion. Waste-derivedfuelsalsoaddressanotherenvi-ronmental problem: solid waste accumulation. The worlds annualmunicipal solid waste (MSW) productionreached more than1.3 billion tonnes in 2012 and the annual production is projected todouble by 2025 due to growing prosperity and urbanization [33].Waste-derived fuels are produced fromMSWthrough amanufacturing process, which involves the removal of non-combustibles such as glass and metals. The remaining combustiblefractionisusedtoprocessthewaste-derivedfuel andincludesrenewablebiogenic wastes (paper, cardboard, textiles or wood) [29],andmixedplastics suchas lowandhighdensity polyethylene (LDPE,HDPE), polypropylene (PP), polystyrene (PS) and polyvinylchloride(PVC), in the form of foils or hard plastic pieces [19]. While thesematerials may be recyclable in theory, it is often nanciallyunfeasible suchthat these materials are sent to landll [12].Waste-derivedfuelsareclassiedeitherasrefuse-derivedfuels(referred to as RDFs) or engineered fuels (EFs), depending on theirmanufacturing process. Unlike RDFs, EF manufacturing involves theseparationof the individual waste components (biogenic andplasticfractions) to allowa better control of fuel composition, fuel particlesize and contaminants content (chlorine in PVC, for example) suchthat EF are considered higher quality fuels compared to RDF. Thecontrol of EF characteristics (composition, particle size, etc.) is keytotheoptimizationofco-feedingapplicationsinexistingandnewthermal processes (combustion, gasication and pyrolysis). EFphysical characteristicsandchemical composition(ashcontent,caloric value, etc.) canbecontrolleddirectlythroughthe EFcomposition, whilethe control ofthe EF reaction prole is morecomplex as it requires the characterization of its reaction kinetics.* Corresponding author. Tel.: +1 514 340 4711x4034.E-mail addresses: odile.vekemans@polymtl.ca (O. Vekemans),jamal.chaouki@polymtl.ca (J. Chaouki).http://dx.doi.org/10.1016/j.tca.2014.12.0070040-6031/ 2014 Elsevier B.V. All rights reserved.Thermochimica Acta 601 (2015) 5462ContentslistsavailableatScienceDirectThermochimica Actaj ournal homepage: www. el sevi er . com/ l ocat e/ t caConsidering that EF characteristics can be specic to each reactor, arobust characterization of its reaction kinetics requires thecharacterization of the EFs individual components kinetics as wellas the interactions between the components(chemical or physicalinteractions).Devolatilization is the rst step for any thermal process,especiallyfor RDFandEF whichgenerallyhave highvolatilecontent [10]. Because of their heterogeneity, RDF and EFdevolatilizations arecomplex. Bymeanof simplication, it iscommon to approximate their devolatilization as the linearcombination ofthe devolatilization of each oftheir componentsmultiplied by their mass fraction. This approximation, referred toas the linear combination (LC) model, is made under theassumption that no interactions between the components occurduring pyrolysis. In the scientic literature, numerous studies onthedegradationofmixtures ofcellulosicmaterial (paper, wood,sawdust, cardboard, etc.) and plastics (HDPE, LDPE, PS and/or PP) atlow heating rate (_100C/min) have shown that the devolatiliza-tion process of the mixtures seems to result from the independentthermal degradation of its individual constituents (e.g., Grieco andBaldi [17]; Kim et al. [21]). However, it has been reported in manyscientic article that the LC model cannot satisfactorily predict thedegradation of the mixtures (e.g., Dong et al. [11], Mani et al. [24],Sharypov et al. [27], Zheng et al. [35]) .When deviations from the LC model are observed, one of theexplanations suggestedbyauthorsisthepresenceof chemicalinteractions. Asummaryof studies identifyingchemical interactionsis given in Table 1. Chemical interactions, as the reason fordiscrepancies between the LC model and the experimental curves,are difcult to model, and no clear proof of changes in the sampleshave been measured to support these hypotheses. However,interactions amongthevolatiles mayhappen, impactingtheproductdistribution, as described by Dong et al. [11], Marin et al. [25] andSharypov et al. [27], but having no effect onthe devolatilizationrate.Another explanation given by authors to account for thedeviations from the LC model are heat transfer limitations, or moregenerally thermal effects. Zheng et al. [35], for example, studied thedevolatilization of waste paper mixed with PE in a TG apparatusspecially designed to heat the samples at up to 850C/min. Theyobserved that, at high heating rate (>100C/min), the presence of PEweakened the reaction intensity of both the paper and the plasticfraction, observed as a shift towards high temperature compared totheLCmodel, andattributedit tothermal effects. Theysuggestedthatthe endothermicity of the PE pyrolysis might lead to a lower localtemperature of the mixture sample and to a corresponding lowerweight loss kinetic. Furthermore, several authors described a shifttowardhightemperatureof thedevolatilizationcurvesofgivenmaterialswithanincreaseinheatingratefrom40to160C/min[26],Nomenclaturea Relative sample mass or % of initial sample mass (wt%,d.b.)aFinal relative sample mass (wt%, d.b.)CpHeat capacity (J/kgC)e Porosity ()e Thickness (m)F Carrier gasow (kg/s)HvolHeat of devolatilization (J/kg)k Thermal conductivity (W/mK)m Sample mass (mg)m0Initial sample mass (mg)S Sample holder bottom surface (m2)t Time (min)T Temperature (C)TendDevolatilization ending temperature (C)TFFurnace temperature (C)TinInlet gas temperature (C)ToutOutlet gas temperature (C)TpmaxDegradation peak maximum temperature (C)TstartDevolatilization starting temperature (C)TSSample temperature (C)U Global heat transfer coefcient (W/m2K)xc,iInstantaneous mass fraction of char in the ber at timeti ()xjInitial mass fraction of constituent j in the EF ()xj,iInstantaneous mass fraction of constituent j in the EFat time ti ()Sub-/superscriptsc Charf Fiberi At time tij EF constituent (ber, hard plastic or soft plastic)N2Nitrogenpf PureberTable 1Summary of studies citing chemical interactions as the explanation for deviation of experimental results from the linear combination model.Author Material + HR Observation Explanation given Proof[28] 50% wastepaper +50%HDPE/LDPE/PP/PS/PVC10C/minShift towards lower temperature only forPVC+paperCatalytic presence of PVC (acid hydrolysis type ofreaction reducing the stability of the cellulosic material)None[16] TetraPack (78%cardboard+21%LDPE)RDF (88% paper + 4%LDPE + 8% PVC)20C/minShift towards lower temperature of the plasticpeak of Tetra Pack and of the hole RDF curveLow extent interactions between the ash contained inthe paper fraction and the plasticFor RDF: catalytic presence of PVCNone[20] 50% PS/PE+50%cellulose/wood/lignin10C/minShift towards higher temperature of the plasticpeak in all cases +changes in the productdistributionPropensity of char to promote the hydrogenation of theunsaturated products, increasing there thermal stabilityTheoretical model notveriedexperimentally[1] 50% oliveresidue +50% HDPE/LDPE/PP/PS2, 10, 20 and 50C/minShift towards higher temperature of plastic peak Inuence of the products formed during the cellulosicmaterial degradation on the plastic degradationNoneO. Vekemans et al. / Thermochimica Acta 601 (2015) 5462 55from 5 to 100C/min [14,31] but even from 5 to 20C/min [1,24].These authors suggested two phenomena to account for this shifttowardhighertemperature:limitationinheattransferfromthefurnace to the sample, leading to a lag between the furnace nominaltemperatureandtheactual sampletemperatureat highheatingrate,and inuence of the heating rate on the decomposition rate, alsocalled the Kissinger effect [6,22]. However, none of these studiesproposed a model to quantify the limitations of the heat transfer inTGapparatus, andthereforeverifyif theorderofmagnitudeofthermal effects in TGA could explain the shifts toward hightemperature.In the present study, pyrolysis ofber, plastic and commercialEF samples are performed in a TG apparatus under various heatingrates. TheeffectsofEFcompositionandheatingrateontheEFdevolatilizationbehavior aremeasured. Furthermore, a modelbased on heat transfer limitations is presented to account for thediscrepancies between theEF andits constituents volatilizationbehavior across the range of experimental conditions tested. Theexperimental data is compared to the modelling results todetermineif heattransferlimitationscanexplaintheobservedeffects of the devolatilization proles.2. Experimental2.1. Single componentsThe engineered fuel used in the present study is produced frommunicipal solidwastesprocessedandsortedintobers(paper,cardboard, wood and some textiles), hard plastics (HDPE, PS andPVC) and soft plastics (LDPE and LDPP). It should be noted that inthis document hard plastics and soft plastics always refer to thisclassication. Samplesofeachof thesefractionswereobtainedfromthe ChittendenSolidWaste District (ChittendenCounty,Vermont, US) in the form of shredded materials with average sizesof 2mm, 3mm and 5mm for the hard plastic, the soft plastic andthebers, respectively. Table2showsthecharacteristicsoftheindividual components: results of proximate and ultimate analysisas well as other relevant properties.2.2. Engineered fuelTheengineeredfuel usedinthisworkwasprovidedbythecompany Accordant EnergyTM, (Rutland, Vermont, US). Thebersandplasticsusedtoproducethefuel wereprocessedtoensureuniformsizeandweightandtoincreaseenergydensity. TheEFprovidedhad anaverage particle size of 3mmandwas composed, inaverage, of 80%ber, 10% soft plastic and 10% hard plastic. It wasproduced for co-combustion with pulverized coal in conventionalboilers.2.3. Thermogravimetric analysisThe thermal behavior of the EF andits components werestudied with a Q5000 apparatus fromTA Instruments (NewCastle,Delaware, US). DuringeachTGexperiments, thesampleswereplaced in a platinum pan and held under a nitrogenowof 20mL/min. Due to the variation in density and size of the materials, theinitial mass sampledifferedfromone component toanother,varying between5mg and 15mg for the bers, 15mg and 30mg forthe hard plastics, 2mg and 10mg for the soft plastics and 20mgand 35mg for the EF. Blank tests were also performed to ensureaccurate measurement (error was less than 0.02mg and 102wt%/C). Atleasttworepetitionswereperformedforeachexperi-ment, showing reproducibility to within 5C. The results presentedhereafter are the average for a set of repetitions.During experiments, the TGA heating prole consisted ofrstheating to 150C with a heating rate of 50C/min, where itremained for 15min to remove all moisture. Then, the sample washeated to 900C at a specic heating rate (5, 20, 30, 50, 80, 100, 150,200, 300 or 400C/min). Finally, the temperature was maintainedat 900Cfor 15mintomake surenofurther mass reductionoccurred.The temperature used to control the TGA was measured from athermocouple mounted on a platinumdisk under the pan. The TGAreference temperature was therefore the temperature of theplatinumdiskinthe TGAfurnace, referredtoas the furnacetemperature (represented symbolically as TF), rather than the truesampletemperature. Anevaluationof thedifferencebetweenthose two temperatures is presented in the next section.TheTGAresultsarepresentedintheformofnormalizedorrelative mass on a dry basis (wt%, d.b.), dened as the ratio of themass sample at a given time t over the initial mass evaluated at150C. Thistemperaturewaschosentoavoidanyeffectof thevariablemoisturecontent of thematerials ontheparametersevaluation.For each set of experimental conditions, the followingparameters were evaluated:-the temperature (C) at which the devolatilization started (Tstart)and ended (Tend), dened as the rst and the last temperature atwhichtherateofweightlosswithrespecttotemperatureissignicant, i.e., da/dTmax> 2102wt%/C;-theduration(inmin) of thedevolatilization, denedasthedifference between therst and the last time at which rate ofweight loss is signicant;-the char yield (wt%, d.b.), dened as the ratio ofnal to initialmass on a dry basis;-the derivative thermogravimetric peaks:-the maximum weight loss (da/dTmax with units of wt%/C);-the maximum weight loss temperature (Tmax with units of C).Theerrorreportedforeachparameterwasevaluatedasthehighest relativeerrorcalculatedamongthedatafor thegivenparameter and the given material.Table 2Characterization of the wastebers, hard plastics and soft plastics.Fiber Hard plastic Soft plasticProximate analysis (5%)Moisture, wt% 4.9 0.2 0.2Volatile mater, wt% 75.0 95.9 93.8Fixed carbon, wt% 7.7 2.0 0.1Ash wt% 12.4 1.9 5.9Ultimate analysis (0.01%)Carbon, wt% 40.17 79.79 70.32Hydrogen, wt% 5.45 12.00 11.45Nitrogen, wt% 0.98 1.06 1.12Oxygen, wt% 42.02 2.66 14.50Neutron activation analysis (5%)Ca, wt% 2.12 1.17 1.40Na, wt% 0.46 0.12 0.07Cl, wt% 0.44 5.41 0.06Caloric value (5%)HHV, MJ/kg 14.1 34.2 39.6Densityrtrue, g/cm3(1%) 1.587 1.120 0.990rbulk, g/cm3(5%) 0.072 0.37 0.05956 O. Vekemans et al. / Thermochimica Acta 601 (2015) 54623. Heat transfer model3.1. Evaluation of the sample temperatureAs previously mentioned, the temperature reported by the TGAapparatus is measuredonaplatinumdisk, under thesampleholder. The temperature measured by the TGA apparatus istherefore different fromthe temperature of the sample. Toevaluate the sample temperature, a heat transfer model isproposed. Arepresentationof theTGAsystemconsideredforthe model is shown in Fig. 1.During the experiments, heat is provided to the sample holderby the infrared furnace, which then heats the sample viaconduction. The inletow of nitrogen (Fin) also provides heat tothe sample. On the other hand, heat exits the system via the hotexhaust gas ow(Fout) andis consumedviatheendothermicpyrolysis reactions (sample devolatilization). The temperature ofthe platinumdisk (TF) measured by the TGA is assumed to be equalto the temperature of the sample holder, which is also in platinum.Thesampleis consideredtohaveauniformtemperature(TS)because of its small size. Therefore, considering that the samplesare dry, that the volume of gas emitted by the devolatilization isnegligible compared to the nitrogen inlet owand that thevariations of the thermo-physical properties of the materials withtemperature are also negligible, the following heat balance can bewritten:FCpN2TinTout( ) US(TFTS) = mCpdTSdtHvoldmdt(1)InEq. (1),Fisthemassowofnitrogen(kg/s), CpN2istheheatcapacity of nitrogen (J/kg K), Tinand Toutare the inlet and outlet gastemperature (C), S the surface of the bottom of the sample holder(7.85105m2), TFthefurnacetemperature(C), TSthesampletemperature(C), mthesamplemass(kg), Cpthesampleheatcapacity (J/kg K), t the time (s), Hvol the heat of devolatilization (J/kg) and U the global heat transfer coefcient (W/m2K). The globalheat transfer coefcient (U) is used to estimate the conductive heatowthroughthesampleholderandthesampleitselfanditisapproximated at equilibrium by the following term [5]:U =1(epan=kpan) (esample=ksample)(2)In Eq. (2), epan is the thickness of the pan (2104m), kpan is thethermal conductivity of the pan (72W/mK), esample is the samplethickness (1103m)andksampleisthethermalconductivity ofthe sample. Because of the porosity of the sample, ksample can beexpressed as [32]:ksample =1 e ( )kf ekN2(3)withetheporosityof thesample(0.5forall samples), kjthematerial thermal conductivity itself and kN2the nitrogen thermalconductivity (0.024W/mK).The thermal parameters of the different materials are given inTable 3. Note that the thermal conductivity of biomass char is lessthan half the one ofber and therefore may change the thermalbehavior of the sample during the devolatilization. Knowing thenal char yield of theber samples, the instantaneous fraction ofchar in theber samples at time i (xc,i) can be expressed as:xc;i =100 af;iaf;iaf;100 af;(4)where af,iis the relative ber mass (wt%, d.b.) at time i and af,thenal relative mass of the ber sample (wt%, d.b.). The ber thermalconductivity (kf) varies then with time according to:kf(t) = 1 xc;i kpf xc;ikc(5)where kpf is the pureber conductivity (0.25W/mK) and kc is thechar conductivity (0.095W/mK).The model resolution procedure is summarized in Fig. 2.Regarding the thermal parameters of the EF, one should note thatthey are dened according to an instantaneous mass fractions (xj,i).Indeed, whilebersinitiallyformthebulkoftheEF, theysoondevolatilize. Oncethis occurs, theplastics instantaneous massfraction increases. However the char frombers ultimatelyexceedsplasticchar, meaningthe berswill enduphavingahighermassfractionaschar. Thisinstantaneousmassfractiontherefore takes into account changes of the mass fractions of the EFconstituentsduetotheirdegradation. Furthermore, duetothesmall size of the sample and the inhomogeneity of the EF, the initialmass fractions of its constituents, xj, may vary from one sample toanother.3.2. EF devolatilizations prediction based on the heat transfer modelIn the linear combination model, one of the underlyinghypothesesisthat, forasamefurnacetemperaturehistory, allthe samples degradations at time t happen at the sametemperature, nomatterthesampletypeorsize. However, duetodifferentthermal parametersandreactionkinetics, differentmaterialscanbeatdifferentsampletemperatureforthesamefurnace temperature. Therefore, we suggest that the EF predictionis made, not based on the furnace temperature, as it is in the LCmodel, but based on the sample temperature of both theconstituents and the EF for every given temperature history. Moreprecisely, in order to predict the EF degradation curve,once thesample temperatures of all the samples have been evaluated, theweightlossofeachconstituentcorrespondingtoeachgivenEFsample temperature are multiplied by their respective massfractionsandthensummed. Thismethodcouldbeextendedtoindustrial furnaceswheretheEFfuel temperaturecanalsobeevaluated with a heat balance, and this temperature can be used toevaluate the degradation kinetic of the EF constituents.4. Results and discussion4.1. Devolatilization analysisThe TG curves of all materials (bers, hard plastic, soft plasticandEF)areshowninFig. 3andtheparametersdeningtheir

Fig. 1. System considered for the heat transfer model.Table 3Heat transfer parameters.Fiber Hard Plastic Soft PlasticCp(J/kg K) 1.4103[30] 2.30103[15] 2.85103[2]Hvol(J/kg) 210,000 [4] 300,000 [8] 469,000 [2]kj(W/mK) 0.25 [4] 0.4 [7] 0.22 [2]0.095 [18]O. Vekemans et al. / Thermochimica Acta 601 (2015) 5462 57devolatilization are given in Table 4 for 10 different heating ratesranging from 5C/min to 400C/min.4.1.1. TGA of EF constituentsDegradation ofber, hard plastic and soft plastic samples wereinitiated at furnace temperatures of about 250275C, 400C and400530C, respectively. Theplasticsamplesexhibitedahigherthermal stability than thebers.Thebers presented two distinct degradation steps, indicatingtwodifferent devolatilizationprocesses, whiletheplastics presentedasingledegradationstep. The rst ber degradationstep(referredtoas P1) occurred between 300C and 400C and was responsible for

Fig. 2. Heat transfer model resolution procedure.

200 400 600 800020406080100Furnace Temperature (C)Normalized mass (%) a.5 20 30 50 80 100 150 200 300 400C/min200 400 600 800020406080100Furnace Temperature (C)Normalized mass (%)b.200 400 600 800020406080100Furnace Temperature (C)Normalized mass (%)c.200 400 600 800020406080100Furnace Temperature (C)Normalized mass (%)d.Fig. 3. TG curves of (a)ber, (b) hard plastic, (c) soft plastic and (d) EF samples, in N2 for HR of 5, 20, 30, 50, 80, 100, 150, 200, 300 and 400C/min.58 O. Vekemans et al. / Thermochimica Acta 601 (2015) 546280% of its total weight loss, while the second step (P3) occurringbetween 600C and 800C resulted in another 10% loss. Betweenthese two steps, a slow weight loss of approximately 2wt%/C wasalsoobserved. Theseobservations areconsistent withpreviousstudies onbers and wood pyrolysis up to 600C [13,16,23,28,34].The rst peakhas beenassociatedwiththe combineddegradationofcellulose, hemicellulose and lignin [9], while the slowand constantweight loss upto600Cwas relatedtothedegradationof ligninalone[16,28]. At temperatures above 600C, Cozzani et al. [10] observed athird degradation step that they attributed to the thermal degrada-tionof calciumcarbonate(CaCO3), usedas aninorganicller inpapermanufacture.Thedevolatilizationofthesoftplasticsamplesoccurredasasingle peak (P2) between 400C and 530C, as shown in Fig. 3(c).Similarly, thehardplasticsdegradedinonemajorstepbetween400C and 550C. Softplastic degradation occurred more rapidlythan the hardplastic one, likely due to the type of plastic composingeach of the waste fractions. The soft plastic sample was composedprimarilyof LDPEandPP, whichrespectivelyyieldmaximumweightloss at 475C and 450C when heated at 10C/min [28] and at 503Cand495Cwhenheatedat 50C/min[1]. Incontrast, thehardplasticssample was mainlycomposed of HDPE andPS, whichdegrade over awider temperature range, reaching their maximum degradation at495C and 506C for HDPE and 420C and 463C for the PS whenheated at 20C/min [16] and at 50C/min [1] respectively. However,theweightlosscurvesofsoftplasticandhardplasticpresentedalmost the same trends, indicating that they had a similar pyrolysisbehaviorduetosimilarchemical structure[1]. It shouldalsobenotedthat another less signicant degradation step (up to 12wt%) wasobserved in hard plastics at a temperature around 300C (displayedinFig. 3(b)). Thisstepischaracteristicof PVCdegradation[16,28], andwas most likely due to PVC content in hard plastic samples.Table4lists several parameters that characterize the devolati-lization process of the samples: devolatilization starting tempera-ture (Tstart), devolatilization ending temperature (Tend) anddevolatilizationtimeaswell asmaximumtemperature(TP1max,TP2maxandTP3max)and da/dTmaxforeachof thethreepeaks.Despite its crowdedness, Table 4 is givento allowrawdatacomparison with future studies. The parameters are reported forseveral heating rates: it is observed that increasing the heating rateresulted in an increase in all temperatures, indicator of the delay ofthe devolatilization process to higher furnace temperatures. As theheating rate was increased from5C/min to 80C/min, Tstart, TP1max,TP3maxandTendof the bers increasedfrom226Cto258C, 347Cto389C, 658Cto763Cand679Cto782C, respectively. For thehardplastics, increasingtheheatingratefrom5C/minto80C/minincreasedTstart, TP2maxandTendfrom242Cto293C, 468Cto524Cand 497C to 554C, respectively. Tstart increased further to 335Cwith increasing heating rate (up to 200C/min), which was likelydue to variations in PVC content between hard plastic samples. Asimilar effect was also observed for the soft plastics: as the heatingratewas increasedfrom5C/minto100C/min, Tstart, TP2maxandTendincreased from335C to 385C, 463C to 495C and 485C to 551C,respectively. One should note that, even if the main degradationTable 4Characteristics of the devolatilization of theber, hard plastic, soft plastic and engineered fuel samples.HRC/min (*) (%) 5 20 30 50 80 100 150 200 300 400TstartC F 3 226 246 253 258 259 260 263 256 258 259HP 4 242 273 269 290 293 289 300 335 322 318SP 4 335 366 374 376 371 385 380 378 370 371EF 3 203 222 239 240 241 253 250 263 272 268TP1maxC F 1 347 367 373 382 389 388 395 385 388 386EF 3 344 363 370 382 375 377 393 396 399 394TP2maxC HP 3 468 465 473 504 502 492 512 505 523 524SP 1 463 483 485 487 481 498 504 496 493 487EF 1 463 488 494 503 506 513 508 514 508 509TP3maxC F 1 658 698 722 727 763 759 770 779 783 772EF 1 683 727 736 764 771 784 798 817 818 811TendC F 1 679 723 744 756 782 777 782 795 800 790HP 3 497 521 527 530 554 541 544 542 554 563SP 2 485 513 518 535 551 528 534 530 527 533EF 2 701 748 761 781 796 804 826 842 852 839P1 da/dTmaxwt%/C F 5 1.20 1.21 1.16 1.09 0.99 1.02 0.94 0.97 0.87 0.83EF 10 0.92 0.73 0.61 0.71 0.62 0.62 0.67 0.63 0.56 0.48P2 da/dTmaxwt%/C HP 15 1.36 1.24 1.22 1.67 1.21 1.21 1.33 1.39 1.34 1.07SP 2 1.93 1.72 1.61 1.37 1.13 1.59 1.53 1.44 1.40 1.33EF 30 0.48 0.55 0.63 0.52 0.49 0.50 0.40 0.31 0.33 0.44P3 da/dTmaxwt%/C F 20 0.077 0.064 0.053 0.051 0.057 0.046 0.037 0.058 0.037 0.030EF 15 0.069 0.056 0.051 0.061 0.057 0.053 0.059 0.054 0.047 0.042Devol. time min F 2 90.70 23.85 16.39 9.98 6.55 5.18 3.47 2.70 1.81 1.33HP 5 51.15 12.40 8.63 4.81 3.27 2.52 1.62 1.03 0.77 0.62SP 5 30.20 7.38 4.83 3.19 2.25 1.43 1.03 0.76 0.52 0.40EF 2 98.56 26.33 17.44 10.82 6.95 5.52 3.85 2.90 1.94 1.43Char yield wt%, d.b. F 10 23.9 22.2 22.3 25.3 26.1 25.1 25.9 23.2 24.2 23.9HP 20 7.1 5.3 7.9 4.5 6.1 7.7 5.6 4.7 5.0 8.7SP 20 5.4 6.1 7.1 6.3 5.5 6.9 5.0 5.3 6.0 6.1EF 10 19.1 19.3 17.6 19.3 19.8 19.9 19.5 19.9 19.2 18.6(*) F:ber samples, HP: hard plastic samples, SP: soft plastic samples, EF: engineered fuel samples.O. Vekemans et al. / Thermochimica Acta 601 (2015) 5462 59step of the hard plastic started at around the same temperature asthe one of soft plastic, due to the early degradation associated withthe PVC, the devolatilizationTstartof the hard plastic was evaluatedbetween 240C and 340C. For all samples, at higher heating rates,the parameters reached a plateau and remained constant.As was explained in the above sections, higher heating rates canlead to error in TGA measurements due to heat transfer limitation.ThisisapparentfromFig. 3, in whichdevolatilizationoccursatprogressivelyhighertemperatureswithincreasingheatingrate.However, despite possibleheattransfer limitationsbetween theTGA and the sample, the rate at which heat was transferred to thesampleincreasedacross theentirerangeof heatingrates (5400C/min) since the devolatilization time decreased continuouslyas observed in Table 4.Several phenomena seemto explain the behavior of thedegradationcurvesonthewholerangeofheatingrates. Atlowheating rate, the signicant sample residence time at lowtemperature can be sufcient to meet the criteria for Tstart(denedin Section 2.3). As the heating rate is increased, though, the sampleresidence time at lowtemperature decreases such that thecontribution of the low-temperature kinetics to the sampledevolatilization decreases: Tstart, and therefore TP1max, TP2max,TP3maxandTend, willtendtoincrease. Butastheheatingrateisincreased further (approximately 80C/min for theber and thehard plastic samples and 100C/min for the soft plastic samples),Tstart, TP1max, TP2max, TP3maxand Tendare no longer affected. This canbeexplainedbytheexistenceofacritical temperature(Tcritical)above which the devolatilization kinetics are extremely fastcomparedtothetimescalesoftheheattransferfromtheTGAto the sample. Above a specic heating rate, the sample residencetime at a temperature below Tcritical became sufciently low suchthat the contribution of the low-temperature kinetics (T