1

Model based development and calibrationInnovative ways to increase calibration quality within the limits of acceptable development effort!

2

MODEL BASED DEVELOPMENTChallanges

3

Next steps EC/ JRC:• Definition of PEMS boundaries (temperature, idling, altitude, inclination,…)• Beginning of 2013: Screening of small fleet• PEMS most likely to come, methodology to be defined until mid 2013• Definition how EC will survey ISC (in addition to OEM tests)

EUROPEAN EMISSION LEGISLATIONPASSENGER CARS

4

ENGINE SPEED / LOAD DISTRIBUTION EXAMPLES OF REAL WORLD DRIVING VS. TEST

Rela

tive L

oad

, %

Engine Speed, rpm500 1000 1500 2000 2500 3000 3500 4000

0

10

20

30

40

50

60

70

80

90

100

RDEPEMS

<100 km/hmoderate

Driver

Rela

tive L

oad

, %

Engine Speed, rpm500 1000 1500 2000 2500 3000 3500 4000

0

10

20

30

40

50

60

70

80

90

100

NEDC

Rela

tive L

oad

, %

Engine Speed, rpm500 1000 1500 2000 2500 3000 3500 4000

0

10

20

30

40

50

60

70

80

90

100

WLTP

<100 km/hdynamic

Driver

5

Load Collective NEDC vs. RDE

N - upm

Md -

Nm

Random Test

RandomTest

PEMS: all modes possible

Options for RDE:

• Random test cycle:Chassis DynoSimulation

• PEMS:Measurement in customer driving with PEMS

���� Decision openNEDC

6

MODEL BASED DEVELOPMENTIntention

7

Challenges in the Powertrain Developmentand AVLs Solutions

Model-Based Development

CO2 / Fuel Consumption

Real Driving Emissions

Broad Vehicle Portfolio

AVL Solutions

Reduction of

development costs€

Reduction of

development time

Keep quality standards1

Clustering of hardware

and engineering activities

Increased system

complexity (EAS, OBD, Hybridization)

Multi-variant simulation with the

same engine family (RDE, OBD, EAS)

Reduced test facilities/variant

through front-loading

Improved quality and robustness

through additional virtual validation1

Independence of environmental testing

from seasonal conditions and vehicles

availability

8

Model Based DevelopmentWhat is it?

• Model based development using a real time capable engine model

• Starting from concept phase until SOP calibration

• Engine model based on semi-physical modeling approach

���� empirical model components derived from AVL experience and test bed data

���� physical components increase the range of application due to better extrapolation

• Easy usability due to the use of suitable simulation environments

Increasing system

robustness within given

development duration

and budget by transferring

development from

real to virtual testing

9

Definitions - Model Accuracy Levels

Maturity Level Description Use Cases

Level 1 Only the main geometrical

data of the engine are used

as input for model set-up

• Concept study and decision

• ECU algorithm design

• Exhaust gas aftertreatment (EAS)

concept

Level 2 Measurement data is used

to make a refinement of the

model to increase accuracy.

• Pre-Calibration: the possible

calibration tasks depends on focus

of the model parameterization

• Used for specific calibration tasks

Level 3 Model is adapted to steady

state and transient data,

measured at AVL. Highest

accuracy which is needed

for model based calibration.

• Variant calibration support

• Ambient correction calibration

(altitude/hot/cold)

• EAS calibration strategy

• OBD calibration support

• Robustness investigations

• ECU algorithm verification

10

Virtual

Real

Te

stb

ed

re

su

lts

EAS System (DOC,

Semi-physical

Thermodynamic

NOx-Emission

EAS System (DOC, DPF, SCR, NLT)

Empirical

static global

HC, CO, Soot, SPL,…

DoE Measurements

number of engine

Combined-model

Increased number of engine specific outputs

HiL Setup

MiLSetup

Baseengine testbed development

Emission validation

Environmental validation

Field data

Modelrefinement

Pre

-ca

libra

tio

n

Do

E T

est

Re

su

lts

Fie

ld d

ata

Robustness analysis

Pre

-ca

libra

tio

n

First engine run

MoBEO basic model

setup

MoBEO refined model

setup

Model-based calibration of various variants

Variant specific hardware change(e.g. intake piping, …)

(No combustion HW change)Semi-physical

Basic Model without

measurement data

Model based DevelopmentModelling Process

11

Zyk

lus

Erg

eb

nis

NO

x [

mg

/km

]

400

420

440

460

480

500

520

540

560

580

600

Kra

fts

toff

verb

rau

ch

[L

/km

]

6.0

6.2

6.4

6.6

6.8

7.0

7.2

7.4

7.6

7.8

8.0

0 100 200 300 400 500 600 700 800 900 1000 1100

time [s]

To

urq

ue [

%]

0

20

40

60

80

100

En

gin

e s

peed

[%

]

0

20

40

60

80

100

Te

mp

. tu

rb.

inle

t [°

C]

100

200

300

400

500

600

Inta

ke a

ir m

ass f

low

[%

]

0

20

40

60

80

100

NO

x C

co

ncen

trati

on

[p

pm

]

0

100

200

300

400

500

Measurement

Simulation

Cycle

re

su

lts

NO

x [m

g/k

m]

Cycle

fu

el c

on

su

mp

tio

n [m

g/k

m]

Cycle results - NEDC

Model AccuracyHigh model accuracy as base for model based calibration

Typical deviations of the cycle emissions and fuel consumption as well as achievable temperature accuracy:

• Fuel Consumption < 3%

• NOx Emission < 5%

• Insoluble Particulate Emission < 10%

• Temperature Intake Side < 10°C

• Temperature Exhaust Side < 20°C

12

Model AccuracyHigh model accuracy as base for model based calibration

Typical deviations of the cycle emissions and fuel consumption as well as achievable temperature accuracy:

• Fuel Consumption < 3%

• NOx Emission < 5%

• Insoluble Particulate Emission < 10%

• Temperature Intake Side < 10°C

• Temperature Exhaust Side < 20°C400 450 500 550 600 650 700 750 800

Time [s]

T.

Tu

rbin

e-I

nle

t [°

C]

300

350

400

450500

550

NO

x C

on

ce

ntr

ati

on

[p

pm

]

0

200400

600

800

1000

Op

ac

ity

[%

]

010

20

3040

50

En

gin

e S

pe

ed

[%

]

050

100

To

rqu

e [

%]

0

20

4060

80

100

Inta

ke

Air

Ma

ss

flo

w [

%]

0

20

4060

80

100

Measurement

Simulation

13

MODEL BASED DEVELOPMENTApplication Environment

14

Model Based Development Virtual test beds – Comparison Simulation Environment

Model in the Loop (MiL)

Advantages

+ Simulation faster than real time (app. 5 times)

+ No hardware parts needed

+ Simulation on normal PC possible

Disadvantages

- Setup of software ECU time consuming

- typically not all ECU functionalities available

- Hard to achieve equal control behaviour as real engine

Hardware in the Loop (HiL)

Advantages

+ All ECU functions available

+ Pre-Calibration of all ECU functions possible

+ Possibility of ECU software and dataset validation

Disadvantages

- Only real time simulation possible

- Need of hardware in the loop test bed

- Need of hardware parts

���� Both environments can be used for pre-calibration of specific tasks

15

Standard Hardware-in-the-Loop test bed

� HiL operation system

� HiL automation system

� Calibration software (INCA,..)

Suitable Application Environment – As Prerequisite to Integrated a Model Based Calibration Methodology in an exciting Application Team

Advanced AVL Hardware-in-the-Loop test bed extended with

� Advanced semi-physical powertrain model

� PUMA & CAMEO

Same interface for the calibration engineer as real test bed

� PUMA Host Connection

Simulation results stored in same format as real test data

� Post-Processing

Same CONCERTO Layouts can be used for data analyzing

� Calibration Software (INCA,…)

16

Application – From Virtual Test Bed to SOP Virtual Test Beds as Extension of Real Test Facilities

Test Bed Virtual Test Bed Vehicle Validation

Calibration

Load Profiles

Environment

Production Tolerances Aging Effects

17

MODEL BASED DEVELOPMENTUse - Cases

18

Model Based Development Concept Investigations

Model based concept investigations

� Assessment of technology route

� Simulation of transient behaviour of engine in early concept phase on MiL

environment

� Definition of possible concepts considering the interaction between

� engine

� exhaust aftertreatment system

� software and calibration

� Sensors and actuators

� environmental conditions

Vehicle & drivetrain simulation

19

DPF Loading Limit

active

Regenera

tion

passiv

eR

egenera

tion

HDDE 2ltr/cyl. class

EGR + DPF

Boundary conditions:Tier4i engineNOX/soot ratio < 10 [g/g]

If NOX/soot ratio increasesDPF soot load in balance pointwill be lowered.

Low engine out total soot emissionsallows also for conditions were no CRT effect is observable a long operationwithout DPF regeneration.Soot Model accuracy can be reduced to a worst case scenario.

Model Based DevelopmentConcept Investigations – DPF Soot Loading

Simulation of specific duty cycles for different applications with respect to DPF soot loading behavior on HiL system � Calibration validation

20

700

800

900

1000

1100

1200

1300

1400

1500

1600

1700

1800

1900

2000

Engine Speed [1/min]

Tem

p. u

pstr

.Tu

rbin

e [

°C]

0

200

400

600

800

Pre

ssu

re u

pstr

. T

urb

ine [

kP

a]

0

125

250

375

500

LP

TC

Sp

eed

[rp

m]

0

50000

100000

HP

TC

Sp

eed

[rp

m]

50000

75000

100000

125000

150000

Tem

p. d

s. C

om

pre

sso

r [°

C]

0

100

200

BM

EP

[kP

a]

800

1600

2400

Model Based DevelopmentCalibration of Ambient Corrections

Simulation of full load altitude operation for validation of ambient correction and engine protection functions

970mbar = 350m (Graz)750mbar = 2500m660mbar = 3500m540mbar = 5000m

Limits for component protection

Limit temperature upstream turbine

Limit temperature downstream compressor

Limit pressure upstream turbine

Limit LP turbochargerspeed

Limit HP turbochargerspeed

No derating up to 2500 m

21

700

800

900

1000

1100

1200

1300

1400

1500

1600

1700

1800

1900

2000

Engine Speed [1/min]

Tem

p. u

ps

tr.T

urb

ine [

°C]

0

200

400

600

800

Pre

ss

ure

up

str

. T

urb

ine [

kP

a]

0

125

250

375

500

LP

TC

Sp

eed

[rp

m]

0

50000

100000

HP

TC

Sp

eed

[rp

m]

50000

75000

100000

125000

150000

Tem

p. d

s. C

om

pre

sso

r [°

C]

0

100

200

BM

EP

[kP

a]

800

1600

2400

Model Based Development Calibration of Component Protection Functions

Simulation of engine failure at full load for validation of engine protection functions

Limit temperature upstream turbine

Limit temperature downstream compressor

Limit pressure upstream turbine

Limit LP turbochargerspeed

Limit HP turbochargerspeed

Limits for component protection

22

Pre-calibration

� Pre-calibration of thresholds, enable/releaseconditions, debouncing

� Simulation of fault-parts (e.g. EGR-orifice, broken DPF, etc.)

� Multi-Variant calibration (e.g. adjustment fromlead-calibration to a follow-up variant)

� …

Ideas for Application of Model-Based-Development in OBD Calibration

Validation of calibration

� Functional check of calibration (Tested-Flag, P-Codes, etc.)

� Check of IUMPRs at different driving profiles (e.g. using same calibration for a commercial truck and a city bus)

� Robustness investigation and tolerances� …

Evaluation and R&D projects currently ongoing

23

Calibration on Hardware-in-the-Loop test beds Virtual Test Beds as Extension of Real Test Facilities

Calibration

Driving Cycle

Environment

Production Tolerances Aging Effects

Powertrain calibration tasks for HiL test bed

� Pre-calibration of different calibration work packages

� Calibration for non-standard ambient conditions

� Calibration of component protection

� Vehicle/Engine derivate calibration

� RDE – Real Driving Emission evaluation

� Real world fuel consumption optimization

� Sensitivity studies taking into account system interactions

� Software and dataset validation

24

Front-LoadingExample Facilities

HiL

Engine Testbed

Chassy Dyno

Road

Ideal Lead Variant Calibration Project (i.e. no

relevant H/W changes)

25

FacilitiesHiL

Engine Testbed

Chassy Dyno

Road

Ideal Lead Variant Calibration Project (i.e. no

relevant H/W changes)

Multi-variant projects can be addressed by: an extension of the test environment through HiL (MiL/SiL) Testing• Keep calibration quality through additional HiL testing, though high number of variants• Multi-variant simulation (calibration clustering, RDE, EAS, OBD)• Keep test facilities usage by a feasible level• Make environmental testing more flexible and efficient

Front-LoadingExample

26

AV

L D

iesel-C

alib

ration

Qualit

y G

ate

s

Time

Cal. QualityS

Ot

-12 -9-15-18 -6-21-24

Month before SOP

G1 - Runnable

G2 - Refined

G3 – Target Achievement

G5 – Validated

G4 – Robust

conventional Calibration process

-27

model based Calibration process

Calibration Process and Dataset Management

Validation by:

1. Dataset Management (Dataset Merging, Clustering, Tracing)

2. Automated HiL Dataset validation und consideration of

different production tolerances in advance to dataset freeze

3. A extensive model based fleet validation with active search for

critical events.

-30

27

Da

tas

et

-Q

ua

lity

model based

Concept Vehicle developmentTest bed development Fleet validationVehicle developmentTest bed development Fleet validation

� Shifting of development tasks in earlier phases and well proven concept decisions

� Reduction of project duration due to additional virtual test facilities

• Independent from environmental conditions

• Independent from vehicle availability

• Higher efficiency for real testing

�High dataset quality due to good

monitoring and solid validation

Concept Vehicle developmentTest bed development Fleet validation

Vehicle developmentTest bed development Fleet validation

Conclusion - Innovative ways to increase calibration quality within the limits of acceptable development effort!

28

AVL LIST GMBHYour Independent Engineering Partner

Thank you for your attention