LUND UNIVERSITY

PO Box 117221 00 Lund+46 46-222 00 00

Combustion Chamber Wall Temperature Measurement and Modeling During TransientHCCI Operation

Wilhelmsson, Carl; Vressner, Andreas; Tunestål, Per; Johansson, Bengt; Särner, Gustaf;Aldén, Marcus

Published: 2005-01-01

Link to publication

Citation for published version (APA):Wilhelmsson, C., Vressner, A., Tunestål, P., Johansson, B., Särner, G., & Aldén, M. (2005). CombustionChamber Wall Temperature Measurement and Modeling During Transient HCCI Operation. (SAE TechnicalPaper Series).

General rightsCopyright and moral rights for the publications made accessible in the public portal are retained by the authorsand/or other copyright owners and it is a condition of accessing publications that users recognise and abide by thelegal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of privatestudy or research. • You may not further distribute the material or use it for any profit-making activity or commercial gain • You may freely distribute the URL identifying the publication in the public portal

2005-01-3731

Combustion Chamber Wall Temperature Measurement andModeling During Transient HCCI Operation

Carl Wilhelmsson, Andreas Vressner, Per Tunestal and Bengt JohanssonDivision of Combustion Engines, Lund Institute of Technology

Gustaf Sarner, Marcus AldenDivision of Combustion Physics, Lund Institute of Technology

Copyright c© 2005 Society of Automotive Engineers, Inc.

ABSTRACT

In this paper the combustion chamber wall temper-ature was measured by the use of thermographicphosphor. The temperature was monitored over alarge time window covering a load transient.

Wall temperature measurement provide helpfulinformation in all engines. This temperature is forexample needed when calculating heat losses to thewalls. Most important is however the effect of thewall temperature on combustion. The walls can notheat up instantaneously and the slowly increasingwall temperature following a load transient will affectthe combustion events sucseeding the transient. TheHCCI combustion process is, due to its dependenceon chemical kinetics more sensitive to wall tempera-ture than Otto or Diesel engines. In depth knowledgeabout transient wall temperature could increase theunderstanding of transient HCCI control.

A “black box” state space model was derived whichis useful when predicting transient wall temperature.To produce a model the engine is run with the loaddescribed by a Pseudo Random Binary Sequence(PRBS). Standard system identification methodologywas then applied to acquire a state space modelwhich calculate the combustion chamber wall tem-perature given IMEPn. Such a model is useful whencontrolling HCCI combustion and makes it possibleto compensate the impact of wall temperature delayfollowing a load transient.

INTRODUCTION

The HCCI engine is still less known than its relativesthe Otto and the Diesel engine. Never the lessit seems to be one of the most promising engineconcepts of the future, combining high efficiency withultra low emissions of the unwanted nitrogen oxides.The first report of the concept of HCCI combustionwas presented by Onishi et al. [1] and was initiallyprimarily an attempt to reduce emissions of unburnedhydrocarbons (HC) and improve part load efficiencyfor two stroke engines. The working principle of theHCCI engine is best understood as a hybrid betweenthe Otto and Diesel engines. The HCCI engine op-erates with a premixed charge as in the Otto engine,and it operates unthrottled with compression ignition,controlling the load via the global air/fuel equivalenceratio (lambda) like the diesel engine. This operationprinciple makes it possible to increase the efficiencycompared to the Otto engine due to the avoidanceof throttling losses. At the same time the high sootand nitrogen oxide emissions of the diesel engineare avoided. Soot emissions is avoided bechauseof the homoginety of the mixture and the avoidanceof localy rich combustion zones. Nitrogen oxideemissions on the other hand is avoided because ofthe decrease in peak in-cylinder temperature due tothe lean operation of the engine.

HCCI combustion can be achieved in numerousways, both in two and four stroke engines. The com-pression ratio can be increased, the inlet air can bepre heated, the HCCI combustion can be triggered bythe use of retained hot residual gas and the octanenumber of the fuel can be altered [2]-[8]. Since

1

the HCCI combustion process is very sensitive tonumerous parameters, feedback combustion controlis needed to operate an HCCI engine in parts of itsoperation range as shown by Olsson et al. [14]. Suchcombustion control can be performed in numerousways using different actuators [10]-[13].

Even though it has many good features, the HCCIengine also has some limitations. The operationalprinciple unfortunately suffers from very high com-bustion rates. This high combustion rate causesnoise as well as wear of engine hardware. Anotherproblem with the HCCI principle is low combustionefficiency at low load. This causes high emissionsof unburned hydrocarbons and carbon monoxide [9].The lower exhaust temperature will cause problemswhen trying to fit a catalytic converter to an HCCIengine.

One of the factors that affects HCCI combustionis combustion chamber wall temperature. The HCCIcombustion is as shown by Olsson et al. [14] verysensitive to the temperature history of the engine;this paper focuses on the behavior of the wall tem-perature when a sudden change of engine load (aload transient) occurs. Investigations of the walltemperature behaviour of an HCCI engine have pre-viously been performed by the use of thermocouplesby Chang et al. [15]. Chang however focuses on thewall temperature as it is changing within the enginecycle and the main purpose is the development ofa heat transfer model suitable for HCCI engines.The conducted work has focused on how the walltemperature changes from cycle to cycle when theHCCI engine is exposed to an increase in load.

The technique used to measure wall tempera-ture is called Laser Induced Phosphorescence (LIP)and has previously been used to measure tempera-tures in internal combustion engines [16]-[19]. TheLIP technique has the advantage that it measuresthe temperature almost intrinsically. It is also possibleto measure temperature remotely using the LIPtechnique. The down side with this technique is thatoptical access has to be provided to the measure-ment spot.

To improve transient HCCI control a model de-scribing wall temperature is very useful. The outputof the model can be used for feed-forward actionin the controller structure compensating for theeffects that wall temperature has on the combustion[24]. Such a model has the potential to significantlysimplify the combustion feedback controller neededto operate an HCCI engine. A model describing walltemperature was identified using standard systemidentification techniques [22]. The engine was run

Figure 1: The test engine used in the experiments.

Table 1: Engine geometric properties.Displaced Volume 1966 cm3

Bore 127.5 mmStroke 154 mmConnecting Rod Length 255 mmNumber of Valves 4Compression Ratio 16.0:1Exhaust valve open 34 ◦ BBDC @ 0.15 mmExhaust valve close 6 ◦ BTDC @ 0.15 mmExhaust valve lift 14.1 mmExhaust valve diameter 41 mmInlet valve open 2 ◦ BTDC @ 0.15 mmInlet valve close 29 ◦ ABDC @ 0.15 mmInlet valve lift 14.1 mmInlet valve diameter 45 mmFuel Supply PFI

on a load described by a Pseudo Random BinarySequence (PRBS) and wall temperature was mea-sured. From the measured wall temperature andIMEP a wall temperature model was identified usingthe subspace algorithm [23]. The model performancewas cross validated by the use of data not used inthe identification process.

EXPERIMENTAL APPARATUS

EXPERIMENTAL ENGINE The engine used in theconducted experiments was a 12l Scania heavy dutydiesel engine converted to single-cylinder HCCI op-eration. The engine is fitted with a piston glass anda piston extension to enable optical access from un-derneath the glass piston. With the help of a mir-ror it is possible to lead laser pulses through the pis-ton glass and the combustion chamber to the cylinderhead. Besides the piston glass no other glass wallsare present in the engine; all the other combustionchamber walls consist of metal.

2

CONTROL SOFTWARE The engine was operatedmanually without feedback combustion control. Injec-tion control software was however used to control theinjection timing and other important operational para-meters such as inlet temperature. The control soft-ware is basically a variant of the system presented byOlsson et al. [10], modified to fit the single-cylinderengine used. The feedback controllers featured inthe software are switched off and no feedback con-trol of the combustion is applied. Prior to these ex-periments, functionality was added to enable a PRBSdisturbance on the total fuel heat feed in each injec-tion. The PRBS sequence provides the possibility toperform system identification and obtain a dynamicstate space model of the wall temperature as a func-tion of the IMEPn time history.

TEMPERATURE MEASUREMENT

LIP, LASER INDUCED PHOSPHORESCENCELaser Induced Phosphorescence (LIP) is a techniquewhere phosphorescence from thermographic phos-phors is utilized for determination of temperature.Wall temperature measurement can be made fromroom temperature up to 2000 K, remotely, close tonon-intrusively and with high accuracy as shownby Allison et al. [20]. The lifetime, i.e. the time forthe phosphorescence intensity to decrease to 1/eof its original intensity, is temperature dependent forthermographic phosphors and can be calibrated toprovide absolute temperature readings.

In this work the thermographic phosphor La2O2S:Euwas used for its temperature sensitivity in the regionof interest, 150 - 300 C. The lifetime of its phos-phorescence was calibrated to temperature in acontrolled environment before the measurements.The phosphor was painted as a spot on the cylinderhead, the binder used was HPC. The spot waspositioned close to the hole originally used by thediesel fuel injector on the side facing the exhaustvalves. No cooling water or oil channels are presentin the direct vicinity of the phosphor spot.

The phosphor spot was excited by 355 nm laserUV-light produced by a pulsed Nd:YAG-laser workingat its third harmonic. The pulsed 10 Hz laser lightwas directed to the phosphor spot via the 45◦ mirrorbelow the piston window as shown in Figure 2. Theemission signal was collected by a quartz glass lenswith f = +100 mm and d = 50 mm and focused onthe Photo Multiplier Tube (PMT). An interference filtercentered at 538 nm was placed in front of the PMTfor spectral filtering to the spectral line of interest.The amplified signal from the PMT was sampled and

Figure 2: Laser based temperature measurementsetup.

saved by a LeCroy 3 GHz oscilloscope.

This temperature measurement methodology haspreviously been performed in a single cylinder HDDiesel engine running under partly premixed condi-tions by Husberg et al. [17]; they however measuredthe temperature on the piston top evaluating differentmethods to solve the optical access and comparingthis technique with the conventional thermocouples.

LIP MEASUREMENT ACCURACY The accuracyof the LIP temperature measurement technique isaccording to Childes et al. [21] estimated to 1%,which in this case means an accuracy in the rangeof 1.5-2 ◦C. The standard deviation of the measuredtemperature was by Sarner et al. [16] found to be0.73 ◦C at 57 CAD BTDC. Since the temperaturemeasurements are performed at 50 CAD BTDC thestandard deviation is probably slightly higher than0.73 ◦C, but it is still below 1 ◦C.

Note that the selected timing of the wall temper-ature measurement (50 CAD BTDC) is used since itis within a crank angle interval found by Sarner et al.[16] to show very low variance between the cycles,this to suppress the effect of the cyclic variability. Thecyclic variability has its lowest impact if the wall tem-perature measurements are performed somewherebetween 47 and 67 CAD BTDC. The objective inthis study is not to investigate the cyclic variability ofthe wall temperature and hence the timing with leastsensitivity to cyclic variability is selected.

TEST PROCEDURE

In all these tests the engine was run using pure iso-octane as fuel. When performing the tests the enginewas first started up and run in steady state for quite

3

some time. When the engine finally reached thermalequilibrium the transient tests were carried out. A typ-ical step was performed in the manner that the enginewas first run in steady state for a short while. Thelaser and the data collection systems were started, toget a well defined output intensity from the laser thetransient was not started until at least 20 s after thelaser was started. The injected total fuel heat wasthen drastically increased. Since the change in fuelheat can be considered to have changed betweencycles IMEPn can also be considered to have doneso. There will of course be a latency due to inlet portwetting. This latency is however evidently very smallconsidering the time it takes the wall temperature torespond to the load change. The engine was then rununtil the desired number of cycles had been captured.

The laser setup was synchronized with the en-gine data acquisition system and it should be pointedout that the measured temperatures were synchro-nized with the corresponding engine cycles. Sincethe engine is run at 1200rpm the engine performs 10cycles each second. This means that even thoughthe temperature often is presented time based it issynchronized with IMEPn which is presented cyclebased.

EXPERIMENTAL RESULTS

STEP DISTURBANCE The first reasonable investi-gation to carry out is an investigation of the responseof the wall temperature. The pure step disturbancegives an answer about the rise time of the wall tem-perature and it basically gives the answer to the re-sponse of the combustion chamber walls to that par-ticular transient. It is important to remember thatthe experiments are carried out on an optical enginerun at moderate loads and a fixed engine speed of1200rpm to match the 10Hz pulsation of the laser.The wall temperature can be expected to be higherin an all metal engine running at higher loads. It is dif-ficult however to make a general statement regardingthe difference in thermal inertia between an opticalengine and an all metal engine. The thermal inertiais likely to depend on the design of the combustionchamber cooling system and so on. It hence proba-bly differs substantially between different engines.

POSITIVE STEP DISTURBANCE Figure 3 showsthe IMEPn corresponding to a representative singlepositive step disturbance. Figure 4, Figure 5 andFigure 6 show the calculated combustion phasing ,calculated combustion duration and measured walltemperature. As seen in the figures the walls, asexpected respond very slowly to the sudden increase

0 500 1000 1500 20000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

Figure 3: A positive step disturbance of IMEPn.

in IMEPn. Note that IMEPn, combustion duration andcombustion phasing are filtered.

The rise time (10%-90%) of the temperature inFigure 6 is found to be about 110 seconds (1100cycles as the engine is run at 1200 rpm). The aver-age rise time considering the five different availablepositive step disturbances is 108 seconds. Thereexists a slight spread in the temperature which is aresult of cycle to cycle variations of the combustionprocess. When investigating combustion phasing andduration (Figure 4 and Figure 5) we realize that com-bustion duration decreases significantly after the loaddisturbance has occurred, the combustion also takesplace earlier in the cycle. Initially however the suddenincrease in the fuel amount feed to the engine retardsCA50 due to the fuel vaporization-related decreaseof charge temperature. CA50 then occurs earlierand earlier as the walls (and engine as such) heatup. The fact that CA50 initially occurs later affectscombustion duration which is increased, some cyclescompletely misfire because of the late combustionphasing. When CA50 advances beyond its previous“steady state” value of 6 CAD the occasional misfiringends. After the initial steep decrease of combustionduration it is easily noticed that combustion durationslowly decreases as the wall temperature increasesand as the combustion phasing advances everfurther. The misfiring due to late combustion phasingcan not be seen in Figure 3 because of the filteringapplied to the data presented in this figure.

NEGATIVE STEP DISTURBANCE A question thatarises when investigating the positive step distur-bance (Figure 3 and Figure 6) is if the fall timeof the wall temperature equals the rise time. Theinvestigation is performed in the same manner as the

4

0 500 1000 1500 20000

2

4

6

8

10

Cycle [−]

CA

50 [C

AD

]

Figure 4: CA50 (CAD ATDC) corresponding to Fig-ure 3.

0 500 1000 1500 20002

3

4

5

6

7

8

9

10

Cycle [−]

Dur

atio

n [C

AD

]

Figure 5: Combustion duration corresponding to Fig-ure 3.

0 50 100 150 200150

160

170

180

190

200

Time [s]

Wal

l tem

pera

ture

[°C

]

Figure 6: The temperature corresponding to the stepdisturbance shown in Figure 3.

investigation of the positive step response. Whenthis is done we realize that the fall time has the samemean average value as the rise time. The averagevalue of both the rise time and the fall time is 108seconds

Even though the wall temperature needs thesame time to stabilize both after the positive andthe negative step disturbances, the combustionphasing and duration behave totally different. Thewall temperature are about to decrease and werealize from Figure 8, Figure 9 and Figure 10 thatthe latency of the wall temperature affects the com-bustion less than it did in the case with the positivestep disturbance. This seems surprising but whenthe fuel heat decreases there is no extra cooling dueto fuel vaporization. Combustion phasing does notchange very dramatically, it initially changes rapidlyabout one CAD because of the sudden change oflambda. However as the wall temperature decreasesthe combustion occurs later and later in the cycle andCA50 (almost) stabilizes at around 4 CAD ATDC.

Onset of HCCI combustion is governed by chemicalkinetics which in turn depends strongly on tempera-ture and this explains why the combustion phasingchanges substantially as the wall temperature de-creases. The combustion duration is mostly limited bythe current lambda which is changed instantaneouslywhen the fuel heat is decreased. Since currentair/fuel ratio limits the combustion duration the walltemperature only slightly affects the combustionduration as the walls cool off and the combustionphasing occurs later. Hence there is only a barelyvisible increase of the combustion duration after theinitial instantaneous change.

5

0 500 1000 1500 20000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

Figure 7: A negative step disturbance of IMEPn.

0 500 1000 1500 2000−1

0

1

2

3

4

5

Cycle [−]

CA

50 [C

AD

]

Figure 8: CA50 (CAD ATDC) corresponding to Fig-ure 7.

0 500 1000 1500 20002

3

4

5

6

7

8

9

10

Cycle [−]

Dur

atio

n [C

AD

]

Figure 9: Combustion duration corresponding to Fig-ure 7.

0 50 100 150 200150

160

170

180

190

200

Time [s]

Wal

l tem

pera

ture

[°C

]

Figure 10: The temperature corresponding to the stepdisturbance shown in Figure 7.

0 500 1000 1500 20000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

#1#2#3#4#5

Figure 11: Repeated positive IMEPn step distur-bance.

REPEATABILITY To confirm that the acquired tem-perature is valid, it is essential to be able to repeatthe behaviour on several different occasions. To real-ize the repeated steps the actual combustion durationwas manually monitored and the measurement wasstarted when the combustion duration had approxi-mately the same value. The different transients werealigned to be able to put them on top of each otherin one single figure. Five different positive step dis-turbances are shown in Figure 11 and as seen thebehaviour on the different occasions is identical. Thisstatement is also valid for the repeated negative stepdisturbances shown in Figure 13. Clearly visible isalso the fact that the rise/fall time is equal, and around110 seconds (1100 cycles). Both combustion phasingand combustion duration are also repetitive in theircharacter.

6

0 50 100 150 200150

160

170

180

190

200

Time [s]

Wal

l tem

pera

ture

[°C

]

#1#2#3#4#5

Figure 12: Temperatures corresponding to Figure 11.

0 500 1000 1500 20000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

#1#2#3#4#5

Figure 13: Repeated negative IMEPn step distur-bance.

0 50 100 150 200150

160

170

180

190

200

Time [s]

Wal

l tem

pera

ture

[°C

]

#1#2#3#4#5

Figure 14: Temperatures corresponding to Figure 13.

COMBUSTION INSTABILITY In Figure 6 it is no-ticeable that the temperature never reaches a stablesteady state; we hence have a slightly unstablesystem. The lack of feedback combustion controlmakes it impossible to break the instability. Com-bustion instability is a known phenomenon in HCCIengines, discussed by Olsson et al. [14]. Olssonet al. find that HCCI combustion phasing duringsome circumstances drifts. This drift either causesever earlier combustion phasing, finally damagingthe engine, or ever later combustion phasing finallyresulting in misfire. The factor thought to set thisinstability off in this case is the glass piston windowwhich is not able to conduct heat as well as a normalsteel piston. An experimental engine with pistonglass is thus more prone to instability than a “normal”HCCI engine without optical access.

To explain and show the mentioned combustioninstability CA50 was investigated. Figure 4 showsCA50 as it changes after a positive step disturbanceand it is easily noticed that the combustion occursearlier and earlier even about 1700 cycles after thestep disturbance. Early combustion gives rise tohigher peak pressure heating the engine furtherand the next combustion hence occurs even earlier.Finally the high peak pressure and/or the pressureoscillations caused by very early combustion willdamage the engine.

MODELLING To obtain a model of the wall tem-perature the engine was run with the total fuel heatdictated by a Pseudo Random Binary Sequence(PRBS) [22]. Wall temperature (Figure 16) wasmeasured and IMEPn (Figure 15) was calculatedduring this PRBS experiment. Since IMEPn isroughly proportional to the total fuel heat it is possibleto identify a model that holds IMEPn as input and thewall temperature as output. The MATLAB R© function;“n4sid” was used to perform system identification onthe data collected in the experiment,“n4sid” makesuse of the subspace algorithm [23].

To reduce the impact of “noise” (cyclic variabil-ity), both IMEPn and measured wall temperaturewere filtered using a low pass Butterworth filter.To improve the performance of the identificationalgorithm, mean and trend of measured wall temper-ature and filtered IMEPn were removed before theidentification algorithm was applied. Mean and trendwere removed by subtracting a first order polynomialobtained from a least-squares fit to the correspondingdata. Mean and trend were then again added to theoutput of the model to obtain correct temperaturevalues for the plots.

7

The order of the model, minimum pulse durationand duration of the PRBS experiment were selectedaccording to ad hoc rules [22] based on the resultsof the step disturbance experiments. The minimumpulse duration was chosen as 150 engine cycles, theduration of the experiment was chosen to 4000 en-gine cycles (200 extra cycles are added to allow thelaser to heat up). The amplitude of the PRBS signalwas chosen as high as possible with the combustioninstability as a limiting factor. It was possible to usean amplitude of roughly 1.7 bar IMEPn. A third ordermodel was used.

The wall temperature model obtained from theidentification process is shown here below. HereTemp(t) is the current wall temperature, x(t) is theinternal state of the state space model, x(0) are theinitial state of x(t). The internal state of the modelhas no physical interpretation.

x(t+Ts) = A ∗ x(t) + B ∗ IMEPn(t)

Temp(t) = C ∗ x(t)

A =

⎛⎝

0.99582 −0.019355 −0.0042260.009134 0.80608 −0.45149

−0.0040296 0.16399 −0.40477

⎞⎠

B =

⎛⎝

0.0013590.0412020.097997

⎞⎠

C =(

203.2 −1.7954 −0.0061066)

x(0) =

⎛⎝

0.0074859−0.024105−0.028688

⎞⎠

In order to validate the model it was used to esti-mate wall temperature on data from a PRBS exper-iment. The experimental data for the validation werenot used in the system identification process, i.e. themodel was cross validated against a reference dataset. The result from one such cross validation can beseen in Figure 17 and Figure 18. The model is asrealized from Figure 18 able to very accurately pre-dict wall temperature. The identification and crossvalidation process were repeated using different setsof data and the model was always able to accuratelycapture the wall temperature behaviour.

DISCUSSION

The conducted experiments show that combustionchamber wall temperature has a strong impact on

0 1000 2000 3000 40000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

Figure 15: IMEPn PRBS experiment used in the sys-tem identification process.

0 1000 2000 3000 4000140

145

150

155

160

165

170

175

180

Cycle [−]

Wal

l tem

pera

ture

[°C

]

Figure 16: Measured wall temperature during thePRBS experiment used for system identification (cor-responding to Figure 15).

8

0 1000 2000 3000 40000

0.5

1

1.5

2

Cycle [−]

IME

Pn

[bar

]

Figure 17: IMEPn PRBS sequence used in the crossvalidation of the state space model.

0 1000 2000 3000 4000140

145

150

155

160

165

170

175

180

Cycle [−]

Wal

l tem

pera

ture

[°C

]

TempModel

Figure 18: Cross validation, modeled and actual tem-perature corresponding to the sequence in Figure 17.

HCCI combustion after a sudden increase of theload. The fact that the wall temperature does notchange instantaneously affects the combustionevents succeeding a load transient. When load ofan HCCI engine is increased, CA50 is expected tooccur earlier in the cycle and combustion duration isexpected to decrease. This also happens but neitherthe combustion duration nor CA50 changes immedi-ately after the load has changed. The time neededfor the combustion duration and CA50 to changeto the new “steady state” value corresponds verywell with the time it takes the combustion chamberwall to change its temperature. This implies that itis combustion chamber wall temperature that affectthe combustion in terms of phasing and duration.Wall temperature change during a load transientcounteracts the change of the combustion causedby a change in the load. The reason for this is thatthe onset of HCCI combustion is mainly governed bychemical kinetics. The chemical kinetic processes de-pend in turn on the combustion chamber temperature.

The rise and fall times of wall temperature areequal but the effect on combustion is different de-pending on if the load is increased or decreased.When load is increased wall temperature limits CA50and combustion duration. CA50 is limited by walltemperature after a negative step disturbance aswell. When load is decreased it is however the locallambda that limits combustion duration. In the nega-tive step disturbance case, wall temperature howeverstill affects combustion duration. After the initialchange, combustion duration is slowly increasing asCA50 occurs later in the cycle.

When controlling an HCCI engine, the effect ofwall temperature will cause problems during tran-sients. Control theory contains numerous methodsused to compensate a controller for a process non-linearity (such as the wall temperature dependence).The modeling experiments show a possible methodthat could be used to obtain a model of the wall tem-perature in an engine. In the case investigated themodel did accurately predict wall temperature. Sincewall temperature is repetitive in its nature the modelcould be used to compensate for wall temperature inthe combustion control. The model identified hencehas the potential of simplifying transient HCCI control.

If the HCCI engine were under combustion con-trol and if combustion phasing and duration wereaffected by this control, wall temperature would de-pend on the phasing and duration of the combustionas well. In this case when the wall temperaturedoes not only depend on the current IMEP, a morecomplicated Multi Input Multi Output (MIMO) modelwould be needed to capture the behaviour of wall

9

temperature. Suitable inputs to such a MIMO walltemperature model would be IMEP, CA50 and com-bustion duration. The design of a MIMO model wouldbe an interesting area of future work.

CONCLUSION

• When performing load transients in an HCCI en-gine the wall temperature will affect the combus-tion in the succeeding cycles. Combustion phas-ing and combustion duration depend heavily oncombustion chamber wall temperature. The walltemperature shows a very slow rise time, in theconducted experiments the mean average risetime was in the range of 110 seconds.

• The wall temperature rise time equals the falltime when the increase/decrease of load is per-formed between the same values of IMEP.

• Repeated load step disturbances gave rise to thesame rise/fall time of the combustion chamberwall temperature. This statement is valid if thecombustion phasing and duration has the sameinitial values before the steps.

• The temperature effect on the combustion dif-fers depending on if the disturbance is performedin positive or negative direction. When the dis-turbance is performed in the positive directionwall temperature has a strong and limiting effecton the combustion duration. When the distur-bance is performed in the negative direction walltemperature has a very limited influence on thecombustion duration. The gas properties have amuch stronger impact on combustion in the neg-ative case limiting the influence of wall tempera-ture.

• Combustion phasing depends heavily on walltemperature both when the step disturbance isperformed in the positive and in the negative di-rection.

• It is by the use of PRBS experiments possible toidentify a state space model describing the com-bustion chamber wall temperature as a functionof current IMEP. The model that was identified isable to accurately predict wall temperature.

REFERENCES

[1] S. Onishi, S.H. Jo, K. Shoda, P.D. Jo, S.Kato, “Active Thermo-Atmosphere Combustion(ATAC)-A New Combustion Process for InternalCombustion Engines”, SAE 790501

[2] P. Najt, D.E. Foster, “Compression-Ignited Ho-mogeneous Charge Combustion”, SAE 830264

[3] R.H. Thring, “Homogeneous-ChargeCompression-Ignition (HCCI) Engine”, SAE892068

[4] M. Stockinger, H. Schapertons, P. Kuhlmann,“Versuche an einem gemischansaugenden mitSelbszundung”, MTZ 53 (1992).

[5] M. Christensen, P. Einewall, B. Johansson,“Homogeneous Charge Compression Ignition(HCCI) Using Isooctane, Ethanol and NaturalGas A Comparison to Spark Ignition Operation”,SAE 972874

[6] M. Christensen, B. Johansson, P. Amneus, F.Mauss, “Supercharged Homogeneous ChargeCompression Ignition”, SAE 980787

[7] M. Christensen, B. Johansson, “Influence of Mix-ture Quality on Homogeneous Charge Compres-sion Ignition”, SAE 982454

[8] M. Christensen, B. Johansson, “HomogeneousCharge Compression Ignition with Water Injec-tion”, SAE 1999-01-0182

[9] G. Haraldsson, J. Hyvonen, P. Tunestal, B. Jo-hansson, “HCCI Combustion Phasing in a MultiCylinder Engine Using Variable CompressionRatio”, SAE 2002-01-2858.

[10] J-O. Olsson, P. Tunestal, “Closed-Loop Controlof an HCCI Engine”, SAE 2001-01-1031

[11] G. Haraldsson, J. Hyvonen, P. Tunestal, B.Johansson, “HCCI Combustion Phasing withClosed-Loop Combustion Control Using VariableCompression Ratio in a Multi Cylinder Engine”,SAE 2003-01-1830.

[12] J. Martinez-Frias, S. M. Aceves, D. Flowers, R.Smith, R. Dibble, “HCCI Engine Control by Ther-mal Management” SAE 2000-01-2869

[13] N. Milovanovic, R. Chen, J. Turner, “Influence ofthe Variable Valve Timing Strategy on the Con-trol of a Homogeneous Charge Compression(HCCI) Engine” SAE 2004-01-1899

[14] J-O. Olsson, P. Tunestal, B. Johansson, S. Five-land, R. Agama, M. Willi, D. Assanis, “Compres-sion Ration Influence on Maximum Load of aNatural Gas Fueled HCCI Engine” SAE 2002-01-0111

[15] J. Chang, O. Guralp, Z. Filipi, D. Assanis, T-W. Kuo, P. Najt, R. Rask, “New Heat TransferCorrelation for an HCCI Engine Derived fromMeasurements of Instaneous Surface Heat Flux”SAE 2004-01-2996

10

[16] G. Sarner, M. Richter, M. Alden, A. Vress-ner, A. Hultqvist, B. Johansson, “Cycle to CycleResolved Wall Temperature Measurements us-ing Laser-Induced Phosphorescence in an HCCIEngine”, Submitted to: SAE Powertrain & FluidSystems Conference & Exhibition 2005, MarriottRivercenter, San Antonio, Texas, USA

[17] T. Husberg, S. Gjirja, I. Denbratt, A. Omrane, M.Alden, J. Engstrom, “Piston Temperature Mea-surement by Use of Thermographic Phosphorsand Thermocouples in a Heavy-Duty Diesel En-gine Run Under Partly Premixed Condition”,SAE 2005-01-1646

[18] A. Omrane, G. Juhlin, M. Alden, G. Josefs-son, J. Engstrom, T. Benham, “Demonstration ofTwo-Dimensional Temperature Characterizationof Valves and Transparent Piston in a Gdi Opti-cal Engine”, SAE 2004-01-0609

[19] J. S Armfield, R. L Graves, D. L. Beshears, M.R. Cates, T. V. Smith, S. W. Allison, “PhosphorThermometry for Internal Combustion Engines”,SAE 971642

[20] S. W. Allison, G. T. Gilles, “Remote thermome-try with thermographic phosphors: Instrumenta-tion and applications”, Rev. Sci. Instrum., Vol 68,No.7, 2616, 1997

[21] Childes, Greenwood, Long, “Review of tem-perature measurement”, American Institute ofPhysics, vol 71, nr 8, 2000

[22] R. Johansson, “System Modeling and Identifica-tion”, Prentice Hall, Englewood Cliffs, NJ, 1993

[23] L. Ljung, “System Identification, Theory for theuser”, Prentice Hall, 1999

[24] K. J. Astrom, B. Wittenmark, “Computer-Controlled Systems”, Prentice Hall, 1997

CONTACT

Carl Wilhelmsson MSc E. E.(e-mail: carl.wilhelmsson@vok.lth.se)Gustaf Sarner MSc(e-mail: gustaf.sarner@forbrf.lth.se).

DEFINITIONS AND ABBREVIATIONS

CA50: Crank angle 50% burnedCAD: Crank Angle DegreeIMEP: Indicated Mean Effective PressureIMEPn: Net Indicated Mean Effective Pressureλ : Relative air/fuel ratioHCCI: Homogeneous Charge Compression IgnitionHR: Heat ReleasePFI: Port Fuel InjectionPRBS: Pseudo Random Binary SequenceTDC: Top Dead CenterBDC: Bottom Dead CenterBBDC: Before Bottom Dead CenterBTDC: Before Top Dead CenterATDC: After Top Dead CenterMIMO: Multiple Input Multiple Output

11