Formacion de Las Emisiones en Diesel

7/26/2019 Formacion de Las Emisiones en Diesel

1/55

Measuring and Predicting Transient DieselEngine Emissions

TRITA MMK 2009:07ISSN 1400-1179

ISRN/KTH/MMK/R-09/07-SE

Licentiate thesisKTH CICERO

Department of Machine Design

Royal Institute of TechnologySE-100 44 Stockholm

ANDERS WESTLUND


2/55

TRITA MMK 2009:07

ISSN 1400-1179

ISRN/KTH/MMK/R-09/07-SEMeasuring and Predicting Transient Diesel Engine Emissions

Anders Westlund

Licentiate thesis

Academic thesis, which with the approval of Kungliga Tekniska Hgskolan, will be presented

for public review in fulfilment of the requirements for a Licentiate of Engineering in MachineDesign. The public review is held at Kungliga Tekniska Hgskolan, Brinellvgen 23 in room

B3, 24th of March 2009 at 10:00.


3/55

Measuring and Predicting Transient Diesel Engine Emissions 2009Abstract

Dueto

its

impact

on

human

health

and

the

nature

surrounding

us,

diesel

engine

emissions have been significantly reduced over the last two decades. This

reduction has been enforced by the legislating organs around the world that

gradually have made the manufacturers transform their engines to todays

complex hightech products. One of the most challenging areas to meet the

legislationsistransientoperationwheretheinertiaingasexchangesystemmakes

transitionfromoneloadtoanotherproblematic.

Modernengineshavegreatpotential tominimize the problemsassociatedwith

transient operation. However, their complexity also imposes a great challenge

regardingoptimizationandsystematical testingoftransientcontrolstrategies in

anenginetestbedcouldbebothexpensiveandtimeconsuming.

The objective of this project is to facilitate optimization of transient control

strategies.Thisshouldbedonebyidentifyingappropriatemeasurementmethods

forevaluationoftransientsandbyprovidingmodelsthatcanbeusedtooptimize

strategiesoffline.

Measurement methods for evaluation of transients have been tested in several

experiments, mainly focusing on emission but also regarding e.g. EGR flow.

Applicableinstrumentsfortransientemissionmeasurementshavebeenidentified

andused.However,nomethodtomeasuresootemissionscycleresolvedhasyet

been found. Other measurements such as EGR flow and temperatures are

believedtohavesignificantlydecreasedaccuracyduringtransients.

Amodel

for

prediction

of

NOx

emissions

have

been

used

and

complemented

with

a new approach for soot emission predictions that has been developed in this

project. The emission models have been shown to be applicable over a wide

rangeofoperatingconditionswithexception forhighlypremixedcombustion. It

has also been shown that models developed for steadystate conditions can be

usedfortransientsoperation.


4/55

Measuring and Predicting Transient Diesel Engine Emissions 2009

2

Contents

1 Introduction............................................................................................................................. 5

2 Projectobjectives.................................................................................................................. 13

3 Emissionformationindieselengines.................................................................................... 16

3.1 NOxformation................................................................................................................ 17

3.2 Sootformationandoxidation........................................................................................ 18

3.3 Emissionreduction......................................................................................................... 20

4 Dieselcombustionmodeling................................................................................................. 21

4.1 Simplifieddieselemissionmodeling.............................................................................. 23

4.2 Simplifiedcombustionmodeling................................................................................... 23

4.3 Simplifiedchemicalreactionschemes........................................................................... 25

4.4 Simplifiedemissionmodeling........................................................................................ 26

4.4.1 Phenomenologicalmodels...................................................................................... 26

4.4.2 Semiempiricalmodels............................................................................................ 30

4.4.3 Statisticalmodels.................................................................................................... 31

5 AdvantagesanddisadvantagesofdifferentemissionmodelcategoriesMotivationforchosenmodels.............................................................................................................................. 33

6 Measuringandevaluatingtransientoperation..................................................................... 35

6.1 Gasexchangesystemmeasurements............................................................................ 35

6.1.1 TransientEGRmeasurements................................................................................ 36

6.1.2 Transientairflowandair/fuelratiomeasurements..............................................37

6.2 Emissionmeasurements................................................................................................ 38

6.2.1 NOxmeasurements................................................................................................ 38


5/55


3

6.2.2 Sootmeasurements................................................................................................ 38

6.3 Postprocessingemissionpredictions............................................................................ 42

7 Introductiontopapersandkeyresults................................................................................. 43

7.1 Paper1,EvaluationofTechniquesforTransientPMmeasurements........................43

7.2 Paper2,PredictionsandMeasurementsofTransientNOEmissionsforaTwostageTurbochargedHDDieselEnginewithEGR.............................................................................. 44

7.3 Paper3,FastPhysicalPredictionofNOandSootinDieselEngines..........................45

7.4 Paper4,FastPhysicalEmissionPredictionsforOfflineCalibrationofTransientControlStrategies.................................................................................................................... 46

8 Summary................................................................................................................................ 47

9 Futurework........................................................................................................................... 48

10 Acknowledgements............................................................................................................ 49

11 References......................................................................................................................... 50


6/55


4


7/55


5

1 Introduction

In 1892, a German engineer named Rudolf Diesel patented a new kind of

combustionengine.

The

diesel

engines

advantage

over

other

internal

combustion

engines was, and still is, its efficiency. The engines spreading vary in different

markets depending on the geographical position, fuel supply and sometimes

tradition.Italsovariesindifferentmarketsegments;itismoredominatinginhigh

mileagesegmentswhere the improved fueleconomysurpasses itshigherprice.

The diesel engine is dominating in heavy application such as busses and trucks

and itsmarketshare in thepassengercarsegment issteadily increasing,mainly

duetoincreasingfuelpriceandimprovedemission.In2007,53%ofthesoldcars

intheEuropeanUnionwerepoweredbyadieselengine[1].

Figure1Anearlydieselengine,[2].


8/55


6

Inadieselengine,thefuelisinjecteddirectlyintothecylinderduringthelastpart

ofthecompressionstrokeandthefirstpartoftheexpansionstroke.Theinjected

fuelmixeswiththeairinthecylinderandignitesduetothehightemperatureof

thecompressedgasesinthecylinder.Theignitionusuallyoccursduringtheinitial

partoftheinjectionandthefuelthathasbeeninjecteduptothatpointcombusts

rapidly.Thispartofthecombustioniscalledthepremixedcombustionperiodand

can be a significant fraction of the total combustion period, typically when the

injectiondurationisshort.Thefuelthatisinjectedafterthispointburnsasithas

evaporatedandmixedwithair to incombustableproportionsafteronlyashort

reaction time compared to the initial ignition delay. This combustion period is

thereforecalled

diffusion

combustion.

The

combustion

process

is

both

turbulent

and instationary but a time averaged, conceptual model for the diffusion

combustionperiodhasbeenproposedbyDec([3])andthisprovidesasimplified

way to describe the process. Decs model is shown in a schematic in figure 2

togetherwithapictureofarealcombustioninanopticalaccessengine.

Figure2Dieselcombustionprocess,[3],[4]


9/55


7

The disadvantage with diesel combustion, compared to e.g. Otto combustion

(normally petrol also called Spark Ignited (SI) combustion), is that it is not

homogenous throughout the cylinder, i.e. parts of the combustion takes place

with a shortage of oxygen, so called rich conditions. Where there is rich

combustion, soot is formed and although most of it is oxidized as the fuel is

furthermixedwith thesurroundingair,someof itwillremainas thegasesare

expelledoutof the cylinderand through the tailpipe.The regionsweresoot is

formedaccordingtoDecsmodelisthebrownregionshowninFigure2.Soot is

emitted from the engine in the form of particles on which water and

hydrocarbons are condensated as the exhaust gases are cooled down. These

particlesare

together

with

other

solid

and

liquid

material

in

the

exhaust

gases

calledParticulateMatter(PM).

Particulatematterisahazardforhumanshealthandtotheenvironmentandis

togetherwithNOxthemainemissionfromthedieselengineandtherebyoneof

its greatest challenge. Figure 2 also show a green region where most NO is

formed; in the flame front where the air/fuel mixture is approximately

stoichiometricandtheflame isashottest.NOx isalsoformed inSIenginesbut

since

these

are

operated

without

excess

air,

they

can

take

care

of

the

emissions

relatively easy with a threeway catalytic converter. This is not possible in the

diesel engine where overall air excess is necessary to achieve acceptable

combustionefficiencyandsootemissions.

PMemissions have been shown to increase respiratorical symptoms, decrease

lungfunctionand increaseuseofasthmaticmedicine.Increasedhealthcareand

mortalityhasalsobeenprovedregardingcardiovascularandlungdiseases[5].

NOxhas

been

shown

to

cause

increased

sensitivity

to

the

respiratory

passages

[5]

and is also the cause of smog. By the influence of sunlight, NOx forms smog

togetherwithhydrocarbonemissions.


10/55


8

The maximum allowed quantities of these emissions are limited by the local

emission legislations. In Europe, these have been defined by European Union

directivesin1992andtightenedinseveralstepsafterwards(Euro1,Euro2etc.).

Asof2008,theEuro5legislationiscomingintoeffectandthenextplannedstep

isin2013,howeverthelevelsforEuro6arenotfixedyet.Thelegislatedlevelsare

showninFigure3whereeachmarkerrepresentsaEurolevel(Euro2wasupdated

onetimeandthereforehastwomarkers). Itshouldbenotedthatsincethefirst

Eurolevel,themaximumallowedPMemissionshavedecreaseby95%andNOx

with 75%. When the suggested levels for Euro 6 come into effect, the

correspondingdecreasewillbe97%and95%respectively.

Figure3EmissionlegislationevolutioninEurope,[1]

To control that the legislation is met, the engine manufacturers have to certify

theirenginesaccordingtoproceduresalsodefinedbyEuropeanUniondirectives.

Beforeyear2000,thecertificationprocedureonly includedstationaryoperation

points, i.e.theenginewasoperatedatthirteendefinedstationaryconditions (%

of maximum speed and load) and these were weighted to represent typical

engineusage.After the year2000, sinceEuro 3, transientengineoperationhas

beenincludedinthecertificationproceduretobetterrepresentrealworldusage.

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0

2

4

6

8

10

1990 1995 2000 2005 2010 2015

PM[g/kWh]N

Ox[g/kWh]

Year

NOx

PM


11/55


9

TheEuropeanTransientCycle(ETC)hasbeenusedforEuro3to5andisshownin

Figure 4. This cycle will be replaced by a World Harmonized Transient Cycle

(WHTC)forEuro6.TheWHTChasbeencreatedtocovertypicaldrivingconditions

intheEU,USA,JapanandAustraliawherelocalcertificationcyclesareusedtoday.

ThestationarypartofEuro6willalsobeharmonized.

Figure4EuropeanTransientCycle,ETC

AscanbeseeninFigure4,theETCcycleishighlyvaryingregardingbothloadand

speed and there are significant consequences when adding this part of the

certificationtothestationary.

Insteady

state

conditions,

different

combinations

of

load

and

speed

correspond

to different conditions in the engine regarding the gas flow in the engines gas

exchange circuits and regarding the temperature of all gases, liquids and

components.Whenatransitionfromoneloadandspeedtoanotherisdemanded

from the engine, the steady state flows and temperature will not be obtained


12/55


10

directlyduetherotational inertiaoftheturbochargerandthethermal inertiaof

the components in the engine. The period between when the new load and

speed were issued until when the corresponding steady state flows and

temperaturearereached iswhat isusuallyreferredtoastransientconditionsor

just transients. In transients, thebiggestproblem is the rotational inertiaof the

turbochargerwhich isthepropertyresponsible for thesocalled turbo lagwhen

thepowerdemandisincreased.Whenmorepowerisdemandedfromtheengine,

the turbine and compressor needs time accelerate its rotational speed and air

flowtomatchthedemand.

InanOttoengine,theinjectedamountoffuelismoreorlessproportionaltothe

airflowsincetheenginehastorunstoichiometrictomakethecatalyticconverter

workproperlyandthereforethetransientpowerdeliveryfromaturbochargedSI

engine is governed by the turbo (after the initial power increase gained by

openingthethrottle).

Atransient inadieselengine isabitdifferent; theenginealwaysoperateswith

excessairandwhenmorepowerisdemanded,theexcessamountisminimizedto

allowtheinjectedamountoffueltoincreasefaster.Theengineisthenranatthis

minimumexcess

amount

of

air

until

the

demanded

amount

of

fuel

is

reached

and

thereaftertheair/fuelratiowill increaseagainuntilitreachesthecorresponding

steadystatevalue.

Thisperiodduringapowerouttakeincrease(e.g.acceleration),whentheair/fuel

ratio is lower than the corresponding steady state value, causes increased soot

emissions due to the impaired conditions for oxidation. The peak values of the

transient soot emission can be more than ten times higher than the

correspondingsteady

state

values

[6].

An

(a

perhaps

abit

extreme)

example

of

suchoccurrencecanbeseeninFigure5.


13/55


11

Figure5Sootcanbeanissueduringacceleration[7]

There are several different measures to decrease the transient soot peaks but

transient engine control strategies also has to consider other aspects, mainly

engine response (power increase rate) and NOx emissions. Unfortunately the

requirements of the different aspect are often contradictory. An effective and

widespreadwaytodecreaseNOxemissionistheuseofExhaustGasRecirculation

(EGR). However, the engine response is dramatically slower when EGR is used

undertransientsandthesootemissionsaresignificantlyincreasedasEGRsteals

energy from the turbine and thereby slows down the boost build up. Another

contradictionistheminimumallowedair/fuelratioallowedduringthetransient;

a

lower

value

means

that

the

fuel

injection

rate

can

be

increased

faster

andtherebyincreasestheengineresponsebutatthesametime,thismeansahigher

sootemissionpeak.


14/55


12

TheEGRrateandtheminimumallowedair/fuelratioarecontrolparametersthat

should be optimized as well as VGT strategy (Variable Geometry Turbine) and

injectionstrategy(injectionangle,injectionpressureandmultipleinjections).

Eventhoughaftertreatmentsystemsarebecomingbetter,furtherdecreaseofthe

engineoutemissionsisnecessaryforfutureemission legislationlevels.Insteady

state operation, these parameterscanbe systematically optimized in anengine

testbed toobtain thebestcompromisebetweensoot,NOxand fuelefficiency.

Fortransientoperationontheotherhand,whennotonly loadandspeedhasto

beconsideredbutalsothetransientstate,systematicallytestingcanbeextremely

timeconsumingduetotheveryhighnumberofcombinationsofconditionsand

controlsettings.Transientcontroloptimizationisthereforeagreatchallenge.


15/55


13

2 Projectobjectives

The aim of this PhD project is to improve the understanding of transient

phenomena

and

to

develop

tools

for

transient

control

optimization.

In

more

concreteterms,theobjectivesaretobeableto:

measureNOxandsootduringtransients

measuretransientflowandgastemperatures

simulatetransientturboperformance

simulatetransient

EGR

flow,

temperature

and

mixing

ratios

simulateNOxandparticleformation

Eventhoughthetransientcyclehasbeen included inthecertificationprocedure

sinceyear2000,mostoftheemissionmeasurementtechniquesthatarestillused

stem from instruments developed for steady state testing. As an effect of this,

mostinstrumenthavegivenprioritytopropertiessuchasmeasurementstability

andreproducibility and long service intervals. Very few instruments therefore

have the response time necessary to study the cycle to cycle phenomena that

occurduringtransients.

Thefirstpartoftheprojecthasbeenmostlyaboutfindingemission instruments

thataresuitable for transientmeasurementsand thishasbeendoneby testing

both instruments developed for steady state measurements and also some

instrumentwithnewertechnologydevelopedfortransientmeasurements.

Gasflows

such

as

EGR

flow

and

the

air

flow

into

the

engine

are

also

difficult

to

measurewithhightimeresolutionandsoaremosttemperatures.Thereasonfor

this is also that the established techniques for these measurements are

developed forsteadystatemeasurements.EGR isusuallymeasuredbyemission

instrumentswhichalreadyhavebeendiscussed.Airflow isoftenmeasuredabit


16/55


14

upstream from the engine, meaning that e.g. pressure build up after this point

mustbeaccountedforseparately.

Temperatures

are

often

measured

with

thermocouples

which

function

by

assumingthegastemperatureandtheresponsetimehere isdeterminedbythe

thermal inertia of the thermocouple. The transient responseofa thermocouple

can be improved by making it smaller but it cannot be made to small since it

wouldbecometoosensitivefortheroughenvironmentintheengine.

While the two first objectives are all about increasing the understanding and

quantifyingtransientphenomena,thethree lastobjectivesareaboutdeveloping

toolsfor

transient

control

optimization.

One

way

to

get

around

the

problem

with

highlytimeconsumingtesting inenginetestbeds istoodeveloppredictivetools

sothatallcontrolsettingscanbeoptimizedbyofflinesimulations.Oneversionof

this solution would be to use somewhat simplified models for shorter

computationaltimesandthendothefinetuninginatestbed.

Suchpredictivetoolscanbedividedintothreecategories:

Firstlyengineperformancemodels.Theseshouldbeabletopredictthebehavior

ofthecompletegasexchangesystemincludingturbochargers,EGRcircuitetcand

howitrespondstodifferentcontrolsettingsduringtransients.Themodelsinthis

categoryareoften1Dmodelsormeanvaluemodels.

The secondpart is prediction of the incylinderconditions. The simplest way to

modelsthisistoassumethattheincylindermixtureishomogenous.Thisstrongly

simplifiedapproachcannotpredictanydetailsaboutthecombustionbutcanbe

enough to use in the first category; engine performance models. The most

realisticway

to

model

the

in

cylinder

conditions

is

with

CFD

(Computational

Fluid

Dynamics), where many details in the combustion process can be resolved by

dividingthecylinderintoafinemeshofsubvolumes.ThedisadvantagewithCFD

isthat it isverytimeconsumingandthat itsometimescanbehardtoverifythe


17/55


15

detailsofthecombustionsincethesearedifficulttomeasure.Therealsoexista

number of approaches that lie somewhere between these two approaches

regardingthelevelofdetailsthatcanberesolved,e.g.2zonecombustionmodels

andmultizonecombustionmodels.

Thethirdandfinalpartistheemissionformationcalculations.Emissionformation

rate is calculated with chemical reaction schemes. These schemes occur in

literaturewithawiderangeofcomplexityandtheydescribehow fuelandair is

turned intowater,carbondioxideandbyproductsviaanumberof intermediate

reactions and species. The most detail schemes include thousands of reactions

butsometimesonlyusingafewofthemcanbeuseful.

Thedifferentmodelsandmodelcategorieswillbedescribedmorethoroughly in

thenextchapters.


18/55


16

3 Emissionformationindieselengines

Figure6Emissionformationinconventionaldieselcombustion,[8].

Figure6showsasocalledphiTmap.Inthismap,thetypicalregions, interms

of temperature and equivalence ratios (phi), for NO formation and soot (PM)

formation/oxidationare

shown.

The

figure

also

contains

amixing

line

which

represents the typical path of a fuel package. There are also lines that show

whereinthephiTmapdifferentsectionsoftheflametypicallyare.

A phiT map is commonly used to describe how different combustion concepts

affecttheemissionformation.Oneimportantandwidelyacceptedtheoryshown

insuchmap is thatsoot is formed in the rich regionsof the flameand that its

onlyformedwithinacertaintemperatureinterval.HCCIcombustionforexample

avoidssoot

formation

by

not

having

rich

regions

and

the

newer

concept

Low

Temperature Combustion (LTC) avoids soot formation by moving the mixing

linetothecoldsideofthesootformationpeninsula.


19/55


17

The figurealsoshowsthatNOx formationandsootoxidationoccur inthesame

region,i.eunderthesameconditions.Thisistheexplanationtowhymeasuresto

decreaseoneoftheemissionsoftenleadtoanincreaseintheother.Thisresults

inthewellknownsootNOxtradeoffcurveshowninfigure7.

Figure7ExampleofasootNOxtradeoffcurve.[9]

3.1

NOxformation

ThemostimportantNOxformationmechanismistheZeldovichmechanism,also

calledthethermalmechanismsincethisisresponsibleforthestrongtemperature

depedenceofNOxemissions.ThemechanismwasproposedbyZeldovichin1946

[9,10]:

NNONO ++2

(1)

ONOON ++2

(2)


20/55


18

Themechanismwaslaterextendedtoincludeathirdreaction:

HNOOHN ++

(3)

As longasthemaximumcombustiontemperature isfairlyhigh(above~2000K),

theZeldovichmechanismisdominatingtheNOproduction.Otherwise,e.g.when

high EGR rates are used or with HCCI combustion, other mechanisms can have

significantcontribution.BesidestheZeldovichmechanism,therearethreeother

suggestedmechanisms[9]:

The

promt

(Fenimore)

mechanism

in

which

NO

is

formed

from

an

intermediatespecieintherichzonesoftheflame

Thenitrousoxide(N2O)mechanism

NOformationfromfuelnitrogen

Anautomotivedieselenginecantypicallynotavoidhightemperatureregions in

its entire operational range (of loadand speedand in transients)and therefore

theZeldovichmechanismisoftenconsideredtobedominating.

3.2

Sootformationandoxidation

Sootformationandsootformationmodelingis,inseveralaspects,morecomplex

than NOx formation. One reason is that NOx is formed only via gasphase

reactions while soot formation also includes solid phase reactions. Another

reason is that the exact mechanisms are not known and no simple dominating

soot formation mechanism have been identified (comparable to the Zeldovich

mechanism

for

NOx

formation).

However,

several

reaction

schemes

and

mechanisms have been proposed and some of the most widely accepted are

brieflydescribedhere:


21/55


19

Theprecursors forsoot formationare formedvia thegasphasereactionduring

fuel pyrolysis. One important precursor is believed to be PAHs (PolyAromatic

Hydrocarbons) and the growth process of these precursors appears to involve

additionofC2,C3orothersmallunitsas,e.g.,C2H2,acetylene[11].

The reaction step when these gas phase species becomes particles is called

nucleation. Several models for this has been proposed and one of the most

popularmodels isthePAHmodel inwhichthePAHsstarttosticktoeachother

after they have reached a certain size. These PAH clusters then evolve to solid

particles[11].

Whenthese

particles

have

been

formed

they

increase

in

size

in

what

is

called

the

mass growth process. This process can be divided into three subprocesses;

coagulation, surface growth and dehydrogenation. Coagulation starts

immediately after the particle formation and can be described as a sticking

collision between particles which significantly increases particle mass and

decreases particle number without affecting the total particle mass [12]. The

surfacegrowthprocess iswhengasphasespeciessuchasacetyleneandPAHs is

adsorbed to the particles surface [11]. In the final stage of soot formation, the

particlesare

dehydrogenated

to

progressively

more

graphitic

carbon

material.

At

the same time, the number of active sites on the particle for surface growth is

reducedandthereforethedehydrogenation leadstothe formationofchain like

structures[11].

Simultaneouslyasthesootformationprocessesandinmanycasesuntilthegases

areemittedfromthecylinder,thesootoxidationprocessistakingplace.Themain

oxidating reactants are OH under fuel rich conditions and O2 under fuel lean

conditions[11].


22/55


20

3.3

Emissionreduction

The trade off relationship between NOx and soot (Figure 6) is fortunately only

fixedfor

aspecific

engine

and

ways

to

reduce

emissions

has

continuously

moved

this tradeoff curve towards origo to meet new legislations (Figure 3). New

combustionconceptsuchasHCCI,PCCIetc.canbreakthetradeoffbyalteringthe

mixing and ignition parts of the combustion. However, conventional diesel

combustion is still dominating and the greatest contribution to the emission

reductioncomes fromcomponentdevelopment.Oneclear trendover theyears

hasbeento increasethe injectionpressureandthereby improvethemixingand

reduce the rich regions that cause soot formation. Exhaust Gas Recirculation

(EGR)had itsbreakthrough forEuro4.TheuseofEGRdecreasesthemaximum

combustion temperature and thereby the NOx formation. Lower combustion

temperature means worse soot oxidation but this can be compensated for by

increasingtheboostpressureandtherebytheair/fuelratioorby increasingthe

injectionpressurefurther.

SincetheEuro4emissionlevel,manyenginemanufacturershavebeenforcedto

equiptheengineswithaftertreatmentsystemssincetheywerenolongerableto

sufficientlyreduce

the

engine

out

emissions.

The

most

common

aftertreatment

systemsareParticulateTrapsandSelectiveCatalyticReduction(SCR)whereNOx

isreducedwithareductante.g.ammonia.However,theengineoutemissionsstill

need tobereducedandcontrolled.AnSCRaftertreatmentsystemonly reduces

7095%sotheengineoutemissionscanstillnotbetoohigh.Anotherfactoristhe

costof the reductant thathas tobeadded.Forallaftertreatmentsystems,size

and service interval are also strong arguments to reduce the engine out

emissions.


23/55


21

4 Dieselcombustionmodeling

Figure7CFDModelingofthedieselcombustionprocess[12]

Since the combustion process in a diesel engine is very complex, the only

possibilitytomodel itrealistically isbyusingCFDcoupledtoachemicalreaction

model.Arealisticmodelhastobeabletodescribeincylinderflowinducedduring

the intakestrokewith properties such as swirl and tumble. It has to be able to

predict how the fuel spray breaks up, evaporates and mixes with the air in the

cylinder,howthefirstreactionsstartandthemixtureautoignitesandthereafter

the diffusion combustion. It also has to be able to predict the incylinder flow

after combustion and heat transfer to the cylinder walls and roof and to the

piston. Many of these processes are difficult to predict separately and are

extremelychallengingcombined.

In CFD combustion modelings, the combustion chamber is divided into sub

volumes.Thenumberofsubvolumesthatareusedrangesfromthousandsupto

threeordersofmagnitudemore.Thegreatestadvantageoverothermodelsisthe

highlevelofdetailthatcanberesolved.Ifamodelhasbeenthoroughlyvalidated,


24/55


22

itcan increase understanding of thecombustionprocessand the model can be

usedtopredicttheinfluenceofe.g.injectorpropertiesandpistonbowldesign.It

canalsoshowwherethegreatestcontributiontotheemissionoriginandhowitis

avoided.

The formation of soot and NO emission can be predicted via chemical kinetics

reactionschemesthatoftenareextensive.Theseschemesdescribeshowfueland

air are converted to carbon dioxide and water via thousands of reactions and

hundredsof intermediatespeciesandarealsousedtopredictautoignitionand

heat release rate. NOx formation can be included in the schemes while soot

formation can only be partly included since soot formation also consist of solid

matterreactionsandnotonlygaseous.

Someexamplesofsootmodelswherethesecomplexgasphasechemicalreaction

schemesareused togetherwithparticlegrowthmodels inCFDmodelsare [13

16].

The greatest disadvantage of using these complex models is the computational

time.Dependingonthenumberofsubvolumes inthemodel,thecomplexityof

themodel,

computer

capacity,

etc.,

computing

one

engine

cycle

require

from

a

couple of hours up to days. If one considers engine transients; one large load

transient in a heavy duty engine has a duration of over 50 cycles and to find

optimumcontrolsetting forthistransient,hundredsofcombinationsofsettings

would have to be simulated. In practice, this means that it isnt feasible to use

thesemodelsfortransientcontroloptimization,i.e.simplifiedmodelsareneeded

forthatapplication.


25/55


23

4.1

Simplifieddieselemissionmodeling

Todecreasethecomputationaltimeneededforemissionpredictions,themodels

canbe

simplified

in

three

aspects:

how

the

combustion

process

is

modeled,

the

chemicalreactionschemesandthemodelsthatdescribestheemissionformation

processes.

4.2

Simplifiedcombustionmodeling

Simplificationsofthecombustionprocesscanbefoundinmanydifferentdegrees

inliterature.Onecommonsimplificationthatsignificantlyreducesthecomplexity

ofthemodelistodisregardthespatialresolution, i.e.use0Dmodels.Themost

common0Dmodelsare:

Homogenousincylinderconditions[17]

Twozone combustion models [1820]: One zone with reactants and one

zonewithproducts.

Multizone combustion models [2124]: One zone with reactants and one

zonewithproductsforeachtimestepafterthestartofcombustion.

Thetwo

zone

approach

is

considered

to

be

afairly

accurate

way

to

describe

SI

combustion [10] since the fuel/air mixture is homogenous throughout the

cylinder and the flamefront propagates outwards from the sparkplug in the

centerofthecylinderwithproductsbehinditandreactantsinfrontofit.Thetwo

zone approach is however less straight forward when it comes to diesel

combustiondueto inhomogenousincylinderconditionsandamuchlessregular

combustion process. The common way to solve this is to assume that the

combustion takes place at stoichiometric conditions and that the combustion

productsarethengraduallymixedwiththeexcessairinthecylinder.Thus,theincylindermixturebecomeshomogenousbeforetheexhaustvalvesopen[18,20].

Themaindrawbackofusingthetwozonemodelfordieselcombustionmodeling

isthatthemaximumtemperatureisunderestimatedsincethetemperatureinthe


26/55


24

burnedzoneatevery instant isanaverageofallburningandpreviouslyburned

elements.Thisunderestimationhassignificant impactonthehighlytemperature

dependentNOxformationprocess.

To better imitate the real combustion process, multizone models have been

developed.Inthesemodels,thecombustionprocessisdividedintodiscretetime

steps and a new zone with combustion products is formed for each one. The

temperature and the mixing processes are still averaged but to a lower extent.

TheincylindertemperatureswiththedifferentmodelsareshowninFigure8.

Figure

8 Incylinder temperatures. Tmean is the homogenous temperature. Tu is the

temperatureintheunburnedzone,i.e.thezonecontainingthereactantsinatwozone

model.Tbistheburnedzone,i.e.thezonecontainingtheproductsinatwozonemodel.

Tzone is the temperature in the product zones in a multizone model. In this case four

zonesareusedbutnormallyonezoneperdegreecombustiondurationisused,i.e.the

numberofzonesattheendofthecombustionistypicallyaround50.


27/55


25

4.3

Simplifiedchemicalreactionschemes

As previously mentioned, the combustion chemistry in a diesel flame is very

complexwith

hundreds

of

species

and

thousands

of

intermediate

reactions.

On

the lean side of the flame, where the combustion is almost completed, the

chemistry is dominated by only a few species and is thereby much simpler to

model.Thetenspecieswithhighestconcentrationare;CO2,CO,H2O,H2,O2,N2,H,

O,NO,andOH [10,21].As the localchanges in temperatureandairsupplyare

slowon the leansideof the flameunlike in the richandstoichiometricsidesof

the flame, the concentration of these species can be assumed to equal the

equilibriumconcentrations.Thefractionsofthedifferentspeciesarefoundwith

carbon,hydrogen

,oxygen

and

nitrogen

balance

and

by

finding

the

equilibrium

balanceforthefollowingreactions:

(4)

(5)

(6)

HH 22

1 (7)

(8)

(9)

Thiscombustionchemistrycanbeusedtopredictprocessesontheleansideof

theflamesuchasNOxformationandsootsurfaceoxidation.

222 COHOHCO ++

2221 COOCO +

OO 22

1

HOOOH ++2

NONOON ++22


28/55


26

4.4

Simplifiedemissionmodeling

Simplifiedemissionmodelingcanbedividedintothreecategories:

Phenomenologicalmodels

Semiempiricalmodels

Statistical(cycleresolved)models

4.4.1

Phenomenologicalmodels

A

phenomenological

model

is

one

which

relates

empirical

observations

of

phenomenatoeachother,inawaywhichisconsistentwithfundamentaltheory,

butisnotdirectlyderivedfromtheory. Thiscategoryofmodelingisawaytodeal

with the complexity of soot formation where important and representative

species and mechanisms for the soot formation process are modeled. One

exampleofaphenomenologicalmodelwasproposedbyFuscoetalin1994[17].

Each part of the soot formation and oxidation process was represented by one

equationreactionasillustratedinFigure9.

Figure9IllustrationofFuscoetalsSootmodel[17].Schematicfigurefrom[25]


29/55


27

Thereactionscanbesummarizedinfourequations:

(4)

(5) (6) 5 (7)andare the concentrations of soot precursor species and growthspecie (acetylene).andare the stoichiometric coefficient and reaction ratecoefficients. and are soot volume fractions and particle numberconcentration. , , , and are the soot mass, molarmass,density,averagediameterandsurfaceareaconcentration(cm2/cm3).isAvogradosnumber. isthesootoxidationratecalculatedwiththeNagleandStricklandConstable(NSC)model:

1 (8)WhereRNSCistheoxidationrate[molCatoms/(cm2*s)],kAkBandkZ,aswellaskT

whichisusedinthenextequation,aretemperaturedependentratefunctions,

pO2isthepartialpressureofoxygen.ThemolarfractionofO2wasapproximated

bytheequilibriummolarfraction,whichisalreadycalculatedfortheNO

formationrate.xAisthefractionofsootwithamorereactivesurfacesite.

xAis

calculated

with

the

following

equation:

1

(9)


30/55


28

WhenFuscoetalpresentedthismodeltheyusedaverage incylinderconditions

[18],butithasalsobeenusedtogetherwithCFD(ComputationalFluidDynamics)

modeled incylinderconditions [three].Extendedversionsof themodelhasalso

beenused,e.g.[26,27]andtherearealsoalternativemodelsproposed,e.g.[28,

29].

Anotherexampleofphenomenologicalmodelsismodelswheredifferentreaction

ratesarecurvefittedtootherparameterssuchastemperature.Forexample,the

complexcalculationsforsootformationratethatareshowninFigure6onaphiT

maphavebeencurvefittedasafunctionofphi(fuel/airratio)andtemperatures

andusedinCFD[8].

A common method to predict NO formation is to assume that the thermal NO

formationisdominatingandthatthethermal(Zeldovich)mechanismcanbeused

topredictthetotalNOformation.FromtheextendedZeldovichmechanism,eq.1

3,therateofNOformationcanthenbewrittenas:

(10)Where [ ] denotes concentration of the different species and ki denotes the

forward and reversed rate coefficients. The rate constants are strongly

temperaturedependentandcanbefoundine.g.[10]

The concentration of the species included in the equation, and thereby the

formationrate,canbeobtainedifcomplexgasphasereactionschemesareused.

Examples

can

be

found

in

in

cylinder

CFD

combustion

and

emission

models,

e.g.

[13,14].Ifreactionschemesarenotused,theequilibriumassumption,[10],can

beusedsincemostNO isformedonthe leansideoftheflame.Thisapproach is

usedbye.g.[21]andleadstothefollowingsimplifiedexpression:


31/55


29

(11)

Where[]edenotesequilibriumconcentrationand (12)

(13)

(14)


32/55


30

4.4.2 Semi-empiricalmodels

Semiempirical models can be described as a further simplification of the

phenomenologicalmodels.

The

most

popular

example

is

the

two

step

Hiroyasu

modelwhichhaswaspresentedin1983[30]withonereactionforsootformation

andoneforoxidation:

.exp (15) .exp (16)Where

and are the mass of soot and fuel vapor.andare thepreexponentialconstantsforsootformationandsootoxidationrespectivelyandtheyareadjustedempirically. andaretheactivationenergiesandarealsoadjusted empirically. is the oxygen molar fraction.,andare incylinderpressureandtemperatureandthegasconstant.

Thismodelhasbeenwidelyused,thankstoitssimplicity.Adisadvantageisthatit

only is valid in a narrow range of operating conditions. The model can be used

withone

,two

,and

multizone

models

as

well

as

with

CFD

models

[8,

18,

24

and

28].


33/55


31

4.4.3

Statisticalmodels

Statisticalmodels([22and3133])areusuallycycleresolvedblackboxmodels

wherethe

emissions

are

predicted

by

curve

fitting

cycle

averaged

and/or

cycle

characteristicdatatoemissions.Thistypeofmodelingisthefastestanditdoesnt

require any deeper understanding of the underlying mechanisms. The

disadvantagesarethathighaccuracyusuallycantbeobtainedoverwideranges

ofoperatingconditionsandthattheyareunreliableforextrapolation.

Statistical modeling could be a useful tool for rough offline optimization of

transientcontrolstrategies thanksto thesimplicityof themodelsandtheshort

computational

time.

The

problem

is

that

more

or

less

all

measurements

are

somewhatdelayedandfilteredorsmoothened.Puredelaycanoftenbeidentified

and corrected for rather easy but the filtering caused by the instruments rise

timesishardertoidentifyandcorrectfor.Thismakesitmoredifficulttoidentify

whatparametersthatarethemostimportantandtheircauseeffectrelationship.

To illustrate this, an artificial signal and a filtered version of the same signal is

shownintheupperpartofFigure10.Thefloatingaveragefilterthatisusedhere

representsaninstrumentrisetimeof~0.7second.Areasonableexamplefor,e.g.,

thesoot

measurements.


34/55


32

Figure10Effectsofinstrumentrisetimeonstatisticalmodels.

The lower part of Figure 10 shows the correlation between the signal and its

filtered version and shows why it is difficult to create statistical models for

transientoperation,especiallyiftheinstrumentsrisetimeisunknown.Thecaseis

often that both cause and effect parameters have significant (and different)

filtering.

Itwouldbeanoptiontodevelopthemodelsduringstationaryoperationwhere

theinstrumentsrisetimesisntaproblem.However,theextremeconditionsthat

occurduringtransients(extremelylowair/fuelratio)aredifficulttoachievein

stationaryoperation.

-0.5 0 0.5 1 1.5 2 2.5 3 3.50

2

4

6

Time

signal

filtered signal

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

2

4

6

signal

filtered

signal


35/55


33

5 Advantagesanddisadvantagesofdifferentemissionmodel

categories-Motivationforchosenmodels

The typeof modelsused for emission predictions in thisproject was chosen by

evaluating the advantages anddisadvantages of the differentmodelcategories.

The CFD models with complex chemistry and particle growth models were

eliminatedatanearlystageduetocomputationalcost.Thecycleresolvedmodels

on the other hand have a very low computational cost but do not give any

physicalunderstandingoftheunderlyingprocessesandthereisalsothepractical

problem with establishing causeeffect relationships due to the rise times of

different

measurement

methods.

The

model

category

that

was

chosen

for

this

projectwasthereforethesimplifiedphysicalmodelcategory.

One specific work that influenced the model development in this project was

EgnellsNOxmodels[21].Inthismodelamultizoneapproachisusedtomodelthe

combustion process, the equilibrium assumption is used to predict the

composition of different species in the cylinder and the NO formation rate is

calculatedwiththeextendedZeldovichmechanism.

Fromthis

NOx

model,

anew

approach

for

soot

predictions

was

developed.

Soot

oxidationontheleansideoftheflameisbelievedtobedominatedbyO2surface

oxidation [11] and can thus be predicted by using the Nagle and Strickland

ConstablemodelintheframeworkoftheNOxmodel,i.e.amultizonecombustion

modelandequilibriumcombustionchemistry.

Sincethemultizonemodelonlydescribestheproductsofcombustion,i.e.starting

inthestoichiometricandleanregionsofthecombustion,thesootformationand

OHoxidation

processes

cannot

be

included

since

these

occurs

at

rich

conditions.

To get around that problem, a new approach was attempted based on two

observations; firstly in a phiT map, e.g. Figure 6, it can be seen that NOx

formation and soot oxidation occur under the same conditions, i.e. high


36/55


34

temperatureandwithavailableoxygen.Thesecondobservation istheNOxsoot

tradeoffcurve,whichclearlyshowsthatoperatingconditionsthatresult inhigh

NOxemissionsalsoresultinlowsootemissionsandviceversa. Theconclusionof

these observations is thatsince it ispossible to predict NOxemissionswith the

multizonemodelitshouldbepossibletopredictsootemissionsaswell.

Asecondaryproductofthatconclusionisthatthevariationsinthesootformation

processontherichsideoftheflamearelesssignificantfortheemissionsthanthe

variationsintheoxidationprocessontheleansideoftheflame.Thereby,agood

estimation of the emitted soot should be found if a typical soot formation

process for the specific engine can be found and if the oxidation process is

describedcorrectly.

In thisapproach it wasassumed that the typicalsoot formationprocess for the

specific engine could be represented as a typical fraction of fuel, with a typical

particlediameter,survivingthroughtheflameandtothestoichiometricandlean

sideswherethemultizonemodelandsootoxidationmodelarevalid.

The typical fraction and diameter for the soot model was found together with

typicalmixing

rate

and

typical

combustion

air/fuel

ratio

for

the

multizone

model

by running an error minimization script for the models over a wide range of

steadystateconditions.

One of the fundamental assumptions in the multizone approach is that the

combustiontakesplaceinadevelopedflameandasaconsequence,theaccuracy

ofthemodeldecreaseswhensignificantpartofthecombustionispremixed.

Moredetailsofthemodelsarefoundinthearticlesintheappendix.


37/55


35

6 Measuringandevaluatingtransientoperation

As described in the introduction section, measuring and evaluating transient

operationis

in

most

aspect

more

difficult

compared

to

stationary

operation

as

it

demands information with high time resolution from instruments and

measurement methods that were developed for stability, repeatability and

durability. The extra precautions and data processing that is necessary to

evaluatetransientoperationwillbedescribedinthissection.

6.1

Gasexchange

system

measurements

Theperformanceofthegasexchangesystem, includingthe intake andexhaust

systemsand theEGRcircuit, isa very important factor duringengine transients

and is therefore very important to be able to measure. The performance

parametersthataremostinterestingtoevaluateareintakeairflowandEGRrate.

Theyaretwoofthemostimportantparametersfortheincylinderconditionsand

arethereforeessentialasinputsforemissionpredictions.


38/55


36

6.1.1

TransientEGRmeasurements

Insteadystateconditions,EGRrate is typicallymeasuredbycomparing theCO2

fractionin

the

intake

with

the

exhaust

concentration

but

since

CO2

is

measured

with a conventional instrument, this cannot be used during the transients.

Instead, the EGR rate was estimated from air based volumetric efficiency (only

fresh air flow is included) via a steady state correlation. This correlation origin

fromawiderangeofoperatingconditionsandisshowninFigure11.

Figure11SteadystatecorrelationbetweenairbasedvolumetricefficiencyandEGRrate

forawiderangeofoperatingconditions

AscanbeseeninFigure11,themaximummeasuredvolumetricefficiencyisonly

0.85.Sincethevolumetricefficiency iscalculatedfromthesweptvolume,100%

efficiency will not be achieved as the intake valves are closed somewhat later

thanthe

bottom

dead

position

of

the

piston.

A

few

percent

is

also

lost

due

to

residualgasesinthecylinder,i.e.internalEGR.VariationsininternalEGRrateare

whatcausetheerrorsintheexternalEGRestimationsinthefigureandwhenthe

estimationisusedduringtransients.

y=114.01x+92.809R=0.9574

10

0

10

20

30

40

50

0.4 0.5 0.6 0.7 0.8 0.9

EGRrate[%]

Airbasedvolumetricefficiency

Measured Linear(Measured)


39/55


37

6.1.2

Transientairflowandair/fuelratiomeasurements

Sincetheairflowtotheengineismeasuredbeforeitscompressors,therewillbea

measurement

error

when

the

engine

is

operating

transient.

This

is

due

to

accumulation effects in the intake system during pressure build up. This can

correctedforwiththefollowingequation:

(17)Thiscorrectionshouldbedoneforallsubvolumes intheengines intakesystem.

P,V,RandTarepressure,volume,gasconstantandtemperature.

Toget thecorrectglobalair/fuel ratio in thecylinder, theexcessair in theEGR

alsohastobeaccountedfor.Thesetwocorrectionshavesignificantimpactonthe

estimatedair/fuel ratio.Air/fuel ratiowasalsomeasuredwithawidebandunit.

However,thisalsohastobecorrectedfortheairintheEGR.

Figure12Correctionoftheestimatedair/fuelratio

-10 0 10 20 30 40 50 600

0.5

1

1.5

2

2.5

Lambda

Cycle

-10 0 10 20 30 40 50 60

0

15

30

EGR[%]

Air & Fuel flow based

A & F flow + Accumulation corr.

A & F flow Accumulation + EGR corr.

Wideband unit

Wideband +EGR corr

lambda min. setpoint


40/55


38

6.2

Emissionmeasurements

6.2.1

NOxmeasurements

Instationaryoperationconditions,NOx ismeasuredwithaECOPhysicsCLD700

chemiluminescencedetector.This instrumentmeasuresthetotalofNOandNO2

measurements but since it has a rise time in the order of 20 seconds it is not

useful for transientevaluation.Forthatapplication,aCambustion fNOx400Fast

CLD chemiluminescence detector was used instead. This instrument has a rise

time

of

4

milliseconds

but

only

measures

NO.

That

is

however

an

acceptable

compromise since significant NO2/NOx ratios only occurs at high global air/fuel

ratiosandthustypicallynotundertransientconditions[10].

6.2.2

Sootmeasurements

ParticulateMatter (PM)andsoothasbeenmeasured inanumberofaspects in

thisproject;

PM

mass

concentration

[mg/m3],

opacity

(%),

Filter

Smoke

Number

(FSN),particlenumberconcentration[N/cm3]andsizedistribution.Thedifferent

measurement methods has been evaluated and compared [34] and it was

concluded that a good characterization of the PM emissions can be achieved if

some kind of mass related measurements are performed as well as some size

distribution related measurement. Mass related measurements are besides PM

mass concentration also opacity and FSN. The most repeatable measurements

have been found to be FSN. However, FSN has to be averaged for about 30

secondsand

are

thereby

not

useful

during

transients.

Therefore

opacity

was

used

for transient measurements instead. Opacity is a bit less stable since it can be

affectedbyliquidPM,humidityandlightabsorbingspeciesintheexhaust,but it

hasarisetimeoflessthanonesecond.


41/55


39

Size distribution measurements and particle number concentration

measurementscanprovideverydetailedinformationabouttheemittedPM.One

ofthereasonswhythesetypesof instrumenthavegainedpopularityrecently is

that researchers have found correlations between human health and particle

number concentrations [5]. It is believed that the reductions in PMmass

emissionshavebeenaccomplishedbydecreasing theaverageparticlemassand

notthenumberofparticles.Thenextemission legislation level,Euro6,willalso

includeparticlenumbermeasurements.

Figure13RisetimeevaluationforPMinstruments

The rise times for the different instruments are showed in Figure 13. This was

testedbydecreasingtheair/fuelratiosignificantlyfromonecycletotheother.In

this figure, normalized opacity and particle number measurements in three

different

size

ranges

are

shown.

The

particle

counters

CPC

3025

and

CPC

3010

measuredthetotalnumberofparticleslargerthan3and50nmrespectively.The

size distribution measured with the EEPS was integrated and shows the total

numberofparticleslargerthan6nminthefigure.Itcanbeseeninthefigurethat

-1 0 1 2 3 4 50

0.1

0.5

0.9

1.1

Time [s]

Normalizedsignal

CPC 3025

Opacimeter

EEPS total

CPC 3010


42/55


40

theconcentrationof largerparticlescorrelatesbetterwithopacity (and thereby

PMmass)thanwhensmallerparticlesalsoareincluded.

The

CPC

3025

was

chosen

to

measure

particle

number

concentrations

during

transientssince it is the fastest instrument. ItwasequippedwithaParticleSize

Selectorsothatitonlymeasuredparticleslargerthan20nm,i.e. tocomplywith

theproposedEuro6legislation.

Figure14CorrelationbetweenPMmeasurementmethods

Figure14showsopacityandparticlenumberconcentration,vs.FSN.Itshowsthat

OpacityandFSNcorrelatesverywellwhiletherearesomespreadwhenparticle

number concentration and FSN is compared, probably due to particle size

variations. This data comes from a wide range of steady state operating

conditionsandcurvefittedlineartrendlineswaslaterusedtocorrelatetransient

measurementstosteadystateandtoestimatesizechangesduringthetransients.

AnexampleisshowninFigure15.

0

5000

10000

15000

20000

25000

02468

10121416

1820

0 0.5 1 1.5 2 2.5

ParticleNumberConcentration

[N/cm3]

Opacity

FSN

Opacity CPC3025


43/55


41

Figure15Variationsinmassaveragedparticlediameterduringatransientcanbe

estimatedbycomparingopacityandparticlenumbermeasurements.

Figure 15 shows how measured opacity and particle number concentration

correlatesduringatransient.Fromthiscorrelation,variationsintheaveragedsize

ofthe

emitted

particles

can

be

estimated.

The

ratio

between

opacity

and

particle

numberconcentrationaboutthreetimeslargeraftercycle20thanbefore. Ifthe

mass per particle is increased three times, that corresponds to a diameter

increaseof~50%.

Thismethodforestimationofhowtheaverageparticlesizevariescanbeusedas

astepinthevalidationofsootpredictionmodels.

-10 0 10 20 30 40 50 600

20

40

60

Opacity[%]

Cycle

-10 0 10 20 30 40 50 60

2

3x 10

4

Particlenumberconcentration[N

/cm3]


44/55


42

6.3 Postprocessingemissionpredictions

Asdescribed,mostinstruments,andespeciallysootinstruments(figure13),have

significantrise

times.

This

rise

time

is

often

caused

by

internal

volumes

in

the

instruments and the result is an averaging/filtering effect. Some averaging also

occursbeforetheexhaustgasesreachestheinstruments,asgasesfromdifferent

cylindersandcyclesaremixedwitheachother.Sincethemeasurementsarenot

cycleresolved,somepostprocessingisnecessarytomakethepredictedemission

comparable.

Figure15Postprocessingofpredictedsootemissionstoimitateexhaustsystemand

instrumentaveragingandtodecreaseeffectsofnoisyinputdata.

Figure15showsthepredictedsootemissionswithandwithoutanaveragingfilter

togetherwiththemeasurements.Thestrengthofthenecessaryfilteringdepends

onthe instrumenttowhichthepredictionsarecompared. Noisy inputdatacan

alsoincreasetheneedforfiltering.

-10 0 10 20 30 40 50 600

0.05

0.1

0.15

0.2

0.25

0.3

0.35

Cycle

Soot[%o

ffuel]

Measured

Predicted

Predicted + filtered


45/55


43

7 Introductiontopapersandkeyresults

This section contains very brief summaries of the papers and some of the key

results

presented

in

them.

The

lessons

learned

from

each

paper

have

been

carriedontothenextoneandthelongtermsgoalhasbeentofulfilltheproject

objectives:

measureNOxandsootduringtransients

measuretransientflowandgastemperatures

simulatetransientturboperformance

simulatetransientEGRflow,temperatureandmixingratios

simulateNOxandparticleformation

7.1

Paper1,EvaluationofTechniquesforTransientPM-measurements

Asthe

title

reveals,

paper

1presents

an

evaluation

of

different

PM

instruments

regarding their suitability for transient measurements. The tested instruments

were a TEOM and an opacimeter for PMmass and two CPCs and an EEPS for

particlenumberconcentrationandsizedistribution.Thetestswereconductedon

asinglecylinderresearchengine.Themostimportantconclusionfromthispaper

was that the measurements with the highest time resolution were achieved by

one of the CPC and the opacimeter, with rise timesjust below 1 second (see

Figure 13). It was also found that the proportionality between PM mass and

particle

number

concentrations

vary,

mainly

due

to

variations

in

particle

size

distribution and that these variations can be decreased if the very smallest

particlesaredisregarded.


46/55


44

7.2

Paper2,PredictionsandMeasurementsofTransientNOEmissions

foraTwo-stageTurbochargedHDDieselEnginewithEGR

In

paper

2,

an

engine

load

transient

was

evaluated

with

different

control

strategies, or more specifically, with different EGR rates during the first part of

the load increase.Thetestedheavydutyenginewasequippedwithatwostage

turbochargingsystemandahighpressureEGRroute.Theaimwastopredictthe

enginebehaviorwiththedifferentstrategieswitha1Dmodelandtocouplethat

modeltoamultizoneNOpredictionmodel. Theobjectivesthatwerefocusedon

inthispaperweretransientNOxmeasurementsandsimulationoftransientturbo

performance,EGRflowandNOxformation.

ThemainconclusionwerethatthemultizoneNOxmodelpredictedtheemissions

wellandthatthismodeleasilycouldbecoupledtothe1Dmodelwithoutslowing

itdownsignificantly.However,duetothehighpressurelevelsinthisengine,itis

difficult predict the EGR flow in the high pressure routeand thishassignificant

influenceontheNOxpredictions.InthisexperimenttheEGRratewasmeasured

by measuring the NO in the intake and in the exhaust. Examples of results are

showninFigure16.

Figure16PredictedandmeasuredIMEPandEGRratewiththreedifferentEGRvalvesettingsduringtheinitialpartofatransient.


47/55


48/55


46

7.4

Paper4,FastPhysicalEmissionPredictionsforOff-lineCalibrationof

TransientControlStrategies

Inpaper

4,

the

models

developed

in

paper

3were

used

to

predict

the

emission

duringtransientoperation.Similarlytopaper2,a fullengine loadtransientwas

evaluatedwithdifferentcontrolstrategiesbutthistimewithmorestrategiesand

also includingsoot. Inthisexperiment,onlymeasurementswereusedas inputs.

Withoutmakinganychangestothemodelsfrompaper3,thepredictedemissions

agreedwellwithmeasurements,asshowninfigure18.Inmostcaseswherethere

are discrepancies in one of the predicted emissions, there is a corresponding

discrepancyintheotherandthisprobablymeansthattheseshortcomingsorigin

frominput

data

or

modeling

the

in

cylinder

conditions.

Figure18Predictedemissionsduringaloadtransientwiththreedifferentstrategies.Thestrategiesareidenticalincase1and2besidestheEGRvalvepositionduringthefirst10cycles.Case2and3areidenticalbesidestheVGTpositionduringthefirst20

cyclesofthetransient

-10 0 10 20 30 40 50 60

0

100

200

300

400

500

600

700

C cle

NO

[ppm

]

-10 0 10 20 30 40 50 60

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

C cle

Soot[%of

fuel]

Pred. 1

Meas. 1

Pred. 2

Meas. 2

Pred. 3

Meas. 3


49/55


47

8 Summary

The long term objective for this project is to develop models that allow

optimizationof

engine

control

strategies

for

transient

operation

via

simulations.

Themotivationforthis istheneedfor improvedunderstandingofthisoperating

conditionandalsoawishtominimizethetimeneededforcalibrationinanengine

testbed.

Thesteppingstonestowardsthatgoalareidentifyingmeasurementmethodsthat

canresolve thetransientbehaviorof theengine in termsofgasexchange flows

and temperatures and in terms of emissions. Also to develop models that can

predictthat

behavior.

Due

to

the

high

number

of

possible

control

settings,

an

additionalrequirementforthemodelsislowcomputationaltime.

At this point, halfway into the project, a range of measurement methods and

simulation tools have been tested. Methods for measuring emissions with

acceptabletimeresolutionhavebeenidentifiedaswellasacceptablemethodsto

estimatetheair/fuelratioandEGRrateduringtransients. However,thereisstill

room for improvements. No effort has yet been made to improve temperature

measurementsand

this

makes

it

impossible

to

resolve

the

dramatic

variations

in

intaketemperaturethatoccurastheEGRvalve isclosedandopenedduringthe

transient.

Regardingthemodelingobjectives,modelsforsootandNOxemissionshavebeen

validatedwithawiderangeofstationaryoperatingconditionsand for transient

conditions and the models appear to be applicable for their purpose. However,

simulatedinputshaveonlybeenusedinlesserextentwhena1Dmodelwasused

topredict

an

engines

transient

behavior

and

provide

inputs

to

amultizone

NOx

modelandmoreworkisneededbeforethelongtermsobjectivescanberegarded

asfulfilled.


50/55


48

9 Futurework

Examplesofareaswhereadditionalworkisneededand/orwouldbeinteresting:

Toprovethevalidityoftheemissionmodelingapproach,itisnecessary

toverifyitonotherengines

To test themodels in their intended application, i.e. to use them with

engine models, 1D and/or meanvalue models, to optimize transient

controlstrategies.

Since the models predict the incylinder emissions during the engine

cycle, it would be interesting to compare the models with incylinder

measurements

Oneofthefundamentalassumptionsinthesootmodelisthatatypical

soot formation ratecanbe found foraspecificengine.However, ifan

engine with variable injection pressure was used, the typical soot

formationratewouldmostlikelybedifferentatdifferentpressuresand

therebyalsointerestingtotrytomodel.

The largestsourcesoferror for thetransientemissionspredictionsare

believed to be measured inputs such as EGR rate, air/fuel ratio and

temperatures. It would be interesting to see how the emission

predictionsimprovedifthesewereimproved.

Investigatehowthemodelscanbeadjustedorextendedtoimprovethe

accuracyforpredictingemissionsinpremixedcombustion.

Try more time steps/number of zones in the multizone to see how it

affectsthetimerequirementforthemodelanditsaccuracy.


51/55


49

10Acknowledgements

ThisworkwascarriedoutwithinCICERO(CentreforInternalCombustionEngine

ResearchOpus),

acompetence

centre

at

KTH,

sponsored

by

the

Swedish

Energy

Agency,vehicle industryinSwedenandKTH.CICERO isgreatlyacknowledgedfor

thesupport.

I would also like to thank my supervisor HansErik ngstrm and all my friends

andcolleaguesatthedivisionofinternalcombustionengines.


52/55


50

11

References

1.

http://www.dieselnet.com/news/2008/02acea.php, 200901202. [peswiki.com/.../DieselStationaryOnecylinder.jpg], 200901203.

Dec J. E., A conceptual model of DI diesel combustion based on lasersheetimaging, SAE 970873. 19974. Courtesy of Mikael Lindstrm, KTH Royal Institute of Technology andSCANIA, 200901205.

Jungnelius S. and Svartengren M., Hlsoeffekter av trafikavgaser, 20006. Hagena J. R., Filipi Z. S. and Assanis D.N., Transient Diesel Emissions:Analysis of Engine Operation During a TipIn, SAE paper 20060111517.

http://www.hankstruckpictures.com/bullnose_rmtf2001.htm, 200901208.

Kaario O., Antila E. and Larmi M. Applying soot phiT maps for engineering

CFD applications in diesel engines, SAE 2005013856, 20059. Heywood J. B. Internal combustion engine fundamentals, McGrawHill,199810.

Rente T., Injection strategies for heavy duty DI diesel engines, doctoralthesis, Chalmers University, 200411.

Xi J. and Zhong BJ. Review: soot in diesel combustion systems. ChemicalEngineering Technology, 29 No. 6, 200612. https://reader009.{domain}/reader009/html5/0330/5abdcecf5fa4e/5ab,20090120


53/55


51

13.

Akihama K., Takatori Y. Inagaki K. Sasaki S. and Dean A. M. Mechanism

of the smokeless rich combustion by reducing temperature, SAE 2001010655, 200114. Kitamura T, Ito T, Senda J and Fujimoto H. Mechanism of smokelessdiesel combustion with oxygenated fuels based on the dependence of theequivalence ratio and temperature on soot particle formation.International Journal for Engine Research, Vol 3. No 4, 200215.

Saric F. A detailed kinetic model for calculation of the soot partile sizedistribution function. Doctoral thesis, Lund University, 200616. Netzell K. Development and application of detailed kinetic models forthe soot particle size distribution function, Doctoral thesis. LundUniversity, 2006

17.

Fusco A., KnoxKelecy A.L. and Foster D.E. Application of aphenomenological soot model to diesel engine combustion, Internationalsymposium Comodia 94, 199418.

C.D. Rakopoulos, D.C. Rakopoulos and D.C. Kyritsis, Development andvalidation of a comprehensive twozone model for combustion andemissions formation in a DI diesel engine, International journal of energy,2003:27, 200319. Ericson C., NOx modelling of a complete engine/SCR system, licentiatethesis, Lund University, 2007

20.

Lapuerta M, Hernandez J J and Armas O. Kinetic modelling of gaseousemissions in a diesel engine, SAE paper 2000012939, 200021. Egnell R. On zerodimensional modelling of combustion and NOxformation in diesel engines. Doctoral thesis, Lund University, 2001


54/55


55/55

Measuring and Predicting Transient Diesel Engine Emissions 200930.

Hiroyasu H., Kadota T. and Arai M., Development and Use of a Spray

Combustion Modeling to Predict Diesel Engine Efficiency and PollutantEmissions, Bulletin of the JSME, Vol. 26, 198331. Desantes J. M., Arregle J. and Molina S. Influence of the EGR rate, oxygenconcentration and equivalent fuel/air ratio on the combustion behaviourand pollutant emissions of a heavy duty diesel engine, SAE 2000011813,200032.

Boulouchos K. and Kirchen P. A phenomenological mean value sootmodel for transient engine operation, MTZ 070812008, Vol. 69, 200833. Wenzel S. P., Modellierung der Russ und NOxEmissionen desDieselmotors, Doctoral thesis, OttovonGuerickeUniversitt Magdeburg,2006

34.

Westlund A and ngstrm HE, Evaluation of techniques for transientPMmeasurements, SAE Paper 2008011680, 2008

Formacion de Las Emisiones en Diesel

Documents

Transcript of Formacion de Las Emisiones en Diesel