Misfire Detection of Spark Ignition Engines Using a New Technique Based on Mean Output Power

Abstract

Control and detection of misfire are an essential part of on-board diagnosis (OBD) of modern spark ignition (SI) engines. This study proposes a novel model-based technique for misfire detection for a multicylinder SI engine. The new technique uses a dynamic engine model to determine mean output power, which is then used to calculate a new parameter for misfire detection. The new parameter directly relates to combustion period and is sensitive to engine speed fluctuations caused by misfire. The new technique requires only measured engine speed data and is computationally viable for use in a typical engine control unit (ECU). The new technique is evaluated experimentally on a four-cylinder 1.6-l SI engine. Three types of misfire are studied including single, continuous, and multiple-event. The steady-state and transient experiments were done for a wide range of engine speeds and engine loads, using a vehicle chassis dynamometer and on-road vehicle testing. The validation results show that the new technique is able to detect all three types of misfire with up to 94% accuracy during steady-state conditions. The new technique is augmented with a compensation factor to improve the accuracy of the technique for transient operations. The resulting technique is shown to be capable of detecting misfire during both transient and steady-state engine conditions.

Full text

M. Boudaghi
Department of Mechanical Engineering,
K. N. Toosi University of Technology,
Mollasadra Street,
Tehran 19991-43344, Iran
e-mail: mahdi_boodaghi@sina.kntu.ac.ir
M. Shahbakhti
Department of Mechanical
Engineering-Engineering Mechanics,
Michigan Technological University,
1400 Townsend Drive,
Houghton, MI 49931
e-mail: mahdish@mtu.edu
S. A. Jazayeri
Department of Mechanical Engineering,
K. N. Toosi University of Technology,
Mollasadra Street,
Tehran 19991-43344, Iran
e-mail: jazayeri@kntu.ac.ir
Misfire Detection of Spark
Ignition Engines Using a New
Technique Based on Mean
Output Power
Control and detection of misfire are an essential part of on-board diagnosis (OBD) of mod-
ern spark ignition (SI) engines. This study proposes a novel model-based technique for mis-
fire detection for a multicylinder SI engine. The new technique uses a dynamic engine
model to determine mean output power, which is then used to calculate a new parameter
for misfire detection. The new parameter directly relates to combustion period and is sensi-
tive to engine speed fluctuations caused by misfire. The new technique requires only meas-
ured engine speed data and is computationally viable for use in a typical engine control
unit (ECU). The new technique is evaluated experimentally on a four-cylinder 1.6-l SI
engine. Three types of misfire are studied including single, continuous, and multiple-event.
The steady-state and transient experiments were done for a wide range of engine speeds
and engine loads, using a vehicle chassis dynamometer and on-road vehicle testing. The
validation results show that the new technique is able to detect all three types of misfire
with up to 94%accuracy during steady-state conditions. The new technique is augmented
with a compensation factor to improve the accuracy of the technique for transient opera-
tions. The resulting technique is shown to be capable of detecting misfire during both tran-
sient and steady-state engine conditions. [DOI: 10.1115/1.4029914]
Introduction
Throughout the last three decades, engine control and monitor-
ing strategies have significantly improved to ensure desirable per-
formance of engines in various operating conditions. The OBD
regulations including OBD-II require that the ECU continuously
monitor engine behavior to prevent harmful emissions and iden-
tify defective parts. Among the strategies, misfire detection strat-
egy is an important part of engine OBD.
The word “misfire” is defined as a lack of in-cylinder combus-
tion, which occurs due to lack of ignition, too lean/rich air–fuel-
ratio, and cold-start, to name but a few. The effects of missing com-
bustion include a sudden decrease in output torque and a sudden
increase in unburned hydrocarbon emissions. Ignition of unburned
fuel by hot spots in the exhaust tailpipe can seriously damage the
catalytic convertor. According to the instruction by OBD-II, misfire
as well as faulty cylinders should be identified on-board.
In general, engines normally operate in a certain manner, in
terms of output torque, vibration level, and exhaust gas tempera-
ture variation, which changes by unexpected events such as mis-
fire. Different methods of misfire detection based on the impact
from missing combustion on engine behavior have been investi-
gated since 1980. Engine misfire detection (EMD) methods can be
grouped into two types including in-cylinder combustion diagno-
sis and postcylinder combustion diagnosis as shown in Fig. 1.
Typical in-cylinder combustion has three main phases including
delay period, rapid combustion or flame spread, and direct burning
of mixture [1]. These three phases can be identified by direct
measurements of in-cylinder pressure, ion current and optical sen-
sor, which can be installed for each individual cylinder. The first
primary method, which is common for off-line combustion
diagnosis, is postprocessing the in-cylinder pressure data measure-
ments by conventional piezoelectric sensors [1–3]. Measuring the
ratio of in-cylinder mixture components and ionization [4–9] are
other ways to monitor and control misfire in engines. Another
method includes using optical sensor integrated with a spark plug
for misfire diagnosis [10]. Other techniques include high speed
imaging methods [11–14]. Misfire measurements by these afore-
mentioned techniques provide accurate results, but they are not
practical for on-board misfire diagnosis mainly due to high cost
and required computational loads.
The second category of EMD methods is based on measure-
ments of postcombustion parameters including exhaust gas prop-
erties, engine block vibration, or engine speed fluctuation.
Analysis of exhaust gas pressure and temperature measurements
before and inside a catalytic converter has been used for misfire
detection [14–16]. Wide-band oxygen sensors installed at the
confluence point of the exhaust ports have been used to measure
variation in oxygen concentration for misfire diagnosis [17,18].
Due to unburned air–fuel mixture in misfired cylinder, the mixture
outflow has lower temperature, pressure, high oxygen ratio as well
as high unburned hydrocarbon ratio in comparison to a normal
combustion cycle. The EMD methods based on exhaust gas prop-
erties typically have low response time; thus, they are not practi-
cal for on-board misfire diagnosis [19].
EMD techniques based on measurements of vibrational effects
and engine speed fluctuation [20–26] are common for OBD of
misfire in engines. EMD based on engine speed measurements is
the most common technique used in production engines and is the
focus of this study. Engine speed based EMD techniques can be
grouped into three categories as shown in Fig. 2. These include
physics-based, signal processing based, and observer/estimator-
based EMD techniques.
Studies in Refs. [27], [28], and [29] include examples of
physics-based EMD techniques, where in Ref. [27] an energy
model of an engine in angular domain is developed for on-line
monitoring of the in-cylinder processes including misfire. Study
in Ref. [22] represents an example of a signal processing based
technique where multivariate statistical analysis is employed to
instantaneously process the engine speed signal to locate multiple
misfire events. Studies in Refs. [23] and [30] are an example of an
estimation-based technique where engine power per cylinder is
Contributed by the Combustion and Fuels Committee of ASME for publication in
the JOURNAL OF ENGINEERING FOR GAS TURBINES AND POWER. Manuscript received July
8, 2014; final manuscript received February 9, 2015; published online March 24,
2015. Editor: David Wisler.
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 091509-1
Copyright V
C2015 by ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

estimated using measurements from shaft encoder and acoustic
emission sensors. The estimated power per cylinder is then used
to determine whether a cylinder is operating poorly.
In this study, a new physics-based EMD method is developed
and a new parameter for misfire monitoring is proposed according
to engine speed fluctuations and combustion period measurements
by magnetic pickup sensor facing the flywheel. The proposed
method can be implemented on-board and can identify misfire and
faulty cylinders with low processing time requirements. The ex-
perimental validation tests are performed for a vehicle with a
four-cylinder 1.6-l engine, using a chassis dynamometer as well
as vehicle on-road testing. This study is a refinement of our earlier
preliminary work [31].
The structure of the paper is as follows. The Powertrain Rota-
tional Dynamics Model section explains the theoretical principle
of the proposed EMD method. Then, the experimental setup to
collect misfire data is explained. Then, experimental evaluation of
the EMD method and discussions to improve detection accuracy
will be provided. The results from the proposed EMD method are
then compared with those from a common industrial EMD
method. Finally, conclusions are reached and a summary of results
is provided.
Powertrain Rotational Dynamics Model
Powertrain rotational dynamic based on Newton’s second low is
Jsys €
usys ¼X
n
k¼0
Tk(1)
where Jsys is the inertia for powertrain rotating parts, €
usys is
engine rotational acceleration, and Tkis the engine torque for each
cylinder. System dynamics depend on the values of produced tor-
que and wasted torque. In combustion stroke, each cylinder gener-
ates a definite value of positive torque denoted as T
ind
. Other
torques including reciprocating torque (T
r
), frictional torque (T
f
),
and road load torque (T
L
) are all negative. Equation (1) can be
written as
Jsys €
usys ¼Tind uðÞTruðÞTLuðÞTfuðÞ (2)
J
sys
is powertrain integrated momentum of inertia including par-
tial moments of inertias of crankshaft (Jcrankshaft), flywheel
(Jflywheel), clutch-system (Jclutch ), and the remaining powertrain
system’s moment of inertia (Jremainings). Some of these inertias are
constant, while the others vary depending on the driving condi-
tions. As a result, powertrain’s total moment of inertia can be
expressed as
Jsys ¼Jcrankshaft þJflywheel þJclutch þJremainings (3)
Tind in Eq. (2) is determined by Eq. (4) as the summation of
generated and consumed torques from the engine’s cylinders
Tind uðÞ¼AprX
Z
m¼1
PmuðÞfuðÞ (4)
where A
P
is piston crown surface area, ris crankshaft radius, and z
is the number of cylinders. PmuðÞis the distribution of pressure
over engine rotational position u, and fuðÞis the sinusoid func-
tion which encompasses the effects of engine rotational position
expressed as follows:
fuðÞ¼sin uðÞþ ksinð2uÞ
2ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
1k2sinðuÞ2
q(5)
where kis the ratio of crank radius to connecting rod length. Other
acting torques have negative values. TruðÞin Eq. (2) demonstrates
the torque consumed by reciprocating mass of inertias. For each
individual cylinder, the reciprocating torque is calculated by [29]
Trui
ðÞ¼Meq r2fui
ðÞfui
ðÞ
€
uiþdfui
ðÞ
du
_
u2
i

(6)
where M
eq
is the reciprocating total equivalent mass. Equation (6)
can be simplified by neglecting the acceleration term in the
bracket, leading to a total of less than 0.2% error [29]. Equation
(6) is simplified to
Trui
ðÞ¼Meq r2fui
ðÞ
dfui
ðÞ
dui
_
u2
i(7)
In normal engine operating conditions, the net value of torque
is positive in acceleration and negative for deceleration
Fig. 1 EMD techniques in literature
Fig. 2 Common approaches in literature for EMD based on
crankshaft speed measurements
091509-2 / Vol. 137, SEPTEMBER 2015 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

conditions. In case of misfire, Tind is negligible; hence, the total
torque is suddenly negative. Moreover, this leads to a sudden
deceleration which reduces engine speed and is used for diagnos-
tic purposes.
The conversion of the torque to power yields
P¼T_
u(8)
Plugging Eq. (2) into Eq. (8) results in
Jsys €
u_
u¼Tind uðÞ
_
uTruðÞ
_
uTLuðÞ
_
uTfuðÞ
_
u(9)
Trend of engine speed variation during the expansion stroke of a
cylinder is indicative of the combustion situation. Integrating Eq.
(9) over the period of expansion stroke (i.e., Du¼180 deg crank
angle degrees for a four-cylinder engine) leads to
1
DuðDu
0
Jsys €
u_
udu¼1
DuðDu
0
½Tind uðÞ
_
uTruðÞ
_
u
TLuðÞ
_
uTfuðÞ
_
udu(10)
Equation (10) is then converted from rotational displacement (du)
to time domain
1
Duðt
0
Jsys €
u_
u2dt ¼Pind uðÞPruðÞPLuðÞPfuðÞ (11)
In misfiring condition, right-hand side of Eq. (11) becomes neg-
ative because of lack of combustion. Thus, the indicated power-
Pind uðÞis not sufficiently large to overcome waste/resisting
power; hence, the left-hand side (LHS) of Eq. (11) is negative dur-
ing misfire. The LHS of Eq. (11) is expressed as
LHS ¼1
Duðtiþ1
ti
Jsys €
u_
u2dt ¼J
Du
_
u3
iþ1_
u3
i
3

(12)
The engine mean rotational speed ð_
uÞduring the combustion
period for the ith cylinder is
_
ui¼Du
Dti
¼Du
si
(13)
where siis the time period for the engine rotation of
Du¼180 deg. Using Eq. (13), Eq. (12) can be expressed as
LHS ¼Du2J=3s3
is3
iþ1
s3
is3
iþ1
 (14)
LHS expresses the net power of the rotational parts of engine in
the Dutime period. Term Du2J

=3 is constant for any particular
engine. Therefore, the term inside the bracket expression is
denoted as the mean power (MP) value, which is based on the
expansion stroke time duration (e.g., 180 deg of crankshaft rota-
tion for a four-cylinder engine) for each individual cylinder. The
Dufactor should be adjusted according to the number of cylinders
in an engine in order to cover the main power generation period
without overlap among cylinders. MP is defined as the main
parameter for EMD in this study. The engine’s misfire diagnosis
parameter for ith cylinder is defined as
MPi¼s3
is3
iþ1
s3
is3
iþ1
 (15)
During normal engine operation, MP value is close to zero;
however, during misfire, the MP value suddenly changes, indicat-
ing the occurrence of misfire. A critical threshold (C
T
) is defined
for MP to distinguish between normal and misfire cycles. MP and
C
T
values depend on engine speed and engine load as will be dis-
cussed later in the “Results” section.
Experimental Setup
A vehicle with a 1.6-l four-cylinder bi-fuel SI engine is used to
conduct the required misfire experimental tests. The major specifi-
cations of the engine are listed in Table 1. The experimental setup
is shown in Fig. 3.
A common production magnetic pickup sensor connected to the
flywheel is used for measuring engine speed. Bosch’s ME7 ECU
system, coupled by INCA software (ETAS GmbH, Stuttgart, Ger-
many), is used to collect the required experimental data. The INCA is
an interface software for basic measurements and ECU calibration.
INCA has the capability of generating various tests for calibration
and vehicle testing (Fig. 3). Using INCA, three types of misfire are
generated according to certain sequence and period of occurrence:
Type I: single misfire with sequence of cylinders 1, 2, 3, and 4.
Type II: continues misfire with sequence of cylinders 1, 2, 3,
and 4.
Type III: multiple misfires with sequence of cylinders 1–3, 2–4,
1–2, and 3–4.
Figure 4illustrates the platform of the three types of misfire
studied in this work. The first type (Fig. 4(a)) includes misfires
occurring with a certain period (i.e., every 38 cycles) in different
cylinders. The second type demonstrated by Fig. 4(b)includes
one misfire per cycle in consecutive engine cycles. Figure 4(c)
Table 1 Engine Specification
Displacement 1.65 l
Architecture I-4
Bore 78.6 mm
Stroke 84 mm
Compression ratio 11.2
Fuel types Gasoline and CNG with
CNG as the base fuel
Fuel system MPFI
Maximum power (CNG/petrol) 76/85 kW at 6000 rpm
Maximum torque (CNG/petrol) 137/155 Nm at 3500–4500 rpm
Number of flywheel teeth 59
Fig. 3 Overview of the experimental setup for misfire
experiments
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 091509-3
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

shows misfire patterns with two consequent cycles including cyl-
inders 1 and 2 (illustrated in part C-1 for ten segments) and cylin-
ders 1 and 3 (illustrated in part C-2 for ten segments).
Misfire experiments include vehicle testing on a chassis dyna-
mometer setup and vehicle testing by driving the vehicle under
urban traffic conditions. This embodies a wide range of load and
engine speeds. Vehicle/engine operating conditions will be pro-
vided in the Results and Discussion section.
Results and Discussion
The misfire tests are done for two different engine conditions
including: (1) 3250 rpm with 50% engine load
1
and (2) 1250 rpm
with 10% engine load. Measured engine speed and calculated MP
values for 50% load at 3250 rpm conditions are shown in Figs.
5–8. For the sake of brevity, the plots for 10% load at 1250 rpm
are shown in the Appendix.
By reviewing the engine speed plots, two posteffects of missing
combustion are observed. (1) There is a drop in spontaneous
engine speed. This drop is used in the designed model-based
EMD method to identify faults as it affects MP values. (2) There
is a continuation of engine speed deviation as a result of engine
dynamics stiffness and damping properties. This continued devia-
tion can adversely affect misfire detection accuracy.
Figure 5indicates engine speed deviations for single misfires
and effects on MP detection parameters; moreover, it is clear
that deviations continued to the next misfire with lower MP mag-
nitude. After each single misfire, it takes a short period for the
MP oscillations to disappear and return to normal engine opera-
tion values (Fig. 5). This settling time for MP oscillation is
referredtoasdamping time in this study. Increasing engine load
and engine speed leads to higher amplitude of rotational engine
speed deviations. By reviewing a wide range of the operating
data for the engine, damping time is found to have an inverse
relation with engine load and direct relation with engine speed.
For diagnostic purposes, damping time plays a significant role.
Smaller damping time decreases the possibility of false misfire
detections.
Increasing engine speed at partial load conditions is found to
lead to higher measurement noise. Conversely, increasing load at
a fixed engine speed decreases the noise. Accordingly, damping
of postmisfire deviation is faster at high load and this causes the
MP to be more accurate in high-load operations.
At normal operation, the MP must have a mean value close to
zero. A misfire event is detected when MP exceeds the C
T
thresh-
old which depends on engine speed and load. Table 2lists the C
T
values used for different engine speeds and loads. It is clear that
larger threshold values are required at higher loads and higher
engine speeds.
Figure 9illustrates the misfire generation signal (MGS) which
is generated systematically by the engine control unit to cause
misfire by lack of in-cylinder spark, followed by peak values of
MP misfire detection signal. To quantize the accuracy of misfire
detection, a metric for misfire detection capability (dc) is defined
as
dc¼tmmdfa
tm
100 (16)
Fig. 4 Patterns of misfire generation including: (a) single mis-
fire, (b) continuous misfires, and (c) multiple periodic misfires
Fig. 5 Engine speed deviations and MP values for single mis-
fires (type I) at 3520 rpm and 50% load
Fig. 6 Engine speed deviations and MP values for continues
misfires (type II) at 3520 rpm and 50% load
Fig. 7 Engine speed deviations and MP values for multi-
misfires (type III—pattern 1, 4 cylinders) at 3520 rpm and 50%
load
Fig. 8 Engine speed deviations and MP values for multi-
misfires (type III—pattern 1, 3 cylinders) at 3520 rpm and 50%
load
1
Load is estimated by ECU based on throttle opening angle.
091509-4 / Vol. 137, SEPTEMBER 2015 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

where t
m
is the total number of misfires, m
d
and f
a
represent num-
ber of missed detections and false detections, respectively. Tables
3and 4show the performance of the proposed EMD technique for
detecting different types of misfire at two different engine speeds.
The results show that dcby the EMD technique ranges from 71%
to 91% depending on misfire types. Single misfires are harder to
detect with 25% possibility of false detections (Table 3). As
expected dcis lower at a smaller engine load, but dcremains
almost constant for both engine loads for type II (i.e., continues)
misfires.
The results in Figs. 5–8represent steady-state operation. One
major challenge is to ensure that an EMD method works both in
steady-state and transient operations. Here, the proposed EMD
method is also tested in highly transient engine operation in vehi-
cle urban driving conditions. A vehicle equipped with the pro-
posed misfire detection technique is used to carry out vehicle tests
in city driving conditions. Figure 10 shows the engine operating
conditions. Engine speed varies from 1500 rpm to 4000 rpm, and
engine load varies from 0% to 40% during a 60 s test period. This
represents a highly transient engine operation. Three sets of mis-
fire events in two types of continuous (type II) and multiple mis-
fire (type III) are applied as shown in Fig. 11.
A close-up of MP values versus MGS is shown in Fig. 12. Sev-
eral false misfire detections can be noticed. Inadequate MP devia-
tions decrease the overall reliability of the proposed technique; in
this case, using the MP value for detection purposes requires elim-
ination of postmisfiring oscillations in normal operation condition.
The MP value based on calculation of consequent power segments
and cylinder-by-cylinder deviations is infinitesimal; thus, the
mean value of consequent segments value is close to zero. To
increase detection accuracy, the factor Cvis derived; so MP is
compensated as follows:
MPc¼MP Cv(17)
where Cvis the compensation factor defined as
Cv¼X
iþ3
k¼i3
MPk
7(18)
Cvis the mean value of MP for seven segments (i.e., two engine
cycles). Cvis used to eliminate the MP deviations for improving
misfire detection accuracy. Using the compensated MPc, the mis-
fire detection results will be smoothed by removing misleading
MP deviations.
Compensated MP values based on Eq. (17) are shown in Fig.
13.MP
c
values are smoother than original MP values in Fig. 11.
By applying C
v
, all these primary oscillations are eliminated. This
results in improving dc(detection capability) from 80% to 90%
for the misfires in Fig. 11. Thus, the new EMD technique can be
used for both steady-state and transient operating conditions.
Applying C
v
to the steady-state test conditions also results in
the elimination of postmisfiring deviations. Figure 14 shows the
compensated MP values for the steady-state test conditions previ-
ously presented in Figs 5–7. By comparing the MP and MP
c
val-
ues in these figures, a significant reduction in the number of false
and missed detections is observed. For the case of misfire type I
(i.e., single-misfire), no false misfire detection and no missed
detection are observed in the new results. In addition, both false
detection and missed detection are reduced by 10% in misfire type
III (i.e., multiple misfires).
Comparison With LU Method. There are other EMD meth-
ods which use concepts which are comparable to the MP
method. Among these methods, the LU method in Ref. [32]is
the most related to the proposed method in this work; thus, the
results from this work are compared to those from the LU
method. The LU method was developed by evaluating crank-
shaft speed loss as a result of misfire (see Ref. [29] for details).
Fig. 9 MGS (filled bar lines) and MP values for 3250 rpm test data
Table 2 C
T
threshold values (1/s
3
)
Engine load
Engine speed 10% 50% 70%
1250 rpm 150 400 600
3500 rpm 1000 3500 4000
Table 3 Misfire detection capability for the tests at 3250 rpm
and 50% engine load
Single (type I) Continuous (type II) Multiple (type III)
Generated, t
m
20 280 280
False detected, f
a
4520
Missed, m
d
210 20
d
c
(%) 70 91 90
Table 4 Misfire detection capability for the tests at 1250 rpm
and 10% engine load
Single (type I) Continuous (type II) Multiple (type III)
Generated, t
m
20 280 280
False detected, f
a
7662
Missed, m
d
210 5
d
c
(%) 55 90 75
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 091509-5
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

The LU method is based on the engine roughness detection pa-
rameter which is augmented by other additional factors. LU
i
is
defined as
LUi¼siþ1si
s3
i
(19)
where siis the combustion duration time period related to ith
cylinder. LU method is applied to the both steady-state and tran-
sient data collected in this work. Figure 15 shows a comparison
between calculated LU and MP
c
values for the third set of mis-
fires in transient driving conditions from Figs. 10 and 11.The
general pattern in LU and MP
c
is similar since both methods use
a similar principle based on engine power-train dynamic model.
However, MP
c
offers some advantages. In steady-state opera-
tion, the value of [(Max cylinder to cylinder deviation)/(fault
signal)] ratio is 0.16 and 0.08 for LU and MP
c
, respectively. In
transient driving conditions, this ratio varies for different condi-
tions, yet the mean ratio is, respectively, 0.16 and 0.04 for LU
and MP
c
. This advantage causes the MP
c
to be less sensitive to
noise; thus, more resistant to false detections. In addition, MP
c
offers better misfire detection capability (dc) as shown in Table
5for transient driving data from Fig. 10. The results indicate
3–8% improvement in dcby using MP
c
method compared to the
LU method. dcis anticipated to improve if the signal to noise ra-
tio is increased by using the methods of signal processing
[33–37].
Fig. 10 Vehicle transient test conditions: (a) engine speed and
(b) engine load
Fig. 13 Compensated MP (MPc) values for transient test data
of Fig. 10
Fig. 11 MP value overlapped by MGS in transient urban driv-
ing conditions
Fig. 14 Compensated MP (MPc) values for misfire types (type
I-a) and (type III-b) at 3520 rpm and 50% load
Fig. 12 Close-up of calculated MP values (continues stair lines) overlapped by MGS (filled bar lines) for the third misfire gener-
ation set in Fig. 11
091509-6 / Vol. 137, SEPTEMBER 2015 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

Summary and Conclusions
A new physics-based EMD technique was developed for detect-
ing misfires in SI engines. The new technique is based on MP
value which is calculated using engine speed measurements. The
proposed technique was tested for a four-cylinder SI engine in
steady-state and transient operating conditions. Steady-state mis-
fire experiments were carried out on a vehicle chassis dynamome-
ter, while transient experiments were done by on-road vehicle
testing. Three different types of misfire were investigated. These
include single, continuous, and multiple misfires. While accurate
in detecting misfires for steady-state conditions, the EMD tech-
nique could not successfully identify transient misfires, which are
shadowed by engine postmisfire oscillations.
A smoothed version of the EMD technique was proposed and
tested. The results indicated that the modified EMD technique
could remove posteffects of misfires, thus could successfully
detect misfire for transient conditions too. Overall, the new EMD
technique exhibited misfire detection capabilities up to 94% for a
range of engine speeds and loads for tested four-cylinder engine.
Finally, the new technique was compared with a common indus-
trial EMD method known as LU. The comparison results con-
firmed that the new method outperforms the LU method by
providing better detection accuracy and lower ratio of max cylinder-
to-cylinder/fault signal.
Although the results were presented for a four-cylinder engine,
similar misfire detection capability is anticipated for six-cylinder
engines. But the extension of the proposed EMD technique for 8
Fig. 15 LU and MP
c
patterns for the third set of misfire genera-
tion in transient operation
Fig. 18 Engine speed variation and MP values for multiple mis-
fires (type III-pattern 1,4 cylinders) at 1250 rpm and 10% engine
load
Fig. 17 Engine speed variation and MP values for continuous
misfires (type II) at 1250 rpm and 10% engine load
Fig. 19 Engine speed variation and MP values for multiple mis-
fires (type III-pattern 1,3 cylinders) at 1250 rpm and 10% engine
load
Table 5 Percentage of total misfire detection capability for LU
and MP
c
in transient driving conditions from Fig. 10
Total generated, t
m
600
False detected, f
a
(%) MP
c
1.67
LU 4.8
Missed, m
d
(%) MP
c
3.8
LU 7
d
c
(%) MP
c
93.5
LU 88
Fig. 16 Engine speed variation and MP values for single mis-
fires (type I) at 1250 rpm and 10% engine load
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 091509-7
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

and 12-cylinder engines is expected to require extensive calibra-
tion phase before implementation due to smooth power generation
from 8 and 12-cylinder engines.
Acknowledgment
The authors gratefully thank Mr. Mohammad Ghafuri and
Dr. Rasoul Salehi from IPCO for their help in collecting the
misfire data which were used to evaluate the proposed EMD
method.
Nomenclature
Ap¼piston surface (m
2
)
C
T
¼threshold value (1/s
3
)
C
v
¼compensation factor
CNG ¼compressed natural gas
d
c
¼detection capability (%)
fuðÞ¼crank–slide mechanism function
Jsys ¼general rotational inertia (kg m
2
)
L¼connecting-rod length (m)
Meq ¼equivalent mass of the sliding parts of crank slide mech-
anism (kg)
MP ¼misfire detection parameter (1/s
3
)
MP
c
¼compensated misfire detection parameter (1/s
3
)
MPFI¼multipoint fuel injection
P¼power (W)
PmuðÞ¼in-cylinder pressure of mth cylinder (Pa)
R¼crank–slide mechanism radius (m)
T
f
¼frictional torque (Nm)
T
ind
¼indicated torque (Nm)
T
L
¼load torque (Nm)
T
r
¼reciprocating torque (Nm)
Du¼expansion stroke angular interval (rad)
k¼ratio of crank radius to connecting rod length
si¼ith combustion period (s)
u¼crankshaft angular position (rad)
_
u¼crankshaft angular speed (rad/s)
€
usys ¼rotational acceleration (rad/s
2
)
Subscripts
c¼compensated
Eq ¼equal
f¼friction
i¼related to ith cylinder
ind ¼indicated
L¼load
m¼misfire
p¼piston
r¼reciprocating
sys ¼related to integrated powertrain system
Appendix
The following figures show the experimental data and calculated
MP values for the test results at 1250 rpm with 10% engine load.
References
[1] Heywood, J. B., 1989, Internal Combustion Engine Fundamentals, McGraw-
Hill, New York.
[2] Luj
an, J. M., Berm
udez, V., Guardiola, C., and Abbad, A., 2010, “A Methodol-
ogy for Combustion Detection in Diesel Engines Through In-Cylinder Pressure
Derivative Signal,” Mech. Syst. Signal Process.,24(7), pp. 2261–2275.
[3] Pyari, F., Luj
an, J. M., Mart
ın, J., and Abbad, A., 2010, “Digital Signal Process-
ing of In-Cylinder Pressure for Combustion Diagnosis of Internal Combustio n
Engines,” Mech. Syst. Signal Process.,24(6), pp. 1767–1784.
[4] Sellnau, M. C., Matekunas, F. A., and Battiston, P. A., 2000, “Cylinder-Pressure-
Based Engine Control Using Pressure-Ratio-Management and Low-Cost
Non-Intrusive Cylinder Pressure Sensors,” SAE Paper No. 2000-01-0932.
[5] Fan, Q., Bian, J., Lu, H., Tong, S., and Li, L., 2014, “Misfire Detection and
Re-Ignition Control by Ion-Current Signal Feedback During Cold Start in
Two-Stage Direct-Injection Engines,” Int. J. Engine Res.,15(1), pp. 37–47.
[6] Vandyne, E. A., Burchmyer, C. L., Wahl, A. M., and Funaioli, A. E., 2000,
“Misfire Detection From Ionization Feedback Utilizing the Smart-Fire-Plasma
Ignition Technology,” SAE Paper No. 2000-01-1377.
[7] Dong, G., Yu, S., Li, L., and Univ, T., 2009, “Misfiring Control in Current
Cycle at Engine Start Employing Ion Sensing Technology,” SAE Technical
Paper No. 2009-01-2713.
[8] Mayhew, C. G., Knierin, K. L., Chaturvedi, N. A., Park, S., Ahmed, J., and
Kojic, A., 2009, “Reduced Order Modeling for Studying and Controlling Mis-
fire in Four-Stroke HCCI Engines,” 48th IEEE Conference on Decision and
Control and 28th Chinese Control Conference (CDC/CCC 2009), Shanghai,
Dec. 16–18, pp. 5194–5199.
[9] Wang, Y., and Zhou, L., 2003, “Investigation of the Detection of Knock and
Misfire of a Spark Ignition Engine With the Ionic-Current-Method,” Proc. Inst.
Mech. Eng., Part D,217(7), pp. 617–621.
[10] Gazis , A., Panousakis, D., Chen, R., and Chen, W. H., 2006, “Computationally
Inexpensive Methods of Ion Current Signal Manipulation for Predicting the
Characteristics of Engine In-Cylinder Pressure,” Int. J. Engine Res.,7(3), pp.
271–282.
[11] PierniKarski, D., and Hunicz, J., 2000, “Investigation of Misfire Nature Using
Optical Combustion Sensor in a SI Automotive Engine,” SAE Technical Paper
No. 2000-01-0549.
[12] Peter son, B., Reuss, D. L., and Sick, V., 2011, “High-Speed Imaging Analysis
of Misfires in a Spray-Guided Direct Injection Engine,” Proc. Combust. Inst.,
33(2), pp. 3089–3096.
[13] Tornatore, C., Sementa, P., and Merola, S. S., 2011, “Optical Investigations of
the Early Combustion Phase in Spark Ignition Boosted Engines,” Proc. Inst.
Mech. Eng., Part D,225(6), pp. 787–800.
[14] Sick, V., 2012, “High Speed Imaging in Fundamental and Applied Combustion
Research,” Proc. Combust. Inst.,34(2), pp. 3509–3530.
[15] Ye, J., 2009, “Application of Extension Theory in Misfire Fault Diagnosis of
Gasoline Engines,” Expert Syst. Appl.,36(2), pp. 1217–1221.
[16] Tamura, M., Saito, H., Murata, Y., Kokubu, K., and Morimoto, S., 2011,
“Misfire Detection on Internal Combustion Engines Using Exhaust Gas Tem-
perature With Low Sampling Rate,” Appl. Therm. Eng.,31(17–18), pp.
4125–4131.
[17] Chiavola, O., 2003, “Combus tion Anomalies Detection in SI Engines From
Exhaust Pressure Signal Processing,” Proc. Inst. Mech. Eng., Part A,217(5),
pp. 537–546.
[18] Chung, Y., Bae, C., Choi, S., and Yoon, K., 1999, “Application of a Wide Range
Oxygen Sensor for the Misfire Detection,” SAE Paper No. 1999-01-1485.
[19] Chung, Y., Kim, H., Choi, S., and Bae, C., 2000, “Flow Characteristics of Mis-
fired Gas in the Exhaust Manifold of a Spark Ignition Engine,” Proc. Inst.
Mech. Eng., Part D,214(4), pp. 373–381.
[20] Lai, S. H.-Y., 1993, “Engine System Diagnosis Using Vibration Data,” Comput.
Ind. Eng.,25(1), pp. 135–138.
[21] Vulli, S., Dunne, J. F., Potenza, R., and Richardson, D., 2009, “Time-Frequency
Analysis of Single-Point Engine-Block Vibration Measurements for Multiple
Excitation Event Identification,” J. Sound Vib.,321(3–5), pp. 1129–1143.
[22] Hu, C., Li, A., and Zhao, X., 2011, “Multivariate Statistical Analysis Strategy
for Multiple Misfire Detection in Internal Combustion Engines,” Mech. Syst.
Signal Process.,25(2), pp. 694–703.
[23] Douglas, R. M., Steel, J. A., Reuben, R. L., and Fog, T. L., 2006, “On-Line Power
Estimation of Large Diesel Engines Using Acoustic Emission and Instantaneous
Crankshaft Angular Velocity,” Int. J. Engine Res.,7(5), pp. 399–410.
[24] Kiencke, U., 1999, “Engine Misfire Detecti on,” Control Eng. Pract.,7(2),
pp. 203–208.
[25] Ball, J. K., Bowe, M. J., Stone, C., and McFadden, P., 2000, “Torque Estima-
tion and Misfire Detection Using Block Angular Acceleration,” SAE Technical
Paper No. 2000-01-0560.
[26] Barel li, L., Bidini, G., Burrati, C., and Mariani, R., 2009, “Diagnosis of Internal
Combustion Engine Through Vibration and Acoustic Pressure Non-Intrusive
Measurements,” Appl. Therm. Eng.,29(8–9), pp. 1707–1713.
[27] Tinaut, F. V., Melgar, A., Laget, H., and Dominguez, J. I., 2007, “Misfire and
Compression Fault Detection Through the Energy Model,” Mech. Syst. Signal
Process.,21(3), pp. 1521–1535.
[28] Stotsky, A. A., 2007, “Statistical Engine Misfire Detection,” Proc. Inst. Mech.
Eng., Part D,221(5), pp. 641–649.
[29] Ponti, F., 2008, “Development of a Torsional Behavior Powertrain Model for
Multiple Misfire Detection,” ASME J. Eng. Gas Turbines Power,130(2),
p. 022803.
[30] Chen, B. C., Wu, Y. Y., and Hsieh, F. C., 2005, “Estimation of Engin e Rota-
tional Dynamics Using a Closed-Loop Estimator With Stroke Identification for
Engine Management Systems,” Proc. Inst. Mech. Eng., Part D,219(12),
pp. 1391–1405.
[31] Boudaghi, M., Shahbakhti, M., and Jazayeri, S. A., 2013, “A New Physics-
Based Misfire Detection Technique for a SI Engine,” ASME Internal
Combustion Engine Division Fall Technical Conference, ASME Paper No.
ICEF2013-19096.
[32] Klenk, M., Moser, W., Mueller, W., and Wimmer, W., 1993, “Misfire Detection
by Evaluating Crankshaft Speed: A Means to Comply With OBDII,” SAE
Paper No. 930399.
[33] Wang, Y., and Chu, F., 2005, “Application of Non-Linear Observers to On-
Line Estimation of Indicated Torque in Automotive Engines,” Proc. Inst. Mech.
Eng., Part D,219(1), pp. 65–75.
091509-8 / Vol. 137, SEPTEMBER 2015 Transactions of the ASME
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms

[34] Gu, F., Jacob, P. J., and Ball, A. D., 1999, “Non-Parametric Models in the Mon-
itoring of Engine Performance and Condition—Part 2: Non-Intrusive Estima-
tion of Diesel Engine Cylinder Pressure and Its Use in Fault Detection,” Proc.
Inst. Mech. Eng., Part D,213(2), pp. 135–143.
[35] McDow ell, N., McCullough, G., Wang, X., Kruger, U., and Irwin, G. W., 2007,
“Fault Diagnosis for Internal Combustion Engines: Currently and Future
Techniques,” SAE Technical Paper No. 2007-01-1603.
[36] Mohammadpour, J., Francheck, M., and Grigoriadis, K., 2012, “A Survey on
Diagnostic Methods for Automotive Engines,” Int. J. Engine Res.,13(1),
pp. 41–64.
[37] Boudaghi, Kh.-N. M., 2011, “The Effect of Distances Between the Vehicles on
Their Fuel Consumptions in Highways,” Proceedings of the 19th Annual Con-
ference on Mechanical Engineering, Birjand, Iran, May 10–12, Paper No.
ISME2011-0984.
Journal of Engineering for Gas Turbines and Power SEPTEMBER 2015, Vol. 137 / 091509-9
Downloaded From: http://gasturbinespower.asmedigitalcollection.asme.org/ on 03/24/2015 Terms of Use: http://asme.org/terms