关 键 词：

东方红 75 拖拉机 变速 箱体 端面 钻孔 专用 机床 设计 CAD

资源描述：

东方红—75拖拉机变速箱体三端面钻孔专用机床设计含8张CAD图,东方红,75,拖拉机,变速,箱体,端面,钻孔,专用,机床,设计,CAD

内容简介：

A partial shrinkage model for selective laser sinteringof a two-component metal powder layerTiebing Chen, Yuwen Zhang*Department of Mechanical and Aerospace Engineering, University of Missouri-Columbia, Columbia, MO 65211, United StatesReceived 8 April 2005; received in revised form 6 September 2005Available online 23 November 2005AbstractA partial shrinkage model for selective laser sintering of a metal powder mixture that contains two kinds of metal powders with sig-nificantly different melting points is developed. Laser-induced melting accompanied by partial shrinkage, liquid metal flow driven by cap-illary and gravitational forces, and resolidification of the metal powder layer are modeled using a temperature transforming model. Theeffect of volume fraction of the gas in the sintered region on the sintering process is investigated.? 2005 Elsevier Ltd. All rights reserved.1. IntroductionSelective laser sintering (SLS) of metal powder is a lay-ered manufacturing method that creates solid, three-dimen-sional objects by fusing powdered materials with a directedlaser beam 1. Melting and resolidification are the mecha-nisms to bond metal powder particles to form a layer ofpart and to bond different layers together to form a func-tional part. Fundamentals of melting and solidificationhave been investigated extensively and detailed reviewsare available in the literatures 2,3. Significant densitychange induced by shrinkage accompanies melting in theSLS of the two-component metal powders since the highmelting point powder alone cannot sustain the structureof the powder layer. In addition, the liquid flow of the mol-ten metal in the liquid pool due to capillary and gravita-tional forces also needs to be considered.Pak and Plumb presented a one-dimensional thermalmodel of melting of the two-component powder bed, inwhich the liquid flow driven by capillary and gravitationalforces is considered. Zhang et al. 5 developed a three-dimensional thermal model of SLS of two-componentmetal powder bed and convection driven by capillary andgravitational forces was taken into account. It was assumedthat all the gas initially in the powder bed is driven outupon melting of low melting point metal powder and theheat affected zone (HAZ) was fully densified. The thicknessof the powder bed used in Ref. 5 was very large, whichapproximated the sintering process of the first layer withthe complete shrinkage. Melting and resolidification of a3-D metal powder layer with a finite thickness heated bya moving Gaussian laser beam was investigated numeri-cally by Chen and Zhang 6.Since the lifespan of the liquid pool in the SLS is veryshort, the powder bed may not have sufficient time toachieve complete shrinkage. The rate of the shrinkage inthe SLS process may be between the complete shrinkageand no shrinkage. A partial shrinkage model of SLS oftwo-component metal powder will be developed and theeffects of partial shrinkage on the SLS process will beinvestigated.2. Partial shrinkage modelThe physical model of the problem under considerationis shown in Fig. 1. A Gaussian laser beam scans the surfaceof a two-component metal powder layer with a constantvelocity, ub. A coordinate system whose origin is fixed at0017-9310/$ - see front matter ? 2005 Elsevier Ltd. All rights reserved.doi:10.1016/j.ijheatmasstransfer.2005.09.027*Corresponding author. Tel.: +1 573 884 6936; fax: +1 573 884 5090.E-mail address: Zhangyu (Y. Zhang)./locate/ijhmtInternational Journal of Heat and Mass Transfer 49 (2006) 14891492the center of the laser beam is employed and the problemappears to be steady-state in the moving coordinate system.The initial temperature of the powder layer, Ti, is below themelting point of the low melting point powder, Tm. As thelaser beam interacts with the powders, the temperature ofthe powders is brought up to Tmand then melting occurs.A liquid pool is formed under the laser beam and the meltinfiltrates the unsintered powders driven by capillary andgravitational forces.The loose powder bed contains high melting point pow-der (H), low melting point powder (s), and gas (g,s; gas inloose powder); the volume fraction of all components addup to unity, i.e., uH+ us+ ug,s= 1. Upon melting, theliquid pool contains high melting point powder (H), liquid() of low melting point component, and gas (g, ; gasin liquid region) and their volume fractions satisfiesuH+ u+ ug,= 1; this relationship is also applicable inthe resolidified region if ualso represents volume fractionof the resolidified low melting point component. Theporosity, e, defined as volume fraction of void that canbe occupied by either gas or liquid, is equal to ug,sin theloose powder and it becomes u+ ug,after melting. Ifthe volume of the gas being driven out from the powderbed is equal to the volume of the liquid generated duringmelting, the porosity of the powder bed before and aftermelting will be the same, i.e., e = ug,s= us+ ug, whichis referred to as constant porosity model 4,5. If the vol-ume fractions of high and low melting point powdersbefore sintering satisfy us/(uH+ us) = ug,s, the powderbed can be fully densified (ug,= 0) under constant poros-ity model 5,6. On the other hand, if there is no shrinkage,one would expect that ug,= ug,sand the porosity, e, willincrease because uincreases during melting. The completeshrinkage under constant porosity model and no shrinkagerepresent two extremes in SLS of two-component metalpowders. Under the partial shrinkage model to be devel-oped, the different shrinkage rates during melting isrepresented by the volume fraction of gas in the liquidpool, ug,.The problem is formulated using a temperature trans-forming model 7. The dimensionless energy equationand corresponding initial and boundary equations areub qyx unsintered powderzLow melting point powder sintered regionliquid pool High melting point powder Fig. 1. Physical model.NomenclatureBiBiot number, hR/kHCdimensionless heat capacity, C0=C0HI0laser intensity at the center, (W/m2)Kdimensionless thermal conductivity, k/kHNidimensionless laser intensity, aaI0R=kHT0m?T0i?NRradiation number, eerT0m? T0i3R=kHNttemperature ratio for radiation, T0m=T0m? T0iRradius of the moving laser beam at 1/e (m)ssolidliquid interface location (m)s0location of surface (m)sstsintered depth (m)Tdimensionlesstemperature,T0? T0m=T0m?T0iT0dimensionless surface temperatureT0temperature (K)Ubdimensionlessheatsourcemovingvelocity,ubR/aHVdimensionless velocity vector, vR/aHX, Y, Z dimensionless moving horizontal coordinate,(x,y,z)/RGreek symbolsathermal diffusivity (m2s?1)aaabsorptivityDdimensionless powder layer thickness, d/Reporositygdimensionless location of the solidliquid inter-face, s/Rg0dimensionless location of the surface, s0/Rgstdimensionless sintered depth, sst/Rsdimensionless false time, aHt/R2uvolume fractionSubscriptsggas(es)Hhigh melting point powderiinitialliquid or sintered regionLlow melting point powderssolid of low melting point powder1490T. Chen, Y. Zhang / International Journal of Heat and Mass Transfer 49 (2006) 14891492r ? uVCLT ? UbooXuH usCLT? WsooZuH usCLT? r ? KrT ?oosu usS? r ? uVS?WsooZusS ? UbooXu usS?1? KoToZ Niexp?X2? Y2 ? NRT Nt4? T1 Nt4? BiT ? T1Z g0X2oToZ 0;Z D3oToX 0;jXj ! 14oToY 0;jYj ! 15where the dimensionless heat capacity, C, source term, S,and thermal conductivity, K, and the dimensionless veloci-ties of the liquid phase, V, are available in Ref. 5,6 andwill not be repeated here.The dimensionless continuity equation of the liquid isouos? UbouoX r ? uV _UL6where_ULis dimensionless volume production rate ofliquid.Assuming shrinkage occurs in the z-direction only, thecontinuity equations for the solid phase of the low meltingpoint powder and high melting point powder areousos? UbousoXouswsoZ ?_UL7ouHos? UbouHoXouHwsoZ 08and the following relationship is valid in all regionse us uH 19Adding Eqs. (7) and (8) together and considering Eq. (9),an equation about volume production rate is obtained:_UL ?o1 ? eos Ubo1 ? eoX?ooZ1 ? eWs?10Under the constant porosity model discussed in Ref. 6,only the third term on the right-hand side of Eq. (10) waspresent. Under the partial shrinkage model, the porosity, e,is not constant and the first two terms on the right-handside of Eq. (10) are necessary. Since the shrinkage onlyoccurs at the interface between liquid pool and the unsin-tered region, the velocity induced by shrinkage, Ws, is aconstant in z-direction within the liquid pool and is equalto zero in the unsintered region. The velocity induced byshrinkage in the liquid pool can be obtained by integratingEq. (8) and the result isWs0;Z S1?e?uH;i1?eoSos? UboSoX?;Z S(11where uH,iis initial volume fraction of the high meltingpoint powder, and e= u+ ug,is the porosity in theliquid pool (including resolidified region).The governing equations are discretized by a finite vol-ume method 8 and solved numerically. The powder bed,which includes unsintered powder, a liquid pool, and sin-tered region, has an irregular shape since the upper surfaceof the powder bed recedes due to shrinkage. A block-offapproach 8 is employed to deal with the irregular geomet-ric shape and the thermal conductivity in the empty spacecreated by the shrinkage is zero. The computation was car-ried out using a non-uniform grid in the X and Y-directionsand uniform grid in the Z-direction. The grid number usedin the numerical simulation was 168 37 22.3. Results and discussionThe effects of the different volume fraction of the gas inthe liquid and resolidified regions, ug, on the shape ofFig. 2. Three-dimensional shape of the HAZ and surface temperaturedistribution (D = 0.25, ug,= 0.2, ug,s= 0.42).T. Chen, Y. Zhang / International Journal of Heat and Mass Transfer 49 (2006) 148914921491HAZ are investigated. The initial porosity of the powder,ug,s, was chosen to be 0.42 based on a simple mass/volumemeasurement procedure by Zhang et al. 5. The three-dimensional shape of HAZ and corresponding surface tem-perature distribution for ug,= 0.2 are shown in Fig. 2(a)and (b), respectively. There is still gas remaining in HAZwhen the partial shrinkage occurs. The void space abovethe top surface is generated due to the shrinkage of thepowder layer during melting. Fig. 2(b) shows the surfacetemperature distribution of the powder layer. The temper-ature at the top surface of the liquid pool is above the melt-ing point of the low melting point powder and decreasesrapidly when the distance from the center of the laser beamincreases.The effects of ug,on the shape and size of HAZ aredemonstrated using the longitudinal and cross-sections ofHAZ shown in Fig. 3. The cases of the complete shrinkage(ug,= 0), no-shrinkage (ug,= 0.42) and partial shrink-ages in between were investigated. For giving dimensionlesslaser beam intensity and scanning velocity, the sinteringdepth increases with decreasing ug,. The size of the liquidpool is also growing with decreasing ug, because lowerug,results in higher thermal conductivity in the liquid.The largest sintering depth and size of HAZ are obtainedwhen ug,is equal to zero, i.e., the complete shrinkage. Inorder to obtain the same sintering depth, the larger laserpower or lower scanning velocity are needed for largerug,. The comparison between the numerical solution forthe complete shrinkage and experimental result was per-formed by the authors 6; the cross-section of the HAZin the simulation result approximately 20% was larger thanthat of the experimental result because the HAZ was por-ous in the experimental results. The cross-sectional areaof the HAZ predicted by using the partial shrinkage modelwith ug,= 0.2 agreed very well with the experimentalresults.4. ConclusionA partial shrinkage model for selective laser sintering ofthe two-component metal powders is developed and theeffects of the volume fraction of the gas in the liquid or sin-tered regions, ug, on the shape and size of heat affectedzone (HAZ) were investigated. The results indicate thatthe sintering depth and volume of HAZ significantlyincreases with decreasing ug,.AcknowledgementSupport for this work by the Office of Naval Research(ONR) under grant number N00014-04-1-0303 is gratefullyacknowledged.References1 J. Conley, H. Marcus, Rapid prototyping and solid freeform fabrica-tion, J. Manufact. Sci. Eng. 119 (1997) 811816.2 R. Viskanta, Phase change heat transfer, in: G.A. Lane (Ed.), SolarHeat Storage: Latent Heat Materials, CRC Press, Boca Raton, FL,1983.3 L.C. Yao, J. Prusa, Melting and freezing, Adv. Heat Transfer 25 (1989)196.4 J. Pak, O.A. Plumb, Melting in a two-component packed bed, J. HeatTransfer 119 (1997) 553559.5 Y. Zhang, A. Faghri, C.W. Buckley, T.L. Bergman, Three-dimensionalsintering of two-component metal powders with stationary and movinglaser beams, J. Heat Transfer 122 (2000) 150158.6 T. Chen, Y. Zhang, Three-dimensional modeling of selective lasersintering of two-component metal powder layers, J. Manufact. Sci.Eng. 128 (2006), in press.7 Y. Cao, A. Faghri, A numerical analysis of phase change problemsincluding natural convection, J. Heat Transfer 112 (1990) 812816.8 S.V. Patankar, Numerical Heat Transfer and Fluid Flow, McGraw-Hill, New York, 1980.X-3-2-1030.00.40.5stX-4-200.00.40.500st(b) 2142(a)Fig. 3. Effect of ug,on the HAZ.1492T. Chen, Y. Zhang / International Journal of Heat and Mass Transfer 49 (2006) 14891492Materials Science and Engineering A 435436 (2006) 172180Effects of processing on microstructure and properties of SLS Nylon 12H. Zarringhalama, N. Hopkinsona, N.F. Kampermanb, J.J. de VliegerbaRapid Manufacturing Research Group, School of Mechanical and Manufacturing Engineering, Wolfson Building,Loughborough University, Loughborough LE11 3TU, United KingdombTNO Science and Industry, De Rondom 1, P.O. Box 6235, 5600 HE Eindhoven, The NetherlandsReceived 13 February 2006; received in revised form 9 June 2006; accepted 21 July 2006AbstractThere currently exists the requirement to improve reproducibility and mechanical properties of SLS Nylon parts for rapid manufacturing(RM). In order to achieve this, further fundamental research is needed and this paper addresses this need by investigating effects of potentialsources of the lack of reproducibility and reports effects in relation to crystal structure, microstructure, chemical structure (molecular weight) andmechanical properties. Different ? crystal forms were identified and related to the unmolten particle cores and the melted/crystallised regions of themicrostructure.Themeltpointofthe?-formvarieddependingonprocessingconditions.Observabledifferenceswerealsopresentwhencomparingthe microstructure of the parts. Molecular weight of parts was significantly higher than virgin powder but used powder also showed an increase inmolecular weight. This was related to improved elongation at break of parts built from the used powder, consistent with previous studies. Tensilestrength showed some increase with machine parameters selected for improved strength but Youngs modulus values were broadly similar. 2006 Elsevier B.V. All rights reserved.Keywords: Selective laser sintering; Nylon 12; Crystal structure; Microstructure; Molecular weight1. IntroductionRapidmanufacturing(RM)isafamilyoftechnologieswhereproducts are made in an additive way directly from 3D CADdata, without the need for tooling. The technology was origi-nally known as rapid prototyping (RP) since the properties ofparts produced were generally only suitable for prototype parts.However, the technology has evolved to the point where RM isaviablemanufacturingtechniqueinnumerousapplicationareas1,2.Selectivelasersintering(SLS)isacommonlyusedtechniquein RM 3 and has proved to be suitable for various applicationsincluding the creation of bespoke hearing aids 4 and parts forFormula 1 racing cars 5. The aerospace industry in particularhas recognised the advantages of SLS for RM where it is cur-rentlythemostwidelyusedtechnique6.Someaircraftalreadyhave numerous SLS production parts as standard 3,7 howeverthe applications are still limited, partly due to the mechanicalproperties of parts produced.Corresponding author. Tel.: +44 1509 227568; fax: +44 1509 227549.E-mail address: mmhzlboro.ac.uk (H. Zarringhalam).Selective laser sintering produces parts by using a laser toselectively sinter individual layers of a material in a powderform (polymers, metals, ceramics). SLS systems are currentlyavailablecommerciallyfromtwodifferentmanufacturerswhichare3DSystemsoftheUnitedStates(previouslyDTM)andEOSof Germany. Prior to build a CAD model must be created andthen processed which includes slicing the CAD model into0.10.15mm thick 2D layers. After this the data is sent to themachineforpartbuildingwhichisdescribedinTable1.Follow-ing build the part is removed from the unsintered powder andloose particles brushed and/or sprayed off gently using com-pressed air.Semi-crystalline polymers (predominantly Nylon 12 and 11)can be successfully sintered with superior mechanical proper-ties than amorphous polymers 8,9. However, shrinkage duringcrystallisationhindersproductionofaccurateparts10.ForthisreasonitisessentialthatformaterialstobeprocessedbySLSthemelt temperature be considerably higher than the crystallisationtemperature so that crystallisation can be delayed and reducedduring the build process 11 to allow new layers to bond to pre-vious layers with a more homogenous microstructure. A highenthalpy of fusion is also preferable to prevent melting of pow-der particles local to the particles targeted by the laser due toconduction of heat. In addition, during laser sintering, a narrow0921-5093/$ see front matter 2006 Elsevier B.V. All rights reserved.doi:10.1016/j.msea.2006.07.084H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180173Table 1Part build stepsMain part build steps1. Thin layer of powder deposited across part bed2. Laser draws cross-section matching corresponding layer in STL file,bonding particles and adjacent layers3. Platform in part-build cylinder moves part downwards a layer4. Steps 13 repeated until build completed5. Cooling down of part cakemelttemperaturerangeincombinationwithalowmeltviscosityisrequiredtoachievethenecessaryleveloffluidityveryquicklywithout inputting excess energy 11.Molecular weight is well known to be a critical characteristicwhendescribingpolymersingeneral12andtheweightaveragemolecular weight (Mw) directly affects the melt viscosity 10.The melt flow procedure is used to characterise melt viscosityby the melt flow index (MFI). Low MFI is associated with highmelt viscosity and high molecular weight. Gornet et al. 13 andShi et al. 10 investigated the effects of powder melt viscosityand related lower MFI (and therefore higher melt viscosity andMw) to improved mechanical properties particularly elongationat break (EaB). It is standard practice in industry to blend virginpowder with used to maintain acceptable part properties whichdeterioratewithcontinueduseofusedpowder13.Thethermalload during the build process is presumed responsible for theincrease in molecular weight and thus increased melt viscosity.Virginpowderisnormallynotusedonitsowninordertoreduceassociatedcosts.ForRMconsistencyinmechanicalpropertiesisparticularlyimportantandsoitisimportanttotackletheexistingpotential for lack of reproducibility of properties.Nylon 12, as with other aliphatic Nylons, has been widelyshown to form stable crystal structures of either the ?- or ?-form 14,15. For Nylon 12 the ?-form, which is more stable,always melts at a higher temperature than the ?-form 16. Ithas been demonstrated that under the majority of processingcondition the liquid phase will crystallise to the ?-form, whichis therefore the most commonly found crystal form in Nylon 12.However, certain conditions do result in the ?-form. These havebeen reported as solution casting (at atmospheric or reducedpressure, below 90C) 16, drawing just below the melt pointand high pressure crystallisation 21. It has been demonstratedthattheproportionsof?-and?-formcanbealtered(from0%to100% for each) by controlling the thermal history 16,17. Dif-ferent variations of the basic ?- and ?-crystal forms have beencreated depending on processing conditions and these can betransformed from one to another by numerous methods depend-ing from which and to which crystal form the material will betransformed 1821.Studies reporting melting points for the ?-form werereviewedbyAharoni15andvaluesrangedfrom172to185Cwith most data points close to 179C. The melt point of theless common ?-form has been reported as ca. 173C 16. SLSNylon is obtained by dissolving polyamide-12 in ethanol underpressure at elevated temperatures followed by slow crystalliza-tiontherebyformingcrystalswitharelativelyhighmeltingpointof ca. 190C and a relatively high heat of melting 9,11. Thecrystal structure of SLS Nylon powder or parts has not beenreported on however the slow cooling rate during productionwould likely result in the virgin powder being comprised of rel-atively large crystals. The production process for SLS Nylon 12does not exactly meet the requirements for formation of ? crys-tal form and the high melt temperatures measured indicate the?-form.Published information relating to alternative methods formanufacture of SLS powder describes how standard Nylon 12can be modified to increase the melt temperature with insignifi-cantincreasetothecrystallisationtemperature22.Theprocessdescribed involves heating powder or granules of the material insteamforextendedtimeduration(upto100h).Itisclaimedthatthis increases the melt temperature by allowing the molecularchainstoberearranged.ItisknownthatevennumberednylonssuchasNylon12canonlyachievefullinter-molecularhydrogenbonding and thus crystallisation by the molecular chains align-inginananti-parallelarrangement14anditisthoughtthatthisis how the chains are rearranged.Youngs modulus and tensile strength of selective laser sin-teredNylon12arecomparablewithvaluesforstandardinjectionmouldedsamples23.HoweverEaB(whichindicatesductility)isatleastanorderofmagnitudelower.Gornetetal.13demon-strated how mechanical properties can vary between differentmachines and even for the same machine in different builds.Salehetal.24andGibsonandDongping25showedconsider-ablevariabilityfordifferentpartorientationandpartsbuiltindif-ferent locations in the build volume within the same build. Ton-towi and Childs 26 investigated the effect of powder bed tem-perature (ambient build powder surface temperature) on densityand showed that small variations in temperature have a markedeffect on part density which can affect mechanical properties.TheSLSmachinemanufacturerssupplypowderandtheirkeypublished mechanical (and other) properties, for their respec-tive SLS Nylon 12 powders, are summarised in Table 2. Sincethe machine parameters used to generate the data are not givenby both manufacturers and because the data is generated usingdifferent standards, this cannot be compared directly. It shouldtherefore only be used to roughly gauge the potential of SLSNylon 12. Where manufactures quoted a range of values, theaverage has been given.From the application of SLS as a production process, therequirement to improve reproducibility and mechanical proper-ties of SLS Nylon parts for RM exists. In order to achieve this,furtherfundamentalknowledgeoftheunderlyingmechanismsisneeded. This paper addresses this need by investigating effectsTable 2Manufacturers published materials properties for SLS Nylon 12 27,283D SystemsEOSTensile strength (MPa)4445Tensile modulus (MPa)16001700Tensile elongation at break (%)920Part melting point (C)184184Particle size, average (?m)5858Particle size, range 90% (?m)2592Not availablePart moisture absorption, 23C (%)0.410.52174H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180of build procedure, parameters and powder blend and reportseffects in relation to crystal structure, microstructure, chemicalstructure (molecular weight) and mechanical properties.2. MethodologyDifferent specimens of SLS Nylon 12 powder and partswere analysed. Powders with different thermal histories wereanalysed and parts were produced from these. These are sum-marised in Table 3 and explained further in Section 2.1. Partsand powder were obtained from TNO Science and Industry andthe Rapid Manufacturing Research Group, Loughborough Uni-versity. These are referred to as TNO and Lboro, respectively.Different machines were used which were already configuredwith contrasting parameters. This was intended to provide awiderrangeofdata.EachmachinewasusedwithSLSNylon12powder supplied by the respective machine manufactures. Thedifferences are explained further in Section 2.2.The powder specimens were analysed for molecularweight distribution by gel permeation chromatography (GPC),thermal properties and crystal structure by differential scan-ning calorimetry (DSC), and microstructure by optical micro-scopy. Part specimens were also tensile tested for mechanicalproperties.2.1. Preparation of powderPowder was prepared in different ways prior to building, asfollows:1. Virgin: Powder that has never been processed in a laser sin-tering machine.2. Used: Part cake and overflow powder that has been througha complete SLS run.3. Refreshed: A mix of 67% used and 33% virgin powder.2.2. Part productionStandardtensiletestspecimens(asshowninFig.1)werepro-duced according to ISO 527-2 1A. The parameters of the TNOmachine were set for optimized mechanical properties whilethose for the Lboro machine were configured for optimizedaccuracy which was anticipated to give reduced mechanicalproperties.Bothmachinescanbeconfiguredforeitherimprovedmechanical properties or improved accuracy and so the choiceTable 3Specimen historySourceSpecificsSpecimensPowderParts fromTNOEOS PA2200 powder, EOS P380machine (tuned for mechanicalproperties)VirginVirginUsedRefreshedRefreshedLboro3D Systems Duraform PA powder,3D Systems Vanguard machine(tuned for accuracy)VirginVirginUsedUsedRefreshedRefreshedFig. 1. Build setup.of configuration for this research was simply based upon theexisting configurations already in use by the different operators.ThepowderbedtemperatureoftheLboromachinewasadjustedfor each type of powder while the powder bed temperature ofthe TNO machine was kept constant for all powder types. Dif-ferencesbetweenthetwodifferentmachinesandtheirparticularbuild setups are listed in Table 4. In order to ensure thermallystable conditions a base layer of two centimetres and a top layerof one centimetre were used for each run. Parts were built flat,laid parallel to the machine x-axis, at fixed places in the middleof the build area as shown in Fig. 1.2.3. Molecular weightThe molecular weight was determined by GPC by drying25mg samples at 120C for 1h. The samples were then dis-solved at room temperature in 5ml chloroform (stabilized withambilene) and 0.5ml trifluor acetic acid anhydride (TFA). After2h dissolving the solution was diluted with 20ml chloroform.This solution was filtered without heating and analysed usingtwo WatersTMStyragelHT6E, 7.8mm300mm columnsin combination with a differential refractive index detector(WatersTM410) and a tuneable absorbance detector (WatersTM486) at wavelength 264nm. Polystyrene was used as the refer-ence material.Table 4Machine differencesTNOLboroParameter optimisationMechanical propertiesAccuracyMachineEOS P3803D Systems VanguardMaterialPA2200 Nylon 12Duraform PA Nylon 12Layer thickness0.15mm0.1mmPowder depositionmechanismHopperRollerFeed powderCold (50C)Heated (+100C)Build platformHeatedUnheatedLaser power (fill)45.7W (90%, 50W)11WLaser power (outline)10.9W (20%, 50W)5WScan speed4000mm/s5000mm/sScan spacing0.3mm0.15mmH. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 1721801752.4. Thermal propertiesDifferential scanning calorimetry (DSC) was used to deter-mine thermal properties in order to study melting and crystal-lizationbehaviourandtoidentifycrystalforms.Samplesofmassca. 10mg were analysed in a Thermal Analysis DSC machineunder nitrogen flow. Samples were heated from room temper-ature at 10C/min to 200C and then immediately cooled at10C/min.2.5. MicrostructureThe microstructure of the processed materials was analyzedvia optical microscopy. Samples of 35?m thickness were cutat room temperature by microtome. Images were taken in brightfield polarized illumination.2.6. Mechanical propertiesTensiledatawasrecordedusingaclip-onextensometer.Eachtensile test was executed with five individual specimens andresults for tensile elongation at break, ultimate tensile strengthand Youngs modulus were recorded.3. Results and discussion3.1. Powder3.1.1. Virgin powder: comparison of Lboro and TNOpowder by thermal analysis and molecular weight analysisFig. 2 shows DSC results for Lboro and TNO virgin pow-der. Both show relatively sharp single endotherms with meltingpeaks at 189C for TNO virgin powder and 188C for Lborovirgin powder. These correspond with known values for virginpowder 29,11 however they are at least 3C higher than thepreviously reported values for crystal forms of non-SLS pro-cessed Nylon 12 (172185C).Fig. 2. DSC results for Lboro and TNO virgin powder.Fig. 3. GPC results for Lboro and TNO virgin powder.As mentioned previously, the conditions in the production ofSLSNylon12powderdifferfromthoserequiredtoform?-formcrystal structure and this suggests that the endotherms obtainedfrom the virgin powder relate to the common ?-form. On thebasis that prolonged molecular mobility increases the melt tem-peratureitcanthereforebeinferredthattheendothermsrelatetothe?-formwithfarlargerandmoreperfectcrystalstructure(duetoincreasedhydrogenbonding)thanpreviouslyanalysedNylon12 specimens. Additionally, Ishikawa and Nagai 16 demon-stratedthatasingleendotherminaDSCtestonNylon12isonlycharacteristic of a sample containing 100% ? crystal form sincesamples consisting of 100% ? crystal form invariably develop asecond melt peak during the DSC run due to recrystallisation ofa portion of the specimen to the ? crystal form. The test param-eters were the same as those used in the present investigationsupporting the view that the crystal structure present in virginSLS Nylon 12 material is of the ? crystal form.Fig. 3 shows GPC results for Lboro and TNO virgin powder.The relative weight average molecular weight is 90,000g/molfor Lboro virgin powder and 70,000g/mol for TNO virgin pow-der. The polydispersity index is 2.26 for Lboro virgin powderand 2.34 for TNO virgin powder. The small peaks on the posi-tivetailofeachcurvearepossiblycharacteristicofanti-oxidants,whichareknowntobeblendedwiththevirginNylon12powderfrom EOS 28 and therefore presumed to be the same with 3DSystemssuppliedpowder.HighermolecularweightoftheLboropowdercouldtheoreticallyhinderprocessingandthusproperties30 however the difference is too small to be significant.3.1.2. Sub-melt temperature heated powder: comparison ofrefreshed and virgin powder by thermal analysis andmolecular weight analysisFig. 4 shows DSC results for virgin and refreshed TNO pow-der. The refreshed powder shows a very slight increase in melttemperature(1C)indicatingmoreperfectcrystalsduetocrys-talline reorganisation. Crystallinity of the virgin and refreshed176H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180Fig. 4. DSC results for virgin and refreshed TNO powder.powder was measured as 55% and 51%, respectively withmelt enthalpies of 115.7 and 106.6J/g, respectively. The meltenthalpy for 100% crystalline material was taken as 209.3J/g31.Theslightdecreaseincrystallinitywiththerefreshedpow-der instead of an increase supports the view that the increase inmelt temperature is due to crystalline reorganisation rather thanincreased crystallinity. The endotherm for the refreshed powdershows a slight bulge to its left indicative of a polymorph likelydue to the 1/3 virgin fraction of the blend.Fig. 5 shows GPC results for virgin and used TNO powder.The relative weight average molecular weight is 170,000g/molfor used TNO powder and 70,000g/mol for virgin TNO pow-der. The polydispersity index is 2.84 for used TNO powder and2.34forvirginTNOpowder.Thisincreaseinaveragemolecularweight and molecular weight distribution indicates that poly-merization occurs in the solid state powder. This may be oneof the mechanisms leading to modified sintering characteristicswith the reuse of SLS powder. For example, reduced molecularmotionduetoincreasedmolecularweightcouldhindercrystalli-sation 18. Since the objective of this analysis was to determineeffects of heating the powder without melting, only the TNOFig. 5. GPC results for virgin and used TNO powder.Fig. 6. Microstructure of Lboro part.material was analysed. This was because it was expected thatany change would be more pronounced than with the Lboromaterial due to the increased thermal load on the material.3.2. Parts3.2.1. Optical analysis of part microstructureFig.6showsamicrographofthecross-sectionofaLboropartwith annotations on the right hand side. The annotated featuresare also found on TNO specimens. The top of the image showsthe edge of the part to which unmolten particles have fused.Moeskopsetal.32havedemonstratedhowthepartmicrostruc-ture is composed of particle cores surrounded by spherulites.These cores were described as the unmolten central regions ofparticles occurring when particles do not receive enough heat tofully melt. Fig. 6 shows this and also shows spherulites withoutcores which formed from particles which were small enough tofully melt. Moeskops et al. 32 stated that the presence of thesecores is believed to be critical to the properties and mechanicalresponse of SLS processed materials.3.2.2. Comparison of virgin powder and parts by thermalanalysis and molecular weight analysisFig.7showsthreesuperimposedsetsofDSCresultsforTNOmaterial. The results are for virgin powder, a part built fromthe virgin powder and a part built from virgin powder whichwas exposed twice (a second exposure of the same area priorFig. 7. DSC results for TNO virgin powder, part and double exposed part.H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180177Fig. 8. Identification of crystal forms.to deposition of the next layer) and which therefore receivedincreased energy input. The curve for virgin powder is the sameas in Fig. 4 and shows one melt peak at 189C. The curve forthe normal part shows two peaks, smaller than the single peakfor virgin powder. The larger of these two is at 185C and thesmaller at 190C. The curve for the double exposed part onlyshows one peak which is at 185C.TheinterpretationofFig.7isthatforpartsthepeaksat185Crelate to a ? crystal form, arising from the melting and crystalli-sation of virgin powder. The small peak at 190C for the partsrepresents a reduced total fraction of the initial ?-form of thevirgin material and demonstrates a shift of 190189C, similarto that shown in Fig. 4 (comparing virgin and used powder).Fig. 8 shows a DSC curve for a Lboro part built from virginpowder and an optical image of one particle core with surround-ing spherulite. This illustrates the concept that the DSC peakat 190C directly represents the unmolten particle core (whichis essentially therefore used powder material) and that the sur-rounding spherulitic region relates to the material which hasbeen melted and then crystallised.Fig. 9 shows GPC results for TNO virgin powder and apart built from TNO virgin powder. A distinct increase of rel-ative molecular weight and molecular weight distribution dueto polymerization is observed as quantified by an increase from70,000 to 230,000g/mol weight average molecular weight andan increase from 2.34 to 3.32 for the polydispersity index. Thisincrease in molecular weight with sintering might be desirablesince higher molecular weight generally results in improvedmechanical properties. Comparing Fig. 9 with Fig. 5 the molec-ular weight of the used powder falls in between those for virginpowder and the part built from virgin powder due to the natureof the polymerisation (i.e. solid state and liquid state for usedpowder and part, respectively).3.2.3. Comparison of Lboro and TNO parts built fromvirgin powder by thermal analysis and optical microscopyFig.10showsDSCresultsforTNOandLboropartsbuiltfromvirgin powder. The curve for the TNO part is the same curve forFig. 9. GPC results for TNO part and virgin powder.the Normal part in Fig. 7 showing two peaks, one at 185 and asmaller peak at 190C. The curve for the Lboro part also showstwo peaks, at 189 and 181C. Both of the smaller peaks are ca.1C higher than their respective peaks for virgin materials asshown in Fig. 2 demonstrating that the unmolten fraction in thepart modifies in a similar manner to how the unmelted virginpowder modifies to its used state. The difference between thelarger peaks indicates significant difference in crystal size of thematerial melted and crystallised for the Lboro and TNO setups.Thisdifferencecouldarisefromthedifferentparameterschosenfor this study (such as heated/unheated build platform) and/orinherent differences between the machine setup.Fig. 11 shows microtomed sections of parts built from TNOvirgin powder (a), Lboro virgin powder (b) and parts injectionmoulded from virgin TNO powder (c). Comparing the imagesfor TNO and Lboro virgin material the Lboro specimen showsunmoltencoresclearlycontrastedagainstthesurroundingmate-rial which consists of an evenly distributed spherulitic structure.These cores are also present in the TNO specimen but are lessclearly defined. There is a clear step change between the coreand the surrounding phase for the Lboro specimen comparedwith a more continuous change from one to the other for theFig. 10. DSC results for TNO and Lboro parts built from virgin powder.178H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180Fig. 11. Microstructure of TNO and Lboro parts built from virgin powder and injection moulded specimen.TNO specimen. The Lboro specimen also has larger cores andtherefore the increased volume fraction of crystals (with a pre-dominant 189C melting peak) within the material may alsocontribute to variations in mechanical properties.The presence of cores may be inherent to SLS processedmaterial but higher energy input (for example TNO speci-mens) reduces the core size. A higher degree of meltingoccurring in the TNO specimen (Fig. 11a) could indicate bet-ter fusion/sintering of particles and hence superior mechanicalproperties. The injection moulded specimen clearly shows noparticle cores and a much more even and fine microstructuredue to the feedstock being fully melted (and presumably morefluid) and followed by faster cooling.3.2.4. Mechanical propertiesFig. 12 shows the results for elongation at break (EaB) forparts made with different powders on the different machines.Fig. 12 shows (for both TNO and Lboro parts) a significantincreaseinEaB(anindicatorofductility)forrefreshedandusedpowder compared with virgin powder. Distinct trends can beobserved relating increased EaB to higher proportions of usedpowder and parameters selected for improved strength. Theseresults match those of previous studies 13.The increased molecular weight of the used powder may beresponsible for the improved ductility since it would directlyrelate to the molecular weight of the particle cores and mayhavehinderedcrystallisationofthesurroundingmeltedandcrys-Fig. 12. Tensile elongation at break.tallised regions. The increase in EaB of Lboro refreshed partsfrom Lboro virgin parts is 94% of the virgin-part value. Theincrease for the TNO parts is only 36%. This may be due tothe increased powder bed temperature for Lboro parts due to theadjustmentforeachpowdertype.Howeveritcouldalsorelatetothe larger cores and hence building with more used powder (i.e.refreshed) would have a more significant effect on mechanicalproperties. To reiterate, the larger the core the greater the signif-icance of the pre-melting powder history (e.g. powder blend).Likewise,thesmallerthecore(i.e.greaterfractionofmeltedandcrystallised region) the greater the significance of build relatedfactors (e.g. build parameters).Fig. 13 shows the results for ultimate tensile strength of partsmade with different powders on the different machines. Fig. 13indicates a difference in mechanical stress response betweenpartsbuiltfromvirgin,refreshedandusedpowder.Thevariationinultimatetensilestrengthbetweenvirginandrefreshedmaterialis small however there is a marked increase in tensile strengthfor used powder. Tensile strength values are higher for the TNOparts reflecting the parameters used (see Table 4).Fig.14showstheresultsforYoungsmodulusforpartsmadewith different powders on the different machines. Fig. 14 showsparts from refreshed powder having almost the same Youngsmodulus as parts from virgin powder for TNO parts but withLboro parts the parts built with refreshed powder show a lowerYoungs modulus. Youngs modulus values are slightly higherfor the TNO parts.Fig. 13. Tensile strength.H. Zarringhalam et al. / Materials Science and Engineering A 435436 (2006) 172180179Fig. 14. Youngs modulus.4. Conclusions and recommendationsMaterials research in the context of improving reproducibil-ity of properties in SLS was conducted by building parts usingdifferent machines, with contrasting build parameters and dif-ferent powders with varying blends. This study cannot be usedtodirectlycomparethetwomachinessincedifferentparametersand powder were used. To compare the machines against eachother a study could be performed using identical powder (fromthesamebatch),buildingpartsoneachmachineusingtheircom-plete range of parameters, i.e. strong parts and accurate partscould be built on each machine and compared.Effects of the various factors on the microstructure of SLSparts were analysed. Parameters chosen for improved accuracybuilt on a 3D Systems machine (Lboro) resulted in well definedunmoltenparticlecores(virginmaterial)andparameterschosenfor improved mechanical properties built on an EOS machine(TNO)gaveasmoothertransitionbetweentheunmoltenparticlecores and the melted/crystallised region. These differences maybe due to the processing conditions, but could also be affectedby inherent thermal conditions unique to each machine or to thespecific powder material properties.The crystal structure of the material was investigated usingDSC to identify the crystal forms. Unmolten particle cores werematched to used powder material and have, most likely, the? crystal form. The melting point was measured between 188and 190C which is higher than commonly reported values fornon-SLS processed Nylon 12 indicating particularly large crys-tals. The melted/crystallised regions were observed with lowermelting points (181 and 185C) than virgin powder which liewithin the known range for the ?-form. Both ? crystal formswere matched to distinct regions within the materials observ-able microstructure. The difference between the melting pointsfor the melted/crystallised regions are presumably due to differ-ent thermal processing conditions. Slower cooling for the TNOparts (due to parameters and/or inherent machine behaviour)likely resulted in larger crystals hence the higher melt point.Analysis by X-ray diffraction would allow detailed characteri-sation of the crystal structure.Relative molecular weight of powder and parts was reportedon. A significant increase in molecular weight was observedwith building parts and it was also shown to increase with heat-ing of the powder below the melt point (used powder) due tosolid state polymerisation. Lboro virgin powder had a slightlyhigher molecular weight than TNO virgin powder though itssignificance is uncertain. It should be noted that batch to batchvariation in ostensibly identical powder (e.g. different batchesof Duraform) has not been considered and so the significance ofthe variation between the Lboro and TNO powder observed iscurrently unknown.The most significant trends observed with mechanical prop-erties were the changes to EaB. TNO parts showed higher EaBthan Lboro parts and, within each group, building with usedpowder resulted in improved EaB. It is important to note thatother important mechanical properties of SLS parts such as ten-sile strength and Youngs modulus are similar to those obtainedfor traditionally processed samples such as injection mouldingwhileEaBofSLSparts,however,lagsfarbehind.Thishighlightsthe importance of investigating SLS material characteristics inrelationtoEaB.Inthisstudyincreasedmolecularweightofusedpowder was related potentially to improved EaB of parts. Otherfactors could also be considered however. This paper demon-strated the variability typically encountered in SLS of Nylon12 material and provided commentary and analysis. Molecu-lar weight had been considered indirectly in other studies usingthemeltflowproceduretodeterminemeltviscosity.Thispresentstudyprovidedaveragemolecularweightvaluesforthefirsttimehoweverfurtherstudyisrequiredtounderstandproperlytherela-tionship between molecular weight and the variability observedin mechanical properties. In depth study in this specific areashould comment on such aspects as changes in lamellae thick-ness distribution, chain entanglement density and tie-moleculedensity.The porous nature of SLS parts is understood to be at leastpartly responsible for the low EaB relative to that of conven-tionally produced parts 33. Another potential issue is that ofvarious boundaries and interfaces within the part. The pres-ence of unmolten particle cores leads to a composite-like struc-ture. As such, inter-crystal-form and, perhaps, inter-spheruliteinterfaces may have varying influence on mechanical proper-ties.Astronginter-crystal-forminterface(betweenparticlecoreand surrounding melted/crystallised region) could lead to thehighly crystalline ?-form being locked in a ?-form matrix ofthe melted/crystallised region with the micro-level mechanicalproperties of each ?-form particle core having a strong positiveeffect at the macro-level, i.e. the mechanical properties of theSLS part. A weak inter-crystal-form interface however couldleadtoruptureofthecorefromthematrixunderrelativelyweakstress resulting in rapid crack formation and propagation andthus inferior mechanical properties such as low EaB, i.e. brittle-ness.Further work could focus on in depth characterisation of thecrystal forms present in SLS processed material and the preciseeffects of different processing and other parameters on these.It would be interesting to investigate methods to quantify theamount of particle cores in specimens. Material interfaces, par-ticularly that between the particle core and melted/crystallisedregion, could also be investigated. Effects of molecular weightonmeltviscosityandtheireffectsonprocessingandmechanicalproperti