Analiza tokovnih struktur v vstopnem sistemu motorjev

Strojni�ki vestnik - Journal of Mechanical Engineering 51(2005)12, 744-756

744 Rodman Opre�nik S. - Prijatelj S. - Katra�nik T. - Trenc F. - Bizjan F.

Analiza tokovnih struktur v vstopnem sistemu motorjev -primerjava rezultatov meritev in numeriène simulacije

Analysis of the In-Cylinder Flow Structures Generated by the Engine InductionSystem � a Comparison of Experimental and Simulated Data

Samuel Rodman Opre�nik - Sa�o Prijatelj - Toma� Katra�nik - Ferdinand Trenc - Franèi�ek Bizjan

V prispevku so prikazani rezultati numeriène analize - simulacij na osnovi raèunalni�ke dinamike tekoèin(RDT) in meritev hitrostnega polja v vstopnem sistemu motorja z notranjim zgorevanjem z uporabo laserskegaDopplerjevega anemometra (LDA). V ta namen je bil izdelan stekleni osmerokotni model valja, ki je po geometrijskiobliki podoben valju resniènega motorja. Premièen bat je tokovne razmere na modelu �e bolj pribli�al resniènimrazmeram pri pravem motorju. Ugotovljeno je dobro ujemanje izraèunanih in izmerjenih rezultatov pri doloèitvitokovnih struktur, hitrostnega polja v valju in pri doloèitvi pretoènega koeficienta v vstopnem ventilu motorja.Raziskava je tudi pokazala, da je mogoèe z numerièno analizo RDT-jem razmeroma dobro napovedati dogajanjev valju motorja in izbolj�ati razumevanje pojavov, ki spremljajo polnjenje valja motorja.© 2005 Strojni�ki vestnik. Vse pravice pridr�ane.(Kljuène besede: motorji ZNZ, raèunalni�ka dinamika tekoèin, CFD, polja hitrostna, anemometri Dopplerjevilaserski)

A comparison and analysis of laser Doppler anemometry (LDA) measurements and a computedfluid dynamics (CFD) simulation of the in-cylinder flow structures generated by an engine induction systemis presented in this paper. An octagonal glass model with a moving piston complementary to a real enginecylinder was applied in the analysis. Good agreement between the measured and simulated patterns of theflow structures and the inlet-valve discharge coefficients was obtained. In general, this study shows that in-cylinder CFD predictions yield reasonably accurate results that help to improve the knowledge of the air-flow characteristics during the intake stroke. CFD therefore represents an effective design tool for developingless-polluting and more efficient internal combustion engines.© 2005 Journal of Mechanical Engineering. All rights reserved.(Keywords: internal combustion engine, computational fluid dynamics, CFD, air flow characteristics, laserDoppler anemometry)

Strojni�ki vestnik - Journal of Mechanical Engineering 51(2005)12, 744-756UDK - UDC 621.43.01:004.94Izvirni znanstveni èlanek - Original scientific paper (1.01)

0 UVOD

Kakovost zgorevanja v valju ottovega indizelskega motorja z notranjim zgorevanjem moènovpliva na nastanek �kodljivih emisij v ostankihzgorevanja, zato je dobro poznavanje tokovnih razmerv valju motorja z notranjim zgorevanjem neobhodno.Strukture turbulentnega toka v valju motorja sopomembne tako za zaèetek kakor tudi za celotno obdobjetrajanja zgorevanja in njegovo uèinkovitost [1].Velikostni red turbulentnih struktur je razlièen: odnajveèjega, ki ga omejuje geometrijska oblikazgorevalnega prostora, pa vse do najmanj�ih, pri katerihprevladuje molekulska difuzija. Vrtinèni tok zgorevalnezmesi, ki ima poudarjen vrtinec v obodni in/ali vzdol�ni

0 INTRODUCTION

In order to reduce harmful exhaust emissionsand increase the efficiency of internal combustionengines, fundamental knowledge of in-cylinder fluidmotion is required, since it significantly affects thecombustion process in both SI and CI engines. It iswell known that the structure of the turbulent flowfield of the fresh charge is a determining factor in theinitiation, the rate of propagation, and the efficiencyof the combustion process in an internal combus-tion engine [1]. The turbulent length scales rangefrom the largest, which are limited by the physicalconstraints of the combustion chamber, to the small-est, which are governed by molecular diffusion. A


745Analiza tokovnih struktur v vstopnem sistemu motorjev - Analysis of the In-Cylinder Flow Structures

smeri valja je stabilnej�i, razpada poèasneje in zatokoristno pomaga pri uèinkovitem me�anju goriva inzraka ter zgorevanja tudi v kasnej�ih obdobjihzgorevanja ([2] in [3]). Zato je za razvoj sodobnihmotorjev, �e posebno glede zmanj�anja �kodljivih emisijv izpu�nih plinih, izredno pomembno dobro razumevanjeèasovnega in krajevnega razvoja tokovnih struktur medvtekanjem in stiskanjem delovne snovi v valju [4].

Pri doloèanju oblike toka v valju, ki je predvsemposledica oblike vstopnega kanala in pogojev delovanjamotorja, so bili do sedaj najveèkrat uporabljeni rezultatipreizkusov na mirujoèem valju oziroma modelu valja (npr.v objavah [5] in [6]). Kasneje so bili osnovnim podatkomdodani �e dodatni parametri, ki so bolje opisovaliintegralni tok v valju [7]. Z rezultati eksperimentalnegadela, ki je bilo podprto s preprostimi fenomenolo�kimimodeli, podobne sta razvila za potrebe svojih raziskavDavis in soavtorji [8] in Murakami in soavtorji [9], je bilomogoèe doloèiti osnovne parametre, ki doloèajo tokovnopolje v valju motorja. Pomembno prednost omenjeneganaèina kombiniranega re�evanja odlikuje enostavnostin majhna poraba raziskovalnega èasa, medtem kopomeni manj�a podrobnost opisa tokovnih strukturveèjo pomanjkljivost, ki nas lahko pripelje tudi donapaènih rezultatov in sklepov. Velik korak prinatanènem preizkusnem doloèanju hitrostnega polja vmirujoèem valju pomeni uvedba laserske Dopplerjeveanemometrije (v nadaljevanju LDA) ([10] in [11]).Rezultati so bistveno natanènej�i od rezulatov, ki sobili izmerjeni s starej�imi ustaljenimi preizkusnimimetodami. Vpogled v tokovno dogajanje v valjuomogoèijo tudi sodobne 3D numeriène simulacije (vnadaljevanju modeli RDT), s katerimi lahko re�imoustrezne tokovne enaèbe in izraèunamo podrobnevrednosti povpreènih hitrosti toka in znaèilnostiturbulentnega hitrostnega polja. Pri uporabi modelovRDT moramo upo�tevati posebnosti nestalnega toka,vpliv visokega Reynoldsovega �tevila toka in izjemnozamotane oblike trdnin - sten pretoènih kanalov, ventilovin valja.

V zadnjem èasu lahko v strokovni literaturi zapotrebe doloèitve tokovnega polja v valju (tudidelujoèega motorja) zasledimo (npr. [12] in [13]) tudipogosto uporabo metode doloèanja hitrosti delcevtoka z analizo optiènih posnetkov - DHD (PIV). Njenaprednost je, v primerjavi z drugimi preizkusnimimetodami, v trenutnem in ne samo èasovnopovpreèenem prikazu celotnega tokovnega stanja vvalju. Ima pa metoda DHD tudi pomanjkljivosti:zahteva velike moèi laserjev, sorazmerno velikedodatne delce za sipanje svetlobe, ki zaradi svoje

well-defined swirl and/or tumble flow structure ismore stable than other large-scale in-cylinder flowsand, therefore, it can break up later during the cycle,giving higher turbulence during combustion ([2] and[3]). Therefore, a good understanding of the evolu-tion process of fluid motion in internal combustionengines is critical when the development of advancedengines with the most attractive operating and emis-sion characteristics is concerned [4].

Simple experimental tests in steady-state flowrigs have been widely used to determine the flowmotion generated by induction systems (see, forexample, [5] and [6]). Some supplementary param-eters to evaluate the bulk motion of in-cylinder flowwere defined more recently [7]. In conjunction withsimple phenomenological models, such as those de-veloped by Davis et al [8] and Murakami et al [9],these experimental procedures helped to determineseveral fundamental variables that define the flowinside the cylinder. Simplicity and a moderate con-sumption of time are the advantages of this type ofapproach. However, the information provided is notdetailed enough, and the assumptions made can leadto inaccurate results. A more accurate experimentaltechnique is the measurement of the velocity field ina steady-flow test rig using laser Doppler velocimetry(LDV) ([10] and [11]). This method provides high-quality results and more information than other con-ventional methods. Another approach for gettingan insight into the in-cylinder flow is the applicationof three-dimensional numerical simulation codes(CFD models), that are efficient in solving the gov-erning flow equations, and thus provide a detaileddescription of the mean velocity and the turbulentvelocity fields. When applied to internal combus-tion engines, the CFD models have to consider thespecific problems linked to the unsteady flows, thehigh Reynolds numbers involved, and the highlycomplex geometry of the solid boundaries.

Many authors (see, for example, [12] and [13])have applied the PIV (Particle Image Velocimetry)method to determine the flow patterns in cylinders,since it produces whole field data relating to a par-ticular instant of the observed time rather than aphase-average. However, high laser-power levelsand relatively large seeding particles are required toachieve satisfactory results; on the other hand, largeparticles introduce measurement errors due to iner-tial effects [14].

The aim of this study was to validate the ac-curacy of the CFD simulations of the in-cylinder



vztrajnosti ne sledijo toku in tako vna�ajo veèjomerilno negotovost [14].

Namen predstavljenega dela je, da ovrednotimonatanènost uporabljene simulacije RDT in razi�èemonjeno uporabnost ter omejitve pri doloèitvi tokovnegapolja v valju motorja z uporabo preizkusne metode LDA.Zahtevnost in razliènost oblike vstopnega sistema zarazliène motorje onemogoèa preprost prenosekstrapolacije dobljenih rezultatov opravljenihpreizkusov na druge motorje [15], paè pa je zelo primerenin pogosto uporabljan pri optimizaciji oblike vstopnega,izpu�nega sistema in zgorevalnega prostoraopazovanega motorja. Zato je bil izdelani modelsteklenega valja motorja ustrezno obdelan in pripravljenza uporabo v simulacijski kodi CFX [16]. Opravljena inprikazana analiza ima dva cilja: 1) da preveri uporabnostin natanènost izraèunanega pretoènega koeficienta in2) da preveri uporabnost rezultatov raèunske simulacijeRDT tokovnih struktur v valju motoja.

Pretoèni koeficient vstopnega ventila motorjapomembno vpliva na kolièino v valj motorja vnesenesve�e snovi in s tem tudi na moè in emisijo �kodljivihsnovi v izpu�nih plinih. Obenem je tudi pomembenvstopni podatek pri raèunanju z niè in enorazse�nimisimulacijskimi modeli. �al je njegova raèunskadoloèitev praktièno nemogoèa zaradi izrednezapletenosti toka skozi ventil.

1 PREGLED UPORABLJENIHPREIZKUSNIH METOD

1.1 Opis preizkusne naprave

Merilna naprava, katere posnetek je prikazanna sliki 1 in s katero so bili opravljeni vsi predstavljenipreizkusi, sestoji iz pet pomembnej�ih sestavnih delov:- sistema LDA za merjenje hitrosti v modelu valja,- raèunalni�ko vodenega polo�ajnega sistema za

uporabljeno merilno leèo LDA,- sistema merilnikov diferenènega tlaka v modelu

valja,- enote sesalnika za prah z elektronsko nastavljivo

frekvenco vrtenja - nastavljivim tokom zraka innastavljivim poljubnim tlaènim padcem v vstopnemventilu,

- kovinske - izvirne glave primerjalnega motorja, kije bila pritrjena na stekleni model valja; omogoèenaje poljubna nastavitev pretoène �pranje v ventilu.

Za potrebe raziskav je bil iz optiènega steklaizdelan osmerostrani model valja dizelskega tlaènopolnjenega motorja MAN D0926 LOH15 z delovno

flows with laser Doppler anemometry (LDA) meas-urements and explore the applicability and the limi-tations of the CFD representation of the in-cylinderflow pattern. The complexity of the fluid motion inIC engines makes it difficult to extrapolate experi-mental results from one engine geometry to another[15]; therefore, validated and calibrated CFD simula-tion codes are gaining popularity when the optimi-zation of the combustion-chamber geometry, the in-duction and exhaust system geometry, etc., is con-cerned. Therefore, a cylinder model with optical ac-cess was built, and its geometry was simultaneouslyimplemented into the CFX (AEA Technology) simu-lation code [16]. The presented analysis has twogoals: 1) to analyze the accuracy and applicability ofthe simulated intake-valve discharge coefficient and2) to analyze and explore the applicability of the CFDsimulations to predict the in-cylinder flow structures.

The intake-valve discharge coefficient is afundamental variable that substantially influencesthe quantity of fresh air in the cylinder, and there-fore the engine performance and the emission of pol-lutants. It is also an important input parameter for allzero- and one-dimensional engine-simulation mod-els. An analytical determination of the flow coeffi-cients is not possible due to the complexity of thefluid flow.

1 EXPERIMENTALMETHODS

1.1 Experimental equipment

The measurement equipment used for all theexperiments is presented in Fig.1; it consists of fivegeneral components:- an LDA system for in-cylinder velocity measure-

ments- a computer-controlled positioning system for an

applied optical probe- a gauge-pressure measurement system- a vacuum-cleaner suction unit to set the appropri-

ate air flows and pressure drops across the cylin-der and cylinder head

- a real engine cylinder head attached to the glassmodel of the cylinder; the setting of an arbitraryintake-valve clearance and any piston-crown po-sition is possible.

An octagonal model cylinder made of opticalglass (similar to the cylinder of the four-stroke 6.871 lturbocharged and aftercooled MAN D0826 LOH15



prostornino 6,871 dm3. Podrobnosti lahko najdemov magistrskem delu Prijatelja [17]. Natanènaraèunalni�ko vodena polo�ajna naprava s koraènimimotorji (natanènost ± 0,01 mm) je omogoèalanatanèno nastavitev, ponovljivost lege izbranihmerilnih toèk in meritev tretje komponenteprostorskega vektorja hitrosti. Trorazse�no hitrostnopolje je bilo doloèeno v 165 izbranih toèkahnotranjosti modela valja za razliène nastavitve odprtjavstopnega ventila. Z uporabo elektronskokrmiljenega ventilatorja sesalnega stroja je bilomogoèe nastaviti razliène pretoène kolièine zraka priizbranem tlaènem padcu v vstopnem ventilu in takoumetno ustvariti razmere, kakr�ne so pri dejanskemmotorju. Za merjenje hitrosti toka je bil uporabljendvo�arkovni LDA za soèasno merjenje dvehkomponent hitrosti izdelovalca TSI z imensko moèjo4 W, ki pa je deloval dejansko z 1 W. S pomoèjopolo�ajne naprave in zavrtitvijo optiène leèe za 90°je bilo mogoèe doloèiti tudi tretjo komponentohitrosti v isti merilni toèki. Dodatno sipanje svetlobein obravnavo veèjega vzorca podatkov v vsakimerilni toèki je zagotavljalo dodajanje drobnih kapljicglicerina v vstopni zraèni tok. Natanènost LDA jebila preverjena pri ustaljenem naèinu delovanja zvrtljivo plo�èo; pri hitrosti 25 m/s je bil odstopekmanj�i od 0,2 %. Razporeditev merilnih mest vopazovanih vodoravnih prereznih ravninah modelavalja bo prikazana v poglavju 3.

Payri in soavtorji [18] so ugotovili, da oblikaèela bata, oziroma oblika zgorevalnega prostora vbatu, nimata pomembnega vpliva na obliko toka vèasu celotnega sesalnega in vsaj v zaèetku taktastiskanja. O podobnih ugotovitvah poroèajo tudiZhu in soavtorji [19], ki poleg omenjenega tudiugotavljajo, da je vpliv oblike kotanje v batu prihitrotekoèih dizelskih motorjih pomemben �ele vzadnji fazi postopka stiskanja delovnega zraka, kose pojavi preèno iztiskanje zraka iz èela bata vnotranjost zgorevalne kotanje v batu. TudiArcoumanis in soavtorji [20] ter Wakisaka in soavtorji[21] ugotavljajo, da oblika zgorevalnega prostoranima vpliva na obliko tokovnega polja pri polnjenjuvalja motorja. Tako lahko ugotovimo, da zaradipoenostavitve oblike èela bata (ravna povr�ina) priopravljenih preizkusih ni trpela kakovost dobljenihrezultatov numeriène simulacije RDT niti rezultatipreizkusnih raziskav v modelu valja.

V tem prispevku je bil pretoèni koeficientpolnilnega ventila izmerjen v pogojih ustaljenegatoka in mirujoèega bata. Podobni naèin so uporabili

diesel engine) with a moving piston and with an at-tached real metal cylinder head was applied for theexperimental analysis. Details of the design and theexperimental procedure can be found in the work ofPrijatelj [17]. An accurate (accuracy ± 0.01 mm) posi-tioning system controlled by a PC with stepper mo-tors ensured the repeatability of the selected meas-urement points and the measurement of the third flow-velocity component. The velocity field in the wholecylinder was determined for selected intake-valve lifts.The speed-controlled fan of a vacuum cleaner madepossible settings of the pressure drop in the inlet valveas well as air mass-flows that corresponded to realis-tic engine operation under diverse running conditions.A two-component laser Doppler anemometer (LDA)for simultaneous measurement of two velocity com-ponents (blue and green beams), manufactured byTSI with a rated power of 4 W, but operating at ap-proximately 1 W, was applied for the experiments. Thethird velocity component was measured by a simpleturning of the optical probe by 90 degrees. Glycerindrops were introduced in the air-flow as seeding par-ticles to ensure an appropriate number of sampleddata to determine the flow-velocity components. Theaccuracy of the applied LDA was previously testedwith a rotating disc: at a velocity of 25 m/s the errorwas smaller than 0.2 %. The determination of the ob-served horizontal measurement planes in the cylinderand the position of the measurement points will begiven in the 3. section.

According to Payri et al [18], the piston-crowngeometry (facing the combustion chamber) has littleinfluence on the in-cylinder flow during the entire in-take stroke and during the first part of the compres-sion stroke; this conclusion also corresponds to theresults published by Zhu et al [19], who reported thatthe effects of the geometry of the piston bowl pip fora high-speed direct-injection diesel engine becomemore pronounced late in the compression stroke dueto the squish. Furthermore, Arcoumanis et al [20] andWakisaka et al [21] reported that the geometry of thecombustion chamber is not significant when the airflow inside the cylinder during the intake stroke isanalyzed. Hence, it can be concluded that the qualityof the results obtained with the simplified geometricalmodel of the piston (flat piston crown) applied forCFD and experimental analyses are not affected bythis simplification.

The results of the measurements and com-putations of the steady-flow discharge coefficientare presented in the second part of the presented



tudi drugi avtorji ([22] do [24]), medtem ko sta Fukutaniin Watanabe [25] potrdila dobro ujemanje medstatiènimi in dinamiènimi vrednostmi pretoènegakoeficienta.

2 NUMERIÈNE SIMULACIJE VTOKAV VALJ

Natanènost izraèunanih rezultatov je moènoodvisna od geometrijske podobnosti modelapretoènih kanalov, ventila in valja, ki so biliuporabljeni pri raèunskem modelu. Z uporaboMicroscribe 3 D kopirne roke in programomraèunalni�ko podprtega naèrtovanja (RPN) so biliobrisi pretoènih povr�in notranjosti valjeve glaveoznaèeni in kasneje s programom za 3 D modeliranjeMechanical Desktop ponovno spremenjeni vdejansko obliko pretoènih kanalov motorja. ki je bilaprimerna za obravnavo v programskem paketu RDT.Ta omogoèa ob upo�tevanju predpisanih robnihpogojev re�itev gibalnih enaèb in izraèun znaèilnostidinamike zraènega toka (RDT). Natanènostizraèunanih rezultatov je odvisna od gostoteraèunske mre�e in izbranega turbulentnega modela.Primer izbrane oblike in gostote raèunske mre�e vizbranem vzdol�nem prerezu modela vtoènega kanalain valja, je prikazan na sliki 2. Payri in soavtorji [18]so na podlagi preizkusnega dela ugotovili, dapredpostavka izotropne turbulence v obièajnem

study. A similar experimental procedure was also pro-posed by other authors ([22] to [24]), while Fukutaniand Watanabe [25] confirmed good agreement be-tween the static and dynamic discharge coefficient.

2 NUMERICAL SIMULATIONS OFTHE INTAKE FLOW

The quality of the computed results stronglydepends on the applied geometry of the engine in-take channel and the cylinder. An appropriate ge-ometry of the engine intake channel was thereforetransferred to the 3D model, applying the CAD code.A 3D contour of the real engine system was pre-cisely copied by the measurement arm Microscribe3D and then transformed into a realistic engine ge-ometry by applying a computer program for 3Dmodeling - Mechanical Desktop. The channel, cyl-inder, valve etc., geometry model was finally trans-ferred into the CFX program package applied for thenumerical evaluation of the flow dynamics (CFD) bysolving momentum equations and by consideringthe prescribed boundary conditions. The accuracyof the computed results mostly depends on the den-sity of the computational mesh and on the suitabil-ity of the applied turbulent model. The details of theapplied mesh configuration in one longitudinal sec-tion of the observed inlet channel and cylinder arepresented in Fig.2. The standard k�e turbulence

Sl. 1. Preizkusna naprava za meritev hitrostnega polja v modelu valjaFig.1. Experimental setup for in-cylinder velocity measurements



modelu k�e ne prina�a veèjih napak pri izraèunutokovnih razmer v valju motorja � posebno primernaje za modeliranje toka v zgorevalnih prostorih brezpoudarjenega vrtinènega toka oziroma brez izrazitepreène komponente hitrosti. S tem je bila potrjenatudi upravièenost uporabe turbulentnega modelak�e tudi v obravnavanem primeru pri re�evanjudiskretiziranih Navier-Stokesovih enaèb.

Tudi izbira robnih pogojev je bila vprikazanem primeru podobna ustrezni izbiri drugihavtorjev (npr. Payri [18]): robni pogojnespremenljivega tlaka je bil predpisan na vstopu inizstopu zraènega toka v model valja, medtem ko stabila za stene valja in vstopnih kanalov predpisanaadiabatni in robni pogoj brez zdrsa (obodna hitrosttekoèine ob steni je enaka 0). Poleg tega je bilapredpisana tudi stalna vrednost temperature zrakana vstopu in predpostavka nestisljivega sredstva.Ta predpostavka ni vnesla v izraèun gostote zrakapomembnej�e napake, ker so bile tudi ustreznespremembe tlaka (podatek) v pretoènem prerezuventila zelo majhne (tlaèno razmerje v ventilu zna�a0,97 do 0,99). Obièajni turbulentni model k�e je biluporabljen tudi pri izraèunu (modeliranju) pretoènegakoeficienta. Vrednosti za turbulentno kinetiènoenergijo k in raztrosno hitrost kinetiène energije e je

model was applied to solve the discretized Navier-Stokes equations, since Payri et al [18] experimen-tally verified that the isotropic turbulence assump-tion of the standard k�e model is reasonably accu-rate for calculations of the in-cylinder flow, particu-larly in quiescent combustion chambers where thereis no strong interaction between swirl and squishand, more generally, in zones of the cylinder wherethe radial flow does not dominate the flow pattern.

In accordance with other publications (seealso [18]), constant pressure boundaries were as-signed to the flow inlet and outlet, whereas no-slip and adiabatic boundary conditions were ap-plied on solid boundaries. The measured tempera-ture was assigned to the intake-flow boundarycondition, and the non-compressible flow wasassumed during the computation procedure dueto a very small air density change being the con-sequence of very small pressure fluctuations inthe intake valve (the pressure ratio in the inletvalve was 0.97 to 0.99). The values for the turbu-lent kinetic energy k and the velocity dissipationof the kinetic energy e should also be prescribedat the flow inlet boundary condition. The option�default intensity and autocompute length scale�was assumed for the performed computations due

Sl. 2. Primer uporabljene raèunske mre�e v vzdol�nem prerezu modela vstopnega kanala in valjaFig.2. Details of the computational mesh in the longitudinal section of the inlet channel and cylinder



bilo treba doloèiti na vstopnem robnem pogoju.Zaradi nepoznavanja teh vrednosti je bila privzetaintenziteta in izraèunana dol�inska skala, kar pa nasreèo ni imelo bistvenega vpliva na izraèunanostrukturo toka.

3 PRIKAZ IN OBRAVNAVA REZULTATOV

3.1 Tokovno polje in potek hitrosti v valju motorja

V sklopu preizkusnih raziskav so bili na 33razliènih merilnih mestih v petih vodoravnih � preènihravninah osmerostranega valja (torej skupno v 165toèkah) izmerjene po tri komponente vektorja hitrostitoka. Veè toèk v vsaki ravnini ni bilo mogoèe izmeritizaradi po�evne postavitve steklenih sten modela. Nasliki 3 je prikazan potek oziroma primerjava izmerjenih inizraèunanih hitrosti v preèni ravnini x � y, ki je oddaljena120 mm od vrha valja. Iz diagrama lahko povzamemo, da:- se lega sredi�èa izraèunanega vrtinca le malo

razlikuje od lege sredi�èa vrtinca toka, ki je bilizmerjen z meritvami LDA,

- se absolutna velikost izraèunanih in izmerjenihvrednosti vektorjev hitrosti v opazovani ravniniprecrej dobro ujema z vrednostmi meritev.

Obodno gibanje (kro�enje v preènih ravninah)sve�e polnitve v valju je zelo pomembno za pravilno

to a lack of any appropriate information. Fortu-nately, their choice has little effect on the flowstructure for the observed case. The standard tur-bulent k�e model was also applied to compute theflow-discharge coefficient.

3 RESULTS AND DISCUSSION

3.1 Flow pattern and velocity field within the cylinder

Three components of the velocity vectorswere measured in the octagonal glass cylinder modelat 33 different measurement points and in 5 differentcylinder transverse planes (altogether at 165 loca-tions). Fig. 3 represents the computed and meas-ured velocity vectors in the x-y plane, 120 mm fromthe cylinder top (z � axis). It can be concluded fromthis diagram that:- the positions of the calculated and measured vor-

tex center nearly coincide- the size of the computed and measured velocity

vectors (length of particular indicated velocityvector) are similar within the observed plane

Tangential motion (in the transverse cylin-der planes) of the fresh charge is very important forappropriate air-fuel mixing, and therefore for effec-tive combustion. Thus, the swirl ratio or the swirl

Sl. 3. Primerjava izmerjenih in izraèunanih vektorjev hitrosti v preèni ravnini valja 120 mm od vrha(smer z); temnej�e pu�èice � meritev, svetlej�e pu�èice � raèunska simulacija

Fig. 3. Comparison of measured and computed velocity vectors in the cylinder transverse section 120mm from the top (direction z); black arrows � experiment, light arrows � numerical simulations



me�anje zgorevalnega zraka z gorivom in poslediènoza èim popolnej�e zgorevanje. Velikost in porazdelitevobodne hitrosti toka v valju moèno vplivata navrtinèno �tevilo toka v valju, ugotavljajo Payri insoavtorji [18]. Zato je na sliki 4 prikazana tudiprimerjava poteka in velikosti obodne komponentevektorja hitrosti v preèni ravnini valja x � y 40 mm odvrha valja. Potek in smer opazovanih hitrosti stapodobna za oba primera. Velikost izraèunanihvrednosti, posebno blizu stene valja (ni prikazano natej sliki) in v preènih ravninah na polovici vi�ine valja,je veèja od izmerjenih vrednosti. Vzrok tièi verjetno vnaslednjih pomanjkljivostih: 1) neupo�tevanjuspremembe lastnosti toka pri pretakanju ob steniosmerostranega modela valja in 2) zaradi nenatanènihmeritev padca tlaka med opazovanim vstopnim inizstopnim robom toka zraka, ki je rabil tudi kot robnipogoj pri numeriènem modeliranju.

3.2 Pretoèni koeficient vstopnega ventila

3.2.1 Doloèitev pretoènega koeficienta z meritvamiin raèunsko simulacijo

Najprej je bil pretoèni koeficient doloèen zuporabo rezultatov meritev. Potem je bila izvedenaraèunska simulacija masnega toka zraka skozi vstopni

number is strongly affected by the size and the dis-tribution of the tangential velocity, F.Payri et al [18].The distribution of the calculated and measured tan-gential velocity component in the transverse plane40 mm from the cylinder top (z = 40 mm), which isperpendicular to the x and y axes is presented in Fig.4. It can be concluded that the orientation of thetangential velocities is similar for both cases. Thecalculated values are larger in comparison with themeasured ones, especially in the transverse planeshalf-way from the cylinder top and in the vicinity ofthe cylinder wall (not presented in this figure). Thereason for this can be found in: 1) a probably incor-rect consideration of the flow characteristic whenthe computation of the flow along the octagonal cyl-inder wall is considered, and 2) incorrect measure-ment of the pressure drop, which represents theboundary condition for the performed computations.

3.2 Discharge coefficient

3.2.1 Experimental and numerical determinationof the discharge coefficient

The discharge coefficient was first deter-mined experimentally, i.e., it was calculated from themeasured in-cylinder mass-flow data by applying

Sl. 4. Primerjava izmerjenih in izraèunanih obodnih hitrosti v dveh izbranih legah v preèni ravnini valja40 mm od vrha (smer z)

Fig. 4. Comparison of the measured and computed tangential velocity components in the cylinder trans-verse section 40 mm from the top (direction z)

vt (r), z=0,04 m



ventil, izraèunani ustrezni pretoèni koeficienti inrezultati primerjani z rezultati preizkusov. Pretoènikoeficient m je bil doloèen z enaèbo:

kjer pomeni Ag

najmanj�o geometrijsko pretoènopovr�ino vstopnega ventila, A

eff pa dejanski

pretoèni prerez v vstopnem ventilu:A

eff je bil potem izraèunan z enaèbo (2), ki

opredeljuje masni tok (podrobnosti najdemo v viru[17]) stisljivega medija skozi zaslonko pri znanihpodatkih: tlaènem razmerju v ventilu p

2/p

1=0,97, ter

temperaturi in tlaku okolice T1 = 295K in p

1 = 100 kPa:

Primerjava izmerjenih in izraèunanihvrednosti koeficienta m je prikazana na sliki 5.Rezultati se zelo dobro ujemajo pri odprtju ventilaza 2 mm. Potem je opazno hitrej�e veèanje inmanj�anje (odmik) izraèunanih vrednostikoeficienta m. V celoti se vrednosti meritev inizraèunov sorazmerno dobro ujemajo. Tudi oblikaspremembe poteka izmerjene kriulje m s sprememboodprtja vstopnega ventila se ujema z napovedmidrugih avtorjev ([22] do [24]) in je potrjena sspremembami toka v �pranji ventila. Pri majhnihodprtjih ventila, pri katerih ne prihaja do odcepitevtoka od trdne stene, vpliv mejne plasti pa je tudizanemarljiv, smer toka sledi obliki ventilne glavein sede�a ventila. Z veèanjem odprtja ventila sezato v prvi fazi odpiranja ventila intenzivnopoveèuje tudi vrednost m. Pri nekoliko veèjihtokovih, veèjih odprtjih ventila, se najprej pojaviodcepitev toka od vodilne povr�ine ventilneglave; dejanska pretoèna povr�ina se zmanj�a in znjo tudi vrednost m. Z nadaljnjim poveèevanjemodprtja ventila opazimo odcepitev toka tudi nastrani sede�a ventila, kar vpliva na nadaljnjezmanj�anje vrednosti m. Potem se usmeritevzmanj�evanja vrednosti m nekoliko umiri zaradiponovnega prilepljenja toka ob vodilno povr�inoventilne glave in s tem nekoliko poveèanegadejanskega pretoènega prereza ventila. Pri velikihvrednostih odprtja ventila opazimo nadaljnje,vendar zmerno zmanj�evanje pretoènegakoeficienta.

the equation of compressible flow through the noz-zle, see also [17]. The discharge coefficient is deter-mined by the equation:

(1),

where Aeff

denotes the effective cross-sectionthrough the inlet valve and A

g the minimum geo-

metrical cross-section of the inlet valve.A

eff was then calculated from Equation (2) by

applying the results of the simulation for the mass-flow m& , from the known pressure ratio across thevalve p

2/p

1=0.97 and from the ambient parameters

T1 = 295K and p

1 = 100 kPa:

(2).

The results of the numerical simulations ofthe discharge coefficient in the inlet valve were com-pared to those obtained by experiments. The resultsof the comparisons are presented in Fig. 5. The re-sults coincide very well for a 2-mm valve lift, whereasit can be seen that the numerical simulation predictsa gradual increasing and a subsequent decreasingof the discharge coefficient earlier than the experi-mental data for approximately 2 mm of the valve lift.Overall, the measured and computed results are inrelatively good agreement. The shape of the dis-charge coefficient curve also coincides well with theresults reported by other authors ([22] to [24]). Fromthe analysis of the flow structure it can be seen thatat very small valve openings the air flow follows theshape of the valve seat and the valve head � no flowseparation occurs and due to the negligible effectsof the boundary layer the discharge coefficient in-creases with increasing valve lift. At larger airflowsthe boundary layer breaks away from the valve head,reduces the effective flow area and thus reduces thedischarge coefficient. Further opening of the valvegives rise to the flow separation, even at the side ofthe valve seat, and substantially decreases the dis-charge coefficient. The decrease of the dischargecoefficient is less pronounced with still further open-ing of the valve, due to the re-attachment of the flowto the valve head � the effective flow area is, there-fore, relatively larger. The discharge coefficient thencontinues to decrease at the largest valve openings,but at a lower rate.

eff

g

A

Am =

12 ê 1

ê ê1 2 2

1 11

2ê 1

ê-1eff

p p pA m

R p pT

-+æ öæ ö

ç ÷æ ö æ öç ÷= × × × × × -ç ÷ ç ÷ç ÷ç ÷è ø è øç ÷ç ÷è øè ø

&



3.2.2 Ocena merilne negotovosti pri doloèitvipretoènega koeficienta

Pri oceni skupne merilne negotovosti zapretoèni koeficient ì, so bili v skladu z enaèbama 1in 2 in ob upo�tevanju ustrezne literature [26]upo�tevane naslednje merilne negotovosti:- ocenjena merilna negotovost izmerjenih

statiènih tlakov ur(p1) = ±1%

- merilna negotovost izmerjene temperature ur(T)

= ±0,17%- merilna negotovost izmerjenega masnega toka

zraka ur( m& ) = ±0,8 %

- merilna negotovost izmerjenega masnega tokazraka u

r() = ±0,8 %

- pri doloèitvi merilne negotovosti geometrijskegapretoènega preseka Ag so bile ocenjeni merilnipogre�ki: pri doloèitvi premera ventila ± 1 mm inpri dvigu ventila ± 0,05 mm; ocenjena merilnanegotovost u

r(Ag) je torej zna�ala ±2,5%.

- Izraèunana je bila tudi merilna negotovostdejanske - efektivne pretoène povr�ine ventilau

r(Aeff) v merilnem obmoèju masnih tokov zraka

(= 0,1 kg/s), ki ne presega ± 1,38%Konèno je bila izraèunana �e skupna merilna

negotovost pretoènega koeficienta:

3.2.2 Assessment of the measurement uncertaintyfor the determination of the discharge coefficient

Taking into account measurement parametersin Equations 1 and 2 and in corresponding literature[26] when determining the total collectivemeasurement uncertainty the following individualuncertainties were considered:- Of the measured absolute pressures u

r(p1) =

±1%- Of the measured temperatures u

r(T) = ±0.17%

- Measurement uncertainty of the measured airmass-flow u

r( m& ) = ±0.8 %

- The following measurement errors wereestimated when assessing the measurementuncertainty of the geometrical valve flow areaAg: ± 1 mm for the valve diameter and ± 0.05 mmfor the valve lift; the assessed measurementuncertainty u

r(Ag) was therefore ±2.5%.

- The measurement uncertainty of the valveeffective flow area u

r(Aeff) was then determined

for the observed mass-flow range of 0.1 kg/sand amounted to ±1.38%.

Finaly the total measurement uncertainty ofthe discharge coefficient was determined for theobserved air mass-flow applying the formula:

Sl. 5. Potek pretoènega koeficienta v odvisnosti od odprtja ventilaFig.5. Discharge coefficient for the intake valve, depending on the valve lift

0,97



ki v opazovanem merilnem obmoèju masnih tokovzraka okrog 0,1 kg/s ne presega ±2,9%

4 SKLEPI

Iz predstavljenih rezultatov raziskave lahkopovzamemo, da je uporaba metode RDT primernain da zagotavlja preprièljive rezultate, ki izbolj�ujejorazumevanje tokovnih pojavov, ki spremljajo vtokmedija v valj motorja. RDT je torej uèinkovito orodjepri snovanju sodobnih motorjev. Omogoèa dobronapoved pojava, mesta in velikosti vrtinènihstruktur v valju. Primerjava med izmerjenimi inizraèunanimi vrednostmi obodnih komponenthitrosti toka v valju modela motorja je pokazala, daso izraèunane vrednosti veèje predvsem v bli�inistene valja in v preènih ravninah, ki le�ijo v podroèjupolovice vi�ine valja. Na to so verjetno vplivali:netoèni izmerjeni podatki padca tlaka v ventilu, kiso obenem pomenili tudi predpisani robni pogoj priopravljenih raèunskih simulacijah in nepoznavanjetokovnih razmer pri kro�enju zraka ob osmerostranihstenah modela valja.

Dobro ujemanje med izmerjenimi inizraèunanimi vrednostmi lahko opazimo tudi prirezultatih pretoènega koeficienta v vstopnem ventilu.Iz rezultatov izraèunov izhaja, da je izraèunzanesljivej�i pri veèjih odprtjih ventila, ko je vplivmejne plasti na tok zraka v �pranji med glavo insede�em ventila manj�i. Gostota raèunske mre�e jelahko zaradi tega pri veèjih odprtjih ventila ustreznomanj�a, s tem pa se ustrezno izbolj�a tudi razmerjemed kakovostjo rezultatov in zmanj�ano potrebnoporabo raèunalni�kega èasa. Analiza dobljenihrezultatov in primerjava z ustreznimi rezultati drugihavtorjev je tudi pokazala, da je vpliv oblike èela bata(oziroma oblike zgorevalnega prostora v batu) in legebata v valju na velikost in potek pretoènegakoeficienta majhna.

(3),

that never exceeded ±2.9%.

4 CONCLUSIONS

The presented study shows that in-cylinderCFD predictions ensure credible results that help im-prove the knowledge of the air-flow characteristicsduring the intake stroke. CFD therefore represents anefficient design tool for developing less-polluting andmore efficient internal combustion engines. Good pre-diction of the vortex center, direction and even thesize of the flow field was made possible by applyingthe described numerical method. A comparison be-tween the measured and computed tangential in-cyl-inder velocities showed that the relatively larger com-puted values � especially close to the cylinder wall �were probably affected by the incorrect measurementof the pressure drop across the intake valve; it wasapplied as a prescribed boundary condition for theperformed computations. On the other hand, it is alsodifficult to precisely determine the flow field close tothe octagonal walls of the model cylinder.

Good agreement between the measured andcalculated inlet-valve discharge coefficients was ob-tained. It can be concluded from the performed com-putations that it is was relatively convenient to cal-culate the discharge coefficients at larger inlet valveopenings, since the boundary layer has a smallerinfluence on the air-flow within the restricted areabetween the valve head and the valve seat. A goodcompromise between the quality of the results andthe computational times could also be achieved forlarger valve openings, since accurate results couldbe obtained with a reasonable computation-meshdensity. The analysis and comparisons with the ap-propriate results of other authors showed that thatthe piston position in the model cylinder has littleeffect on the size of the discharge coefficient.

%9,2029,00253,00138,0)()(()( 2222 =±=+=+= greffrr AuAuu m

5 LITERATURA5 LITERATURE

[1] Rask, R.B. (1979) Laser Doppler anemometer measurements in an internal combustion engine. SAE Paper790094, 1979.

[2] Heywood, J.B. (1987) Fluid motion within the cylinder of internal combustion engines. Journal of FluidsEngineering, 109 (1987), pp.3-35.

[3] Heywood, J.B. (1988) Internal combustion engine fundamentals. McGraw-Hill, New York.



[4] Huang, R.F., Huang, C.W., Chang, S.B., Yang, H.S., Lin, T.W., Hsu, W.Y. (2005) Topological flow evolu-tions in cylinder of a motored engine during intake and compression strokes. Journal of Fluids andStructures, 20 (2005), pp.105-127

[5] Kang, K.Y., Reitz, R.D. (1999) Intake flow structure and swirl generation in a four-valve diesel engine.Spring Technical Conference, ASME 1999, ICE-vol. 32-2 (1999) , Paper no. 99-ICE-182

[6] Snauwaert, P., Sierens, R. (1986) Experimental study of the swirl motion in the direct injection dieselengines under steady state flow conditions (by LDA). SAE Paper 860026.

[7] Yun, J.E. (2002) New evaluation indices for bulk motion of in � cylinder flow through intake port systemin cylinder head. Proc. Instn. Mech. Engrs. Part D, 216 (2002), pp. 513-521

[8] Davis, G.C., Kent, J.C. (1979) Comparison of model calculations and experimental measurements of thebulk cylinder flow processes in a motored PROCO engine. SAE Paper 790290, 1979.

[9] Murakami, A., Arai, M., Hiroyasu, H. (1988) Swirl measurements and modeling in direct injection dieselengines. SAE Paper 880385, 1988.

[10] Desantes, J.M., Rastor, J.V., Doudou, A. (2001) Study of the steady flow produced by direct injectiondiesel engine intake ports. Proc. Instn. Mech. Engrs. Part D, 215 (2001), pp. 285-298

[11] Uzkan, T., Borgnakke, C., Morel, T. (1983) Characterisation of the flow produced by a high-swirl inlet port.SAE Paper 830266, 1983.

[12] Haste, M.J., Garner, C.P., Halliwell, N.A. (1999) A PIV study of the inlet port deactivation on the in-cylinderflow structure in an SI engine. Fall Technical Conference, ASME 1999, ICE-vol. 33-3 (1999), Paper no.99-ICE-240

[13] Li, Y., Zhao, H., Peng, Z. Ladommatos, N. (2002) Particle image velocimetry measurement of in-cylinderflow in internal combustion engines-experiment and flow structure analysis. Proc. Instn. Mech. Engrs.Part D, 216 (2002), pp. 65-81

[14] Chan, V.S.S., Turner, J.T. (2000) Velocity measurement inside a motored internal combustion engine usingthree-component laser Doppler anemometry. Optics & Laser Technology, 32 (2000), pp. 557-566

[15] Payri, F., Desantes, J.M., Pastor, J.V. (1996) LDV measurements of the flow inside the combustion chamberof a 4-valve D.I. diesel engine with axisymetric piston-bowls. Experiments in Fluids, 22 (1996), pp. 118-128

[16] User manual for the program package CFX 5.5[17] Prijatelj, S. (2004) Analysis of flow phenomena in a valve of internal combustion engine. M.SC. Thesis

No. M/1277 (2004). Faculty of Mechanical Engineering, Ljubljana, UL[18] Payri, F., Benajes, J., Margot, X., Gil, A. (2004) CFD modeling of the in-cylinder flow in direct-injection

diesel engine. Computers & Fluids, 33 (2004), pp.995-1021[19] Zhu, Y., Zhao, H., Melas, D.A., Ladommatos, N. (2004) Computational study of the effects of the geom-

etry of the piston bowl pip for a high-speed direct injection diesel engine. Proc. Instn. Mech. Engrs. PartD, 218 (2004), pp. 875-890

[20] Arcoumanis, C., Begleris, P., Gosman, A.D., Whitelaw, J.H. (1986) Measurements and calculations of theflow in a research diesel engine. SAE Paper 861564, 1986.

[21] Wakisaka, T., Shimamoto, Y., Issihiki, Y. (1986) Three-dimensional numerical analysis of in-cylinder flowsin reciprocating engines. SAE Paper 860464, 1986.

[22] Tsui, Y.Y., Lee, S.Y. (1992) Calculation of turbulent flow through engine inlet ports. Int. J. Heat and FluidFlow, 13 (3) (1992), pp. 232-240

[23] Arcooumanis, C., Whitelaw, J.H. (1987) Fluid mechanics of internal combustion engines � a review. Proc.Instn. Mech. Engrs., 201 (1987), pp. 57-74

[24] Baumruk, P., Hatschbach, P., Mare�, B. (2003) Simulation of steady flow through engine inlet port tocylinder. 2003 EAEC European Automotive Congress A, pp. 229-237.

[25] Fukutani, I., Watanabe, E. (1995) Air flow through poppet inlet valves � analysis of static and dynamicflow coefficients (820154); Design of racing and high performance engines PT-53, SAE 1995, pp.111-121

[26] Coleman, H.W., Steele, W.G.. (1998) Experimentation and uncertainity analysis for engineers, 2nd edition.John Wiley & Sons, Inc., New York.



Authors� Addresses: Dr. Samuel Rodman Opre�nikDr. Toma� Katra�nikProf.Dr. Ferdinand TrencDr. Franèi�ek BizjanUniversity of LjubljanaFaculty of Mechanical Eng.A�kerèeva 6SI-1000 Ljubljana, [email protected]@[email protected]@fs.uni-lj.si

Mag. Sa�o [email protected]

Prejeto: 26.8.2005

Sprejeto: 16.11.2005

Odprto za diskusijo: 1 letoReceived: Accepted: Open for discussion: 1 year

Naslovi avtorjev: dr. Samuel Rodman Opre�nikdr. Toma� Katra�nikprof.dr. Ferdinand Trencdr. Franèi�ek BizjanUniverza v LjubljaniFakulteta za strojni�tvoA�kerèeva 61000 [email protected]@[email protected]@fs.uni-lj.si

mag. Sa�o [email protected]

Documents

Analiza tokovnih struktur v vstopnem sistemu motorjev