Upload
haque
View
217
Download
2
Embed Size (px)
Citation preview
STUDY AND APPLICATION OF PREDICTIVE DI-PULSE
MODEL FOR DIESEL COMBUSTION IN GT-POWER
GT-SUITE Conference Italiana, Torino 15/6/2015
A. Gallone, S. Pizza, A. Capra, M. Rimondi - General Motors Powertrain Europe
D. Bellomare - Powertech Engineering
GM CONFIDENTIAL
AG
EN
DA
1. DI-PULSE MODEL OVERVIEW
2. MODEL CALIBRATION DATA
3. RESULTS
4. SUMMARY AND NEXT STEPS
Page 3
Page 7
Page 12
Page 15
2
3
1. DI-PULSE MODEL
What is it?DI-Pulse is a predictive combustion model designed to handle modern injection
strategies: the model predicts the combustion rate and associated physical quantities
for direct-injection diesel engines with single and multi-pulse injection events.
Why?To increase predictive capabilities of the GT-Power model, in order to be more independent from testing data.
DI-Pulse +other features
GT-Power
Testing
GT-Power
Testing
In-cylinder pressure data
are needed to calibrate
combustion in GT-Power
No emission prediction
Possibility to perform
virtual pre-calibration
NOx & Soot emission
prediction
4
DI-Pulse calibration
4 coefficients
100 in-cylinderpressure data from
testing
Combustion profileanalysis
100 Burn rates
Physical quantities andengine parameters of 100engine operating points
~ 20 in-cylinderpressure data from
testing+
~ 20 injection profiles
N injectionprofilesPhysical quantities, engine
parameters and emissionsof N engine operating points
NOx & Sootmodels
calibration
vs.Fast run time: only 5-10% longer
of non-predictive combustion
(for typical engine models)
+
1. BENEFITS OF DI-PULSE MODEL
5
DI-Pulse calibrationThe 4 calibration parameters are:
Entrainment Rate Multiplier
Ignition Delay Multiplier
Premixed Combustion Rate Multiplier
Diffusion Combustion Multiplier
A DoE run was carried out varying the 4 parameters.
DI-Pulse emissions modelsNOx model: based on the extended Zeldovich mechanism (the calibration of 2 parameters
is required by a DoE).
SOOT model: 3 different models (all require the calibration of 2 parameters):
Hiroyasu model
Modified Hiroyasu model
Nagle and Strickland-Constable model
1. CALIBRATION OF DI-PULSE MODEL
Different size/family: SDE, MDE and LDE;
Different turbocharger system: FGT and VGT;
Different EGR type (temperature and quantity): LP and HP;
Different swirl motion: inlet port geometry and swirl flap;
Different peak firing pressure (PFP) level.
6
ENGINE SIZE turbo EGR TYPE swirl flap PFP LEVEL
1 SMALL VGT LP + HP N MEDIUM
2 MEDIUM FGT HP N LOW
3 MEDIUM VGT HP Y HIGH
4 LARGE VGT HP Y MEDIUM-HIGH
1. ENGINES CONSIDERED IN THE STUDY
7
INJECTOR:
hole geometry: diameter (0.116-0.139 mm) and
number (7-8)
discharge coefficient (Cd=1 considered constant)
injection rate profiles
CYLINDER:
initial conditions to calculate the trapped mass that means air
(volumetric efficiency) + fuel + residual gases (residual%)
and boundary conditions for heat transfer (Twall)
measured pressure cycles, to compare with predicted ones
flow characteristics (it seems they not affect the results)
CRANKSHAFT: usual data to link piston position and
then combustion chamber volume to crank angle.
DI-Pulse calibration is carried out on a single cylinder model (same as a CPO model): since this
kind of model doesn’t have valves, initial in-cylinder conditions at IVC have to be imposed.
2. SINGLE CYLINDER TO CALIBRATE DI-PULSE
8
For each engine point the fuel injection rate profile is
required: in this work the measured profiles were chosen.
The raw data coming from EVI measurements are processed
in order to obtain ‘clean’ and more physical injection profiles.
Injection rate differences between Zeuch and EVI methods:
SOI -> EVI signal delayed by 50-70 s (0.7 CA at 2000 RPM);
SHAPE -> distorsion;
EOI -> EVI shows not plausible tails and echos.
2) Signal shifted (60 micros) and compensated (optional)1) Signal filtering: noises, tail and echo removed
2. INJECTION RATE PROFILES PROCESSING
signal currentEVIZeuch
EVIfiltered
EVIfiltered + shifted + compensated
9
2. CALIBRATION POINTS CHOICE
EGR ZONE
The trend in calibration point choice is to reduce the
total number of calibration points, preferring the EGR
area, that plays a key role in the current emissions
cycles.
Furthermore, from a sensitivity study,
the EGR points affect greatly the
calibration results, much more than
full load or other points.
10
DI-PULSE PARAMETER min max
ENTRAINMENT RATE MULTIPLIER 0.95 2.8
IGNITION DELAY MULTIPLIER 0.3 1.7
PREMIXED COMBUSTION RATE MULTIPLIER 0.05 2.5
DIFFUSION COMBUSTION RATE MULTIPLIER 0.4 1.4
SETTING: a “latin hypercube“ DoE run was
carried out varying the DI-Pulse parameters in
the advised ranges.
A total number of 2000 points is suggested, that
means a run time about 3 hours for 18 engine
working points (depending from HW and licences
capabilities).
POST PROCESSING (parameter to optimize):
BURN RATE RMS: an “improved“ BR RMS value is available in the standard RLT results;
IMEP error: not used for optimization, only for check.
_ =( )
2. DOE SETTING AND POST PROCESSING
11
NOx MODEL SETTINGCalibration carried out on EGR points only and based on two main parameters:
NOx multiplier: multiplier to the predicted net rate of NOx formation
N2 oxidation activation energy multiplier: multiplier to the activation energy of the
N2 oxidation rate
Calibration effectuated by a DoE run, in order to minimize the error between simulated and
measured NOx concentrations
SOOT MODEL SETTINGCalibration carried out on EGR Points only and based on two main parameters:
Soot Formation Multiplier
Soot Burnup Multiplier
Calibration effectuated by a DoE run, in order to minimize the error between simulated and
measured SOOT concentrations
2. CALIBRATION OF EMISSION MODELS
12
3. BURN RATES AND CYLINDER PRESSURES
injection rate
measured signals
calculated signals
Full load calibrationFull load prediction
High load prediction
Low load prediction(also calibration affected by same approximation)
13
1000
rpm
3000
rpm
1250
rpm
1500
rpm
1750
rpm
2000
rpm
2250
rpm
2500
rpm
2750
rpm
3. NOx EMISSIONS
LESS ACCURATE
IN GENERAL for the analyzed engines a
goodness prediction chart can be used to
highlight the different zones where the cylinder
pressure cycles are in agreement or not with the
measured ones.
The predictive capabilities for NOx emissions are
affected by the quality of the DI-Pulse predicted
cylinder pressure and temperature: in the points
where the pressure and temperature are well
predicted, also the NOx model seems to work
properly.
measured
calculated (calibration + prediction)
14
#4 - EGR 11.8%EGR SWEEP AT PART LOAD
3. ASSESSMENT OF PREDICTIVE CAPABILITY
Prail SWEEP AT PART LOAD
#1 - Prail 68 MPa
#5 - prail 38 MPa
calibrationcalibration
predictionprediction
NOx: physical trend(a bit sensitive to Prail)
SOOT: physical trend,values improvable
SOOT: physical trend(a bit sensitive to Prail)
NOx: physical trend andvalues as previuos page
15
Low computational and model calibration efforts are required;
The cylinder pressure prediction is satisfying mainly at medium-high engine loads;
The cylinder pressure prediction at low engine load points is less accurate, in particular in
the EGR zone with multiple pilot injections;
The DI-Pulse model is sensitive to different engine operating conditions;
Acceptable predictive capabilities for NOx and SOOT have been obtained, influenced by the
quality of the predicted cylinder pressure and temperature (mainly for NOx).
4. SUMMARY
16
IMPROVEMENTS IN COMBUSTION PREDICTION:
additional efforts are still needed to improve predictive capabilities for multiple pilot injections
and reach satisfactory sensitivity to the swirl.
DEEP ASSESSMENT FOR NOx AND SOOT MODELS:
additional attempt to have an acceptable prediction with engine virtual calibration intent: in
the GT-Power V7.5 version some additional options and improvements are available for NOx
emissions prediction.
4. PERSPECTIVES AND NEXT STEPS
17
INVESTIGATION ABOUT INJECTOR MODEL:
study in deep the effect of the calculated fuel
injection pressure: with a constant Cd approach,
the calculated value is representative of the
pressure within the sac and it is not constant,
whereas using a Cd variable with tip lift, the
calculated pressure could be more constant and
similar to the rail value (to check if there is an
effect on predicted combustion).
4. PERSPECTIVES AND NEXT STEPS
fully detailed injector model (GT)
simple geometric andfluid properties +
empirical behaviour
hemi-empirical/statistic injector model
injection pressure in the sac (calculated)
use of an injector model to
become independent from test
bench data, moving from
measured injection rate
profiles to simulated ones.
injection rate
profileselectrical signal
THANKS FOR YOUR ATTENTION
ANY QUESTIONS?
18