STAR Global Conference 2012
DARS Workshop
March 20, 2012
Amsterdam - NETHERLANDS
An Integrated Approach for Simulating Diesel
Engine Performance using Direct Injection
Stochastic Reactor Model (DI-SRM)
Michal Pasternak, Andrea Matrisciano
Division of Thermodynamics
Brandenburg University of Technology
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
2
Applications
Evaluation of engine performance parameters such as IMEP, fuel consumption, etc.
An interaction between engine components and influence on engine performance
Engine optimisation towards high performance and low emissions
Engine cylinder models for cycle simulations – challenges, demands
In-cylinder performance and emissions formation at different operating points
Low CPU demand, fast calibration and validation procedure
Inputs for engine aftertreatment processes: catalytic converter, DPF, EGR
Detailed information about exhaust emissions – chemical kinetics
HC, CO, NOx, soot
Introduction: 0D Modelling of IC Engines
Integrated simulation methods are needed to take concurrently
into consideration engine performance and kinetic of emission
formation, and to ensure CPU acceptable for cycle simulations
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
3
Specific Objectives & Presentation Outline
0D based simulation method for DI Diesel engine performance
parameters study
Concept of the integrated simulation method
DI-SRM for DI Diesel engines
Chemical sub-model
Optimizer
Exemplary application
1. Model development and calibration
In-cylinder parameters and exhaust emissions validation
2. Simulation of EGR effects on NOx and soot emissions
Outlook and summary
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
4
Integrated approach concept
Integrated Approach for 0D based Engine Simulations
Emissions prediction
Fuels for Diesel engines
Reaction mechanisms
Speed-up of models development
calibration, training and control
Optimization of engine output
parameters
Optimizer
Engine in-cylinder processes
with detailed chemistry
0D, PDF based approach with
low CPU (up to < 5sec per cycle)
DI-SRM DI-SRM 0D Engine
Direct Injection
Stochastic
Reactor
Model
Optimization Algorithms
Processes control and automation
Fuel Model
Complex Reaction
Mechanisms
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
5
Overall engine modelling concept with DI-SRM
Direct Injection Stochastic Reactor Model (DI-SRM)
Temperature and mass fractions – random variables
Chemical kinetics and in-cylinder turbulence modelling
Transport equation for Mass Density Function (F())
0D model of engine in-
cylinder processes
In-cylinder content as
ensemble of particles
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
6
Time evolution of the MDF defined by the transport equation
DI-SRM
Qi (II) chemical reaction, volume changes, heat transfer
fuel injection: particles creation, ensemble update
R.h.s (III) mixing in composition space – modelling
Mixing time model affecting turbulence intensity
Key model parameter during validation
Direct Injection Stochastic Reactor Model (DI-SRM)
1( ; ) ( ( ) ( ; )) ( ; )
i
j
i
i i j
I IIIII
JF t Q F t F t
t x
,
1
RN
ii i j j
j
MQ
1 ,
1 1
1 1( ) ( )
S RN Ngi
S i i j j w
i jv i v v
h AMRT dVQ h p T T
c M mc dt mc
,
finj
i i f i S
mQ Y Y i 1,...,N
m , ,
1
1 Sfinj
S 1 i f i f i
ip
mQ Y h h
c m
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
7
Turbulence treatment – mixing time modelling for engine conditions
Regimes with different mixing intensity
Turbulent time scale (t) must be
given at each time of the cycle (!)
Curl’s mixing model
Mixing in composition space
Mixing frequency
Variable mixing time model
Mixing time based on the mean
conditional scalar dissipation rate history
Direct Injection Stochastic Reactor Model (DI-SRM)
1t
C
1
tmixNp N
( ) ( ) ( ) ( )1( ) ( ) ( ( ) ( ))
2
n m n mt t t t t t
Ilu
1
stτ χ
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
B
z1
A0
soi’
eoi
’
’inj
z2
0 1 2( ) ( , , , , , , , )injt f A B soi eoi z z
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
8
Chemical Kinetics Sub-Model
Reaction mechanisms – exemplary validation
n-Heptane (A)
Basis model for Diesel engine
simulations – skeletal model (121species)
n-Heptane (B)
Small size model (28species)
for CPU oriented applications
Based o
n:
Tsuru
shim
a, T
., P
roc. C
om
b. In
st.
2009.
32:
p.
2835
-2841.
0.01
0.1
1
10
100
0.7 0.9 1.1 1.3 1.5
1000/T (K-1)
Ignitio
n D
ela
y (
ms)
Exp phi 0.5Exp phi 1.0Exp phi 2.0Sim phi 0.5Sim phi 1.0Sim phi 2.0
n-heptane
p=40bar
0.01
0.1
1
10
100
0.8 0.9 1 1.1 1.2 1.3 1.4 1.5
1000/T (K-1)
Ignitio
n d
ela
y (
ms)
Exp.
Calc.
Phi=1.0
p=40bar
Ze
uch,
T.,
More
ac,
G.,
Ahm
ed, S
.S.,
and M
auss,
F.,
Com
bust.
Fla
me 6
51
–674 (
2008).
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
9
Genetic Algorithms (GA) based modelling
Natural evolution based concept
“Stronger properties are inherited to the next
generation”
Remarks, features, issues
Solution accuracy depends on the evaluation
function
Accuracy of the solution increases with time
Main benefits
Parallelisation – easy for computations
Suitable for multi-objectives problems
Suitable for problems with many local
minimum and maximum, i.e. with “noisy character”
Optimizer
Kusia
k e
t al.(2
009)
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
10
Applications – Simulation process and
model performance
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
11
Overall simulation method
Model calibration based on one model parameter – mixing time
Applications: simulation of performance and exhaust emissions formation
Simulation Process
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
12
Self-calibration method for Diesel engines simulations
Fast and simple
procedure for model
validation
Model Calibration
Mixing time
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
13
DI Diesel engine
In-cylinder pressure
results
n-heptane (A)
Settings
100 cycles
200 particles
1CAD time step
Model Performance: In-Cylinder Parameters
Pro
ceedin
gs o
f IC
DE
RS
2011
Engine Type
Diesel Engine
Bore (mm) 81
Stroke (mm) 95.5
Compression ratio (–) 16.3
Fuel n-heptane
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
14
Rate of heat release and mean in-cylinder temperature
Model Performance: In-Cylinder Parameters
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
Pro
ceedin
gs o
f IC
DE
RS
2011
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
15
Engine-out emissions
Integrated results at EVO
CAD based history of emissions formation
High accuracy in predicting NOx and unburned HC
Model Performance: Emissions
Pro
ceedin
gs o
f IC
DE
RS
2011
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary
16
Applications – Simulating NOx and soot
formation
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
17
Diesel engine
In-cylinder pressure
results
n-Heptane
Mixing time
calibration
Baseline Model Validation
Refe
rence d
ata
: N
akov,
G.,
Mauss,
F.,
Wenzel, C
., K
rüger,
C. ”A
pplic
atio
n o
f a S
tatio
nary
Fla
mele
t
Lib
rary
Based C
FD
Soot
Model fo
r Low
-NO
x D
iesel C
om
bustio
n”,
TH
IES
EL 2
010.
Engine DI Diesel
Bore (mm) 83
Stroke (mm) 99
Load (bar) 14
Compression ratio (–) 16.2
Background
Modelling and Method
Model Performance
Exemplary Application
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
18
Model Application to Simulate EGR Effects
Refe
rence e
xperim
enta
l data
: N
akov,
G.
et
al.,T
HIE
SE
L 2
010.
Experiment: solid line
Simulation: dash-doted line
•Effects of EGR variation Constant SOI The same mixing time scales
In-cylinder pressure
•The same fuel mass and injection rate AFR variation The same load
Very similar/same
conditions for
air fuel mixing
Background
Modelling and Method
Model Performance
Exemplary Application
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
19
Mean in-cylinder temperature and soot formation – oxidation
Detailed kinetic soot model: Mauss (1997) & n-Heptane from Zeuch et al. (2008)
Model Application to Simulate EGR Effects
Refe
rence e
xperim
enta
l data
: N
akov,
G.
et
al.,T
HIE
SE
L 2
010.
The same air-fuel mixing conditions for each case
Soot formation and oxidation affected mainly by the local
air fuel ratio – chemical effects simulation 500
700
900
1100
1300
1500
1700
1900
-50 -30 -10 10 30 50 70
Crank angle (deg ATDC)
Mean
tem
pera
ture
(K
)
Sim, Egr18%
Sim, Egr23%
Sim, Egr27%
0.00
0.20
0.40
0.60
0.80
1.00
1.20
-10 10 30 50 70 90 110
Crank angle (deg ATDC)
So
ot
ma
ss
(-)
DI-SRM
Exp
0.00
0.20
0.40
0.60
0.80
1.00
1.20
-10 10 30 50 70 90 110
Crank angle (deg ATDC)
So
ot
ma
ss
(-)
DI-SRM
Exp
0.00
0.20
0.40
0.60
0.80
1.00
1.20
-10 10 30 50 70 90 110
Crank angle (deg ATDC)
So
ot
ma
ss
(-)
DI-SRM
Exp
500
700
900
1100
1300
1500
1700
1900
-50 -30 -10 10 30 50 70
Crank angle (deg ATDC)
Me
an
te
mp
era
ture
(K
)
Sim, Egr18%
Sim, Egr23%
Sim, Egr27%
EGR=18% EGR=23% EGR=27%
AFR=1.51 AFR=1.42 AFR=1.33
Background
Modelling and Method
Model Performance
Exemplary Application
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
20
Engine-out NOx and Soot
Model Application to Simulate EGR Effects
Refe
rence e
xperim
enta
l data
: N
akov,
G.
et
al.,T
HIE
SE
L 2
010.
0.00
0.20
0.40
0.60
0.80
1.00
1.20
EGR 18% EGR 23% EGR 27%
No
rmalized
co
ncen
trati
on
(-)
EXP
SIM
NOx
0.00
0.20
0.40
0.60
0.80
1.00
1.20
EGR 18% EGR 23% EGR 27%
No
rma
lize
d c
on
ce
ntr
ati
on
(-)
EXP
SIM
Soot
•Correct prediction of NOx concentration at EVO for different EGR rates
•Correct trend in predicting mass of soot as a function of EGR changes
•Influence of the baseline model validation accuracy
Background
Modelling and Method
Model Performance
Exemplary Application
Summary
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
21
Single run cost
Solution settings
Particles clustering
100 cycles, 200 particles
1.0 CAD global time step
Overall simulation procedure using the optimizer – validation process
Genetic algorithm, 10 individuals calculated at once
DI-SRM parallelized on 32 CPUs,~9 populations needed
10 individuals in each population
Possible benefits
Reaction mechanism
More complex reaction mechanisms, multi-component fuels
Optimization
Engine performance parameters, engine-fuel matching
Computations
Fuel 2: ~1.5 hour for
complete validation
(p, ROHR, emissions)
Background
Modelling and Method
Model Performance
Exemplary Application
Summary
Fuel Hardware CPU (s)
1) n-heptane – A (121 species)
33s/1cycle
2) n-heptane – B (28 species)
- Parallel on 32 CPUs
- 2.9 GHz
- Parallel on 32 CPUs 5s/1cycle
DARS Workshop 2012
Amsterdam, Netherlands
M.
Paste
rnak, F
. M
auss,
A. M
atr
iscia
no
, B
randenburg
Univ
ers
ity o
f T
echnolo
gy
22
High level of automation of engine simulation processes thanks to an integration
with general purpose engine optimizer
Speed up of the overall simulation process (validation, optimization, application)
Applicable for simulating engine – fuels interaction (optimization, fuels database)
Capability of the DI-SRM to accurately simulate Diesel engine in-cylinder
performance such as pressure, heat release or mean temperature
Correct treatment of the turbulent mixing
High accuracy in simulating NOx
Promising results of simulating soot formation and oxidation
Correct trend in predicting mass of soot at EVO
Predictive behaviour of the model in terms of simulating chemical effects
in soot formation/oxidation and NOx formation (sensitivity on EGR and AFR)
Summary
Background
Modelling and Method
Model Performance
Exemplary Applications
Summary