Aero-Thermal Research Particulars in Ultra-Micro Gas Turbines

Transcription

1 Toshio NAGASHIMA nd Susumu TERAMOTO Deprtment of Aeronutics nd Astronutics University of Tokyo 7-3-1, Hongo, Bunkyo Tokyo JAPAN Chischi KATO Institute of Industril Sciences University of Tokyo JAPAN Sburo YUASA Deprtment of Aerospce Engineering Tokyo Metropolitn Institute of Technology Ashigok 6-6 Hino City Tokyo JAPAN ABSTRACT Aerodynmics in turbo-components nd mtched combustor chrcteristics relting to the design of ultrmicro gs turbines (UMGT) were presented. Some detiled nlysis were provided for the effects of het trnsfer, tip clernce nd geometricl restriction imposed to UMGT flow pssges, s well s leding concept of combustor design to chieve micro flme structure for enhnced efficiency nd stbility. 1. INTRODUCTION 1.1 UMGT for Dynmic Power Conversion Compctness is mndtory for mobile devices, nturlly their power source included. Bttery is most conveniently used in current mobile infocommuniction devices, but not stisfctory for prolonged use becuse of poor specific energy (W-hr /kg). Amongst vrious kinds of power conversion systems (Tble 1) [1], internl combustion engines re superb in specific energy s well s specific power (W/kg), which cn be ttributble to distinguished specific chemicl energy tht fuels contin. It looks therefore very dvntgeous nd on trget to explore the extreme in hrdwre minituriztion of gs turbine system tht hs been predominntly developed for ero-engines nd lrge electricity genertion. But, how smll cn it be, suitble for ctuting robots nd UAV propulsion? Specific power [W/ kg] Smll (Bttery) Mobile Compct (Fuel Cell) Durble Specific energy [W-hr/ kg] Ngshim, T.; Termoto, S.; Kto, C.; Yus, S. (2005) Aero-Therml Reserch Prticulrs in Ultr-Micro Gs Turbines. In Micro Gs Turbines (pp ). Eductionl Notes RTO-EN-AVT-131, Pper 3. Neuilly-sur-Seine, Frnce: RTO. Avilble from: RTO-EN-AVT

2 Menwhile, n innovtive concept of Power-MEMS turbines ws proposed by MIT [2] in 1997, since then lot of work relting to MEMS fbriction technology hve been mde to chieve minituriztion in gs turbines to such n extent s competitive size with disposl btteries, even smller s shirt button. Tble 1: Power Conversion Systems [1] Output Input Electricity Het Chemicl Photons Kinetic Electricity Het Btteries Fuel cells Mgnets (Superconducting) Inductors Thermoelectrics Thermionics Genertors Fuel cells Ohmic heters Het pumps Phse- chnge High Cp Electrolysis Ioniztion / Recombintion Thermo-chemicl electrolysis Chemicl Fuel cells Combustors Propellnts Explosives LED s Dischrges Light bulbs Lsers & Trnsistors Microwves Rditors Chemicl Lsers Motors Thermodynmic cycles Jet Propulsion Photons Photovoltic cells Therml Absorbers Photolysis Electrolysis Resonnt cvities Metstble toms Rdiometers Kinetic Genertors Friction Impct ioniztion Triboluminescence Flywheels Combustion Engine relted 1.2 Performnce Improvement Fig. 1 presents trend of gs turbine efficiency with respect to output for electricity [3], wherein ech plot corresponds to prticulr gs turbine for industril use or t reserch level. Chmpion dt t kW level re the plots of cermic gs turbine (CGT) tested for utomobile use, so tht ll units instlled het exchnger. Micro gs turbines (MGT) re seen to be plotted round CGT, which is hopeful sign s distributed energy sources. The principl reson of deteriortion in efficiency for smller gs turbines is mostly concerned with mteril strength ginst high temperture, since the smller size gs turbines re unble to elborte ir cooling holes inside the limited spce of vnes nd bldes due to mnufcturing difficulty. A mesure needed for improvement is by mens of either thermodynmic cycles or component efficiencies including mechnicl, so tht cermic mterils were one of the importnt choices to rise mximum cycle temperture, nd much efforts were mde to instll n dvnced het exchnger nd to pply ir berings to dispense with hevy oil berings nd uxiliry pump units. Co-genertion s well s combined cycles needs to be conceptulized for prctice to improve efficiency s totl power system, thus, so fr so much efforts towrds the right direction. It is pprent, however, tht further minituriztion to Power-MEMS turbines of tens to hundreds wtts is beyond our chievble trget technology. Then, gin, wht is most criticl to relize UMGT? 3-2 RTO-EN-AVT-131

3 Efficiency(Electricity)% Trget Metl GT Output (kw) Figure 1: Gs Turbine Trend, Efficiency versus Output for Electricity. It needs to be nswered, in the first plce, from the fesibility point of view, tht hs to be exmined crefully s gs turbine system. Finger-top gs turbine engine by UT group [4] gives n nswer tht employs recupertor s the most prcticl choice to lower TIT to utilize conventionl mterils. An dvnced micro chnnel recupertor will be the centrl issue to keep the totl system within llowble compctness. The compctness hs been judged by totl system including fuel tnk. Since 3D design of rotors needs to be dopted for better efficiency t this lower TIT, MEMS fbriction my be not pplicble, so tht rotor dimeter of the minimum 8 mm ws decided for mchining reson. The ltter size is lso bout mtch with the micro recupertor designing. Combustor looks promising s shown lter in this text, but gs berings t high speed rottion beyond 1,000 krpm re yet to be tested. Plm-top gs turbine engine, which is 5 times lrger in size of Finger-top, ws prototyped nd tested s the clibrtion tool for chllenge of such technology explortion. Isomur et l. [5] looked into existing dt on smll turbochrger performnce to conclude tht n engine system is fesible with TIT 1173K nd turbine rotor of 10 mm in dimeter, 870 krpm revolution, compressor nd turbine efficiency, 62.5% nd 64.5%, respectively, nd pointed out need to protect compressor ginst the het from combustor for secure efficiency. MIT group [6] mintin seeking MEMS turbine to run demo engine with lmost double sizing the originl compressor nd turbine rotors which comprise of six silicon wfer lyers, tking further step to rise TIT up to 1600K, yet employing rchitecture of integrted compressor nd turbine disks. Concerning the originl MIT Power-MEMS turbine with rotor dimeter of 4 mm, Ribud [4] mde creful study on detiled het blnces between individul components to check the system energetic integrity, concluding tht doubled dimeter, i.e. 8mm, produces better thn mrginl output, nd tht the het loss t turbine nozzles is most detrimentl to the engine efficiency. It will be very importnt to know tht dibtic flow condition will never be relized in UMGT, simply becuse surfce to flow re rtio is drsticlly incresed, in ddition to incresed het trnsfer rte due to RTO-EN-AVT

4 flows t smll Reynolds number. To mximize output for given fuel input, the key for fesibility exists in het mngement relting the engine rchitecture, which is crucil nd indispensble. The current knowledge bout similrity rules between flow dynmics nd het trnsfer my hve to be re-exmined to pply for precise prediction, control nd improvement of UMGT performnce. 2. RADIAL TURBO COMPONENT PARTICULARS 2.1 Micro Rdil Turbo-Flows Conventionl Rdil Turbo-Flows Designing rdil flow turbo-components with respect to loss sources nd opertion rnge is known to be much dependent upon pst experiences, fr more so thn doing xil ones, becuse of ccentuted flow slip or devition ner blde triling edge, 3D boundry lyer development long blde surfces nd duct wlls, lekge flow through clernce between blde tip nd shroud wll, s well s complexity in flow turn from xil to rdil nd vice vers, ccompnying strong swirl motion. For instnce, pssge vortex develops in rdil turbine rotors resulting in stremline shifts in the rdil direction, i.e. outwrd on the blde pressure surfce nd inwrd on the suction surfce, to chnge the flow pttern inside nd wke (Fig. 2) [7]. Wheres, in centrifugl compressor impellers, smll low momentum region evolves t the shroud corner on the blde suction surfce to develop downstrem into lrge defect in velocity distribution t the impeller exit [8, 9]. Flow choking t modern trnsonic centrifugl compressors lso leds to trouble sometimes, ssocited with mismtch between impeller nd diffuser. In prticulr, the unstedy flow interction t the semi-vneless region, tht is, just prior to the inlet throt of vned diffusers, cuses n error in designing pproprite diffuser flow ngle. Fig. 3 shows results of fully unstedy simultion on such impeller-diffuser flow interction [10-13]. LE Suction Surfce TE LE TE Pressure Surfce Figure 2: Surfce Flow Pttern on Rdil In-Flow Turbine. 3-4 RTO-EN-AVT-131

5 Computtionl grids for viscous interction clcultions Performnce mp of model compressor stge Pressure loss in diffuser Figure 3: Unstedy Flow Interction between Impeller nd Diffuser. RTO-EN-AVT

6 2.1.2 Reynolds Number Correction Phenomen observed bove in conventionl turbo-flows led to eventully erodynmic losses nd opertion rnge in the corresponding turbo-component, which needs to be evluted, in ggregte, s specific component efficiency. Aerodynmic losses my be conveniently ttributed to drg force, mixing, shock wves nd sher work, which re depending upon Reynolds number Re nd Mch number M, bsed on similrity or nlogy rules. A rule of thumb for Re correction on turbo-component efficiency is thus often expressed ccording to the next formul: (1 η)/(1 ηref) = (Dref/D) 1/N = (Reref/Re) 1/N where ηref, Dref nd Reref re the efficiency, rotor tip dimeter nd Reynolds number for reference size. N is number, which is function of Re, nd is usully ssigned to be 5 for lrger rotors nd 4 for smller ones, nd my be ssumed to be 3 for the size of severl millimeters. As shown in Fig. 4, tking reference size of 4 mm nd trget efficiency of 0.7, the formul bove indictes tht 10% increse in the dibtic efficiency is t lest required for lrger rotors with the size of 40 mm. Figure 4: Rule of Thumb for Adibtic Efficiency with Regrd to Rotor Dimeter. CFD simultion lso offers chnce to mke similr check on Re effect, wherefrom the following results were obtined for CIAM designed turbine [4]: The ltter mjor specifiction is listed in Tble RTO-EN-AVT-131

7 Tble 2: CIAM Turbine Design Specifiction rottionl speed 210,000 rpm pressure rtio 2.5 mss-flow rte g/sec speed rtio U/C0~0.63 inlet totl temperture 1223K specific het rtio 1.35 Nozzle outer dimeter 52 mm inner dimeter 42 mm number of bldes 17 blde height 2 mm Rotor outer dimeter 40 mm inner dimeter 24 mm number of bldes 15 blde height 2 mm To relize Re effect upon the turbine performnce, the pressure level ws presently chnged to become 5 times smller thn the design, s listed below, ssigning inlet totl nd outlet sttic pressure, respectively: inlet totl pressure p 0 * = kp kp outlet sttic pressure p = kp kp The clcultion results were summrized for comprison s follows: Higher Re Lower Re Efficiency: Totl to Totl Totl to Sttic Mss flow rte (g/s): It should be reminded tht the tip clernce effect ws not tken into considertion. Due to depressuriztion, therefore, efficiency ws reduced by roughly 10%, which coincided with the bove rule of thumb. Mss flow rte ws lso found to be slightly (3%) lower thn expected, which ws n dditionl informtion. The brekdown of this decrese in efficiency into mjor possible loss mechnisms is further necessry. Concerning Re influence upon blde surfce flows, Myle [14] pointed out tht the flow in conventionl turbines is likely lwys trnsitionl, while it tends to remin lminr for smll size components, thus, lible to seprtion, mening lrge source of loss. Tble 3 compres therefore turbine Re mongst vrious engine output levels including UMGT, together with sketch of Re distribution long the ltter RTO-EN-AVT

8 gs pth, whereby it seems tht flow seprtion cn occur t ny component in UMGT. A possible presence of lrge flow disturbnces like in wke my dd difficulty in predicting such flow seprtion nd trnsition. Tble 3: Representtive Reynolds Number t Typicl Turbine Engines 1.E+06 Fig.1 Micro gs turbine genertor cross-section REYNOLDS NUMBER bsed on inlet velocity & blde chord 1.E+05 1.E+04 1.E+03 1.E+02 compressor impeller compressor diffuser turbine sttor turbine rotor COMPONENT POSITION ALONG FLOW PATH Ctegory Men Turbine Dimeter [m] Engine Core Air Flow [kg/s] Turbine Inlet Temperture[K] Turbine Reynolds Number Big size ,000 Medium size ,000 Smll size ,000 Micro size ,000 Ultr-Micro , Het Trnsfer It hs been lredy pointed out tht dibtic flow condition will never be relized in UMGT, so tht isotherml wll condition, due to very low Biot number, is more pproprite within the flow pssges. It cuses significnt influence upon the engine fesibility fter ll, so tht more ccurte prediction method needs to be employed for estimting het trnsfer coefficients t low Re, in prticulr, where flow seprtion is observed to occur. Similrly, it will be crucil, from therml efficiency point of view, to distinguish pinpoint component/loction within the integrted system, so tht pproprite mesures cn be tken for directing het fluxes towrds the optimum therml control. In fct, for UMGT system, het loss t the turbine nozzle hs to be kept the minimum, while protection of the compressor from hot components is very effective [4, 15]. Het trnsfer effect between turbine nd compressor upon cycle efficiency hs been discussed elsewhere [16, 17]. Menwhile, het genertion due to frictionl force needs to be wtched crefully for the mngement of ir berings nd motor-genertor opertion with micro clernce t ultr high speed shft rottion Tip Clernce Clernce between rotor tip nd csing wll is inevitble for mnufcturing s well s rotor dynmicl reson, so tht it hs to be kept s lest s possible to reduce lekge flow. It will be interesting nd beneficil to compre turbines nd compressors with regrd to the mechnism of lekge flow cross the blde tip [18], since the reltive wll motion ginst rotor is opposite for turbine nd compressor, insmuch 3-8 RTO-EN-AVT-131

9 s there is fundmentl difference in the flow, either ccelerting or decelerting. There will be lso nother wy to divert this lekge vi employing shroud disk, though excessive stress my result from centrifugl force due to the dditionl shroud mss Geometry Restriction 2D geometry restriction comes from the necessity of MEMS fbriction, which mens diversion from pplying technicl dt bse so fr built up for improving performnce of rdil turbo-mchinery, thus rises severe difficulty in designing rotor bldes. According to conventionl design prctice for rdil rotor bldes, the meridionl velocity is kept lmost constnt long the gs pth for better control in diffusion. This is usully chieved by properly djusting the flow pssge re between two neighbouring bldes, e.g. widening blde height towrds the inner rdius for turbines, while reducing it towrds the outer rdius for compressors. MEMS restriction of constnt blde height brings bout excessive speed either t turbine exit or t compressor inlet, wheret both exit nd inlet re connected to either exducer or inducer region tht mkes 90 degrees flow turning from rdil to xil or vice vers. High speed turning often cuses flow seprtion t the corner of flow pssge, thus, hs to be voided, in prticulr, t the compressor inlet, since its influence spreds ll the downstrem pssge to cuse severe penlty in compressor efficiency nd opertionl stbility. There my be nother wy to control flow diffusion in 2D plne by mens of, for instnce, employing blde bckwrd in compressor, which however brings bout chnge in rection degree nd lso increses pssge friction loss, thus, llowble only within limited rnge. Fig. 5 compres rotor design profiles of 40mm in tip dimeter for compressor nd turbine. VKI UT CIAM VKI Impeller Design Turbine Design Figure 5: Illustrtion of UMGT Rotor Profile Design. It is obvious tht 3D profiles give better erodynmic performnce. It is lso fr from proven tht MEMS mnufcturing on silicon bse be the best for UMGT engine system, so tht 2D blde profiles could be thus esily bndoned to be of no design necessity. On the other hnd, such rdil rchitecture with 2D flt rotors my find mtch in thin structure devices with other components like het exchnger. It will be worthwhile to explore n innovtive design philosophy, to produce it s the simplest rotor geometry s possible, but to control inlet streming in 3D fshion letting the flow djust for itself. RTO-EN-AVT

10 2.2 Turbines According to conventionl rdil in-flow turbine dt, efficiency nd mss flow rte re presented ginst rotor exit ngle in Fig. 6 s prcticlly useful informtion. The smller flow ngle t both nozzle nd rotor exit flow leds to better dibtic efficiency, but this lso results in smller mss flow rte, which mens reduction in turbine power. Incresing the height of nozzle vnes nd rotor bldes cn be mesure to mintin mss-flow rte, but mteril stress concentrtion due to centrifugl force tends to occur inevitbly ner the root of blde triling edge s shown in Fig. 7, prticulrly for 2D rotors, therefore, the ltter rottionl speed hs to be definitely limited for mteril strength reson. A counter mesure is thus needed to overcome tht problem for 2D rotors. For conventionl 3D rotors, mteril stress is evenly distributed to the entire hub surfce of the turbine disc, nd such difficulty my not emerge. Nozzle Angle Figure 6: Effect of Exit Angle on Efficiency (left) nd Mss Flow Rte (right). Figure 7: Predicted Stress Distribution for Typicl 2D Rotor. In the followings, UT work [4] on designing, prototyping nd testing rdil turbines for Plm-top gs turbine engine is explined. It is prt of NEDO interntionl project tht will be ddressed in seprte text, therefore, the present explntion is only concerned with computtionl prediction nd comprison of erodynmic performnce mongst vriously designed rdil turbines, where rotor tip dimeter ws fixed to be 40mm. Design specifiction nd the corresponding velocity tringles re given in Tble 4 nd Fig RTO-EN-AVT-131

11 Tble 4: Design Specifiction of 2D Rdil Turbines Model 1 Model 2 Model 3 Pressure rtio Turbine Inlet Temp. [K] Rottionl Speed [rpm] Mss Flow Rte [kg/s] Nozzle Outer Di. [mm] Rotor Outer Di. [mm] No. of Bldes (nozzle) (rotor) Blde Height [mm] Output [kw] Figure 8: Rotor Designs nd Velocity Tringles. Numericl experiments were bsed upon Reynolds-Averged Nvier-Stokes simultions (RANS), wherein the efforts were mde for identifying mjor contributors to the loss genertion mechnism. Fig. 9 shows typicl exmple of the computtionl results, where iso-mch contour lines re plotted together with velocity vectors in the nozzle, blde nd exducer pssges of Model 1 turbine. This computtion ws done RTO-EN-AVT

12 for velocity rtio u/c 0 of 0.7 t the mximum dibtic efficiency condition. As indicted in Fig. 10, nerly hlf of the totl loss genertion occurred in exducer region, wheres qurter in the nozzle wke. In 2D turbines, reltively lrge kinetic energy remins t the exit of the blde pssge, depending upon the outer to inner dimeter rtio, nd it seems very difficult to recover this kinetic energy in 2D exducer pssge. This is one of the essentil resons why dibtic efficiency of 2D turbines is reltively much lower thn tht of conventionl 3D rdil turbines. Figure 9: Predicted Velocity Vectors nd Iso-Mch Number Distribution (Model 1 turbine; nozzle, blde nd exducer, from left to right). Figure 10: Predicted Contributions to Loss Genertion (Model 1 turbine). 2.3 Compressors It is unlikely to experimentlly obtin locl erodynmic nd het trnsfer dt over the blde surfces of UMGT rotors. Precise mesurements wherever vilble long the gs pth, either up- or down-strem of the rotor, re therefore vluble informtion, whereby the key component efficiency nd opertion hs to be inferred crefully nd resonbly with help of CFD simultion. Experiments re very difficult to crry out, in prticulr, to mke initil engine strt nd hot runs, tht is, with combustor in opertion, which necessittes CPU ided control logics nd utomtic sequences. In the following, only brief explntion is provided for preliminry results of the experiments performed by employing CIAM designed compressor for Plm-top gs turbine engine. The objective of experiments is twofold, in the first, to compre erodynmic performnce of 2D rotor ginst the conventionl 3D, nd in the second, to identify the effect of tip clernce. Design specifiction of CIAM impeller is listed in Tble 5, where it should be noticed tht the configurtion presently employed tpered blde height, thus not 2D in strict sense. Photos of tested impeller profiles re shown 3-12 RTO-EN-AVT-131

13 in Fig. 11 s well s set with vned diffuser in Fig. 12 nd the shft connection of rotor nd drive turbine in Fig. 13. Tip clernce ws djusted by fine pitch screw turning movement of the circumferentil ring tht lso comprises the intke bellmouth (Fig. 14). Tble 5: Specifiction of CIAM Impeller nd Diffuser Impeller Diffuser Inlet dimeter [mm] 14 Exit dimeter [mm] 40 Number of bldes 8 Thickness of bldes [mm] mx2 Inlet blde height [mm] 3 Exit blde height [mm] 1 Inlet blde ngle [ ] 35 Exit blde ngle [ ] 76 Inlet dimeter [mm] 45 Exit dimeter [mm] 60 Number of bldes 10 Thickness of bldes [mm] mx1 Conventionl 3D CIAM 2.5D Figure 11: Tested Impeller Profiles. Design Point Rottion: 210 krpm Pressure Rtio: 3.15 Mss Flow: 13.5 g/s Figure 12: CIAM Impeller nd Diffuser. RTO-EN-AVT

14 CIAM Impeller DriveTurbine Figure 13: Impeller nd Drive Turbine. Clernce Figure 14: Clernce Adjustment. Experimentl results re presented in Figs Compressor chrcteristics t design rottion of 210 krpm were not fully obtined in Fig. 15, becuse of filure in bering oil supply during opertion, which resulted in hlt of test runs. The design point PR is indicted fr bove the experimentl dt, prtly becuse of difficulty in mesurements nd prtly since the pressure nd het losses need yet to be clibrted nd compensted in the test rig. Fig. 16 compres the performnce between 2.5D CIAM nd 3D Conventionl, the mss flow rte hving converted into non-dimensionl prmeter, common to ll rottionl speed dt. The pek efficiency looks comprble ech other, but there is vst difference in the rnge of mss flow prmeter. The results obtined by chnging tip clernce re plotted in Fig. 17, which shows deteriortion both in lod fctor nd efficiency s the clernce vlue is incresed. A summry tendency is presented in Fig. 18, wherein the slope of 2.5D CIAM dt in the reltionship between efficiency ginst clernce is lrger thn usul. Tht liner tendency seems observble, s the clernce is rised up to extrordinry vlue of 40%, though unrelible due to yet preliminry test results RTO-EN-AVT-131

15 3.5 PR Design Pt ηp 70% 60% 50% 40% , , , ,000 30% 20% 10% CIAM 3D (Conventionl) 1 90, Mss Flow (Corrected) [g/s] 0% 0.0E E E-02 φ 3.0E E-02 Figure 15: Compressor Chrcteristics. Figure 16: Compressor Efficiency CIAM vs. 3D Conventionl. 2.5E-03 Ψ 2.3E-03 Design Point Design Point 70% ηp 60% 50% Design Point 2.1E-03 40% 30% 1.9E E-03 20% 10% Design Point E E E E E-03 Φ 1.0E-02 0% 0.0E E E E E-02 Φ Ψ: loding prmeter, Φ:mss flow prmeter Figure 17: Tip Clernce Effect on Compressor Loding nd Efficiency. RTO-EN-AVT

16 70% ηp 60% 50% 40% 30% 20% 10% According to Pmpreen 73, ηp/ηp= k(t/b) where k =is n empiricl constnt. Present experiments yield k = 0.65, which is lrger thn the commonly chosen vlue of k = 0.3, indicting more concern necessry for UMGT. Experiment ηp/ηp= k(t/b) 0% -20% 0% 20% 40% 60% t/b 80% 100% t:clernce, b:blde height@tip Figure 18: Compressor Efficiency CIAM vs. 3D Conventionl. 3. COMBUSTOR PARTICULARS 3.1 Specific Problems for Ultr-Micro Combustors Downsizing Combustors Vrious ides for micro-combustors hve been proposed for Power-MEMS, nd thermo-physicl issues relted to micro-scle combustion hve been discussed in previous ppers [19-21]. In downsizing combustor to micro-scle on the order of mm, fundmentlly importnt scling fctors re s follows: Reltive increse in quenching distnce; Lrger het losses due to the incresed surfce-to-volume rtio; Shortened diffusion chrcteristic time of mss nd het; nd Flow lminriztion. The first three fctors induce poor flme stbility nd combustion efficiency. For the quenching distnce, the typicl vlues for methne-ir nd hydrogen-ir t φ=1.0 nd tmospheric pressure re 2.55 mm nd 0.64 mm, respectively [φ: equivlence rtio] [22]. This problem implies the potentil dvntge of ctlytic combustion which mkes use of chmber wll surfce in micro-combustor, in prticulr, for hydrocrbon fuels. The rtio of het loss to het relese rte in combustor (HLR) is proportionl to (α T)/(SHR l ) [α: het trnsfer rte on the combustor, T : temperture difference between combustor outer wll nd the surroundings, SHR : spce heting rte, l : typicl length of combustor][23]. Generlly, SHR is defined s ( m& f h/(vol P c )) [ m& f : fuel mss flow rte, h: het of combustion, Vol: volume of combustion chmber, P c : chmber pressure]. This reltion shows tht HLR of combustor increses with decresing the size of the combustor due to the lrger surfce-to-volume rtio, wheres this vlue is reduced by incresing SHR RTO-EN-AVT-131

17 The diffusion chrcteristic time is represented s L 2 /D [L: typicl length, D: diffusion coefficient]. In downsizing combustor, the shortened diffusion chrcteristic time hs crucil effect on the rpid uniformity in concentrtion distributions ner flme in the combustor. Figure 19 shows typicl exmple of this chrcteristic, tht is, photogrphs of miniture propne flme stbilized on tube of inner dimeter d in =0.1 mm ner n extinction limit t φ =20. The imge on the left is direct photogrph of the flme nd the middle imge is schlieren photo. The imge on the right is schlieren photogrph of the corresponding unburned pre-mixture jet. It is obvious tht the schlieren imge region of the flme, tht is, the higher temperture region, extends to the fr outside the visible flme region nd beyond the injector exit. This my be n evidence tht lrge quntity of het ws trnsferred by conduction from the flme to its surroundings nd lso the injector, resulting in substntil het loss from the flme due to the scle effects of the lrge surfce-to-volume rtio. The schlieren photogrph of the unburned pre-mixture lso showed tht the flow seems to spred rpidly towrd the rdius just fter leving the exit of the tube. This strongly suggests tht some fuel cn diffuse to the surroundings before reching the tip of the flme bse due to lrge diffusion velocity. Therefore, n ultr micro flme in scle of n ordinry flme zone my dilute its fuel concentrtion, expnd its flme zone, nd decrese its temperture, leding to decrese in flme stbility nd combustion efficiency of combustors Figure 19: Micro Flme Configurtions just before Extinction. C 3 H 8 /Air, U je = 1.7 m/s, φ e = 20, d in = 0.1mm Left: Direct photogrph. Middle: Schlieren Photo with combustion. Right: Schlieren Photo without combustion. Unit: [mm] In ddition, the flow in ultr-micro combustors should be lminr. In fct, the Reynolds number in the ultr micro-combustor proposed s the originl MIT Power-MEMS turbine (pressure rtio: 4, ir mss flow rte: 0.15 g/s, combustor height: 1.3 mm, combustor exit temperture 1600 K, fuel mss flow rte m& f = 7 g/hr) [19] ws 100 ~ 200, which is 3 to 4 orders of mgnitude lower thn those in conventionl combustors. This prevents the chievement of rpid mixing between fuel nd ir nd high SHR due to the low trnsport coefficients relted to mss nd het trnsfer in combustors. In ultr-micro combustors, turbulent combustion, which is employed in most conventionl combustors in order to increse SHR, is not possible Performnce Requirements [24] An ultr-micro combustor tht dequtely stisfies the performnces required for UMGT hs not yet been developed, even when hydrogen hs been used. In order to downsize combustor for gs turbines, the following chrcteristic fctors hs to be ccomplished in ddition to the fundmentlly importnt scling fctors described in 3.1.1: RTO-EN-AVT

18 Low pressure loss; Low het loss to other components; High spce heting rte (SHR); nd Premixed combustion. Reducing the pressure loss in combustor is prticulrly importnt to relizing UMGT with low compressor pressure rtios. Conventionl smll gs turbines cn only llow pressure loss of 3 to 5%, bove which their therml efficiency is remrkbly decresed. Consequently, the Swiss roll burners (see lter) re indequte for the present purpose becuse of lrge pressure loss due to their long pths [25]. On the other hnd, cycle clcultion of the gs turbine tht introduced het losses nd trnsfer from the combustion chmber to compressor nd turbine showed tht this het trnsfer remrkbly reduces the therml efficiency, minly due to decrese in the efficiency of compressor nd turbine [26]. Reducing the het trnsfer from the chmber my be the most importnt nd difficult issue in developing n dequte ultr-micro combustor for UMGT. Seprted combustor configurtion my be one of the clever wys to improve the performnce, however, this issue remins the first priority in future reserch. For the combustor of gs turbines, the rtio of chmber volume to ir mss flow rte ( m& ) is proportionl to the chrcteristic length. This mens tht minituriztion of gs turbines requires rising combustor SHR, becuse of reltive decrese in combustor volume ginst m&. SHR, defined previously, cn be simplified s SHR ~φ m& /(Vol P c ) ~φ S v/vol ~φ v/ l [φ: equivlence rtio, S: cross section re of combustor, v : men flow velocity through combustor, l : length of combustor]. When φ of combustors is constnt in ny kind of gs turbines, SHR ~ v/ l 1/ τ b. [ τ b : residence time in the combustor]. This shows tht τ b depends only on SHR irrelevnt to the downsizing. In fct, SHR nd τ b estimted for MIT Power-MEMS turbine re the sme order s those of much lrger gs turbine combustor with hydrogen fuel [27]. Potentil dvntges of UMGT re high energy density nd high power to weight rtio, which is inversely proportionl to chrcteristic length. Therefore, high SHR in compct ultr-micro combustor is essentil to ffecting these dvntges. As mentioned bove, in ultr-micro combustors, lminr flow prevents rpid mixing between fuel nd ir, nd when using micro fuel injector, fuel cn diffuse to the surroundings before reching the flme bse. These phenomen suggest tht premixed combustion, insted of diffusion combustion, should be selected in ultr-micro combustors. Consequently, in ultr-micro combustors, the reliztion of high SHR in lminr flow my be the most chllenging requirement mong the prerequisites for UMGT. 3.2 Burning Methods of Ultr-Micro Combustors Approches According to the originl MIT proposl, the output of Power-MEMS turbine ws 16 W tht required combustion chmber volume of 0.04 cm 3, spce heting rte of 3.3x10 3 MW/(m 3 MP), combustor pressure loss under 5 % nd combustion efficiency over 99.5 %. The MIT group mde two hydrogen/ir micro combustors nd exmined their combustion chrcteristics. One ws fully premixed combustor with slit jet configurtion, nd volume of cm 3, nd the other ws dul-zone combustor with primry nd dilution-zone configurtion nd volume of cm 3. For the former combustor, the premixed hydrogen-ir entered the combustion chmber nd then formed the premixed flmes held by the recircultion zones t the inlets of the chmber. For the ltter, it first burned the premixed hydrogen-ir ner the stoichiometic condition in the primry zone, nd then diluted the combustion products, reducing 3-18 RTO-EN-AVT-131

19 their temperture to desired turbine inlet temperture in the dilution zone, which is similr to conventionl lrge-scle combustors. Although the operting rnge of the dul-zone combustor showed much broder region thn the premixed combustor, it did not provide better combustor efficiencies thn the former. As result, their combustors needed volumes 5 to 9 times s lrge s tht of the originl Power-MEMS turbine concept for better flme holding nd complete combustion. In order to improve poor combustion stbility nd efficiency, the Swiss Roll burner, which pplies excess enthlpy flmes, hs been proposed s micro-scle combustors for Power MEMS devices. This type of burner consists of 2 spirls. The pre-mixture enters through one spirl, forming flme t the center of the spirls, nd the combustion products exit vi the other pssge, exchnging het with the incoming premixture. It is well known tht the quenching limits cn be effectively eliminted through the use of recircultion of therml energy from the combustion products to prehet the incoming pre-mixture. For gs turbine combustors through which reltively lrge mount of ir flows, however, there re severl disdvntges to be overcome for use of this type of burners; lrge pressure losses due to their long pths nd n unvoidble increse of the totl chmber volume including spirls, which prohibits the reliztion of high spce heting rtes. As nother pproch, ctlytic combustors for hydrocrbon fuels hve been developed t severl reserch institutes [20]. An obvious dvntge of these combustors is tht they improve flme stbility nd increse the combustion efficiency, however the problem generlly lies in its low rection rtes resulting in low SHR, nd the pre-heting of the incoming rectnts. While, the reltive increse of surfce re nd the lower tempertures of the ctlytic rection suggest tht micro-scle combustors using ctlytic rections my be esier to be implemented thn those using gs phse rections [20] Concept of Flt-Flme Combustor [28] Fesibility relting to the most suitble combustion methods for UMGT ws minly considered from the viewpoint of premixed combustion nd high SHR. Reynolds number for ultr-micro combustors will be very smll, nd thus, the flows in the combustors would be lminr. In ddition, for UMGT, the current MEMS technology does not llow the use of multi-stge, xis-flow nd three-dimensionl turbo compressor nd turbine in high-performnce gs turbines. MEMS technology is limited to twodimensionl centrifugl turbo design with single stge, cusing the ultr-micro combustors to tke on flt disk shpe. Under this condition, one of the burning methods to form lminr flme in disk-shped combustor is flt-flme burner with porous plte introduced into the incoming pre-mixture. It is well known tht, for pre-mixtures with higher burning velocities thn the incoming flow velocity through the porous plte, het losses by conduction to the porous plte lower the burning velocity of the flme until blnce is obtined between the new burning velocity with het loss nd the incoming flow velocity, nd thus the flt-flme is stbilized on the surfce of the porous plte [29]. Therefore the flt-flme is obtined when the burning velocity without het losses is lrger thn the incoming flow velocity. If the height of the flt-flme micro combustor is equl to the flme zone thickness, SHR would be remrkbly high, thus meeting the level required by UMGT. 3.3 Combustion Chrcteristics of Flt-Flme Ultr-Micro Combustors [24, 28, 30] Prototype Combustor Bsed on the bove concept, flt-flme ultr-micro combustor employing hydrogen ws designed s shown in Fig. 20. The bsic dimensions of the combustor were essentilly the sme s those in the specifictions proposed by MIT group [19]. The combustor hd porous plte mde of stinless steel with n verge filtering ccurcy of 5µm nd nozzle mde of BN (boron nitride). The height of the combustion chmber with n nnulr region ( mm) ws 1 mm, nd the volume ws cm 3. After combustion, the burned gs exited rdilly outwrd to the tmosphere through slit with height of RTO-EN-AVT

20 0.3 mm between the nozzle nd the qurtz tube. Totl pressure loss of this combustor ws bout 5% t the mximum, which stisfied the pressure loss requirement of UMGT. The experiment ws performed t tmospheric pressure nd room temperture. Air mss flow rtes m& were vried from to 0.15 g/s, where typicl m& vlue ws determined so s to gree with the ir volume flow rte between the present experimentl condition nd the prcticl UMGT condition, working t 0.4 MP nd m& = 0.15 g/s. Detils of the pprtus nd the procedure re described elsewhere [28, 30]. Nozzle Combustion chmber φ10.5 φ5 Qurtz tube 1 Porous plte H 2 /Air Premixture Unit:[mm] Figure 20: A Schemtic of the Flt-Flme Micro-Combustor Combustion Chrcteristics Flme Appernce Figure 21() shows typicl direct photogrph of hydrogen/ir flt-flme formed in the ultr-micro combustor t m& = g/s nd φ = 0.4. The sme flme ws photogrphed using n imge intensifier nd CCD cmer is shown in Fig. 21(b). This flme ws very stble, occupied the whole volume of the combustion chmber of cm 3, nd burned well within this smll spce. It is obvious tht the flme zone ws lmost equl to the combustion chmber volume itself, so tht the flme chieved extremely high SHR of 7100 MW/(m 3 MP) t φ =0.4 in lminr flow. Thus, this flt-flme stisfied UMGT burning method requirement. With incresing φ, the flme remined stble nd did not show ny chnge in ppernce up to φ =1.0. () (b) BN nozzle Flt-flme BN 1 0 Unit:[mm] Figure 21: H 2 /ir Flt-Flme Appernce in Ultr-Micro Combustor t () Direct photogrph with exposure time of 180 s. (b) Imge intensifier photogrph with exposure time of 1/30 s. m& = g/s nd ϕ = RTO-EN-AVT-131

21 Flme Stbility Limits Figure 22 presents the flme stbility limit of the present burning. Extinction occurred t low φ conditions, but the flsh-bck did not occur even t φ =1.0. The stble flme region sufficiently stisfied the design opertion region of UMGT combustors (ssuming 16 W power output). The minimum occurred t m& round 0.02 g/s, then rose shrply s m& decresed, nd then slightly incresed s m& incresed s well. Het losses re considered to hve crucil effect on the flme stbility lower limit. In generl, het loss is proportionl to surfce re, nd het genertion is proportionl to fuel mss flow rte ( m& f = m& φ). Thus, the rtio of het loss to het genertion is proportionl to L 2 /( m& φ); tht is, reduction in m& increses this rtio, mking the effects of the het loss greter, resulting in severe temperture drop ner the combustion chmber wlls where the flme quenches. This cn explin the shrp rise of the lower limit for m& smller thn 0.02 g/s. On the contrry, s m& continues to increse, blow-off in the combustion chmber would be inferred due to the limittion of the chemicl rection time in n excessively stretched flme zone, or the incresed incoming velocity tht exceeds the burning velocity of the pre-mixture. Equivlence Rtio, ϕ Stble Flme Region No Flme Region Air Mss Flow Rte, m & [g/s] Figure 22: Flme Stbility Limits of Ultr-Micro Combustor for H 2. Temperture Distribution nd Het Losses Temperture distributions in the combustion chmber were mesured by K-type thermocouple with wire dimeter of 0.05 mm. The thermocouple wires were inserted in the combustion chmber through two slits on the combustion chmber wll, nd junction ws plced in the middle t constnt rdius position of 3.35 mm. Typicl temperture distributions in the combustion chmber for φ =0.35 nd 0.45 t m& = g/s re presented in Fig. 23. Tempertures on the surfces of the porous plte nd the nozzle showed the verges of those of the ctul surfces nd those of the gs phse close to the surfces. The porous plte ws substntilly heted by het conduction from the flme. Of note, the pek temperture positions moved slightly towrd the porous plte surfce t higher φ. This is explined by the fct tht s φ incresed, the flme zone becomes thinner becuse the burning velocity becomes lrger. At φ =0.45, the pek temperture ws lmost the sme s the dibtic flme temperture of hydrogen/ir T d,φ=0.45. These results confirmed the occurrence of complete combustion in the combustion chmber t φ =0.45, nd the existence of substntil het loss to both the nozzle nd to the porous plte. At φ =0.35, however, the pek temperture ws considerbly lower thn the dibtic tempertures T d,φ=0.35. Specultively, the reson for this my be tht the flme zone becomes thicker s φ decreses, nd then the temperture drops due to het losses to the nozzle before it reches the mximum. This mens tht het loss effects increse s φ decreses nd would prevent complete combustion. RTO-EN-AVT

22 Temperture, [ o C] Td, ϕ=0.45 Td, ϕ= Height from Porous Plte, [mm] Figure 23: Temperture Distributions in the Ultr-Micro Combustor for H 2 t m& = g/s. Regrding the het trnsfer from the flme, the het loss to the porous plte ws ssumed not to hve ftl effects on the flme behvior, since the porous plte would exchnge sufficient het with the unburned pre-mixture pssing through. However, the het loss to the nozzle clerly cused temperture drop. The temperture distributions ner the nozzle showed tht the het losses to the nozzle were pproximtely 15% of the het genertion of the hydrogen-ir pre-mixture of m& = g/s t φ = 0.4. This ssumption ws quite close to the het loss estimtion from the temperture distribution in the nozzle mesured by n dditionl experiment. These het losses would not only worsen the combustion efficiency by decresing the rection rte, but would lso impede self-sustining opertion of gs turbine if bout 10% of het generted were lost through conduction to the compressor nd turbine [26]. Combustion Efficiency The combustion efficiencies were clculted from the concentrtions of unburned hydrogen in the exhusted gs collected vi qurtz micro probe of n inner dimeter of 0.2 mm t the exit slit, nd were nlyzed using hotwire-type semiconductor hydrogen detector. Figure 24() shows the combustion efficiency versus φ t constnt m& =0.037 g/s using brss porous pltes. Combustion efficiencies for φ> 0.4 reched more thn 99.2 %. In this region, the hydrogen burned completely despite the het loss trnsferred to the nozzle, which occurred fter the temperture reched the mximum. This suggests tht the het loss to the nozzle hd little effect on the flt-flme combustion on the porous plte itself, nd only cused temperture decrese of the exhusted gs fter complete combustion. However, mesurements indicted tht the combustion efficiency decresed drsticlly for φ< 0.4. This suggests tht the flme zone becomes thicker s φ decreses nd combustion rections would be inhibited by het loss to the nozzle. Finlly, the flme quenched due to severe het loss, precluding self-sustining rection. On the other hnd, the combustion efficiency versus m&, given in Fig. 24(b), lso showed tht complete combustion ws chieved t wide rnge of m&, except ner the flme stbility lower limit. It cn be concluded tht for hydrogen, the excellence of the flt-flme burning method in micro combustor for UMGT ws confirmed experimentlly RTO-EN-AVT-131

23 Combustion efficiency, [%] Lower flme stbility limits Equivlence rtio, φ Combustion efficiency, [%] Lower flme stbility limits Air mss flow rte, m& [g/s] Figure 24(): The Combustion Efficiency versus ϕ t m& = g/s. Figure 24(b): The Combustion Efficiency versus m& t ϕ = Additionl Experiments In the present experimentl conditions, for hydrogen, outstnding combustion chrcteristics relted to flme stbility, SHR, nd combustion efficiency were ttined using the flt-flme ultr-micro combustor with volume of cm 3, which confirmed tht the flt-flme burning method is suitble s n ultrmicro combustor of UMGT. Menwhile, the effects of component mterils nd fuel on the flt-flme burning method re unknown. In order to exmine those effects, dditionl experiments were crried out s preliminry tsk () Mteril of Porous Plte nd Nozzle [30, 31] Even using brss nd het insulting cermics s porous plte mterils, the flme ppernce nd the flme stbility limits were substntilly the sme s those shown in Fig. 21 nd 22. However, for the brss plte fter mny hours of combustion, non-uniform combustion nd red-heting of the porous plte were observed due to n increse of dimeter of ech brss smll prticle. This implies high tempertures on the porous plte surfce tht might produce its oxide or exceed its melting point; therefore, selection of the most suitble mteril of the porous plte will be importnt. On the other hnd, het trnsfer experiment using het insulting nozzle mde of silic fiber showed tht the het losses to the nozzle remrkbly decresed to pproximtely 1% of the het genertion of the hydrogen-ir pre-mixture. In ddition, het trnsfer t the combustor exit slit ws the most dominnt, which cused not only het losses to the outside of the combustor but lso recircultion of the exhusted gs enthlpy by exchnging het with unburned pre-mixture upstrem of the porous plte. This suggests tht the het trnsfer t the combustor exit slit hs to be minimized to increse insultion performnce of the flt-flme ultr-micro combustor. (b) Methne Fuel For UMGT using hydrocrbon fuels, the combustor height of 1 mm shown in Fig. 20 seems to be sufficiently nrrow becuse the quenching distnces re wider thn tht of hydrogen. The flme stbility behvior ws exmined to consider pplicbility of the flt-flme burning concept to methne fuel using similr flt-flme combustor with stgnnt wll with vrible height over porous plte. RTO-EN-AVT

24 The wll height of the flt-flme combustor plyed crucil role in the flme stbility due to het loss nd flme stretch effects. When the wll height ws 4 mm, dilution by ir fter burning t φ =0.9 could be effective nd SHR could be 650 MW/(m 3 MP). Therefore, this burning method without ny ctlysis is effective for reltively lower SHR. However, in order to increse SHR with hydrocrbon fuels, ctlytic components re necessry to ctivte surfce rection or to improve flme stbility. 4. CFD SIMULATION 4.1 CFD Tools for UMGT Appliction [34, 35] The flowfields relted to UMGT hve very smll length scle, nd geometry quite different from those of conventionl turbomchinery. Flow physics tht determine the performnce of UMGT my be different from lrger turbomchinery, so it is not pproprite to pply existent design method directly for the design of UMGT. Millimetre-sized turbomchinery is not yet relized even t component level, so tht CFD is inevitbly of extensive usge, t lest t the initil stge of UMGT development. The flowfields inside UMGT, however, hve severl unfvourble fetures for CFD tht: Blde shpes re restricted from the requirement of mnufcturing process. It is difficult to optimize the blde shpe to the sme level s conventionl turbomchinery, for instnce, seprtion region tends to be lrger within the pssge. Het trnsfer from solid wlls is quite lrge. Clernce height cn be lrger thn 10 percent of blde height, which is considerbly lrger thn tht in conventionl turbomchinery. Reynolds number is s low s order of 10 4 nd the effect of turbulent trnsition my not be negligible. Tble 6 summrizes CFD tools used for the design nd nlysis of UMGT t severl orgniztions. It should be noted tht ll reserch orgniztions pply CFD tools tht hve been developed for the design nd nlysis of conventionl turbomchinery. By ssuming fully turbulent flows, the results re not necessrily stisfctory for the flowfield with turbulent trnsition, so the lst feture, low Re number effect, is still remined for future reserch. Cres should be tken lso in treting the other three fetures. Tble 6: CFD Tools used for the Design nd Anlysis of UMGT orgniztion code grid topology turbulence modelling MIT commercil(fluent) unstructured k-ε VKI TRAF3D structured (H) Bldwin-Lomx CIAM in-house structured (O+H, dptive) Cokley q-ω UT in-house structured (H, overset for clernce region) Bldwin-Lomx 4.2 Chrcteristics of Qusi Two-Dimensionl Turbo-Flows [36, 38, 39] The chrcteristics of flowfield in two-dimensionl micro turbine re discussed bsed on n exmple shown in Fig. 25. Outer dimeter of the rotor is 40mm nd the design pressure rtio is 2.91 t rottionl speed of 24,000RPM. Figure 26 compres the totl-pressure loss t ech prt of turbine, nozzle, rotor, exducer nd the exhust dynmic pressure. The nozzle loss nd exhust dynmic pressure re distinct in the figure. The exducer loss hs only minor contribution t lower speed rtio, but it becomes s lrge s the rotor loss t higher speed rtio RTO-EN-AVT-131

25 Figure 25: Two-Dimensionl Turbine. Figure 26: Totl Pressure Loss Components. () Nozzle (b) Rotor (c) Exducer Figure 27: Velocity Vectors nd Reltive Mch Contours of 2D Turbine. Since the turbine flowfield is ccelerting, the flowfields t nozzle nd rotor re essentilly smooth. No seprtion is observed in nozzle flow, so most of the nozzle loss is ttributble to the wke downstrem of thick triling edge. The rotor flow is lso clen. Smll seprtion region ppers t off-design condition, however since the flowfield is lmost two-dimensionl, it is not so difficult to design cscde so tht seprtion is suppressed to the minimum. Sizes of turbine do not ffect these fetures, so Re number do not hve much influence over the loss t nozzle nd rotor other thn the chnge in the boundry lyer thickness. The mtter is different in exducer, s it hs two prticulrities peculir to two-dimensionl turbine. The cross section t the inlet of exducer is smller thn tht t the inlet of rotor so the flow velocity t exducer is much higher thn tht of conventionl tree-dimensionl rotor. The second point is tht the flow turns 90 degree from rdilly inwrd to xilly. The exducer thus leds to lrge exhust loss. Right-ngled turning of high-speed exhust flow induces seprtion t csing side which is the mjor source of the exducer loss. So fr s turbine is concerned, difficulties minly come from two-dimensionl geometries rther thn the size. The sme hppens t the inlet of two-dimensionl compressor. Figure 28 shows the velocity vector mp t the mid-pitch section of CIAM compressor (Fig ). The flow enters the inlet xilly nd then turns 90 degree towrd the impeller. The lrge curvture of the csing induces seprtion region. RTO-EN-AVT

26 Figure 28: Velocity Vectors & Mch Contour. Figure 29: Velocity Vectors & Mch Contour (idel inlet). Since the seprtion tkes plce t the inlet, it cuses negtive effect over the entire flowfield within the pssge s well s totl-pressure loss due to the seprtion itself. In order to evlute the influence of the inlet seprtion, the chrcteristic curves re compred with those evluted with idel inlet. Figure 30 shows velocity vectors t the mid-pitch meridionl plne of the simultion with idel inlet. Inlet boundry of the computtionl domin ws plced right upstrem of the leding edge of impeller, nd the velocity vectors t the inlet were specified so tht they ligned with wlls to eliminte the seprtion. Figure 30: Chrcteristics Curve of CIAM Compressor (effect of inlet curvture). The chrcteristic curves evluted from the simultions with rdil-inflow re compred with those with xil-inflow in Fig. 31, which indictes tht: The rdil-inflow impeller hs lrger mss flow rte nd wider operting rnge; The pressure rtio of rdil-inflow impeller is lrger thn tht of xil-inflow; nd The improvement with rdil-inflow is remrkble especilly when the mss flow rte is high RTO-EN-AVT-131

27 Figure 31: Polytropic Coefficients Plotted ginst Inlet Flow Angle. 4.3 Non-Adibtic Wll Effects upon Rdil Turbines [36] Reduction in the length scle cuses increse in the surfce-to-volume rtio, hence the increse in the het losses. According to the het trnsfer nlysis by Ribud, het losses t the turbine nozzle of 8mm-sized UMGT reches twice s much s tht t the combustor. The wll het trnsfer induces two different influence over the flowfield. One is the devition of reltive inlet flow ngle of the rotor, nd the other is the influence on the development of wll boundry lyer. Het loss t the nozzle lowers the temperture nd the volume flow rte of fluid inside the nozzle, hence it reduces the bsolute velocity t the rotor inlet. Since the bsolute flow ngle nd the blde rottionl speed remin constnt, the reltive flow ngle decreses. Although this effect is well known even in mcro turbomchinery, it is more importnt in micro turbomchinery where the het loss is significnt. For exmple, nozzle het loss cuses 10 degree of devition in the inlet flow ngle even for wll temperture t 700K. Error in the estimtion of het loss directly result in error in the reltive inlet flow ngle, so quntittive prediction of wll het trnsfer is becoming more importnt in micro turbomchinery. The ltter effect is more complicted since there is no simple method to predict its influence. Figure 31 compres the polytropic efficiency of turbines with nd without wll het loss. Efficiencies re plotted ginst reltive flow ngle, so the first effect is compensted in the figure. The effect of wll het loss is not noticeble t positive inlet ngle, but the efficiency of the turbine with isotherml wll is considerbly lower t negtive inlet ngle (higher U/C 0 ). Figure 33 compres the stremlines t flow ngle of pproximtely -50 degree. The stremlines show seprtion region t the pressure side of the leding edge only for the isotherml wll which indicte tht deteriortion in efficiency shown in Fig. 31 is ttributble to boundry lyer seprtion. Boundry lyer profiles ner the leding edge (Figure 32) show tht fluid ner the isotherml wll decelertes much rpidly thn the dibtic wll. Since the temperture of isotherml wll is lower thn the stgntion temperture of incoming flow, the fluid is cooled ner the wll nd reduces its volume, nd the reduction of the volume flow rte leds to the fster decelertion. So the leding edge seprtion observed for the isotherml wll is consequence of wll het trnsfer. RTO-EN-AVT

28 () Adibtic wll (b) Iso-therml wll Figure 32: Boundry Lyer Profiles ner the Leding Edge. Figure 33: Mid-Spn Stremlines. 4.4 Lekge Flow in Rdil Turbines [18] Tip clernce of turbomchinery lrgely depends on specifiction of berings, so reltive clernce height of micro turbomchinery tends to be lrger thn tht of conventionl turbomchinery. For micro turbomchinery with blde height of 0.5mm, tip clernce s smll s 0.05mm gives 10 percent of reltive tip clernce. Therefore, the clernce flow hve lrge impct on the performnce of micro turbomchinery. Numericl experiments showed tht the tip clernce of 10 percent blde height deteriortes the dibtic efficiency of micro-scle turbine by 15 points RTO-EN-AVT-131

29 Figure 34: Performnce Deteriortion due to Clernce. Figure 35: Re Number Effect on Performnce. Although its reltive clernce is lrge, the chrcteristics of ssocited loss in UMGT re essentilly the sme s the lrger turbomchineries. The clernce loss is well correlted with the lekge flow rte, nd the flowfield inside the clernce gp is determined minly by the pressure difference cross the blde surfces nd the scrping flow induced by the reltive motion of the csing. The scle effect ppers remrkbly in the mgnitude of the scrping flow. Figure 36 shows velocity vectors t the middle of the clernce gp for 2 percent clernce height. All the velocity vectors for 40mm-dimeter rotor direct from pressure side to suction side. On the other hnd, the scrping flow of 4mm-dimeter rotor blocks the clernce flow t forwrd hlf of the blde. Consequently, the lekge flow rte of the 4mm-dimeter rotor t two percent clernce becomes lmost zero, nd the efficiency deteriortion is lmost negligible. Figure 36: Velocity Vectors t Mid-Gp Plne (φ4mm). Figure 37: Velocity Vectors t Mid-Gp Plne (φ40mm). In generl, decrese of Reynolds number leds to deteriortion of efficiency. However s for turbines, decrese of Reynolds number brings reduction of clernce loss, nd this effect is more noticeble in micro-scle. If we mke the most of this effect, for exmple shifting the blde loding forwrd where the scrping flow is more strong, there my chnce tht we cn design micro-scle turbine more efficient. 4.5 Usge of CFD Tools Although CFD codes listed in Tble 6 re similr to ech other (t lest for the tretment of turbulence), ech orgniztion s view of CFD is chrcterized by how they use it. RTO-EN-AVT

30 Sirkov et l. t MIT evluted efficiency of miscellneous rdil impellers with CFD, nd developed models to predict performnce penlty cused by het ddition, csing drg nd chnnel boundry lyer. They estimted tht the het trnsfer could led up to 25 percent efficiency penlty, csing drg to points, boundry lyer to 10 points, nd inlet turning to 5 points. They lso discussed performnce trends using the developed models, nd discussed design guidelines of UMGT. Brembussche et l. t VKI pplied their blde optimiztion system for the design of two-dimensionl impeller. In their optimiztion system, performnce of impeller is evluted by CFD nd Artificil Neurl Network (ANN) dtbse, nd the blde shpe is optimized using Genetic Algorithm (GA). The system is originlly developed for the design of conventionl turbomchinery, but they concluded tht it cn be pplied for the design of smll two-dimensionl impeller with smll modifiction. Ivnov et l. t CIAM lso pplied CFD for the design of two-dimensionl impeller nd turbines, but they optimized the blde shpes mnully. They did not include the clernce effect, but identified the sources of loss from the observtion of CFD results. They then redesigned the blde so s to suppress the identified loss source. Ngshim et l. t Univ. of Tokyo numericlly investigted the impeller designed by CIAM nd turbine designed by their group, nd qulittively discussed the loss sources, but not for quntittive nlysis. Former two pproches clerly differ from ltter two cses. MIT nd VKI seems fully rely on CFD prediction, nd focused on how to dopt CFD into the design process. On the other hnd, CIAM nd UT presume tht uncertinty is inevitble for CFD prediction, nd limit oneself to qulittive discussion. There my be discussions bout relibility nd limittion of CFD, but too much confidence on CFD sometimes cn result in misleding conclusion. For exmple, Sirkov et l. drw conclusion from their numericl experiment tht the efficiency penlty due to inlet turning is pproximtely 5 points, one third of tht cused by the csing drg. This conclusion my be true for their prticulr configurtion, but not lwys pplicble to other cses. Figure 38() shows totl pressure distributions of CIAM impeller. Low pressure region t the inlet which is cused by the inlet turning extends downstrem. It merges with suction surfce boundry lyer to crete lrge low pressure region, nd covers lmost hlf of the pssge t the middle of the blde. Figure 38(b) is totl-pressure distribution of the sme impeller without inlet turning. The low-pressure region ner the csing is considerbly thinner thn the cse with inlet turning. The contribution of the inlet turning is obviously lrger thn the csing drug, nd furthermore the interction between the influence of inlet turning nd the boundry lyer loss is not negligible. The extent of the inlet low-pressure region is influenced by the inlet velocity, so it suggest tht the loss due to inlet turning will be well correlted with specific speed of the impeller RTO-EN-AVT-131

31 () With inlet turning (b) Without inlet turning Figure 38: Totl Pressure Distribution (CIAM Impeller). The limittion of Sirkov s modelling becomes cler only fter the detiled observtion of the flowfield, so tht ny optimiztion or modelling bsed on CFD hs to be ccompnied with detiled observtion of the flowfield, to clrify the limittion of the CFD nlysis. 4.6 Remining Issues [37] CFD tools used tody do no necessrily provide solutions ccurte enough to blindly rely on. There re two remining issues: (1) proper usge of CFD with the premise tht it hs limittion, nd (2) improvement of the ccurcy nd relibility of CFD itself. The former is essentilly the sme s wht is required in CFD nlysis of lrger turbomchinery. On the premise tht CFD is not ccurte enough for quntittive prediction, it is better to use CFD for understnding of flowfield nd qulittive or reltive evlution of the performnce, rther thn direct evlution of loss coefficient. For the quntittive prediction of the performnce, relible reference dt is essentil (this is lso the sme s in the cse of lrger turbomchinery). The lrgest issue in the cse of UMGT is tht there is no relible numericl or experimentl dt tht cn be used s the reference. There is few experiment of micro-scle turbomchinery, nd detiled mesurement is not reported in the litertures. So the requirement of relible numericl results is urgent to estblish the reference mongst conventionl CFD tools. In either wy, CFD method tht gets round difficulties ssocited with current CFD, such s seprtion or turbulent trnsition, is necessry. Recent development of compressible LES code finds n ppliction for trnsitionl flows in low-pressure blde pssge T106 with nd without inflow turbulence tht mimics the nozzle wke, nd their effects on the seprtion nd trnsition of the blde s boundry lyer. Figure 39 shows the computtionl region nd grids used for the LES of T106 cscde flow t the designed condition s shown in tbles 7 nd 8. Notes tht grids re shown here t every ten-grid lines for clrity reson nd pproximtely 6 million grids were used for the computtions. In order to provide inflow turbulence with intensity of 7.1%, seprte LES of compressible homogeneous turbulence ws performed nd instntneous flow field were fed to the inlet boundry of the cscde computtion t every time step fter n pproprite scling ws mde. RTO-EN-AVT

32 Figure 39: Computtionl Grid with every 10-Grid Lines Shown. Tble 7: Design Dt for T106 Turbine Cscde Design Point Condition Isentropic outlet Mch # M 2,th 0.59 Isentropic outlet Reynolds # Re 2,th Inlet flow ngle β [º] Outlet flow ngle β [º] Geometry Chord length C 100 [mm] Pitch/Chord rtio t/c Inlet Tble 8: Free-Strem Conditions P t, 1, T t, 1, [P] [K] Outlet p [P] Time-verge distributions of the sttic pressure long the blde surfces were compred with the cscde test dt s shown in figure 40, which shows n excellent greement between the computtion nd mesurement including the smll seprtion tht tkes plce t round 80% chord length for the clen inflow cse. Figure 41 shows n instntneous distribution of the velocity divergent. For clen inflow cse pressure wves tht generte when the lrge-scle vortices pss through the triling edge propgte in the most portion of the cscde pssge. Although not shown here, lter investigtion indictes tht these pressure wves initite the secondry instbility of the boundry lyer on the blde s 3-32 RTO-EN-AVT-131

33 suction surfce before they re wekened by the interction with the pressure wves tht originte from the triling edge of the neighbouring blde. On the other hnd no strong pressure wves re confirmed for the turbulent inflow cse. Figure 40: Comprisons of Time-Averge Sttic Pressure Distributions on Blde Surfces. Figure 41: Computed Instntneous Distributions of Velocity Divergent in Blde Cscde With (left) nd Without (right) Free-Strem Turbulence. Figure 42 shows time-verge distributions of the skin friction coefficient long the blde surfces. For the clen inflow cse, boundry lyer seprtes t round 80% downstrem from the leding edge (20% from the triling edge) nd short seprtion bubble is formed therefter. The skin friction coefficient rpidly increses fter the rettchment due to the boundry lyer trnsition. Turbulent inflow cse shows sudden reduction in the skin friction coefficient downstrem of the leding edge due possibly to the instntneous seprtions cused by the incidence chnge. But, in this cse no seprtion tkes plce downstrem of the blde becuse of n erlier trnsition cused by the inflow turbulence. Although the detiled investigtion is currently underwy, the results presented here seem very promising nd LES of the low-reynolds number turbine flow will certinly become fesible in the ner future. RTO-EN-AVT