Deliverable Report INGAS

hammercoupleΜηχανική

22 Φεβ 2014 (πριν από 7 χρόνια και 8 μήνες)

664 εμφανίσεις




Deliverable Report

INGAS



Grant agreement N°

218447



Project acronym

INGAS

Project title

Integrated GAS powertrain

=
䱯w=敭楳獩潮猬=

2

optimised and efficient CNG engines for
passengers cars (PC) and light duty vehicles
(LDV)



Instrument

Inte
grated Project

Theme

SST

=
㈰〷2

=
oqa=N
=
“Sustainable Surface Transport”
=


Start date of project

01.10.2008

Duration

36 Months



IP Co
-
ordinator

Massimo Ferrera, CRF

IP Project manager

Stefania Zandiri, CRF




Subproject

SPB0

Sub
-
project Co
-
ordina
tor

Manfred Hoppe



Deliverable

Engine model ability to simulate
the test bench engine, with
comparison between experimental

data and simulations, and
definition of the range for valid
uses


D.B0.5



Due date of deliverable

31/10/2009

Actual submissi
on date

dd/mm/yyyy



Organisation name of lead contractor for
this deliverable

Guillaume Peureux



GDF SUEZ



Report status

Consortium confidential



Revision version

1.0

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

2

Revisions table


Version

Date

Reason

1.0

29.10.2009

First draft

1.1

03.11
.2009

Second draft (light modifications)












Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

3

Table of contents


Revisions table

................................
................................
................................
...........

2

Table of contents

................................
................................
................................
........

3

Executive summary

................................
................................
................................
....

4

1

Research activities

................................
................................
..............................

4

1.1

Introduction

................................
................................
................................
..

4

1.2

Tested engines and experimental facilities

................................
...................

5

1.3

CVUT JBRC model:

................................
................................
...................

11

1.3.1

GT
-
Pow
er Model of 102/110 Engine

................................
...................

12

1.3.2

GT
-
Power model of 102/120 Engine

................................
...................

14

1.3.3

Knock model

................................
................................
.......................

17

1.4

GDF SUEZ model: GNVSim calibration

................................
.....................

19

1.4.1

Description of the model

................................
................................
.....

19

1.4.2

Calibration tools

................................
................................
..................

20

1.4.3

Calibration parameters taken into account

................................
..........

22

1.4.4

Calibration methodology

................................
................................
.....

24

1.4.5

Gas composition

................................
................................
.................

25

1.4.6

Focus on the 102/110 engine

................................
..............................

26

1.4.7

Focus on the 102/120 engine

................................
..............................

31

1.4.8

Formation prediction of combustion products

................................
.....

33

1.4.9

Knock onset prediction

................................
................................
........

34

2

Conclusions

................................
................................
................................
.......

36

2.1

CVUT JBRC model

................................
................................
....................

36

2.2

GDF SUEZ model

................................
................................
......................

41

2.3

Overall conclusion

................................
................................
......................

42

Annex 1: GT
-
Power TPA model results (4


102/110)

................................
.............

44

Annex 2: GT
-
Power model results (4


102/120)

................................
.....................

51

Annex 3: GNVSim
-

operating points for the 102/110 engine

................................
...

53

Annex 4: GNVSim
-

comparison between experimental and s
imulated curves
(102/110 engine)

................................
................................
................................
......

55

Annex 5: GNVSim
-

calculated maps (102/110 engine)

................................
...........

58

Annex 6: GNVSim
-

comparison between e
xperimental and simulated curves
(102/120 engine)

................................
................................
................................
......

61

Annex 7: GNVSim
-

calculated maps (102/120 engine)

................................
...........

66


Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

4

Executive summary

The development of advanced gas v
ehicles has to be supported by a survey of the gas
composition (including trace components) in the relevant EU countries, which is carried out in
WP B0.1. The investigation of the reaction of advanced CNG engines to changing gas
compositions through bench
tests (WP B0.4) and simulation (WP B0.2) will create the basis
for reliable control strategies ensuring an optimal engine operation for the specified range of
gas compositions and avoiding power loss, engine damage, emission increase and
driveability loss
by gas quality fluctuation.

This deliverable gives a description of test facilities at the Czech Technical University in
Prague


Josef Bozek Research Centre and first engine tests operated with two engines
fuelled with transit natural gas available at the

bench (task B0.4.2). These tests were used by
both CVUT JBRC and GDF SUEZ to calibrate constants of their own computational
simulation tools (task B0.2.1) with the purpose to obtain a good agreement between
simulated returns and test bench results. It pre
pares the future use of these models in the
following tasks of the project. Results of these calibrations are given here in terms of
predictive abilities of
engine efficiency, combustion products formation and knock
occurrence
.
As a matter of fact, these w
ill then enable analysis and extrapolation of engine
test campaigns to come.

Calibration of both models gave quite satisfying results with regard to the experimental data
obtained on both engines tested in CVUT JBRC facilities. Next test bench campaigns wi
ll
open the issue on natural gas composition variations, which is the core of SPB0 activities.
Therefore, the following step of simulation activities, in parallel with these further
experimental campaigns, is on the way in good agreement with the formerly
scheduled tasks
of the Sub
-
Project B0.

The calibrated models will be powerful tools, together with the test campaigns to come, to
create the basis for reliable control strategies ensuring an optimal engine operation with
regard to natural gas quality vari
ations.


1

Research activities

1.1

Introduction

As natural gas is distributed to end
-
customers through an international interconnected
network, with various injection ports all along the European gas grid and different origins of
supply, its composition may vary

all along the grid and over the time. The need for a security
of supply requires various sources of natural gas which could lead to an increasing risk of
variation of its composition / quality at the delivery point.

Fluctuations in fuel composition may pa
rticularly affect the combustion quality at lean
operating limit conditions. Thus, the stability of fuel specifications is an important parameter
for engine manufacturers to achieve the best compromise between high level of power, low
consumption, low emis
sions and the knock prevention.

The development of advanced gas vehicles has to be supported by a survey of the gas
composition (including trace components) in the relevant EU countries, which is carried out in
WP B0.1. The investigation of the reaction of

advanced CNG engines to changing gas
compositions through bench tests (WP B0.4) and simulation (WP B0.2) will create the basis
for reliable control strategies ensuring an optimal engine operation for the specified range of
gas compositions and avoiding po
wer loss, engine damage, emission increase and
driveability loss by gas quality fluctuation.

The purpose of this document is to give a description of test facilities at the Czech Technical
University in Prague


Josef Bozek Research Centre and first engine

tests operated with two
engines fuelled with transit natural gas available at the bench (task B0.4.2). Those tests were
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

5

used by both CVUT JBRC and GDF SUEZ to calibrate constants of their own computational
simulation tools (task B0.2.1) so as to assess th
eir prediction ability and prepare the future
use of these models in the following tasks of the project. As a matter of fact, they will then
enable analyse and extrapolation of engine test campaigns to come.

It is significant to underscore that the estima
tion of prediction ability described further in this
document was not performed with the aim of comparing models accuracy so as to choose
one of these tools. Indeed, the use of both models will be complementary and, in any case,
the same work will not be d
one twice with different tools. CVUT JBRC model is best fitted to
explore and elaborate different kinds of control strategies with several natural gas
compositions fuelling the engine, whereas GDF SUEZ model will be used to extrapolate the
results obtained

with the CVUT JBRC model to a much wider matrix of natural gas
compositions, and confirm these strategies are still relevant on such matrix.

Thus, content of DB0.5 is the following:



Presentation of first set of engine tests for preparation gas quality var
iation test
campaigns on these engines,



Model constants calibration to simulate test bench engines behaviours and their
predictive abilities as a result,



Prospects for next tasks: experimental and simulation activities.

The two 1D models that have been tak
en into account are:



GNVSim, previously developed by GDF SUEZ on a AMESim


platform with use of
IFP
-
Engine® library,



A model developed by CVUT JBRC on the GT
-
Power platform.


1.2

Tested engines and experimental facilities

Two engines have been installed on the

test beds in engine laboratory and prepared for use
as a source of experimental data. Both of them are of the size of the engines typically
dedicated as a prime mover of LD trucks. The engine accessory layout and their initial
adjustment are arranged aimi
ng to obtain the range and the assortment of experimental data
enabling to describe impact of
fuel
composition in sufficiently comprehensive way. The
parameters of the engines are briefly summarized in the
Table
1
.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

6

Eng
ine Notation (InGAS)

4


102/110

4


102/120

Basic Engine Geometry

Cylinder #

4

Bore

102 mm

Stroke

110 mm

120 mm

Displacement

3.6 dm
3

3.92 dm
3

Compression Ratio

10

12

Valve # / Cylinder

2

4

Engine Performances

Maximum Speed

2650 min
-
1

2800 min
-
1

Maximum Torque

approx. 300 Nm
*

@ 2100 min
-
1

610 Nm

@ 1500

=
ㄶ〰=min
-
1

Maximum Power

approx. 70 kW
*

@ 2100 min
-
1

120 kW

@ 2100

=
㈸〰=min
-
1

Turbocharger

Make

KKK K16

CZ C12

Control

Uncontrolled

Variable Turbine Geometry

Maximum Boost Pressure

appr
ox. 1.6 bara

2.4 bara

Intercooler

None

Air
-
to
-
Water

Mixture Formation

Arrangement

Common (central) mixer

Excess
-
Air Ratio





1

E捥獳
-
AirR慴楯C潮r潬

M慮畡l(慮yv慬略)潲oCl潳o搠䱯潰(


=ㄩ

E桡畳琠G慳R散er捵c慴a潮

N潮e

C潯l敤;䱯wPr敳獵re

Co
ntrol units

Electronic Control Unit

Distributed:


-
捯cr潬;Ig湩i潮C潮rol;

Mi畲攠桲hl攠䍯湴r潬

i獴ri扵敤:


-
捯cr潬;Ig湩i潮C潮rol;

Mi畲攠桲hl攠䍯湴r潬;

V呇R慣欠C潮r潬;

EGR桲hleC潮r潬,

K湯捫c整散i潮

*

With the uncontrolled tu
rbocharger both the boost pressure and the engine torque
ascends with increased speed continuously. Simultaneously the exhaust gas temperature
and knock attitude increases. The maximum torque and the maximum power as well are
estimated as the maximum value
s at which it is still possible to keep engine running
without excessive overloading.

Table
1
: Description of testing engines

Both engines are 4
-
cylinder ones, turbocharged and they are equipped with central mixer
enabling either i
ntentional adjustment of excess
-
air ratio (

) to any arbitrary value or
maintaining λ = 1 condition being controlled by closed loop λ
-
control. The intake manifold of
the both engines upstream of the compressor is adapted for independent delivery and
meteri
ng of two gaseous components in addition to basic fuel delivery. In this way desired
composition of tested fuel blend can be created on
-
line and engine behaviour can be
investigated along a set of operating points using content of the particular component
in the
fuel blend as an independent variable. Purposely purchased mass flow meters are used for
measurement and closed loop control of added component flow. The devices are
configurable as concern the kind of the metered gas. The range of flexibility cover

whole
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

7

range of gasses which are taken into account as components of fuel blends planed for
investigation.

A 4xØ102/110 engine is of quite conventional (rather obsolete) design. It is dedicated for
(preliminary) testing of fuel composition impact under mod
erate operating conditions. It is
equipped with uncontrolled turbocharger generating relatively low boost pressure and its
compression ratio is CR = 10. Intercooler and exhaust gas recirculation (EGR) duct are not
installed at all. The appearance of the e
ngine is introduced in
Figure
1
.

The simple accessory layout decreases the number of degree of freedom during two way
data exchange between experimental and simulation environments. Simple reciprocal
relationship d
escribes the changes of engine (and its working substance) behaviour along a
set of operating points, not being influenced by interference of any control system.

The moderate values of peak cylinder pressure enables the engine operation at knock
condition
for period of time sufficiently long for acquisition of representative set of
experimental data. Thanks to the low basic level of the engine stress it is even possible to
perform the engine operation with significant knock intensity under steady state cond
ition.

The test bench with AC dynamometer (so
-
called Winter
-
Eichberg kind of rotational machine
-

appropriate for steady state testing) is completed with data acquisition system for sensing
and recording of conventional set of integral physical quantities
.


Figure
1
: 4xØ102/110 Engine at Test Bench

Engine speed is sensed by magnetic pick
-
up facing flywheel teeth and evaluated
chronometrically by an appropriate SW routine. Information about engine torque is
transferred from absolut
e angle encoder located on the shaft of dynamometer scale pointer.
Working substance pressure and temperature are sensed at the set of measuring point
located in relevant part of engine peripheries. Fuel consumption is measured simultaneously
by both a tur
bine
-
based flowmeter and by a metering orifice. Exhaust gas composition is
determined by set of laboratory analysers (CO, CO
2
, CH
4



NDIRA; NO, NO
2



NDUVA, O
2



PMD) taking sample from exhaust manifold upstream of the catalyst generating information
about

raw exhaust gas composition. If desirable a similar set of analysers can be used for
determination of exhaust gas composition downstream of the catalyst. Information about the
throttle position, the lambda
-
control actuator position, ignition timing and ad
ditional fuel flow
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

8

is transferred to the centralised data acquisition system using communication channels with
distributed set of particular ECUs controlling independently the mentioned circumstances.

A distributed part of the data acquisition system se
rves for acquisition of angle resolved
patterns of in
-
cylinder pressure (sensed by piezo
-
electric sensor) and the pressures in both
the intake and the exhaust manifold close to the ports of indicated cylinder (sensed by piezo
-
resistive sensors). In this wa
y data are acquired for “Three Pressure Analysis” (TPA) as
mentioned in description of layout and behaviour of VCJB
-
JBRC simulation tools.

A purposely designed counter card records the time periods between consecutive teeth on
flywheel gear passing a sens
or with resolution of 20 ns synchronously with the angle
resolved pressure records.

The complete data acquisition system generates a set of outputs which consist of:



time depending record of integral parameters with sampling period in the
range adjustable
from fraction of second to several seconds.



sequence of lines from time depending report each containing representative
set of measured quantities for particular operating point (typically recorded as
soon as steady state readings are obtained).



records of

angle resolved (resolution of 0.25 °CA) pressure patterns
describing pressure of the engine working substance inside the cylinder and
just up
-

and downstream of it for each operating point.



counter card output for each operating point.

In the framework of

the off
-
line evaluation the integral data are smoothed (if necessary) using
either median filter or moving average. The complete set of derived quantities is evaluated
numerically and stored together with directly measured ones.

The heat release patterns

are evaluated using in
-
cylinder pressure record in combination
with relevant integral quantities. The in
-
cylinder pressure pattern are combined with manifold
pressure angle resolved patterns into one text file and formatted to use as GT
-
Power TPA
input as

described later.

Instantaneous values of engine torque, crankshaft angular acceleration and instantaneous
crankshaft speed are calculated using both gas force and inertial force as inputs. The
patterns obtained in this way are compared with instantaneous
crankshaft speed evaluated
from counter card output. Acceptable level of cylinder
-
to
-
cylinder variability is tested by this
comparison for each operating point.

Description of the set of regimes where the experimental data were acquired for models
calibrat
ion and verification is introduced in
Figure
2
.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

9

1000
1200
1400
1600
1800
2000
2200
2400
2600
0
10
20
30
40
50
60
70
Pe [kW]
Speed [/min]
0
20
40
60
80
100
120
Throttle [0-100]
Speed
Throttle


Figure
2
:
Definition of Operational Points for Model Calibration and Testing 4


㄰㈯ㄱ〠䕮E楮e

Data were acquired as subsets of points dist
ributed either along speed axis (maintaining the
throttle position constant) or along load axis at constant speed. For each subset constant
ignition timing was adjusted. Engine power was chosen as independent variable in
Figure
2
.
The curves marked with filled points were acquired while engine operated being fed with lean
mixture. All other points were acquired during engine operation on stoichiometric mixture.

A 4

102/120 engine is dedicated for testing of fuel im
pacts under the condition as close to
those planed for SPA targets, as they are reasonably reachable still using conventional
engine design. The engine is equipped with a turbocharger with variable turbine geometry
(VTG
-

maximum boost pressure of approx.

1.5 barg). Compression ratio of the engine is CR
= 12. A powerful intercooler is installed in engine intake manifold and EGR duct is equipped
with powerful EGR cooler.

Controlled engine loading is done by DC dynamometer (appropriate for steady
-
state test
ing
and for testing in simple transients).

Measuring equipment is arranged in similar way to that mentioned in the description of the
4

102/110 engine with few differences.

The value of the engine torque is transferred from dynamometer ECU (simultaneous
ly using
both an analogue channel and an OPC server


client communication). VTG control actuator
position is transferred to the acquisition system through serial communication channel.
Turbocharger speed is sensed by eddy current based sensor facing compr
essor blades.
Both average and instantaneous turbocharger speed is evaluated. Beside the cooling water
temperatures both at engine inlet and outlet the cooling water flow is sensed by paddle
-
type
sensor. Beside continual raw exhaust gas analysis, an additi
onal set of analysers takes
sample from engine intake manifold. The molar fractions of particular components in fresh
mixture are used for correct and accurate determination of EGR rate and they create
additional information concerning mixture composition.


Two piezo
-
resistive pressure sensors are installed in exhaust manifold up
-

and downstream
of the turbine. Their signals are acquired and evaluated simultaneously with in
-
cylinder
pressure patterns (sensed by a piezo
-
electric sensor). During the subsequen
t evaluation the
acquired patterns of exhaust manifold pressures are compared with the corresponding
patterns generated by virtual sensors included into GT
-
Power model layout as it is described
later. Similarly the modelled and measured patterns of instant
aneous turbocharger speed are
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

10

compared reciprocally. Both the exhaust manifold pressure patterns and the patterns of the
instantaneous turbocharger speed are not used for model calibration therefore the
agreement between modelled and measured waveforms cre
ates an additional confirmation
of plausibility of the results.

Similarly to the arrangement applied at 4


102/110 engine the instantaneous crankshaft
speed is acquired and evaluated synchronously with angle resolved pressure patterns in
order to confirm t
he validity of in
-
cylinder pressure record from one cylinder only as
representative enough for whole engine.

Engine knock is evaluated by Digital
-
Signal
-
Processor based unit which evaluates the
vibration sensor signal using Discrete Fourier Transform. It
generates an analogue signal
which is proportional to the knock intensity. The output signal is observable on
-
line (enabling
the test bench crew to be aware of actual knock intensity) and it is recorded by data
acquisition system for subsequent elaboration

off
-
line.

Initial adjustment for full load curve has been established. In
Figure
4

the speed depending
patterns of setpoints for particular adjustments are presented as they were preliminary
determined for each re
gime. The setpoints were established in reciprocal relationship aiming
to reach maximum obtainable power and efficiency keeping at the same time knock
-
free
operation and maintaining the exhaust gas temperature and turbocharger speed just below
the limit of

acceptability. Engine power and turbocharger speed patterns are introduced in
Figure
4

too.

Similarly to the 4


102/110 engine a set of calibration and evaluation data for exchange with
model environment was acqui
red. In case of the 4


102/120 (
Figure
3
) engine certain
limitations had to be taken into account when set of regimes was selected.



Figure
3
:
4


㄰㈯ㄲ〠䕮E楮攠慴⁔敳e⁂敮捨

The VTG rack
had to be adjusted into one of limit position where the turbine characteristics
are available. The subsets of regimes were sensed as a speed depending curves at constant
throttle positions (and constant VTG rack positions) while other adjustable parameters

were
incidentally changed keeping in mind the necessity to protect engine from excessive load.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

11

0
20
40
60
80
100
120
140
160
180
1100
1300
1500
1700
1900
2100
2300
2500
2700
2900
Speed [/min]
Power [kW], Throttle [0-130],
TC speed [1000/min], VGT Rack [0-80]
0
5
10
15
20
25
30
Ign. Timing [°bTDC], EGR [%]
Throttle
Power
VTG Rack
Turbo Speed
Ign. Timing
EGR rate


Figure
4
:
Proposed Full Load Curve 4


㄰㈯ㄲ〠䕮E楮e

The set of operational points for which the model calibration / verification dat
a were
generated is described in
Figure
5

by plot of speed
-
depending patterns of the throttle position
and engine power for each subset. Full load curve is plotted in the
Figure
5

by dotted lines.

All mentioned experimental results were obtained on engines fuelled by Transit Natural Gas.

0
20
40
60
80
100
120
140
1100
1300
1500
1700
1900
2100
2300
2500
2700
2900
Speed [/min]
Power [kW], Throttle [0-130]
Throttle
Power


Figure
5
:
Definition of Operational Points for Model Calibration and Testing 4


㄰㈯ㄲ〠䕮E楮e

1.3

CVUT JBRC model:

The main effort was concentrated on GT
-
Power based (1
-
D) models. They were compiled
and tuned to obtain agreement with experimental results. The assortment and the range of
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

12

the model outputs were selected in order to ensure (among other) their usability as

inputs for
more detailed model used by GDF SUEZ. The models were also used for evaluation of
experimentally unavailable quantities as an instantaneous cylinder charge composition,
residual gas fraction (RGC) and combustion chamber wall temperature estimat
ions.

For model calibration and testing the systematic generation of the calibration / verification
data started. The data from approximately 80 operation points have been acquired using
Transit Natural Gas as a fuel. The selected points cover certain rang
e of engine speed,
throttle position, turbine control rack adjustment, ignition timing and exhaust gas recirculation
rate. The excess
-
air ratio was kept in most cases at


= 1 using closed loop

-
control.
Several points were acquired while engine operated
on lean mixture. The knock
-
free engine
operation was investigated as well as the engine operation with occurrence of various
intensity of knock. The raw experimental outputs were evaluated and pre
-
processed by 1
-
D
model.

Two particular mathematical models

were designed and tuned as correspond to two engines
used as sources of the experimental data.


1.3.1

GT
-
Power Model of 102/110 Engine

The flow of relevant data concerning the 4


102/110 engine is illustrated in
Figure
6
. The
common evaluation of both the integral data and the record of the angle
-
resolved pattern of
in
-
cylinder pressure together with pressure in the intake and exhaust manifold just up
-

and
downstream of the cylinder are the model inputs.

The “Three Pres
sure Analysis” (TPA) elaborates the imposed data structures. The measured
in
-
cylinder pressure patterns were used as input data for recalculation of heat release by GT
-
Power internal routine. The manifold pressure patterns create boundary conditions at the

control volume borders. Mentioned strategy of the model design leaves the turbocharger
outside of the model control volume, therefore model operation is not burdened with time
consuming iterative search for turbocharger


engine equilibrium state. At the
same time the
assortment of model outputs is almost unlimited. Therefore one of the most significant model
exploitation is sophisticated evaluation of the experimentally obtained data.

TEST BENCH (4

102/110)
Sensing & Partial Evaluation
INTEGRAL PARAMETERS
Time
-
Depending Patterns
In
-
cylinder Pressure, Intake
/ Exhaust Pressure
Angle
-
Resolved Patterns
COMMON EVALUATION
GT
-
Power
One
-
cylinder TPA
Engine Geometry
One separate cylinder +
intake and exhaust ports
Angle
-
Resolved Patterns of:
In
-
Cylinder Pressure
Temperature burnt/
unburnt
Heat Release
Mass fraction of components

Partner
Demands
GDF
SUEZ
Instantaneous Crankshaft
Speed
Angle
-
Resolved Patterns
Evaluation of Acceptability of
Cyl
-
to
-
Cyl
Variation


Figure
6
:
Flow
-
Chart


102/110 Engine Data E
laboration

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

13

Measured static pressure waveforms in intake and exhaust manifold are imposed to the
model as an angle resolved boundary condition. An average intake and exhaust
temperatures and both air and fuel mass flow obtained from measurements are used as

integral boundary conditions. The valve mass flows and cylinder pressure are iteratively
calculated in a forward simulation till a convergence occurs. The valve lifts had been
determined by the direct measurement on real specimen. The valve discharge coe
fficients
had been obtained from measurements performed at steady flow test rig.

Heat release pattern is evaluated at the beginning of the each iteration from the backward
simulation using the high pressure part of measured cylinder pressure and mass flow
from
the last iteration.
Figure
7

shows a model layout of one separated cylinder. The flow
components, control elements and virtual sensors are visible in the figure. The virtual
sensors are dedicated for determination of angle re
solved patterns of the charge state
quantities, heat release fraction and the charge composition. Measured pressure patterns
are entered into the model in a form of a text file. Automated experimental TPA input data file
generation is an integrated part o
f an in
-
house developed software package for a multiple
channel pressure indication and its post
-
processing.



Figure
7
: 4


㄰㈯ㄱ〠1
-
Pow敲⁍od敬e污yo攠e数er慴敤cy汩湤敲⁔PA

An illustration of model results as concern the
ir assortment and usability is introduced in the
Figure
8
. In upper part of the figure the angle resolved patterns of relevant pressures are
displayed as they develop with increased engine load (left to right). Bot
h the modelled (solid
lines) and the measured (dotted lines) patterns of the in
-
cylinder pressure are plotted. The
pressure scale is arranged in order to enhance the legibility of patterns in low pressure part.
Typical difference between modelled and acqui
red pattern is highlighted. The derived
integrate value of the residual gas content is plotted in the bottom of the figure depending on
indicated mean effective pressure.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

14



Figure
8
: Illustration of
102/110 GT
-
Power Model results

M
odel results are presented in Annex 1 by plots of iso
-
lines of selected integral model
outputs (to illustrate model results consistency) and by the plots of angle resolved in
-
cylinder
pressure patterns in comparison with experimental results.


1.3.2

GT
-
Power mod
el of 102/120 Engine

Somewhat different strategy was implemented for design of GT
-
Power based model of the
4


102/120 engine. Due to the design features of the manifolds it is impossible to find
appropriate location for installation of the manifold pressur
e sensors enabling the use of their
signals as the boundary conditions. Therefore the model of whole engine layout was
compiled. The procedure of the acquisition and elaboration of relevant data is illustrated in
Figure
9
. Complet
e engine geometry was imposed as it was determined using either the
engine design documentation or measured directly at particular specimen. The steady state
compressor and turbine maps were obtained from the turbocharger manufacturer R&D
department. Only
one “executive” multiplier was used during model tuning in order to obtain
agreement between model and experiment. It influenced average speed of turbocharger
shaft. In fact the multiplier introduces the corrections of the turbine maps towards pulsating
fl
ow in engine exhaust manifold. The iterative search for engine


turbocharger equilibrium
state is demanding for computation time. On the other hand the complete and
comprehensive description of behaviour of whole engine and its accessory enables plausible

extrapolation of the results for further use.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

15

TEST BENCH (4

102/120)
Sensing & Partial Evaluation
INTEGRAL PARAMETERS
Time
-
Depending Patterns
In
-
cylinder Pressure
Angle
-
Resolved Patterns
Instantaneous Crankshaft &
Turbo Shaft Speeds;
Exhaust Pressure
Up/Downstream Turbine
Angle
-
Resolved Patterns
COMMON EVALUATION
GT
-
Power
Engine + Turbo
Whole Engine Geometry
& Turbo Maps
Angle
-
Resolved Patterns of:
Pressure
Temperature burnt/
unburnt
Heat Release
Mass fraction of components

Evaluation of Acceptability of
Cyl
-
to
-
Cyl
Variation;
Cross
-
Check
Model/Experiment
Partner
Demands
GDF
SUEZ


Figure
9
:
Flow Chart


102/120 Engine Data Elaboration

The GT
-
Power model layout of the whole 4


102/120 engine with turbocharger, EGR valve
and cooler and virtual sensors and contro
llers is displayed in
Figure
10
. The virtual TPA
output is highlighted in the upper right corner in the figure.

To complete the insight into the model layout it has to be mentioned that several additional
multipli
er are introduced in order to overbridge the calculations concerning the behaviour of
the flow control elements themselves. In this way the desired flow of particular substance
(both the mixture flow through the throttle and the EGR flow trough EGR control

valve) can
be entered directly and model run is not burdened by calculation of flow value from
characteristic of control element itself. The specific design of control elements and their
characteristics are not the topics of the present research activity.


The same routines for heat release evaluation from measured pressure patterns were used
for multi
-
cylinder model as in the case of TPA on the separate cylinder of the 4


102/110
engine.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

16



Figure
10
:
GT
-
Power model layout of the w
hole 4


㄰㈯ㄲ〠敮g楮e

Example of model results is introduced in
Figure
11
. The content of the total residuals (= sum
of residual gas and EGR) is plotted against engine speed for operation points defined in
Figure
5
. Individual points plot the directly measured EGR values. Dashed lines only confirm
the agreement between model and measurement concerning EGR rate which is obtained by
use of a multiplier applied on EGR valve op
ening. The solid lines are elevated above the
dashed ones by calculated amount of residual gas. The
Figure
11

illustrates the ability of the
model to generate data which are not accessible for direct measurement (a
nd which are
useful for subsequent evaluation).

Set of simulation results namely in
-
cylinder pressure patterns is compared with experimental
data in Annex 2.

Both mentioned models generate wide range of exactly defined results (including those not
access
ible for direct measurement) enabling exact description of even very subtle influence
of fuel composition on engine working cycle behaviour. On the other hand, the validity of
directly acquired values was not doubted significantly and they are still usable

as inputs for
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

17

the kind of evaluation where simulation results cannot be used


especially cycle
-
to
-
cycle
variability.

Resulting sets of both the angle resolved patterns and the integral engine parameters were
archived and simultaneously sent to GDF SUEZ f
or their own model calibration needs for
input data.



Figure
11
:
GT
-
Power model outputs

1.3.3

Knock model

Knock model is based on empirical relationship for calculation of ignition delay. Knock occurs
when the ignition delay time is ful
ly exhausted (exhausted part of the ignition delay is equal
to one). The correlation for methane included as a part of in
-
house code OBEH (Cycle) is
used for subsequent calculation.

The inputs for knock model (angle resolved patterns of in
-
cylinder press
ure and temperature
of unburned mixture) can be obtained either from experimental data using simple two zone
mean temperature model or from OBEH (Cycle) code or from 1
-
D commercial code GT
-
Power.

The simplest method for implementation of two zone approach

is based on assumption that
locally averaged cylinder charge temperature represents a weighting average of
temperatures of both burnt and unburned zones while the burned fraction is the weighting
factor. The unburned zone temperature after start of combus
tion is computed assuming
polytropic compression of the unburned charge


two zone mean temperature model
(2ZMTM). Polytropic exponent is then calculated using relationship developed by Brunt [
1
]
.





[
1
] Brunt, M., F., J., Emtage, A., L.: The Calculation of Heat Release Energy from Engine

Cylinder Pressure Data, Detroit, 1998,
SAE 981052

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

18

4*102/120 engine, RPM 1300, WOT, without EGR, Light Knock,
fuel = TNG
0
5
10
15
20
25
0
5
10
15
20
25
Cycle # [-]
Knock Onset [deg aTDC]
Knock Onset - Experiment
Knock Onset - A = 1
Knock Onset - A given by calibration












u
1
4
T
18508
A
exp
p
10
13
.
8

Figure
12
: Determination of kno
ck onset for knocking cycles

First modelling results do not show good agreement with experimentally determined knock
onsets. An additional calibration procedure has been implemented into knock model. A
multiplier A has been used which significantly improv
es relationship between measured and
simulated position of knock onset. This multiplier affects the ratio between activation energy
and temperature of unburned mixture in above mentioned correlation. As the
Figure
12

shows the results are very promising. The model describes the knock onset positions even
quantitatively. The problem has occurred in case of non
-
knocking cycles. The model with the
tuned value of the multiplier identified certain level of knock also fo
r regimes which are
declared from experiment as knock
-
free in many cases. That is why a different multiplier
tuning strategy has been applied. In this case the simulation starts with regimes where only
knock
-
free cycles were recorded. Certain limit value f
or the multiplier is determined in this
way. Subsequently the knock onset is calculated using data from average cycle for given
regime not exceeding the preliminary determined limit value for multiplier. The
Figure
13

shows an example of the results from simulation after limit calibration procedure was applied.

Experimental value of knock onset for each operating point was averaged from values of
knock onsets evaluated for each one from sequence of 80 individual cyc
les acquired for
given operating point. The value 50 of knock onset denotes cycle without occurrence of
knock in this figure. The intensity of knocking is described also by the output voltage from
vibration sensor evaluation unit (AKR


0 V means knock
-
fre
e operation, 5 V corresponds to
heavy knock) in this figure. It is well visible that the mentioned model is able to describe
occurrence of knock, but there is certain difference between experimental and model results
as concern quantitative expression of k
nock onsets. The results of implementation of two
various methods for determination of unburned fraction temperature are presented.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

19

4x102/120 engine, RPM 1300, without EGR
0
5
10
15
20
25
30
35
40
45
50
14
15
16
17
18
19
20
21
22
23
Ignition Timing [deg bTDC]
Knock Onset [deg aTDC]
0
1
2
3
4
5
6
AKR [V]
knock onset - experiment
knock onset - OBEH
knock onset - 2ZMTM
AKR

Figure
13
: Determination of knock onset for both knock
-
free and knocking cycles


1.4

GDF SUEZ model:

GNVSim calibration

Previously to its involvement in the InGAS project, in order to help car manufacturers to
optimize their natural gas engines, GDF SUEZ developed GNVSim which is a simulation tool
used to predict the impact of natural gas composition on
power output and exhaust
emissions. GNVSim is an appendix of the IFP
-
Engine


library [
2
]

in the commercial software
LMS Imagine.Lab AMESim

.

The main asset of GNVSim is the possibility of taking into account the main characteristics of
natural gas on an AM
ESim


platform,
with a high focus on the impact of quality variations of
natural gas as a multi component gaseous fuel.

All given information on the model itself and how it works is for information only. Since this
simulation tool itself is not a deliverab
le of the InGAS project, all details necessary to relevant
use of the future results to be obtained with GNVSim are given in this report. As further and
more detailed calibration of the model parameters might have to be done in the following
tasks involvin
g use of this tool, all possible changes in the model ability to predict the
simulated engine behaviour will then be added in the deliverables of these tasks.


1.4.1

Description of the model

GNVSim is based upon the IFP
-
Engine® Coherent Flame Model 1D (CFM
-
1D) w
hich is a
one dimension model dividing the combustion chamber into two zones: the burned and
unburned gases. The system of differential equations describing the mass and energy flows
between these two zones is a function of the cylinder pressure (considere
d as homogenous),
the temperature of each zone and the crankshaft angle being the only independent variables.
Closure for the system of differential equations is obtained by a phenomenological model of
turbulent combustion describing the burned gas mass fr
action.




[
2
]

F.
-
A. Lafossas, O. Collin, F. Le Berr, P. Menagazzi, Society of Automotive Engineers, 2005
-
01
-
2107

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

20

The two zones are separated by an infinitesimal flame front whose surface is wrinkled by the
turbulent flow occurring inside the combustion chamber. The flame surface is firstly
hemispherical and the cylindrical when its diameter exceeds the distan
ce between the piston
and the cylinder head. The turbulence model employed is a simplified k
-


model.

Turbulence is mainly created by movement of in
-
cylinder load during the intake stroke. The
impact of turbulence arises in the propagation speed term of th
e flame front which describes
the mass fraction burned. Propagation speed of the flame front is a function of the laminar
flame speed and the turbulence characteristics.

Details on the regular CFM
-
1D model, knock model included, can be found in the
publica
tions [
3
,
4
].

With regard to differences between GNVSim and IFP
-
Engine® gasoline CFM
-
1D model,
natural gas characteristics as a fuel were taken into account through the following main
modifications and developments:



introduction of natural gas composition
in the fuel definition (methane, ethane,
propane, butane, nitrogen and carbon dioxide),



thermodynamic properties calculations,



laminar flame speed calculations with dependency over equivalence ratio, residual
and EGR (Exhaust Gas Recirculation) rate and
the composition of natural gas.

Hydrogen has also been added for Hythane® simulations. Laminar flame speed correlations
are based on calculations with GDFKin® mechanism.

This implementation then resulted in a dedicated CNG library which is used in combinat
ion
with the IFP
-
Engine® library.


1.4.2

Calibration tools

Use of GNVSim requires only the use of the AMESim® platform, the IFP
-
Engine® library and
the GNVSim dedicated submodels. Nevertheless, calibration of the relevant model constants
with comparison to avail
able experimental data is an activity that needs a great number of
run tests with the software so as to lead to the choice of a satisfying set of values for these
parameters. Therefore, another software tool, Scilab, was used with help of a script so as to

control the AMESim® platform and run these numerous calculations.

1.4.2.1

Engine system sketch in AMESim® environment

The
sketch mode
, in which the various components are linked together, is the first step of
modelling a system with AMESim®. The sketch used in th
is study is presented in
Figure
14
.
It focuses on the compression and expansion strokes with combustion phase included. Thus,
accurate description of valve actuation was not taken into account in the combustion
cha
mber modelisation. The valve closing is simulated with the annulment of the flow signal
since the combustion model needs to detect this occurrence to run normally.




[
3
]


Lafossas F.
-
A., Colin O., Le Berr F., Mennegazzi P., Application of a New 1D Combustion Model to
Gasoline Transcient
Engine Operation, SAE Paper 2005
-
01
-
2107, 2005,

[
4
]

S. Richard, S. Bougrine, G. Font, F.
-
A. Lafossas and F. Le Berr, On the Reduction of a 3D CFD Combustion Model to
Build a Physical 0D Model for Simulating Heat Release, Knock and Poll
utants in SI Engines, Oil & Gas Science and
Technology


Rev. IFP, Vol. 64 (2009), No. 3, pp. 223
-
242

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

21

Gas definition
Engine definition
Combustion chamber
Ignition advance
Valve train

Figure
14
: sketch representing the single cylinder engine model

I
n the combustion chamber component most of the parameters, which will be calibrated
during this study like the turbulence characteristics (i.e. turbulence constant and flame
wrinkling parameter), are defined. The engine definition component describes the e
ngine
geometry. The gas definition component is a 12 gas model that enables to take into account
the impact of gas composition in the simulation. The spark timing advance component
enables the model to optionally consider an additional delay as a parameter
, through the
implementation of a constant subtractor. This additional delay would have the same effect as
the later described initial flame volume parameter, and so was not used in this study.

1.4.2.2

Calibration script

A script, supported on Scilab, is used to c
ontrol and run the sketch. Scilab is a numerical
computational package whose syntax is similar to MATLAB®. This script uses a library of
functions that enables the management of input parameters, run of simulations and export of
results data.

The calibrati
on approach is presented in
Figure
15
. First step is initialisation of the model
input data: the operating point characteristics. Its aim is to set the model variables for a
single chosen operating point. Then, eve
ry chosen parameter (each would be one of the so
-
called constants of the model) is calibrated following a given order. A calibration loop has to
be implemented for every parameter to calibrate. An error function, as convergence criterion,
is defined for ea
ch calibration loop. The error function expresses, for a given phase of the
engine cycle, the distance between the experimental and the simulated cylinder pressure
curves. Every calibration loop converges, by dichotomy, to the parameter value that
minimize
s the error function. The resulting value is saved and then used in the following
calibration loop dedicated to next parameter.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

22

Experimental
cylinder pressure points
Data base of
experimental functioning
points
Definition
of
the
simulation
parameters
Initialization
of
the calibration parameters
Simulation RUN
Evaluation
of
the error between simulated and
experimental
data
Storage
of
the calibrated parameter
CALIBRATION LOOP:
modification of
the calibration parameter
POINT LOOP:
incrementation
of
the functioning
point

Figure
15
: data flow of the calibration loop


1.4.3

Calibration parameters taken into account

In this study
, calibrations of the combustion model on the AMESim
®

platform are performed
with closed valve calculations. The parameters that are fitted in this study are the residual
gases ratio (IGR), the turbulence constant of the model, the C1 parameter of Woschni’
s
correlation for thermal losses [
5
] and the flame wrinkling multiplier. Other parameters, like for
instance C
dis

and C
turb

that control turbulent kinetic energy creation and destruction phasing
along the cycle, are kept to the software default values. Wal
l temperature is kept constant to
an arbitrarily chosen value over all the simulated engine tests because of lack of data.




[
5
]

Woschni G., Universally Applicable Equation for the Instantaneous Heat transfer Coefficient in the Internal Combustion
Engine,
Society of Automotive E
ngineers

, SAE 670931

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

23

Cylinder pressure : comparison between experimental data and
simulation result
0
10
20
30
40
50
60
200
260
320
380
440
500
crank angle (°)
cylinder pressure (bar)
P_cyl mesurée
P_cyl estimée
1
2
4
5
3
experimental
simulation

Figure
16
: comparison between the experimental (dotted line) and the simulated with the
calibrated paramet
ers (solid line) cylinder pressure curve
-

simulation key parameters to
reproduce experimental behaviour

Figure
16

shows a comparison between the experimental cylinder pressure curve and the
simulation results. Dif
ference between both curves is on purpose overdone so as to give an
easier understanding of the methodology. The various steps for CFM model calibration that
are highlighted in the graph are:

1.

In
-
cylinder mass:

Mass calibration is achieved through residual
gases rate estimation (IGR). It does not take
into account some particularities of each point such as acoustic waves magnitude.

2.

Start of compression stroke:

At this step, thermal losses for compression stroke have to be calibrated. Thermal loss
estimation
is made with choice of the C1 parameter [
5
]. Error in pressure estimation can be
induced here by in
-
cylinder mass estimation error, and then, also by:



compression ratio uncertainty due to manufacturer’s prod
uction tolerance,



thermal losses, because the calibration is averaged over all experimental data from
the tests,



ideal gases hypothesis may not be representative of real behaviour.

3.

Start of combustion:

Initial flame volume parameter has to be calibrated. I
t represents the flame core at the very
beginning of the ignition. In this study, initial flame volume calibration is performed at the
same time as the turbulence constant, so as to choose relevant values for these parameters.
The turbulence constant has a
n influence upon the kinetic energy of the average in
-
cylinder
load movement. However, this parameter has no physical meaning, i.e. it does not represent
the number of rotations of the gas load around the tumble or the swirl rotation axes inside the
combus
tion chamber.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

24

4.

Turbulent flow:

A calibration of the
flame wrinkling, that controls turbulence description of the early
combustion phase, is also performed.

Combustion speed increases with the value of the
flame wrinkling multiplier.

Final calibration of the

turbulence constant is done when all other calibrations are made.

5.

Expansion stroke:

Thermal losses are also calibrated at expansion stroke. During this phase, all cumulated
errors are gathered. Same reasons as for compression stroke may explain specific e
rrors
(not induced by previous calibrations) in this phase.


1.4.4

Calibration methodology

The choice of an optimal calibration order for the different parameters is crucial with aim to
minimize the error function. The calibration order importance is emphasized
by the
interdependence of each parameter. The approach used in this study is to follow the order of
influence of each parameter upon the cylinder pressure curve (
Figure
16
). According to this
assertion, the calibra
tion order is presented in
Table
2


calibration step

parameter calibrated

1

2

3

4

5

6

in cylinder mass (through residual gas rate)

thermal losses (first set)

initial flame volume and turbulence constant

flame wri
nkling

thermal losses (final set)

turbulence constant (final set)

Table
2
: calibration order used for this study

With regard to assessing the accuracy of the calibration approach, three different relative
errors are calculated for
each operating point:



maximum pressure error



maximum pressure angle error



indicated mean effective pressure error

The two first errors indicate the level of closeness in the pressure peak description, whereas
the third one attests the global correspondence

between the two curves. An example of the
impact of each calibration step upon these errors is presented in
Figure
17
. The maximum
pressure error constantly decreases along the calibration procedure, so does the m
aximum
pressure angle error. The indicated mean effective pressure error decreases along the
calibration procedure but slightly increases with the last step. Nevertheless, it was confirmed
that the final turbulence constant calibration leads to a better co
rrespondence with regard to
the pressure peak description. It is interesting to mention that even though the first thermal
losses calibration does not affect significantly the error, calculations showed that its absence
causes a marked degradation of the c
alibration final results.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

25

0
10
20
30
40
50
60
1
2
3
4
5
6
calibration step (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure error

Figure
17
: evolution of the relative error between the simulation results and the experimental
data after each calibration step

Once the calibration procedure is completed, dependency over load and engin
e speed is
taken into account to establish correlations describing the parameters variation over
experimental data. For a single parameter, for instance the turbulence constant, the
correlation equation will be under the following form:

0
1
...
a
load
a
load
a
turb
n
n










(1)

It is a polynomial function of the load where the coefficients are function of the engine speed
N
:

0
1
max
1
max
)
(
...
)
(









N
N
N
N
a
p
p
n


(2)

N
max

being the maximum engine speed used for correlation implementation.

Then, the final step consists in running si
mulations using directly the parameter values
resulting from the maps. This phase allows to assess the accuracy of the established
correlations estimating the evolution of the relative errors for each operating point. The
prediction ability of the model is

first tested with the experimental data used for model
calibration and then with a new panel of operating points (if available).


1.4.5

Gas composition

Since no accurate data on the transit natural gas composition available at the test bench
facility, its compo
sition was assumed to be an averaged composition from natural gas data
published by the natural gas grid operator during 2006 and 2007, given by CVUT JBRC, see
Table
3
. This data shows some variations of compositio
ns but is consistent with an averaged
composition of Russian gas in 2008, information supplied by E.ON Rurhgas.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

26

assumed transit natural
gas composition

concentration
(mol. %)

CO
2

N
2

O
2

CH
4

C
2
H
6

C
3
H
8

n
-
C
4
H
10

i
-
C
4
H
10

0.07

0.8

-

97.95

0.82

0.26

0.05

0.04

S
um

100

Table
3
: Assumed composition for transit natural gas available at CVUT JBRC test facilities


1.4.6

Focus on the 102/110 engine

1.4.6.1

Calibration results

The characteristics of the different operating points used for CFM calibration for
the 102/110
engine are listed in
Table 3
1

of Annex 3. The comparison between the experimental cylinder
pressure curves and the curves resulting from simulation with the calibrated parameters
value is presented in Annex 4, for eac
h operating point. It appears (
Figure
18
) that the
calibration procedure leads to very acceptable relative errors.

0
1
2
3
4
5
6
7
0.2
0.4
0.6
0.8
1
load (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure error

Figure
18
: presentation of the different calculated relative errors for
operating points simulated
with calibrated values of the parameters

Relative errors average and distribution are presented in
Table
4
. Indeed, only a few treated
operating points lead to relative errors superior th
an 2%. Average error and distribution of the
set of errors show that the calibration procedure is satisfactory in terms of simulation results
with the optimum calibrated parameters.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

27


Average (%)

Proportion of points

with error< 1%

Proportion of points

with

error< 2%

Error on maximum pressure

Error on maximum pressure angle

Error on indicated mean effective pressure

1.4

0.9

1.7

53

58

21

84

100

63

Table
4
: average and distribution of measured relative errors for the simulation cur
ves
obtained with optimum parameters resulting from calibration


1.4.6.2

Maps implementation

The optimum parameter values obtained in the calibration phase are used to establish
correlations according to equations (1) and (2). With aim to increase the prediction a
bility of
the correlation, different degrees of polynomials (1) and (2) are chosen. For instance,
turbulence constant and C1 parameter are described by equation (1) and equation (2)
respectively of third and second degree (
Table
5
). IGR ratio and flame wrinkling multiplier are
represented by equation (1) and equation (2) of respectively third and first degree (
Table
6
)
Initial flame volume is described by two polynomial equa
tions of second degree (
Table
7
).
The different maps obtained are presented in Annex 5. The choice of the polynomial degree
is independent of the parameter nature and only depends on the parameter variation bias.

T
he set of values in
Table
5
,
Table
6

and
Table
7

are presented for the record.


turbulence constant

C1 parameter


2


1


0


2


1


0

a
3

a
2

a
1

a
0

8.5525

-
68.398

57.536

-
6.8722

-
1
6.29

122.67

-
102.54

11.289

7.7379

-
54.133

44.917

-
3.9013

-
274.93

824.41

-
646.81

170.97

510.09

-
1453.4

1135.5

-
299.91

-
235.16

641.63

-
501.34

136.35

Table
5
: parameters the equation (2) obtained during the map implementation for the
turbulence constant and the C1 parameter (thermal losses)


Residual gas rate (IGR)

flame wrinkling


1


0


1


0

a
3

a
2

a
1

a
0

65.0622

-
156.2022

123.4506

-
37.9680

-
86.3042

209.1192

-
169.9096

59.8850

10.3115

-
16.5464

8.7318

-
1.5855

-
12.3667

22.0570

-
12.6528

12.0585

Table
6
: parameters of the equation (2) obtained during the map implementation for the IGR
ratio (in cylinder mass) and the flame wrinkling multiplier


initial flame volume


2


1


0

a
2

a
1

a
0

-
2.2764

0.3749

2.4763

3.1215

-
0.0618

-
4.5755

-
0.6262

-
0.3912

2.8297

Table
7
: parameters of the equation (2) obtained during the map implementation for the initial
flame volume


1.4.6.3

Correlations review

With regard to exploring the limits of the correlation approach
, optimum parameter values
and calculated values are analysed (
Table
8
) for one chosen parameter, in this case initial
flame volume. Indeed, different behaviours arise when the optimum values are compared to
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

28

the va
lues resulting from correlations. For the operating points used for correlation creation,
i.e. 1300 min
-
1
, 1600 min
-
1

and 2250 min
-
1
, relative error between the calculated and the
optimum value is insignificant. However, when the correlation is used for ot
her estimate initial
flame volume for other operating points then relative error dramatically increases.

operating point

initial flame volume


speed (min
-
1
)

load (
-
)

optimum

calculated

error (%)

1301

1301

1300

1599

1601

1601

1900

1900

1900

1900

2090

2099

2100

2101

2251

2252

2249

0.90

0.74

0.39

0.98

0.42

0.29

1.05

0.67

0.37

0.27

1.06

0.66

0.34

0.23

0.64

0.33

0.21

1.08

1.02

0.96

1.01

0.80

0.79

0.60

1.80

1.53

0.67

1.61

1.15

0.70

0.71

0.77

0.73

0.72

1.08

1.02

0.96

1.01

0.80

0.79

0.98

0.79

0.72

0.71

0.93

0.77

0.71

0.70

0.77

0.73

0.72

0

0

0

0

0

0

39

128

113

6

72

50

1

1

0

0

0

Table
8
: optimum flame wrinkling parameter, calculated value and resulting relative error for
each operating point used for calibration

This degradation is due to th
e fact that different biases are observed for the variation of initial
flame volume with the engine load (
Figure
19
) 1300 min
-
1
, 1600 min
-
1

and 2100 min
-
1

points
were chosen because correspond to similar tendencies
. This lack of homogeneity, expressed
by a great variation of the optimum value for different operating points, might be correlated to
the fact that the experimental data comes from distinct bench test campaigns, with variations
of natural gas quality, amb
ient conditions, occasional occurrence of knock or so. The defect
of uniformity of the experimental data could be the consequence of uncontrollable variations
of e.g. gas composition or any slight change in the test rig set up.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

29

0.5
0.7
0.9
1.1
1.3
1.5
1.7
1.9
0.2
0.4
0.6
0.8
1
load (-)
initial flame volume (cm3)
1300
1600
1900
2100
2250

Figure
19
: variation of the initial flame volume for different operating points


1.4.6.4

Correlations prediction ability

For one thing, calculated correlations are used to simulate the cylinder pressure for the
previously used operating points (
Table 3
1

of Annex 3). It appears (
Figure
20
) that the use
of correlations leads to a wider dispersion of the maximum pressure error, whereas the
maximum angle error and the indicated mean effective error are

kept behind 5%.

0
5
10
15
20
25
30
35
40
0.2
0.4
0.6
0.8
1
load (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure error

Figure
20
: various calculated relative errors for operating points simulated with values of the
parameters resulting from the established correlations

Average and distribution of relative errors are presented in
Table
9
. As mentioned before,
correlations seem to cause a marked degradation in pressure peak description in terms of
Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

30

accuracy of the maximum pressure. Indeed, 50% of the analysed points have a maximum
pressure er
ror higher above 10%. With regard to the maximum angle pressure and the
indicated mean effective pressure error, results are more satisfactory.


Average
(%)

Proportion
of points

with

error<1%

Proportion of
points

with
1%<error<5%

Proportion of
points

with
1%<error<5%

Proportion
of points

with

error> 10%

maximum pressure error

maximum pressure angle error

indicated mean effective pressure error

10.8

0.8

2.8

17

67

11

17

33

72

17

0

17

50

0

0

Table
9
: average and distribution of measur
ed relative errors for the simulation curves
obtained with parameters resulting from correlations (points used for correlations)

Secondly, new points (not used for correlations development) are used for correlations
prediction ability estimation. Character
istics of the new operating points are listed in
Table 3
2

of Annex 3. These points are obtained at wide opened throttle.

0
2
4
6
8
10
12
14
0.9
0.95
1
1.05
1.1
load (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure error

Figure
21
: various calculated relative errors for operating points simulated with

values of the
parameters resulting from the established correlations

Comparing the results obtained with optimum calibrated parameters (
Figure
18
) and the
parameters obtained using established correlations (
Figure
21
) it appears that the use of
correlations decreases accuracy but enables simulating tendencies outside the experimental
points field. Average and distribution of relative errors are presented in
Table
10
.


Average (%)

Proportion of
points

with error<1%

Proportion of
points with

1%<error<5%

Proportion of
points with
5%<error<10%

maximum pressure error

maximum pressure angle error

indicated mean effective pressure error

6.2

0.5

3.9

13

87

7

27

13

60

47

0

33

Table
10
: average and distribution of measured errors for the simulation curves obtained with
parameters resulting from correlations (new points)


Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

31

1.4.7

Focus on the 102/120 engine

1.4.7.1

Calibration results

The characteristi
cs of the various operating points used for CFM calibration for the 102/120
engine are listed in
Table 3
3

of Annex 3. In Annex 6 is presented, for each operating point,
the comparison between the experimental cylinder pressure cu
rves and the curves resulting
from simulation with the calibrated parameters value. Globally, it appears (
Figure
22
) that
results from simulation are in good agreement with experimental data.

0
1
2
3
4
5
6
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
load (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure error

Figure
22
: various calculated relative errors for operating points simulated with calibrated
values of the parameters

Average and distribution of relative errors are presented in
Table
11

Only a few treated
po
ints lead to relative errors above 5%. It arises from average error and distribution of the set
of errors that calibration procedure is satisfactory in terms of simulation results with the
optimum calibrated parameters.


Average (%)

Proportion of
points

wi
th error<1%

Proportion of
points

with error<5%

maximum pressure error

maximum pressure angle error

indicated mean effective pressure error

1.5

0.6

2.5

40

90

0

100

100

90

Table
11
: average and distribution of measured relative erro
rs for the simulation curves
obtained with optimum parameters resulting from calibration


Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

32

1.4.7.2

Maps implementation

The optimum parameters values obtained in the calibration phase are used to establish
correlations according to equations (1) and (2). Since only
two points per engine speed are
available (
Table 3
3

of Annex 3), equation (1) is imperatively linear. The different maps
obtained are presented in Annex 7. The experimental data used for the CFM calibration
come from a unique tes
t campaign. Therefore, there is a better homogeneity in experimental
results, i.e. the evolution of calibrated parameters is smooth when varying the load or the
engine speed.

The set of values in
Table
12
,
Table
13

and
Table
14

are presented for the record.


turbulence constant

residual gas rate (IGR)


2


1


0


2


1


0

a1

a0

-
3.6594

9.5904

5.9979

-
17.14

-
2.4299

8.171758

17.131

-
12.888

-
15.159

13.788

-
4.2173

10.982

Table
12
: parameters of the equation (2) obtained during the map implementation for the
turbulence constant and the IGR ratio (in cylinder mass)


flame wrinkling

C1 parameter


2


1


0


2


1


0

a1

a0

-
0.0945

-
1.2069

-
3.5387

4.4174

3.1622

7.00
92

-
56.967

47.396

93.468

-
77.722

-
39.165

34.645

Table
13
: parameters the equation (2) obtained during the map implementation for the flame
wrinkling parameter and the C1 parameter (thermal losses)


initial flame volume


1


0

a1

a0

4.263

-
5.0922

-
3.5525

5.3944

Table
14
: parameters of the equation (2) obtained during the map implementation for the initial
flame volume


1.4.7.3

Correlations prediction ability

As a matter of fact, prediction ability of the model usin
g previously established maps for the
102/120 engine is insufficient (
Figure
23
). Parameters resulting from correlations provide a
dramatic error broadening. Disagreement amplification between simulation results an
d
experimental data can be attributed to inaccurate maps. This lack of exactitude in
parameters variation description is connected to the very small amounts of points available
for correlations formulation. A reduced number of data makes it more difficult
to build
relevant correlations so as to extrapolate engine behaviour to operating points outside of the
range of the points used for calibration when these do not give a sufficiently broad
information on the overall operating range. Therefore, it would be
suitable to have more than
two points per engine speed in order to describe more appropriately parameter variations in
the operating range of the engine.

Deliverable Report

Contract N.:218447

hammercouple_fdd576da
-
88a1
-
478a
-
80eb
-
4f9b547c50eb.doc


INGAS Integrated Project

33

0
10
20
30
40
50
60
0.5
0.55
0.6
0.65
0.7
0.75
0.8
0.85
0.9
load (-)
error (%)
maximum pressure error
maximum pressure angle error
indicated mean effective pressure

Figure
23
: various calculated relative errors for operating points simulate