Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

JSAE 20119021
SAE 2011-01-2045

Development of In-cylinder Mixture and Flame

Propagation Distribution Measurement Device with Spark
Plug Type Sensor
Koshiro Kimura, Sachio Mori, Masato Kawauchi and Rio Shimizu
TOYOTA MOTOR CORPORATION

Copyright © 2011 Society of Automotive Engineers of Japan, Inc. and Copyright © 2011 SAE International

ABSTRACT [2]-[10], but they are limited to just several points of

measurement and there are no examples of this
A new method to measure in-cylinder flame method being applied to the measurement of spatial
propagation and mixture distribution has been distribution.
developed. The distribution is derived from analyzing
the temporal history of flame spectra of CH* and C2*, Therefore, the aim of this work is to establish a
which are detected by a spark plug type sensor with mixture distribution analysis method that can be used
multi-optical fibers. The validity of this method was in an actual engine. The development expanded
confirmed by verifying that the measurement results measurement of the equivalence ratio from the radical
corresponded with the results of high speed flame luminescence at one point to create a spatial
visualization and laser induced fluorescence (LIF) distribution measurement method.
measurement. This method was also applied to
analysis of cyclic combustion fluctuation on start-up in This paper first describes the basic principles of the
a direct injection spark ignition (DISI) engine, and its method to measure the equivalence ratio from the
applicability was confirmed. radical luminescence within the engine cylinders. It
then details the construction of a spatial distribution
INTRODUCTION measurement device by integrating a sensor into the
spark plug. Finally, this paper provides how the
Recent years have seen an increase in demand for validity of the measurement results was investigated
engines with lower fuel consumption (i.e., higher and reports the results of applying this method to
efficiency) and CO2 emissions to help address the measure mixture distribution in a DISI engine.
issues of global warming and the depletion of
petroleum. In order to improve the thermal efficiency EXAMINATION OF EQUIVALENCE RATIO
of engines, it is necessary to optimize combustion. MEASUREMENT METHOD
And, it is extremely important to identify the in-cylinder
mixture distribution, since this has a large influence on PRINCIPLE OF MEASURING EQUIVALENCE
combustion and emission. Measurement of mixture RATIO FROM RADICAL LUMINESCENCE
distribution makes it possible to obtain an indicator of
the design improvement. Previously, a method was proposed for estimating the
equivalence ratio from light emission intensity ratios
Conventional visualization methods, as typified by LIF such as, OH*/CH*, C2*/OH*, C2*/CH*, (* indicates a
[1], enable highly precise two-dimensional mixture radical) and so on [2]-[10]. Figure 1 shows an
measurements. However, these measurements example of the chemiluminescence from a premixed
require the use of a specially designed visualized methane-air flame broken into spectra (i.e., the light
engine, which may restrict the reproduction of actual emission intensities at each wavelength) [2]. The
engine conditions during development or delay radicals, such as OH*, CH* and C2*, which are the
feedback to the engine developers. intermediate species of chemical reactions, emit light
during combustion and their characteristic line spectra
On the other hand, another method is available that can be observed. The major elementary reactions
measures the equivalence ratio from the radical that produce these radicals are shown in Table 1
luminescence of the flame. By detecting the [11]-[14]. If the equivalence ratio is altered, the
self-emitted light from the flame with a miniature probe, equilibrium state of the elementary reactions changes.
it is possible to make simple equivalence ratio The light emission intensity also changes due to the
measurements in an actual engine. Many examples differences in the amount of each radical that is
of the application of this method have been reported produced and quenched mainly by inert components
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

as N2. Therefore, it may be possible to estimate the By subtracting the estimated background from the
equivalence ratio from the light emission intensity of original spectrum, intensity only of the line spectrum
the radicals. was evaluated. Figure 4 shows the light intensity over
crank angle in the spectra from OH*, CH*, and C2*.
㻯㻴㻠㻙㼍㼕㼞㻌㼜㼞㼑㼙㼕㼤㼑㼐 Light is emitted from CH* and C2* during the same
㻝㻜㻜㻜 period, but the light from OH* is emitted relatively later.
䃥 㻩㻌㻜㻚㻥
㻱㼙㼕㼟㼟㼕㼛㼚㻌㼕㼚㼠㼑㼚㼟㼕㼠㼥㻌㻔㼍㻚㼡㻚㻕
㻝㻚㻝 㻯㻴㻖㻔㻜㻘㻜㻕 This is probably because CH* and C2* are mainly only
㻝㻚㻞 㻯㻞㻖㻔㻜㻘㻜㻕 produced in the reaction zone of the flame front and
㻝㻚㻟 are only observed when the flame front passes
through the measurement point, while OH* is
㻻㻴㻖㻔㻜㻘㻜㻕 produced over a wide area from the reaction zone to
the high-temperature burned region[7]. On the other
㻯㻞㻖㻔㻝㻘㻜㻕 * *
hand, luminous intensity ratio C2 /CH is almost
* *
constant in the detecting time periods of CH and C2 .
* *
㻜 In this research, C2 /CH were used to estimate the
㻠㻜㻜 㻠㻡㻜
㻟㻜㻜 㻡㻜㻜
㻟㻡㻜 equivalence ratio of the flame front during combustion.
㼃㼍㼢㼑㼘㼑㼚㼓㼠㼔㻌㻔㼚㼙㻕
Figure 1 Spatially resolved flame spectra at different
equivalence ratios [2]
㻿㼜㼍㼞㼗㻌㼜㼘㼡㼓
* * * 㻵㼚㼠㼍㼗㼑 㻼㼞㼑㼙㼕㼤㼑㼐㻌㼒㼡㼑㼘㻙㼍㼕㼞 㼂㼍㼘㼢㼑
Table 1 Major OH , CH , C2 excitation kinetics 㼂㼕㼟㼡㼍㼘㼕㼦㼑㼐㻌㼑㼚㼓㼕㼚㼑
*
CH + O2  OH + CO 䠄1䠅 㻝㼙㼙
* 㻱㼤㼔㼍㼡㼟㼠
C2H + O  CH + CO 䠄2䠅
*
C2H + O2  CH + CO2 䠄3䠅
*
CH2 + C  C2 + H2 䠄4䠅 㻳㼍㼟㻌㼍㼚㼍㼘㼥㼦㼑㼞㻌
㻲㼘㼍㼙㼑
㻽㼡㼍㼞㼠㼦㻌㼣㼕㼚㼐㼛㼣 㻹㼑㼍㼟㼡㼞㼑㼙㼑㼚㼠㻌㼜㼛㼕㼚㼠
BASIC EXAMINATION 㻸㼑㼚㼟 㻲㼘㼍㼙㼑㻌㼟㼜㼑㼏㼠㼞㼡㼙
㻿㼜㼑㼏㼠㼞㼛㼟㼏㼛㼜㼑
The inside of an engine cylinder is an unsteady field in
which the pressure and temperature are constantly 㻵㼚㼠㼑㼚㼟㼕㼒㼕㼑㼐㻌㻯㻯㻰㻌㼏㼍㼙㼑㼞㼍
changing, and the flame differs substantially from that
Figure 2 Experimental setup of basic examination
in the steady field e.g. Figure 1. Therefore, the
spectral behavior in an actual engine was investigated
and the validity of the current method was examined.

Figure 2 shows the visualized engine and in-cylinder Table 2 Engine specifications
flame spectra measuring system that were used in the 㻮㼛㼞㼑㻌㽢 㻿㼠㼞㼛㼗㼑 㻤㻢㽢㻤㻢㻌㻔㼙㼙㻕
experiment. The visualized engine specifications are 㻯㼛㼙㼜㼞㼑㼟㼟㼕㼛㼚㻌㼞㼍㼠㼕㼛 㻥㻚㻤
shown in Table 2. A homogenous premixed 㻵㼚㼠㼍㼗㼑㻌㼢㼍㼘㼢㼑㻌㼛㼜㼑㼚㻌㼠㼕㼙㼕㼚㼓 㻟㻡㻣䠄㼐㼑㼓㻌㻮㼀㻰㻯䠅
isooctane-air mixture is supplied from upstream of the 㻵㼚㼠㼍㼗㼑㻌㼢㼍㼘㼢㼑㻌㼏㼘㼛㼟㼑㻌㼠㼕㼙㼕㼚㼓 㻝㻝㻠䠄㼐㼑㼓㻌㻮㼀㻰㻯䠅
intake port and fired. During combustion, the light 㻱㼤㼔㼍㼡㼟㼠㻌㼢㼍㼘㼢㼑㻌㼛㼜㼑㼚㻌㼠㼕㼙㼕㼚㼓 㻡㻤㻜䠄㼐㼑㼓㻌㻮㼀㻰㻯䠅
emitted from the flame at a point 1 mm below the 㻱㼤㼔㼍㼡㼟㼠㻌㼢㼍㼘㼢㼑㻌㼏㼘㼛㼟㼑㻌㼠㼕㼙㼕㼚㼓 㻟㻢㻟䠄㼐㼑㼓㻌㻮㼀㻰㻯䠅
ground electrode of the spark plug passes through the
visualization window in the side wall of the combustion
chamber.
㻾㼑㼜㼞㼑㼟㼑㼚㼠㼍㼠㼕㼢㼑㻌㼜㼛㼕㼚㼠㼟㻌㼠㼛㻌㼑㼟㼠㼕㼙㼍㼠㼑㻌㼎㼍㼏㼗㼓㼞㼛㼡㼚㼐
This luminescence is detected by a spectroscope and
is broken into a spectrum. A CCD camera with an 㻱㼟㼠㼕㼙㼍㼠㼑㼐㻌㼎㼍㼏㼗㼓㼞㼛㼡㼚㼐㻌㼎㼥㻌㼘㼑㼍㼟㼠㻙㼟㼝㼡㼍㼞㼑㼟㻌㼙㼑㼠㼔㼛㼐
image intensifier then amplifies and records the 㻠㻜㻜
㻿㼕㼓㼚㼍㼘㻌㼕㼚㼠㼑㼚㼟㼕㼠㼥㻌䠄㼍㻚㼡㻚䠅

spectrum since the light intensity of each wavelength 㻸㼕㼚㼑㻌㼟㼜㼑㼏㼠㼞㼡㼙㻌㼕㼚㼠㼑㼚㼟㼕㼠㼥 㻯㻴㻖㻦㻌㻠㻟㻝㼚㼙

is quite weak. This process is repeated over 50 㻟㻜㻜
cycles and then averaged. Simultaneously, an
exhaust analyzer measures the average equivalence 㻯㻞㻖㻦㻌㻡㻝㻣㼚㼙
㻞㻜㻜 㻻㻴㻖㻦㻌㻟㻜㻢㼚㼙
ratio inside the cylinder from the exhaust composition.
㻝㻜㻜
Figure 3 shows the flame spectrum from actual
measurement. The line spectra for OH* (306 nm), 㻯㻻㻙㻻䠖 㻯㼛㼚㼠㼕㼚㼡㼛㼡㼟㻌㼟㼜㼑㼏㼠㼞㼡㼙
CH* (431 nm), and C2* (517 nm) are all above the 㻜
broad continuous spectrum (300 to 600 nm) caused 㻞㻡㻜 㻟㻜㻜 㻟㻡㻜 㻠㻜㻜 㻠㻡㻜 㻡㻜㻜 㻡㻡㻜
by the bond between CO and O in the burned gas. 㼃㼍㼢㼑㼘㼑㼚㼓㼠㼔㻌䠄㼚㼙䠅
This background radiation (continuous spectrum) is
estimated by least-squares method. Series of points Figure 3 Flame spectrum in cylinder at -15 deg ATDC
away from the line spectrum are used to estimate. (50-cycle average)
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

㻡㻜㻜 㻜㻚㻠 intensity ratio. Thus, it can be stated that it is

possible to estimate the equivalence ratio at the flame

㻿㼕㼓㼚㼍㼘㻌㼕㼚㼠㼑㼚㼟㼕㼠㼥㻌㻔㼍㻚㼡㻚㻕
㻻㻴㻖
㻠㻜㻜 front using the obtained curve in Figure 5. The
㻜㻚㻟
㻯㻴㻖 relationship is dependent on fuel and wavelength

㻯㻞㻖㻛㻯㻴㻖
㻟㻜㻜 transmission characteristics of the system.
㻯㻞㻖㻛㻯㻴㻖 㻜㻚㻞
㻞㻜㻜
㻯 㻞㻖 DEVELOPMENT OF SPATIAL DISTRIBUTION
㻜㻚㻝 MEASUREMENT METHOD FOR ACTUAL
㻝㻜㻜
ENGINES
㻜 㻜
㻙㻟㻜 㻙㻞㻜 㻙㻝㻜 㻜 㻝㻜 㻞㻜 㻟㻜 MEASURING DEVICE
㻯㼞㼍㼚㼗㻌㼍㼚㼓㼘㼑㻌㻔㼐㼑㼓㻚㻭㼀㻰㻯㻕
㻝㻞㻜㻜㼞㼜㼙㻌䃥䠖 㻝㻚㻜㻌㻿㼜㼍㼞㼗㻌㼍㼐㼢㼍㼚㼏㼑䠖 㻙㻟㻤㼐㼑㼓㻌㻭㼀㻰㻯 Figure 7 shows the configuration of the developed
㻵㻹㻱㻼䠖 㻜㻚㻞㻝㻹㻼㼍㻌㻱㼤㼜㼛㼟㼡㼞㼑㻌㼠㼕㼙㼑䠖 㻟㻥㻜䃛㼟 measuring device. In order to capture the light
* * emitted by the flame from multiple directions simply by
Figure 4 In-cylinder emission intensity of OH , CH , and replacing the spark plug, a commercially available
*
C2 (50-cycle average) spark plug-type sensor with built-in optical fibers was
utilized. The optical fibers in this sensor detect light
Figure 5 shows the relationship between C2*/CH* and emitted by the flame from 32 circumferential directions
the equivalence ratio I when the equivalence ratio is inside the cylinder with a fixed solid angle. By
changed. An almost linear relationship is obtained rearranging the light output from the 32 optical fibers
over the approximately range from I = 0.8 to 1.4. in a line (in the vertical direction) at the entry slit of the
Figure 6 shows the change in C2*/CH* when the spectroscope, it is possible to output an image
pressure in the cylinder is changed by load and the containing both the position information in the vertical
temperature by intake air heating. The pressure and direction and the wavelength information in the
temperature when CH* and C2* reach their maxima horizontal direction. In order to measure the
were selected as the representative values. The equivalence ratio of each cycle from the radical
results show that C2*/CH* is almost constant, with a luminescence, a high speed camera coupled with an
fluctuation within the usage range of the pressure and image intensifier was used to take continuous shots of
* *
temperature. Signal intensities of CH and C2 are the image for 100 cycles at an exposure time of 3 deg
influenced by temperature and pressure. However, CA period synchronized with the combustion duration.
the influence is canceled by making it to the luminous This exposure time can be applied without depending
on the engine speed, because light intensity per crank
angle is almost constant even if the speed varies.

㻜㻚㻠

㻜㻚㻟
㻯㻞㻖㻛㻯㻴㻖

㻱㼚㼓㼕㼚㼑
㻜㻚㻞 㻻㼜㼠㼕㼏㼍㼘㻌㼒㼕㼎㼑㼞
㻿㼘㼕㼠
㻜㻚㻝 㻿㼜㼑㼏㼠㼞㼛㼟㼏㼛㼜㼑
㻝㼏㼔㻚
㻜
㻜㻚㻣 㻜㻚㻤 㻜㻚㻥 㻝 㻝㻚㻝 㻝㻚㻞 㻝㻚㻟 㻝㻚㻠 㻝㻚㻡 㻵㼙㼍㼓㼑㻌㼕㼚㼠㼑㼚㼟㼕㼒㼕㼑㼐㻌㼔㼕㼓㼔
㻱㼝㼡㼕㼢㼍㼘㼑㼚㼏㼑㻌㼞㼍㼠㼕㼛㻌䃥 * 㻟㻞㼏㼔㻚 㼟㼜㼑㼑㼐㻌㼢㼕㼐㼑㼛㻌㼏㼍㼙㼑㼞㼍
*
Figure 5 Relationship between C2 /CH and
equivalence ratio 㻞㻡㻜䃛㼙
㻿㼜㼍㼞㼗㻌㼜㼘㼡㼓㻌㼠㼥㼜㼑㻌㼟㼑㼚㼟㼛㼞

㻜㻚㻠 㻟㻞㼏㼔㻚㻌㼒㼕㼎㼑㼞㼟
䃥䠖 㻝㻚㻟㻣 㻼㼛㼟㼕㼠㼕㼛㼚㻌㼛㼒㻌㼒㼕㼎㼑㼞㼟
㻜㻚㻟
㻯㻞㻖㻛㻯㻴㻖

㻯㼞㼍㼚㼗㻌㼍㼚㼓㼘㼑
㻜㻚㻞 䃥䠖 㻜㻚㻥㻣 㻰㼑㼠㼑㼏㼠㼍㼎㼘㼑㻌㼞㼑㼓㼕㼛㼚

㻜㻚㻝
䃥䠖 㻜㻚㻤㻞
㻜 㼃㼍㼢㼑㼘㼑㼚㼓㼠㼔
㻢㻡㻜 㻣㻡㻜 㻤㻡㻜 㻥㻡㻜 㻜㻚㻣 㻝㻚㻝 㻝㻚㻡 㻝㻚㻥 Figure 7 Multi channel flame spectra analyzing system
㼀㼑㼙㼜㼑㼞㼍㼠㼡㼞㼑㻌 㻼㼞㼑㼟㼟㼡㼞㼑
㼕㼚㻌㼠㼔㼑㻌㼏㼥㼘㼕㼚㼐㼑㼞㻌㻔㻷㻕 㼕㼚㻌㼠㼔㼑㻌㼏㼥㼘㼕㼚㼐㼑㼞㻌㻔㻹㻼㼍㻕
* *
Figure 6 C2 /CH vs. temperature and pressure
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018
* *
Figure 8 shows the relationship between C2 /CH of
the constructed system and the equivalence ratio. 㼂㼍㼘㼢㼑
* *
C2 /CH is larger than that of Figure 2, since there is a 㻱㼤㼔㼍㼡㼟㼠 㻵㼚㼠㼍㼗㼑㻌
㼟㼕㼐㼑 㼟㼕㼐㼑
difference of the wavelength characteristics which is 㻿㼑㼚㼟㼛㼞
due to the sensor and the grating of spectroscope.
The error bars indicate the measurement uncertainty
of the system which depends on the signal intensity in 㻰㼑㼠㼑㼏㼠㼍㼎㼘㼑
each I. The uncertainty isr5% max in the range of 㼞㼑㼓㼕㼛㼚㻌
I 1.1-1.6. 㻣㻜㻑㻌㼛㼒㻌㼎㼛㼞㼑㻌 㻟㻞㻌㼏㼔㻚
㻼㼕㼟㼠㼛㼚㻌㼍㼠㻌㼀㻰㻯 㻮㼛㼞㼑
㻜㻚㻣
㻜㻚㻢 㼼䃢
Figure 9 Detectable region of sensor
㻜㻚㻡
㻯㻞㻖㻛㻯㻴㻖

㻜㻚㻠
㻜㻚㻟
㻜㻚㻞
㻜㻚㻝 㻸㼕㼓㼔㼠㻌㼠㼛㻌㼛㼜㼠㼕㼏㼍㼘㻌㼒㼕㼎㼑㼞 㻱㼟㼠㼕㼙㼍㼠㼑㼐㻌㼜㼛㼟㼕㼠㼕㼛㼚㻌
㻜 㼛㼒㻌㼒㼘㼍㼙㼑㻌㼒㼞㼛㼚㼠
㻜㻚㻣 㻜㻚㻤 㻜㻚㻥 㻝 㻝㻚㻝 㻝㻚㻞 㻝㻚㻟 㻝㻚㻠 㻝㻚㻡 㻝㻚㻢 㻝㻚㻣 㻝㻚㻤
㻿㼑㼚㼟㼛㼞 㻲㼘㼍㼙㼑㻌㼜㼞㼛㼜㼍㼓㼍㼠㼕㼛㼚
㻱㼝㼡㼕㼢㼍㼘㼑㼚㼏㼑㻌㼞㼍㼠㼕㼛㻌䃥
* *
Figure 8 Relationship between C2 /CH and equivalence 㻰㼑㼠㼑㼏㼠㼍㼎㼘㼑㻌㼞㼑㼓㼕㼛㼚
ratio㻌 of the developed system 㼛㼒㻌㼑㼍㼏㼔㻌㼒㼕㼎㼑㼞

ESTIMATION OF FLAME FRONT POSITION 㻰㼑㼠㼑㼏㼠㼕㼛㼚㻌㼜㼑㼞㼕㼛㼐㻌㼛㼒㻌㻯㻴㻖㻌㼑㼙㼕㼟㼟㼕㼛㼚㻌

To measure the spatial distribution of mixture using

the equivalence ratio at the flame front position, the 㻯㻴㻖㻌㼑㼙㼕㼟㼟㼕㼛㼚㻌
position of the flame front should first be measured. 㻵㼚㼠㼑㼚㼟㼕㼠㼥 㻰㼑㼠㼑㼏㼠㼕㼚㼓 㻰㼑㼠㼑㼏㼠㼕㼚㼓
㼟㼠㼍㼞㼠 㼑㼚㼐
Figure 9 shows the region inside the cylinder that is
observable by the sensor. Each of the optical fiber 㻯㼞㼍㼚㼗㻌㼍㼚㼓㼘㼑㻌
can detect light in an annular region (approximately
70% of the bore diameter when the piston is at top Figure 10 Estimation of flame front position from
*
dead center) with the spark plug at the center. CH emission intensity

Figure 10 shows a conceptual diagram of the

estimation of the flame front position. In the current
method, the period over which the flame front passed
through each of the circumferential detectable region
was first obtained, based on the period from the start 㼂㼕㼟㼡㼍㼘㼕㼦㼑㼐㻌㼑㼚㼓㼕㼚㼑
to the end of detection of CH* emissions from the 㻭㼠㻌‒㻝㻠㻚㻥㻌㼐㼑㼓㻌㻭㼀㻰㻯㻌
㻲㼘㼍㼙㼑
flame front. In this work, the flame propagation
speed within each detectable region could be 㻽㼡㼍㼞㼠㼦㻌㼣㼕㼚㼐㼛㼣
calculated from the length of the region and the period
that the flame passes it by assuming constant speed. 㻱㼘㼛㼚㼓㼍㼠㼑㼐㻌㼜㼕㼟㼠㼛㼚
Then, the movement of the flame front was estimated
over a series of time. Meanwhile, the period 㻴㼕㼓㼔㻌㼟㼜㼑㼑㼐㻌
between the start and end of CH* emission was 㻹㼕㼞㼞㼛㼞
㼢㼕㼐㼑㼛㻌㼏㼍㼙㼑㼞㼍 㻹㼑㼍㼟㼡㼞㼑㼐㻌㼒㼘㼍㼙㼑㻌
assumed to be the timing at 25% of the peak CH*
㻮㼛㼠㼠㼛㼙㻌㼢㼕㼑㼣 㻱㼟㼠㼕㼙㼍㼠㼑㼐㻌㼒㼘㼍㼙㼑㻌㼒㼞㼛㼚㼠㻌㼜㼛㼟㼕㼠㼕㼛㼚
emission intensity.
㻱㼤㼔㼍㼡㼟㼠 㻵㼚㼠㼍㼗㼑 㻭㼠㻌‒㻤㻚㻥㻌㼐㼑㼓㻌㻭㼀㻰㻯㻌
A visualized engine and direct photography of the 㼟㼕㼐㼑 㼟㼕㼐㼑
flame were used for verification. Figure 11 compares
the measured flame with the estimated front position.
The 32 signals were grouped into 8 directions from the
standpoint of signal noise ratio (S/N). Then the flame
front in each direction was interpolated with straight 㻰㼑㼠㼑㼏㼠㼍㼎㼘㼑㻌㼞㼑㼓㼕㼛㼚
lines. The estimated position was almost consistent 㼛㼒㻌㼟㼑㼚㼟㼛㼞
㻻㼎㼟㼑㼞㼢㼍㼠㼕㼛㼚㻌㼍㼞㼑㼍㻌䃥㻢㻜㻌㼙㼙㻌
with the measured flame front. And, it was confirmed
that the current method ensures sufficient practical 㻝㻞㻜㻜㼞㼜㼙㻌䃥䠖 㻝㻚㻜㻌㻿㼜㼍㼞㼗㻌㼍㼐㼢㼍㼚㼏㼑䠖 㻙㻟㻤㼐㼑㼓㻌㻭㼀㻰㻯
precision for the purpose of simple measurement of 㻵㻹㻱㻼䠖 㻜㻚㻞㻝㻹㻼㼍
spatial distribution of the equivalence ratio within the
detectable region. Figure 11 Verification of estimated flame front position
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

MEASUREMENT REGION OF SPATIAL After conducting tests using these three cases, a
DISTRIBUTION OF EQUIVALENCE RATIO comparison was made between the mixture
distributions obtained using LIF and the equivalence
Figure 12 shows the measurement region of the ratio from the current study. Figure 13 shows the
equivalence ratio distribution when the piston is at top experimental system of LIF. The fuel was mixed with
dead center. The distribution was identified by a fluorescer (diethyl ketone) and was made to
dividing the radial direction into 3 to 6 and fluoresce by an excimer laser (308 nm) emitting a
circumferential direction into 24 to 48 divisions due to sheet-shaped beam. This was photographed with an
the resolution of the estimated flame front position. ICCD camera and the fuel distribution was obtained.
Furthermore, these regions were redefined depending Calibration for LIF was done with premixed fuel-air.
on the position of the piston. At 35 deg ATDC and Linear relationship between I and LIF signal was
after, observation up to the bore wall was possible. confirmed in I 0.9-1.5.

㻱㼤㼔㼍㼡㼟㼠 㻵㼚㼠㼍㼗㼑㻌
㼟㼕㼐㼑 㼟㼕㼐㼑 㼂㼕㼟㼡㼍㼘㼕㼦㼑㼐㻌㼑㼚㼓㼕㼚㼑 㻵㼚㼖㼑㼏㼠㼛㼞
㻭㼚㼓㼘㼑䠖 㻤㻌㼐㼕㼢㼕㼟㼕㼛㼚㼟 㻲㼡㼑㼘㻌㻗㻌㻲㼘㼡㼛㼞㼑㼟㼏㼑㼞
㻿㼑㼚㼟㼛㼞
㽢
㻾㼍㼐㼕㼡㼟䠖 㻟䡚㻢㻌㼐㼕㼢㼕㼟㼕㼛㼚㼟 㻸㼍㼟㼑㼞㻌㼟㼔㼑㼑㼠
㻮㼛㼠㼠㼛㼙㻌㼢㼕㼑㼣

䠙
㻹㼑㼍㼟㼡㼞㼑㼙㼑㼚㼠 㼀㼛㼠㼍㼘㻌㻌㻞㻠䡚㻠㻤㻌㼐㼕㼢㼕㼟㼕㼛㼚㼟 㻱㼤㼔㼍㼡㼟㼠㻌 㻵㼚㼠㼍㼗㼑㻌
㼞㼑㼓㼕㼛㼚㻌 㻽㼡㼍㼞㼠㼦 㼟㼕㼐㼑 㼟㼕㼐㼑
㼣㼕㼚㼐㼛㼣
Figure 12 Measurement region of equivalence
ratio distribution

㻹㼕㼞㼞㼛㼞 㻵㻯㻯㻰㻌㼏㼍㼙㼑㼞㼍 㻻㼎㼟㼑㼞㼢㼍㼠㼕㼛㼚㻌㼍㼞㼑㼍

Figure 13 Laser induced fluorescence method
VERIFICATION OF EQUIVALENCE RATIO
DISTRIBUTION
Figure 14(a) shows the fuel distributions immediately
The distribution of the mixture within a DISI engine after ignition obtained using LIF. In Case A the fuel
was artificially modified for the verification. The engine is unevenly distributed on the intake valve side, while
specifications were the same as those in Table 2. in Case B the fuel is unevenly distributed on the
The fuel injection angle and injection timing were exhaust valve side, and in Case C the fuel has
changed to create the three cases shown in Table 3. become almost homogeneously distributed over both
sides of the cylinder. For comparison, Figure 14(b)
shows the equivalence ratio distributions measured at
Table 3 Engine operation conditions the time of combustion with the newly developed
method. Even though the measurement regions and
㻯㼍㼟㼑㻌㻭 periods were different for both methods, the
㻵㼚㼖㼑㼏㼠㼛㼞 㻵㼚㼖㼑㼏㼠㼕㼛㼚㻌㼠㼕㼙㼕㼚㼓㻌 qualitative distributions tended to correspond to each
㻙㻤㻜㻌㼐㼑㼓㻌㻭㼀㻰㻯 other in all three cases (A, B, and C). In Figure 15,
㻠㻥㼻 㻿㼜㼍㼞㼗㻌㼍㼐㼢㼍㼚㼏㼑 the equivalence ratios from both methods are
㻲㼡㼑㼘 㻙㻝㻢㻌㼐㼑㼓㻌㻭㼀㻰㻯㻌 presented along the line(X) within the dotted circle. A
㻯㼍㼟㼑㻌㻮 good quantitative correspondence was found,
㻵㼚㼖㼑㼏㼠㼕㼛㼚㻌㼠㼕㼙㼕㼚㼓㻌 particularly in the area near the sensor, thus
㻙㻟㻜㻜㻛㻢㻜㻌㼐㼑㼓㻌㻭㼀㻰㻯
㻝㻣㼻 confirming the validity of the results obtained using the
㻿㼜㼍㼞㼗㻌㼍㼐㼢㼍㼚㼏㼑
㻼㼕㼟㼠㼛㼚
㻙㻞㻝㻌㼐㼑㼓㻌㻭㼀㻰㻯㻌
developed method.
㻯㼍㼟㼑㻌㻯
㻵㼚㼖㼑㼏㼠㼕㼛㼚㻌㼠㼕㼙㼕㼚㼓㻌
㻙㻟㻜㻜㻌㼐㼑㼓㻌㻭㼀㻰㻯
㻠㻥㼻 㻿㼜㼍㼞㼗㻌㼍㼐㼢㼍㼚㼏㼑
㻙㻟㻜㻌㼐㼑㼓㻌㻭㼀㻰㻯㻌
㻝㻞㻜㻜㻌㼞㼜㼙㻌㻌㻵㻹㻱㻼䠖 㻜㻚㻞㻟㻌㻹㻼㼍 㻲㼡㼑㼘䠖 㼕㼟㼛㻙㼛㼏㼠㼍㼚㼑
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

APPLICATION TO CYCLIC VARIATION IN DISI

ENGINE

㻔㻙㻕 㻔㻗㻕 㻔㻙㻕 㻔㻗㻕 The developed method can be applied to the analysis
㻜 㼄 㻜 㼄 of combustion fluctuations in a DISI engine. When a
㻼㼕㼟㼠㼛㼚㻌㼍㼠㻌㼀㻰㻯 DISI engine is started, the fuel is injected in the middle
㻯㼍㼟㼑㻌㻭 of compression to retard the ignition timing, which
㻱㼤㼔㼍㼡㼟㼠 㻵㼚㼠㼍㼗㼑
㼟㼕㼐㼑 㻜 㼟㼕㼐㼑 㻜 results in the fast activation of the catalyst fine to the
increased exhaust temperature. Figure 16 shows
the fuel spray behavior during compression stroke
㻔㻙㻕 㻔㻗㻕 㻔㻙㻕 㻔㻗㻕
㼄 㼄 predicted by CFD. The fuel spray moves along the
top of the piston. It swirls up at the cavity and the
mixture gathers around the spark plug. As time
elapses, it diffuses toward the direction of the exhaust
㻙㻝㻞㻌㼐㼑㼓㻌㻭㼀㻰㻯
side. One key factor in realizing stable combustion is
㻯㼍㼟㼑㻌㻮 to generate the preferable mixture distribution and to
control its fluctuation. The developed method was
utilized to analyze the causes of combustion
㼄 㼄 fluctuation, using its ability to figure out the mixture
distribution.

㻙㻝㻢㻌㼐㼑㼓㻌㻭㼀㻰㻯 The engine specifications are the same as those in

㻸㼑㼍㼚 䃥 㻸㼑㼍㼚 䃥 Table 2. Figure 17 shows the coefficient of variance
㻜㻚㻥 㻯㼍㼟㼑㻌㻯 㻜㻚㻥
㻝㻚㻜
(COV) of the indicated mean effective pressure
㻝㻚㻝 (IMEP) at different injection timings. The COV of the
㻝㻚㻝
㻝㻚㻞 IMEP is comparatively small in case D, but it is larger
㻝㻚㻞 㼄 㼄 㻝㻚㻟 in case E due to the retarded injection timing. To
㻝㻚㻟
㻝㻚㻠
analyze the differences in these fluctuations, Figure
㻝㻚㻠
18 compares the 100-cycle averages of the mixture
㻝㻚㻡 㻝㻚㻡
㻾㼕㼏㼔 㻙㻞㻠㻌㼐㼑㼓㻌㻭㼀㻰㻯 㻾㼕㼏㼔 distributions and flame propagation. In case D, the
mixture is rich in the area around the spark plug out to
(a) LIF (b) Current method the exhaust side. In case E, the mixture is even more
䠄20-cycles averaged䠅 䠄50-cycles averaged䠅 rich in this same area. This indicates that the more the
ignition timing is retarded, the more fuel gathers at the
Figure 14 Comparison of equivalence ratio
top of the combustion chamber.
distributions
To further examine the causes of combustion
fluctuation, analysis was conducted at each cycle.
When the developed method was applied, the
㻝㻚㻢 㻰㼑㼢㼑㼘㼛㼜㼑㼐㻌㼙㼑㼠㼔㼛㼐
combustion pressure over 100 cycles was also
㻝㻚㻞 measured and the IMEP was calculated from the
䃥

㻸㻵㻲 combustion pressure. Figure 19 shows these results.

㻜㻚㻤 A comparison of the results from case D and E
㻯㼍㼟㼑㻌㻭 indicated several cycles where the IMEP suddenly
㻜㻚㻠
㻝㻚㻢
became small and the fluctuations became larger.
Figure 20 compares the heat release rates in case E
㻝㻚㻞 for the best and worst four cycles in terms of IMEP.
䃥

In the worst four cycles the rise of the initial heat

㻜㻚㻤 release rate is delayed and the decline in IMEP is
㻯㼍㼟㼑㻌㻮
㻜㻚㻠 thought to be the cause of the differences in the initial
combustion.
㻝㻚㻢

㻝㻚㻞
䃥

㻱㼤㼔㼍㼡㼟㼠㻌 㻵㼚㼖㼑㼏㼠㼛㼞 㻵㼚㼠㼍㼗㼑㻌

㻜㻚㻤 㼟㼕㼐㼑 㼟㼕㼐㼑
㻯㼍㼟㼑㻌㻯 㻙㻝㻢㼐㼑㼓㻌㻭㼀㻰㻯
㻜㻚㻠
㻔㻙㻕 㻜 㻔㻗㻕
㻰㼕㼟㼠㼍㼚㼏㼑㻌㼄 䃥
Figure 15 Comparison of equivalence ratio in LIF and 㻠㼐㼑㼓㻌㻭㼀㻰㻯 㻞㻚㻜
developed method 㻝㻚㻢
㻝㻚㻞
㻜㻚㻤
㻝㻠㼐㼑㼓㻌㻭㼀㻰㻯 㻜㻚㻠
㻜

Figure 16 Behavior of fuel spray

in compression stroke injection (CFD)
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

Figure 21 shows the mixture distribution and flame

propagation of the worst four and best four cycles,
㻝㻞㻜㻜㼞㼜㼙㻌䃥䠖 㻜㻚㻥㻣 along with the frequency distribution of the
㻿㼜㼍㼞㼗㻌㻭㼐㼢㼍㼚㼏㼑䠖 㻝㻜㻌㼐㼑㼓㻌㻭㼀㻰㻯
equivalence ratio, I In the worst four cycles, an
㻞㻜 excessively rich mixture, where I locally exceeds 1.6,
㻯㼍㼟㼑㻌㻱
㻵㻹㻱㻼㻌㻯㻻㼂㻌㻔㻑㻕 㻝㻢 exists around the spark plug. In contrast, in the best
㻝㻞 four cycles, a mixture with I = approximately 1.2 to 1.5
㻯㼍㼟㼑㻌㻰 was distributed almost without excessively rich
㻤 mixture. Consequently, it is thought that the
㻠 generation of local and excessively rich mixture
around the spark plug degrades the initial combustion
㻜
㻙㻡㻜 㻙㻝㻡 and is one of the major causes of the increase in
㻵㼚㼖㼑㼏㼠㼕㼛㼚㻌㼠㼕㼙㼕㼚㼓㻌㻔㼐㼑㼓㻌㻭㼀㻰㻯㻕 fluctuations when the ignition timing is retarded.
Figure 17 Comparison of IMEP COV
CONCLUSION
㻱㼤㼔㼍㼡㼟㼠 㻯㼍㼟㼑㻌㻰 㻵㼚㼠㼍㼗㼑 㻯㼍㼟㼑㻌㻱
䃥 The one-point measurement of the equivalence ratio
㼟㼕㼐㼑 㼟㼕㼐㼑 㻜㻚㻥 from the radical luminescence in the cylinder was
㻞㻞㻭㼀㻰㻯 㻞㻞㻭㼀㻰㻯 extended to the spatial distribution using a spark
㻠㻢㻭㼀㻰㻯 㻠㻢㻭㼀㻰㻯 㻝㻚㻞 plug-type sensor with built-in optical fibers. The
conclusions drawn from the analysis are summarized
㻝㻚㻡 below:

㻝㻚㻤
㻱㼟㼠㼕㼙㼍㼠㼑㼐㻌㼒㼘㼍㼙㼑㻌 㻮㼛㼞㼑 㻿㼑㼚㼟㼛㼞 (1) A simple and effective method to measure the
㼒㼞㼛㼚㼠㻌㼜㼛㼟㼕㼠㼕㼛㼚 distribution of equivalence ratio was developed by
combing the multi-point C2*, CH* measurement and
Figure 18 Equivalence ratio distributions and flame front estimation.
flame propagation (100-cycle average)
(2) The validity of these measurements in an actual
㻜㻚㻟 engine was verified from the points described below:
㻵㻹㻱㻼㻌㻔㻹㻼㼍㻕

㻜㻚㻞 -It appeared that the ratio of radical luminescence

from C2* and CH* was not noticeably affected by a
change in in-cylinder pressure and temperature.
㻜㻚㻝
- The flame front position estimated throughout the
㻯㼍㼟㼑㻌㻰
㻜 luminescent periods of CH* coincided well with the
㻮㼑㼟㼠㻌㻠㻌㼏㼥㼏㼘㼑㼟 results from the visualization.
㻜㻚㻟
- The distribution of equivalence ratio measured by
㻵㻹㻱㻼㻌㻔㻹㻼㼍㻕

㻜㻚㻞 current method agreed qualitatively with most

observations by LIF, despite the differences in
measured region and timing.
㻜㻚㻝
㼃㼛㼞㼟㼠㻌㻠㻌㼏㼥㼏㼘㼑㼟 (3) The cyclic variation of DISI engine on start-up
㻯㼍㼟㼑㻌㻱
㻜 was investigated as an application of current method.
It was found that maldistribution of excessively rich
㻜 㻞㻜 㻠㻜 㻢㻜 㻤㻜 㻝㻜㻜
mixture may result in poor formation of early flame.
㻯㼥㼏㼘㼑㼟
Figure 19 Cyclic variations of IMEP
Further works are under consideration for more
detailed analysis, especially on the improvements in
㻞㻜 special resolution and observable area.
㻴㼑㼍㼠㻌㼞㼑㼘㼑㼍㼟㼑㻌㼞㼍㼠㼑㻌㻔㻶㻛㼐㼑㼓㻚㻕

㻯㼍㼟㼑㻌㻱 㻮㼑㼟㼠㻌㻠㻙㼏㼥㼏㼘㼑㻌㼍㼢㼑㼞㼍㼓㼑
㻝㻡
㻝㻜㻜㻙㼏㼥㼏㼘㼑㻌㼍㼢㼑㼞㼍㼓㼑
㻝㻜

㻜
㼃㼛㼞㼟㼠㻌㻠㻙㼏㼥㼏㼘㼑㻌㼍㼢㼑㼞㼍㼓㼑
㻙㻡
㻜 㻟㻜 㻢㻜 㻥㻜 㻝㻞㻜
㻯㼞㼍㼚㼗㻌㼍㼚㼓㼘㼑㻌㻔㼐㼑㼓㻌㻭㼀㻰㻯㻕
Figure 20 Comparison of heat release rate
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

㻯㼍㼟㼑㻌㻱 㼃㼛㼞㼟㼠㻌㻠㻌㼏㼥㼏㼘㼑㼟㻌
㻠㻜㻭㼀㻰㻯 㻠㻜㻭㼀㻰㻯 㻟㻣㻭㼀㻰㻯 㻠㻜㻭㼀㻰㻯
㻢㻝㻭㼀㻰㻯 㻣㻟㻭㼀㻰㻯 㻢㻝㻭㼀㻰㻯 㻢㻣㻭㼀㻰㻯

䃥䠖㻝㻚㻢
㻜㻚㻠 㻜㻚㻠 㻜㻚㻠 㻜㻚㻠
㻜㻚㻟 㻜㻚㻟 㻜㻚㻟 㻜㻚㻟

㻼㻚㻰㻚㻲㻚

㻼㻚㻰㻚㻲㻚
㻼㻚㻰㻚㻲㻚

㻼㻚㻰㻚㻲㻚
㻜㻚㻞 㻜㻚㻞 㻜㻚㻞 㻜㻚㻞
㻜㻚㻝 㻜㻚㻝 㻜㻚㻝 㻜㻚㻝
㻜 㻜 㻜 㻜
㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥
䃥 䃥 䃥 䃥
㻯㼍㼟㼑㻌㻱 㻮㼑㼟㼠㻌㻠㻌㼏㼥㼏㼘㼑㼟
㻞㻞㻭㼀㻰㻯 㻝㻥㻭㼀㻰㻯 㻝㻥㻭㼀㻰㻯 㻝㻥㻭㼀㻰㻯 㻸㼑㼍㼚䃥
㻜㻚㻥
㻠㻜㻭㼀㻰㻯 㻟㻠㻭㼀㻰㻯 㻟㻝㻭㼀㻰㻯 㻟㻝㻭㼀㻰㻯
㻝㻚㻞

㻝㻚㻡

䃥䠖㻝㻚㻢 㻝㻚㻤
㻾㼕㼏㼔
㻜㻚㻠 㻜㻚㻠 㻜㻚㻠 㻜㻚㻠
㻜㻚㻟 㻜㻚㻟 㻜㻚㻟 㻜㻚㻟
㻼㻚㻰㻚㻲㻚

㻼㻚㻰㻚㻲㻚

㻼㻚㻰㻚㻲㻚
㻼㻚㻰㻚㻲㻚

㻜㻚㻞 㻜㻚㻞 㻜㻚㻞 㻜㻚㻞

㻜㻚㻝 㻜㻚㻝 㻜㻚㻝 㻜㻚㻝
㻜 㻜 㻜 㻜
㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥 㻜㻚㻥 㻝㻚㻝 㻝㻚㻟 㻝㻚㻡 㻝㻚㻣 㻝㻚㻥
䃥 䃥 䃥 䃥
Figure 21 Cyclic variations of equivalence ratio distributions and flame propagation

REFERENCES
10. N IIDA, et al. ”Study of Radical Luminescence of
1. Taketoshi Fujikawa, et al. “2-D Temperature
Premixed Flames - Equivalence Ratio
Measurements of Unburned Gas Mixture in an
Dependence on Radical Luminescence Spectra
Engine by Two-line Excitation LIF Technique”
from Gasoline, Methanol and Methane Flames
SAE Paper 2006-01-3336, 2006
-“ 12th International Symposium of Alcohol Fuels,
2. Jun Kojima, et al. “Spatially Resolved of the
1998
Reaction Zone of Laminar Methane/Air Premixed
11. Devriendt K, et al. “Kinetics of formation of
Flames” Proc. Combust. Inst. 28, 2000
chemiluminescent CH(A2Ǽ) by the elementary
3. Yuji Ikeda, et al. “In-spark-plug Sensor for
reactions of C2H(X2 Ȋ+) with O(3P) and O2(X3
Analyzing the Initial Flame and Its Structure in an
Ȋ g): A pulse laser photolysis study” Chem.
SI Engine” SAE Paper 2005-01-0644, 2005
Phys. Lett., 1996
4. Nobuyuki Kawahara, et al. “Measurement of
12. M Tamura, et al. “Collisional Quenching of CH(A),
Flame Propagation Characteristics in an SI
OH(A), and NO(A) in Low Pressure Hydrocarbon
Engine Using Micro-Local Chemiluminescence
Flames” Comb. Flame, 1998
Technique” SAE Paper 2005-01-0645, 2005
13. M. R. Berman, et al. “Temperature dependence
5. Jacek Hunicz, et al. “Transient In-Cylinder AFR
of CH radical reactions with O2, NO, CO and
Management Based on Optical Emission Signals”
CO2” Symposium on combustion 19, 1982
SAE Paper 2004-01-0516, 2004
14. J Grebe, K. H. Homann “Kinetics of the Species
6. S. Longdill, et al. “Investigation into Air-Fuel Ratio
OH(A2 Ȋ +), OH(X2 Ȉ and CH(X2 Ȉ ) in the
Measurement of a High Performance Two-Stroke
System C2H2/O/H” Ber. Bunsenges. phys. Chem.
Engine by an Optical Method” SAE Paper
86, 1982
2002-01-0741, 2002
7. T Chou, D J. Patterson. “In-Cylinder
Measurement of Mixture Maldistribution in a CONTACT
L-Head Engine” Comb. flame 1995
8. K Sohoma, et al. “Application of Rapid Optical Koshiro Kimura: [email protected]
Measurement to Detect the Fluctuations of the
Advanced Power Train Engineering Div. 2 R&D Group
Air-Fuel Ratio and Temperature of a Spark
2, Toyota Motor Corporation Higashifuji Technical
Ignition Engine” SAE Paper 910499, 1991
Center 1200, Misyuku, Susono, Shizuoka, 410-1193
9. H Sawada, et al. “The Method of Measuring
Japan
Air-Fuel Ratio by Radical Luminescence in High
Combustion Pressure” SAE Paper 1999-01-0507,
1999
 Downloaded from SAE International by Imperial College London, Monday, September 24, 2018

$OOULJKWVUHVHUYHG1RSDUWRIWKLVSXEOLFDWLRQPD\EHUHSURGXFHGVWRUHGLQD 3RVLWLRQVDQGRSLQLRQVDGYDQFHGLQWKLVSDSHUDUHWKRVHRIWKHDXWKRUVDQGQRW

UHWULHYDOV\VWHPRUWUDQVPLWWHGLQDQ\IRUPRUE\DQ\PHDQVHOHFWURQLFPHFKDQLFDO QHFHVVDULO\WKRVHRI6$(7KHDXWKRULVVROHO\UHVSRQVLEOHIRUWKHFRQWHQWRIWKHSDSHU
SKRWRFRS\LQJUHFRUGLQJRURWKHUZLVHZLWKRXWWKHSULRUZULWWHQSHUPLVVLRQRI6$( 6$(&XVWRPHU6HUYLFH
7HOLQVLGH86$DQG&DQDGD
,661
7HORXWVLGH86$
)D[
(PDLO&XVWRPHU6HUYLFH#VDHRUJ
6$(:HE$GGUHVVKWWSZZZVDHRUJ
3ULQWHGLQ86$