TRANSIENT COOLING SYSTEM SIMULATION BY MEANS OF AUTOMATIC ENGINE PARAMETER CALIBRATION Gerald Pichler, Simone Fratti, Gerald Steinwender, Günther Pessl BMW Dieselmotorenentwicklung, Steyr Kuli User Meeting 2013 Linz, Austria, 26 th -28 th June
TRANSIENT COOLING SYSTEM SIMULATION BY MEANS
OF AUTOMATIC ENGINE PARAMETER CALIBRATION
Gerald Pichler, Simone Fratti, Gerald Steinwender, Günther Pessl
BMW Dieselmotorenentwicklung, Steyr
Kuli User Meeting 2013
Linz, Austria, 26th-28th June
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 2
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 3
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr
Updated BMW internal testing guidline for the cooling system
Build up of a Kuli model for transient cooling cycle analysis for a 225kW BMW Diesel engine
Air side and coolant circuit modeling supported by 3D-CFD simulations
Validation of the baseline model with vehicle measurements (hill climb and high transient cycle)
Coupling of ModeFrontier® with Kuli with the aim to optimize the engine parameters
The model should be accurate enough for comparing and optimizing different cooling packages.
0,0
50,0
100,0
150,0
200,0
250,0
0 50 100 150 200 250 300
En
gin
e lo
ad
Vehicle speed
100kph+
Trailer
Kuli User Meeting , 26th-28th June 2013 Seite 4
INTRODUCTION AND MOTIVATION (1)
Hot Idle
Hill Solo
Sporty
US
Clubsport Racetrack
Hill 60kph
Solo Hill 35kph+
Trailer
210kph
Vmax
Hill Solo
Moderate
G. Pichler , BMW Steyr
The BMW six cylinder Diesel engine Vehicle: BMW 535d
The 4 mass engine model of Kuli
Kuli User Meeting , 26th-28th June 2013 Seite 5
INTRODUCTION AND MOTIVATION (2)
6-Cylinder 740d, 535d, X5, X6 xDrive40d
225 kW 600 Nm
Piezo 2000 bar Crankcase aluminum
Two main groups of engine parameters
6 Heat transfer coefficients 4 Heat capacities
In sum 10 parameters of Kuli’s 4 mass engine model have to be optimized
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 6
INTRODUCTION AND MOTIVATION (3)
Thermal engine model
Vehicle: Engine load point, Engine rpm
Component characteristics
Radiator, TOC, OC, CAC, EGR, ...
Track data
Altitude, ambient pressure,
ambient temperature,…
Underhood flow CFD simulation
Air mass flow rates, velocity distribution Heat balance measurements
Automatic transmission: Efficiency map,
TOC flow rates, heat capacities
Controllers
EGR controller
Fan
Thermostat…
Thermal engine capacities
Component weights, fluid volume
CFD coolant circuit
Coolant flow rates
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 7
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 8
HEAT BALANCE MEASUREMENTS (1) WORKFLOW
Steady state measurements (heat balance for oil and water
circuits)
Load step measurements
(test bench)
Set up of a Kuli® model
Thermal engine model
calibration (4 mass model)
Transient cooling
simulations
modeFRONTIER®
Optimization algorithms •Heat capacities
•HTC
•Heat resistances
Manual
calibration
Load steps for different engine speeds:
• 1750 rpm
• 2250 rpm
• 3000 rpm
G. Pichler , BMW Steyr
AirFuelradiation,convectionCACcoolantExhaustemErrorHQQQQHPdB
Kuli User Meeting , 26th-28th June 2013 Seite 9
HEAT BALANCE MEASUREMENTS (2) ENERGY BALANCING
Balancing error Engine power
Exhaust enthalpy
Heat loss to coolant
Heat rejection CAC
Heat loss to ambient Injected fuel power
Air intake enthalpy
emP
ExhaustH
coolantQ
CACQ
radiation,convectionQ
FuelQ
AirH
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 10
HEAT BALANCE MEASUREMENTS (3) ENERGY BALANCING
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Am
ou
nt
of
fue
l an
d a
ir in
take
en
erg
y [
%]
Engine speed [rpm]
BalancingError
Convection & radiation
Heat rejection CAC
Exhaust enthalpy
Engine power
Coolant heat
Full load engine operating point Partial load engine operating point
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
Am
ou
nt
of
fue
l an
d a
ir in
take
en
erg
y [
%]
Engine speed [rpm]
BalancingError
Convection & radiation
Heat rejection CAC
Exhaust enthalpy
Engine power
Coolant heat
Mean balancing error at full load: 3% Mean balancing error at partial load: 2%
Max. error from 850rpm to 1250 rpm up to 14%.
Coolant exit temp.=110°C Coolant exit temp.=110°C
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 11
HEAT BALANCE MEASUREMENTS (4) COMPARISION TESTBENCH VS. VEHICLE
165
103
164
102
0
20
40
60
80
100
120
140
160
180
4400 3750
Pe
[kW
]
U/min
Effective engine power
Testbench
Vehicle
0
10
20
30
40
50
60
4400 3750
Inje
cted
mas
s [m
g/st
roke
]
U/min
Injection quantityTestbench
Vehicle
high
11,458
10,217
11,972
9,929
0
2
4
6
8
10
12
14
4400 3750
Coo
lant
flow
[m
³/h]
U/min
Coolant flow through engine
Testbench
Vehicle
0
10
20
30
40
50
60
70
80
90
100
4400 3750Q
_W
Mot
or [k
W]
U/min
Total coolant heat flow
Testbench
Vehicle
0
5
10
15
20
25
30
35
4400 3750
Q_
WM
otor
[kW
]
U/min
Heat flow charge air
Testbench
Vehicle
15
16
15
15
15
15
16
16
16
16
16
17
4400 3750
dQ_
WM
otor
[kW
]
U/min
Deviation coolant heat flow
High deviation in coolant heat flow between
vehicle and testbench for both operating
points although injected fuel mass almost
same. To be checked:
• Air wind speed on test bench on crankcase
and on oil pan.
• Thermal conditions in the vehicle underhood
(convection and radiation). As a result a heat correction map had to be
implemented to the Kuli vehicle model.
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 12
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 13
KULI MODEL SET UP (1) COOLANT CIRCUIT
To cabin heater
From cabin heater
High temperature radiator
Low temperature radiator
Thermostat
Water pump
Turbo charger cooling jacket
SCR dosing module
Low temperature path
for transmission cooling
Engine oil cooler
High pressure EGR cooler
Transmission oil cooler
EGR cooler valve
G. Pichler , BMW Steyr
Test bench model was built up to simulate load steps.
Open circuit with boundary conditions from load step
measurements (flow rates, temperatures…).
Flow rates through EGR and oil cooler branch were not
measured on testbench. Flow split ratio was taken into
account based on 3D-CFD results.
Oil cooler entry on
coolant side is between cylinder 1-2 warm up
has to be considered
to calculate accurate
entry temperature
differences at the oil
cooler (heat transfer
between oil and water).
Oil circuit N57D30T0 Kuli User Meeting , 26th-28th June 2013 Seite 14
KULI MODEL SET UP (2) COOLANT CIRCUIT
Testbench model EGR cooler branch
Oil-water cooler branch
Oil plate HX
1 2
3
A characteristic map
depending on an average oil
temperature and engine
speed was added.
The map was taken from a
fully detailed oil circuit
model.
Oil flow rate map
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 15
KULI MODEL SET UP (3) COOLANT CIRCUIT
Vehicle model
1
2
3
1 High temperature radiator 9 HP-EGR cooler
2 Low temperature radiator 10 Engine model
3 Thermostat 11 Engine oil cooler
4 Expansion tank 12 Hydraulic oil cooler
5 Cabin heater 13 Charge air cooler
6 Thermal transmission model
7 Transmission oil cooler
8 Water pump
4
5
7 6
Charge air circuit
Hydraulic oil circuit
8
9 10
10
11
12
Engine oil circuit
Transmission oil circuit
13
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 16
KULI MODEL SET UP (4) AIR PATH MODEL
Air mass flow rates taken from 3D-CFD simulation
for stationary operating points (hill climb to Vmax)
Comparison of 3D-CFD air flow rates with flow rates
calculated out of vehicle measurements shows that
simulation is predicting ~10% less flow rate for hill climb
Accurate fan curve(s) w/o air flaps essential for Kuli
Extracting a flow distribution out of CFD for Kuli matrix is a
common workflow. For simulating accurate total air flow
rates, results from measurements were taken into account.
Condenser
Charge air cooler
Hydraulic Oil Cooler
Radiator
Fan System
Heater
Normal velocity profile of condenser and
CAC during 35 kph hill climb
Method for building up the fan
in 3D-CFD is still under evaluation fully rotating fan is today too time consuming in a
development process
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 17
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 18
ENGINE PARAMETER CALIBRATION (1) COUPLING KULI AND MODEFRONTIER
Parameter optimization Kuli engine parameters
DOE done with 1500 designs
used for a sensitivity study for
a load step on testbench for
3000 rpm.
ModeFRONTIER Workflow
Parameters to be optimized:
6 x Heat capacity
4 x Heat transfer coefficients
55
60
65
70
75
80
85
90
95
100
105
0,0 15,0 30,0 45,0 60,0 75,0 90,0 105,0 120,0 135,0 150,0 165,0 180,0 195,0 210,0 225,0
T [
°C]
Time [sec]
Coolant temperature
Results of the DOE
Load step at 3000 rpm
G. Pichler , BMW Steyr
-200
-150
-100
-50
0
50
100
150
200
Eff
ec
tiv
e S
ize
- E
rro
r o
il
-1
-0,8
-0,6
-0,4
-0,2
0
0,2
0,4
0,6
0,8
1
Eff
ec
tiv
e S
ize
- E
rro
r c
oo
lan
t
Kuli User Meeting , 26th-28th June 2013 Seite 19
ENGINE PARAMETER CALIBRATION (2) RESULTS OF SENSITIVITY ANALYSIS
Effective size on error coolant Effective size on error oil
• The height of the bars represents the magnitude of the effect of the variable on the objective. E.g. the HTC of the oil pan has a low
effect on the error on the coolant side.
• A red colored bar means that the variable has a direct effect on the objective, meaning that an increase of the value of the variable
leads to a higher deviation.
• Blue colored bars are an indication that the defined variable space is in an acceptable range.
• Aim of the evaluation was to define an accurate variable range to speed up the following optimization.
Low importance, low effect Low importance, low effect
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 20
ENGINE PARAMETER CALIBRATION (3) LOAD STEP CALIBRATION ON TESTBENCH
Based on the results from sensitivity study the parameter
optimization was only done for the engine oil side for a load step at
1750rpm. The picture on the left side below shows the results of the
coolant side for the simulated load step at 1750rpm.
Load step at 1750rpm
Oil temperature (Optimization target)
Coolant temperature
Optimization results of ModeFrontier
About 60 designs of ModeFrontier were
Enough to find the correct engine parameters.
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 21
ENGINE PARAMETER CALIBRATION (4) LOAD STEP CALIBRATION ON TESTBENCH
Load step at 1750rpm
Oil temperature (Optimization target)
Coolant
Load step at 2250rpm
Parameters from 1750rpm load step optimization
Parameters from 1750rpm load step optimization
Parameters from 1750rpm load step optimization
Parameters from 1750rpm load step optimization
Load step at 3000rpm
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 22
ENGINE PARAMETER CALIBRATION (5) 2 DIFFERENT VALIDATION CYCLES
Miramas BMW test track Mt. Ventoux hill climb solo (moderate)
High dynamic test track cycle done in Miramas (FR). With temperature gaps up to 20K the cycle is ideal for the validation of the thermal engine masses.
The second test was done on Mt.Ventoux without trailer. Max speed for this test is about 80kph (~50mph). Following boundary conditions from tests
were used for transient simulation:
• Engine speed and BMEP
• Vehicle speed and gear
• Ambient conditions
Veh
icle
spe
ed [
kph]
Gea
r
Veh
icle
spe
ed [
kph]
Gea
r Alt
itud
e [m
]
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 23
ENGINE PARAMETER CALIBRATION (6) VALIDATION ON MT. VENTOUX HILL CLIMB
Engine oil temperature Coolant temperature
Results of engine parameter validation is not satisfying. Specially the comparison of the engine oil temperature between test and
simulation during hill climb shows a big deviation. The temperature trend on the coolant side is OK, also the deviation between test and
simulation is in an acceptable range. Based on latest results the approach for parameter optimization was updated.
Parameter optimization for oil and cooant for hill
climb cycle
Validation on Miramas test track
cycle
Validation on test bench load steps
Updated approach
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 24
ENGINE PARAMETER CALIBRATION (7) VALIDATION ON MT. VENTOUX HILL CLIMB
Engine oil temperature Coolant temperature
Parameter optimization for oil and cooant for hill
climb cycle
Validation on Miramas test track
cycle
Validation on test bench load steps
Te
mp
era
ture
[°C
]
Time [sec]
Load step optimized
Te
mp
era
ture
[°C
]
Time [sec]
Cycle optimized
Te
mp
era
ture
[°C
]
Time [sec]
Cycle optimized
ModeFrontier optimization (Pareto frontier)
Err
or
co
ola
nt
[-]
Te
mp
era
ture
[°C
]
Time [sec]
Load step optimized
Error oil [-]
Final design Pareto frontier
The picture above shows the best results of the
optimization regarding the error on oil and
coolant side. Pictures on the left side are
showing an overall improvement on the
temperature trends of oil and coolant.
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 25
ENGINE PARAMETER CALIBRATION (8) VALIDATION ON MIRAMAS TEST TRACK CYCLE
Parameter optimization for oil and cooant for hill
climb cycle
Validation on Miramas test track
cycle
Validation on test bench load steps
Coolant temperature
Engine oil temperature
Some time sections of the cycle are showing
almost perfect correlation between simulation and
measurement specially on the coolant side.
Specially in propulsion engine operation points the
oil temperatures are lower than the simulation is
showing.
The wind conditions during test track
measurement have a high influence on the results
but were not documented directly in the test
report.
Round 1 Round 2
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 26
ENGINE PARAMETER CALIBRATION (9) VALIDATION ON TEST BENCH LOAD STEPS
Parameter optimization for oil and cooant for hill
climb cycle
Validation on Miramas test track
cycle
Validation on test bench load steps
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
110,0
115,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330
Tem
pe
ratu
re [°
C]
Time [s]
Simulation
Experimental
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
110,0
115,0
120,0
125,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330
Tem
pe
ratu
re [°
C]
Time[s]
Simulation
Experimental
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255
Tem
pe
ratu
re [°
C]
Time [s]
Simulation
Experimental
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
110,0
115,0
120,0
125,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255
Tem
pe
ratu
re [°
C]
Time[s]
Simulation
Experimental
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
110,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330 345 360
Tem
pe
ratu
re [°
C]
Time [s]
Simulation
Experimental
55,0
60,0
65,0
70,0
75,0
80,0
85,0
90,0
95,0
100,0
105,0
110,0
115,0
120,0
125,0
0 15 30 45 60 75 90 105 120 135 150 165 180 195 210 225 240 255 270 285 300 315 330 345 360
Tem
pe
ratu
re [°
C]
Time[s]
Simulation
Experimental
1750 rpm 2250 rpm 3000 rpm
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 27
OVERVIEW
Introduction 1
Heat balance measurements 2
Kuli model set up 3
Engine model parameter calibration 4
Conclusion 5
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013 Seite 28
CONCLUSION A transient engine model for cooling system analysis was built up in Kuli and a validation with
two different driving cycles and load step measurements on test bench were done.
Comparison of measured coolant heat balances in the vehicle and on test bench show
significant differences at almost the same engine operating points. Specially convective heat
transfer of crankcase, oil pan and cylinder head is difficult to determine exactly on the test
bench.
Heat transfer coefficients and the thermal inertia (direct/indirect) of the engine model were
optimized automatically by coupling Kuli with modeFRONTIER®
Generation of a vehicle model based on a test bench model to a was partially successful. The
oil temperatures during hill climb showed biggest deviation. Therefore a cycle optimization of
the engine parameters for the hill climb cycle was done showing accurate results on Miramas
test cycle finally.
Absolute media temperatures during driving cycles with deviations below 5K is difficult to
reach but possible if a database of previous cycle measurements is available.
In conclusion the created transient model is accurate enough for comparing and optimizing
different cooling packages, but the proposal of absolute media temperatures during driving
cycles with deviations below 5K is difficult to reach but possible if a database of previous
cycle measurements is available.
G. Pichler , BMW Steyr Kuli User Meeting , 26th-28th June 2013