Application of the New Kinetic
Knock Model to A Downsized TGDI
Engine
Christof Schernus, Carolina Nebbia, Francesco Di
Matteo, Matthias Thewes
prepared for:
GT CONFERENCE 2013
OCTOBER 21-22. 2013, Steigenberger Airport Hotel
Frankfurt, Germany
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
1
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
2
Introduction
Knock is as old as the spark ignition engine. Today, when optimizing new
engines’ efficiency, it needs to be addressed already in concept modeling.
Piston eroded by knock
Pressure profiles
Source: Abschluussbericht des Sonderforschungsbereichs 224 Motorische Verbrennung, RWTH Aachen, http://www.sfb224.rwth-aachen.de/Kapitel/kap3_4.htm
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
3
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
4
Auto-ignition behavior of fuels
Auto-Ignition behavior of fuels and fuel components is characterized by
auto-induction time measured in shock wave tubes and RCM.
Pure paraffins C7H16, C8H18
Auto-Induction time τI /ms
 Higher pressure reduces autoinduction time
 Low-temperature auto-induction
time of iso-Octane 5 to 10 times
longer than n-Heptane
 n-Heptane shows NTC behavior
between 800 and 1000 K
 iso-octane data shows changing
gradient, too.
 Similar auto-induction times for
both species at high temperature
Source: Abschluussbericht des Sonderforschungsbereichs 224 Motorische Verbrennung, RWTH Aachen, http://www.sfb224.rwth-aachen.de/Kapitel/kap3_4.htm
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
5
Auto-ignition behavior of fuels
Primary reference fuels for gasoline octane number rating are mixtures of
iso-Octane and n-Heptane
Obvious n-Heptane impact:
Auto-Ignition time τI /ms
 Adding n-Heptane to iso-Octane
reduces auto-induction time
 NTC characteristic emphasized
with increasing n-Heptane share
Many modeling approaches of
different complexity in literature
 Detailed chemistry
 Macro-kinetics (4- or 5-step)
 Auto-induction time models
based on Arrhenius equations
Source: Abschluussbericht des Sonderforschungsbereichs 224 Motorische Verbrennung, RWTH Aachen, http://www.sfb224.rwth-aachen.de/Kapitel/kap3_4.htm
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
6
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
7
Auto-induction times predicted by models
Two out of four models in GT-SUITE allow for surface hot spots and ON
Douaud & Eyzat (D&E)
I (α ) =
 Single Arrhenius equation
 Strictly monotonic
τ ign
 Influences:
α
1
1
∫ τ ign
ω IVC
dα ,
 ON 
= M 1* C1* 

100


– Pressure (substitute for concentration)
I (α ) = 1 ⇒
C2
 C4 
* p −C 3 * exp

M
2
*
T


p ↑⇒ τ ign ↓, T ↑⇒ τ ign ↓
– Temperature
– Octane number (RON+MON)/2
Kinetics Fit (KinFit)
 Triple Arrhenius
 NTC capable
 Influences as above plus
– AFR
– Dilution by inert gas
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
I (α ) =
1
α
1
∫ τ ign
ω IVC
1
dα ;
τ ign
=
1
1
+
τ1 + τ 2 τ 3
b
 fi 
 ON  i
c
d
e
τi
= M 1* ai * 

 * [ fuel ] i * [O2 ] i * [diluent ] i * exp
M
T
100
2
*




i =1,3
p ↑⇒ [ fuel ], [O2 ] ↑⇒ τ ign ↓, T ↑⇒ τ ign
M1, M2: user fitting parameters
© by FEV – all rights reserved. Confidential – no passing on to third parties
8
Auto-induction times predicted by models
Knock model predictions are compared to auto-induction times from
literature by modeling a Rapid Compression Machine
Characteristic measured pressure trace in a Rapid Compression Machine
Rapid Compression Machine
140
pressure / bar
Mixture : ethanol/O2/inert mixture
120 φ = 1, Ar/O2 = 3.77, Ar : N2 = 70 : 30
pini = 781mbar, Tini=293K
100 pC = 35bar, TC = 840K
p(t) (bar)
p´(t) (bar/ms)
end of
compression
80
60
 Piston rapidly moves into top
position and remains there
 Pressure and temperature
increase due to compression
heat
losses
 Precursor reactions take place
 Auto-ignition when critical radical
concentration achieved
ignition delay tign
 Time from reaching TDC until
ignition := auto-induction time
pistion
deceleration
40
20 initiation
0
0
5
10
15
20
25
30
35
40
time / ms
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
10
Auto-induction times predicted by models
According to the formula, Douaud & Eyzat model is sensitive to ON and
pressure, but does not react to charge dilution
D&E model fits to
 Low-temperature slope between PRF100
and PRF90
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
11
Auto-induction times predicted by models
The Kinetic Fit knock model matches in larger temperature range and
reacts to AFR and dilution.
KinFit model fits to
 Extended temperature range and curvature
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
12
Auto-induction times predicted by models
Maximum end gas temperatures of modern TGDI engines may reach into
the NTC or higher temperature range. This calls for KinFit knock model.
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
13
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
14
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
15
Average combustion model
A typical way of using a knock model is based on target knock index
using an average combustion profile.
Example: Knock index profile for GTDI engine
Knock index formula for KinFit:
Predicting spark timing requires
target Knock Index Profile.
Target Knock Index taken from
reference engine simulation
operating at knock limit
 Similar, ideally identical
combustion system
 Validated combustion model
 EngCylFlame and
EngCylTWallSoln recommended
 M1, M2 could be fitted to
“average” knock crank angle
 Resulting KI profile adopted for
“predictive” application, then.
Source for KI formula: GT-SUITE V7.3 Online Help ENGÍNE (GT-POWER) REFERENCE TEMPLATES – EngCylKnock
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
16
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
17
Alternating combustion rate model
max. Knock Amplitude /bar
Analysis of cyclic variation and knock reveals higher knock propensity
and knock amplitude for short burn delays like for earlier spark timing
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
αSpark = -6 °ATDC
αSpark = -4 °ATDC
αSpark = 0 °ATDC
Boundary Conditions
 RON 95
 Stoichiometric operation
 Intake pressure: 136 kPa
 Here: port fuel injection
Multiple knocking cycles
1 – 2 knocking cycles
No knocking cycles
12
16
20
24
28
32
Burn Delay 0-5% [°Crank Angle]
Measured data from DFG project
“Experimentelle Untersuchung und
Modellierung stochastischer
Klopf-ereignisse bei hochaufgeladenen
Benzin-Motoren“, RWTH Aachen,
2008-2011
Source: M. Budde: “Modellierung zyklischer Schwankungen der Entflammung an der Klopfgrenze von Ottomotoren“, Haus der Technik, March 15, 2013
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
19
Alternating combustion rate model
The simulation approach to capture knock in cyclic variation needs to
employ different combustion profiles.
Cylinder pressure affected by
burn rate variation.
120
Pressure / bar
100
Averaged cylinder pressure
Fastest
burning
cycle
Short
burn
delay, fast
combustion
Simulation with only fastest burn
rate leads to overall wrong
operating point.
80
Average burn profile represents
average operating point
60
 Boost
40
 Performance
 Temperature level
20
0
-40 -30 -20 -10
0
10
20
30
40
50
60
Fast combustion representative
for knocking cycles.
Crank angle / ° CA ATDC
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
20
Alternating combustion rate model
Since knock is correlated with the short combustions in cyclic variations,
the alternative is to evaluate the knock model in fast burn rates
1) Target Knock Index definition
100
Start
Cum. Heat Release /%
90
80
AVG combustion
average – σ
70
60
Reach steady state
average
50
Switch to fast
combustion
40
30
Definition of KItgt
20
10
End
0
-10
0
10
20
30
Crank Angle /°ATDCF
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
40
50
M1, M2 could be fitted knock crank
angle in fast burning cycles
© by FEV – all rights reserved. Confidential – no passing on to third parties
21
Alternating combustion rate model
Since knock is correlated with the short combustions in cyclic variations,
the alternative is to evaluate the knock model in fast burn rates
2) Using Target Knock Index for spark timing
100
Start
Cum. Heat Release /%
90
80
AVG combustion
average – σ
70
Reach steady state
60
average
50
Switch to fast
combustion
40
ST + ∆ϕ
yes
30
KI > KItgt?
no
20
10
AVG combustion
(until steady state is reached)
0
-10
0
10
20
30
40
50
Crank Angle /°ATDCF
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
End
22
Alternating combustion rate model
For the same combustion stability, the alternating method just takes more
time to converge for the same result.
Different KI targets for both
applications
Alternating combustion
 Needs more cycles to converge
 Delivers approx. same result
Conclusion:
 Use average combustion for
cases with similar cyclic variation
 Use alternating combustion if
changes in burn rate COV are
expected. Then include this
effect in combustion model using
the same KI as determined for
fast cycles on base engine.
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
23
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
24
Example results
KinFit model applied to SGT engine allows for good prediction of Octane
number and compression ratio effect on retarding combustion
ON=9590 has similar effect to ∆CR=0.8
 9.8A ~ 9.0B
 10.6A ~ 9.8B
Same effect found
on test engine
A: ON=95
B: ON=90
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
26
Agenda
 Introduction
 Auto-ignition behavior of fuels
 Auto-induction times predicted by models
 How to apply knock models
– Average combustion model
– Alternating combustion rate model
 Example results
 Conclusion
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
27
Conclusion
The new Kinetic Fit knock model improves the torque prediction
in particular at higher end gas temperatures
Kinetics fit has several advantages compared to D&E
 Given sensitivity to AFR excursions and to dilution e.g. by EGR (as far as inert)
 Different gradients over temperature and NTC behavior
Some of the standard prediction problems remain
 E.G. need for empirical data such as target knock profile
Knock models can be used with average burn rate or with cyclic variation
 As long as cyclic variation percentage does not change, both methods deliver similar
results, i.e. more or less same torque and spark timing
 Include known dependency of cyclic burn rate variations to improve torque prediction
KinFit model results close to literature ignition delays and engine test bed results
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
© by FEV – all rights reserved. Confidential – no passing on to third parties
28
Application of the New Kinetic
Knock Model to A Downsized TGDI
Engine
Christof Schernus, Carolina Nebbia, Francesco Di
Matteo, Matthias Thewes
prepared for:
GT CONFERENCE 2013
OCTOBER 21-22. 2013, Steigenberger Airport Hotel
Frankfurt, Germany
Christof Schernus, GTConf2013_FEV_Knock.pptx, 22. Oktober 2013
29