Seminar Report

on

FAULT DETECTION OF
AUTOMOBILE ENGINE
Submitted in the partial fulfillment of the award of the degree
of

Bachelor of Technology
in

Electronics and Instrumentation Engineering

of

Cochin University of Science and Technology

by

REGIL ROY

October, 2016
Department of Electronics Engineering
College of Engineering, Chengannur-689121
Phone: (0479) 2454125, 2451424

Fax: (0479) 2451424

COLLEGE OF ENGINEERING, CHENGANNUR

KERALA

DEPARTMENT OF ELECTRONICS ENGINEERING

CERTIFICATE
This is to certify that the seminar entitled

FAULT DETECTION OF
AUTOMOBILE ENGINE

Submitted by

REGIL ROY
is a bona fide work done by him.

Coordinator

Head of the Department

ACKNOWLEDGMENT
To begin with, I would like to thank GOD ALMIGHTY for providing me with His
blessings to perceive this seminar with interest and vigour.
I would like to extend my deep gratitude to Dr.Jyothiraj V.P, Principal, College of
Engineering Chengannur(CEC), Dr. Nisha Kuruvila, Head of the Department of
Electronics and Mr. Sam K Sumesh, Associate Professor and Staff-in-charge, Electronics
and Instrumentation Engineering, CEC for providing me with the necessary facilities
required to bring the seminar to a successful completion.
I would like to express my special thanks to our seminar coordinators Mr. Liju
Philip, Assistant Professor, Electronics Engineering Dept., Dr. Sarah Jacob, Associate
Professor Electronics Engineering Dept. and Mrs. Maneesha Krishnan, Assistant Professor
Electronics Engineering Dept., whose encouragement, guidance and support from the
initial to the final stage enabled me to develop an immense understanding of the topic. Im
also thankful to them for guiding me about handling the presentation and conducting the
seminar.
Last but not the least, I would like to convey my regards to all those who supported
me through the course of the successful completion of this seminar.

TABLE OF CONTENTS

1.

Introduction

2.

Ohv Engine

3.

Working of ohv engine

4.

Current method for fault detection

5.

Proposed method

6.

What is PCA ?

7.

Example for PCA

8.

Materials and methods used

9.

Results

10.

Results related to use of PCA

11.

Conclusion
Reference

LIST OF FIGURES
2.1

2.2

Digital Thermometer

3.1

Data Acquisition Module

3.2

Thermocouple Temperature Measurement Circuit

4.1

Fitting Curve of Thermocouple Reference Table

13

4.2

Fitting Curve of RTD Reference Table

14

4.3

Error in Fitting PT1000

14

5.1

Effect of Cold end on Hot end

17

5.2

Hot end, Cold end, Reference temperature

18

5.3

Cold end Temperature v/s Temperature Error

18

5.4

Hot end Temperature v/s Temperature Error

18

5.5

Thermometer Stability

20

5.6

Response Time of Thermocouple

21

ABSTRACT

Today condition monitoring using vibration analysis is not only used for simple proposes
like bearings and gears fault detection but also it is used in hybrid motors and cars in
automobile industry for detecting or anticipating faults. The main aim of this paper is

using vibration analysis in automobile industry to detect and categorize faults appeared
due to poppet valve clearance and incomplete combustion sometimes called misfiring
phenomenon in internal combustion engine. So we used four accelerometers on OHV
engine body for registering vibration signals. Then using PCA technique acquiring data
were investigated. At the end we succeeded to classify and detect faults with a high ratio
of efficiency

FAULT DETECTION OF AUTOMOBILE ENGINE

1. INTRODUCTION

Automobile science is progressing increasingly with an astonishing speed and car

manufacturers are competing with each other to reach more target markets. One of the
main factors in costumers point view is delivery of faultless and intact car. Automobile
engine is the main part of it. During manufacturing process after assembly, the engine may
have some faults due to its complexity. Finding these faults is the most professional step of
completion process of manufacturing. Fault detection via vibration analysis is one of
modern methods which have many uses in car industry. In recent years plenty of
investigations of vibrational behavior monitoring were done by means of researchers to
detect and fix possible faults in various industries especially in rotating machines such as
car engines. Manufacturers also can check car engine in last step of manufacturing process
and in case of any fault, they can identify and remove it.

2. OHV ENGINE.

1
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

An overhead valve engine (OHV engine) is an engine in which the valves are placed in
the cylinder head. This was an improvement over the older flathead engine, where the
valves were placed in the block next to the piston. Overheadcamshaft (OHC) engines,
while still overhead valve by definition, are usually categorized apart from other OHV
engines. OHV engine is a four stroke and a linear four cylinder engine with 8 valves. This
engine is manufactured by optimizing cylinder, cylinder head, cam shaft, piston, ring, rod
and valves and has Euro 2 standard. This engine volume as showed in Fig. 1 is 1600 cm2.
OHV engines have some advantages over OHC engines:

Smaller overall packaging: because of the camshaft's location inside the engine
block,OHV engines are more compact than an overhead cam engine of comparable
displacement

Less complex drive system: OHV engines have a less complex drive system for the
camshaft

when

compared

with OHC engines.

Most

OHC

engines

drive

the camshaft or camshafts using a timing belt, a chain, or multiple chains. These
systems require the use of tensioners which add complexity. In contrast, an OHV
engine has the camshaft positioned close to the crankshaft which may be driven by a
much shorter chain or even direct gear connection. However, this is somewhat negated
by a more complex valvetrain requiring pushrod

2
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig 1: OHV engine overview

Fig 2: schematic diagram of ohv engine

3. WORKING OF OHV ENGINE

3
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Ohv engine is an internal combustion (IC) engine in which the piston completes four
separate strokes while turning a crankshaft. A stroke refers to the full travel of the piston
along the cylinder, in either direction. The four separate strokes are termed:
1. Intake: also known as induction This stroke of the piston begins at top dead center
(T.D.C.) and ends at bottom dead center (B.D.C.). In this stroke the intake valve
must be in the open position while the piston pulls an air-fuel mixture into the
cylinder by producing vacuum pressure into the cylinder through its downward
motion.
2. Compression: This stroke begins at B.D.C, or just at the end of the suction stroke,
and ends at T.D.C. In this stroke the piston compresses the air-fuel mixture in
preparation for ignition during the power stroke (below). Both the intake and
exhaust valves are closed during this stage.
3. Combustion: also known as power or ignition This is the start of the second
revolution of the four stroke cycle. At this point the crankshaft has completed a
full 360 degree revolution. While the piston is at T.D.C. (the end of the
compression stroke) the compressed air-fuel mixture is ignited by a spark
plug (in a gasoline engine) or by heat generated by high compression (diesel

engines), forcefully returning the piston to B.D.C. This stroke produces

mechanical work from the engine to turn the crankshaft.
4. Exhaust: also known as outlet. During the exhaust stroke, the piston once again
returns from B.D.C. to T.D.C. while the exhaust valve is open. This action expels
the spent air-fuel mixture through the exhaust valve.

4
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig 3: 4 strokes of internal combustion

4. CURRENT METHOD FOR FAULT DETECTION

5
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

At last point of manufacturing process of overhead valve (OHV) engine, technical tests are
done as usually. These tests include general examination of engine like its sound,
vibration, resonance and other features. At the end of this test all cars which were not in
consistency with standards should be studied carefully to detect faults and repair them.
After detecting and fixing the faults, mentioned cars should be tested and confirmed again.
Today fault detection of these engines at IKCO is done via trial and error. In this method
possibility of incorrect fault detection is high and sometimes one engine would be opened,
repaired and closed many times due to incorrect diagnosis of faulty part. Present method is
not only economic but also is one of problems in manufacturing line. One of the best
methods which can be replaced is vibration and noise analysis of engine. Using vibration
analyzing, one can determine relationships between vibration answer and engine faults.

5. PROPOSED METHOD
6
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Using statistical methods in fault detection has been increasingly developed in last decade
and its reason is that mentioned methods dont need mathematical models. Meanwhile
correct understanding of experimental data will result in more suitable efficiency of these
data based methods. Among performed researches about fault detection of car engine you
can find ones based on data analysis using many sensors via covariance matrix for
anticipating, detecting and following faults. In other words there is not a pattern for online
fault detection. Here we recommend using vibration analysis via Principal Component
Analysis (PCA) for detecting faults of poppet valve clearance and incomplete combustion.

6. WHAT IS PCA ?

PCA is a suitable statistical trick which has many application in face detection and image
processing. It is common technique in observation patterns in data with high range. Recent
7
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

progress in data collection and saving capabilities of them results in large volume of
information in most of sciences. Researchers in different fields such as engineering,
astrology, biology and economics are encountering with more and more observations.
Therefore there are new challenges in data analysis in comparison with today data
platforms. Traditional statistical methods lose their efficiency for two reasons, one for
their increasing of number of observations and the other due to their increasing of related
variables of an observation.
Number of variables which should be measured for every observation is called
dimensions. Term of variable is used more in statistics while in computer and machine
learning, term of feature or property is more common. Data platforms with high
dimensions not only provide us with so many opportunities but they also provide
calculation challenges too. One of their problems is that all features of them are not
necessary and vital for observation hidden knowledge of data. Therefore in many fields
reducing data dimensions is one of considerable discussions.
PCA finds a linear relationship between variables which shows general procedure of data
changes. In mathematical language PCA is based on orthogonal decomposition of
covariance matrix of process variable in direction with most data changes. So PCA can be
described as a linear transformation which converts a set of observations of main
correlated variables into a set of uncorrelated variables. In short, PCA is one of methods
for analyzing multi variable data and its main purpose is reducing dimensions of problem
under study. Using PCA you can replace so many correlated explanatory variables with
small number of new explanatory variables which are called principal components and are
uncorrelated. In this way data dimensions are reduced.

7. EXAMPLE FOR PCA

8
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

As an example in two dimensional mood, first axis (principal component) is in a direction

with more data distribution and maximum data variance and another axis is vertical to first
axis which means less data distribution (see fig.

Fig. 4. Placement of new axes according to principal components

Brief algorithm of PCA is a below:

1- Choosing input data to compute Matrix X
2- Computing covariance matrix for finding eigenvectors and amount of system.
R=(XT*X)/(n1)

(1)

(Rj I ) Pj= 0, (j= 1, 2,3,..., m)

(2)

In which n is number of rows of X matrix, j is eigenvalue and Pj is matrix of orthogonal

eigenvectors.
3- Reduction of matrix dimensions of eigenvectors in a form that mentioned matrix only
include eigenvector correspondent to r, main eigenvalue.
9
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

P= [p1, p2, p3,..pr]

(3)

4- Calculating data related to main system coordinate using reverse transformation in (4)
X* =X*(PPT )

(4)

5- Computing difference amount of data related to main coordinate and primary data in
every moment.
ri(k)=Xi(k)-Xi*(k),(i=1,2,3,.m)

(5)

6- At the end, computing remaining amount in every moment (k) in form of total squares
of every row.
res= i=1 r2i(k)

(6)

Considering changing nature of phenomenon, in case of any fault, we expect to see

considerable changes in variable trend of residuals to detect potential faults.

8. MATERIALS AND METHODS USED

One of the faults which are done deliberately in engine for analyzing its
vibrations is combustion fault. For making this fault we should take wire of
related cylinder out. Doing this, two combustion faults happened. First one
was taking cylinder wire 1 and second fault was taking out wire of cylinder 1
and 2 simultaneously. Vibration signals of healthy and faulty engines were
registered. Comparison of vibration signals of these 3moods (healthy, cutting
wire 1 and cutting wire 1 and2) had considerable results which are described
below:
10
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

The gap between rocker arm and valve seat in closed position is closed
poppet valve clearance or valve filter. If this gap increases very much, course
time of valve will be decreased and it will cause some problems in internal
cylinder combustion. Mentioned problems will increase engine vibrations. To
make this fault, we increase smoke valve filter of third valve from its
standard clearance of 0.3mm to 0.6mm.
To study vibrations in different utilizations, besides vibration meter we need a
system for monitoring, processing and saving output signal of sensor.
Implementing direct accelerometer has similarities to indirect ones in case of
sensor and data processing software but related hardware facilities are
different. For example facilities in a data saver are used separately and in a
modular form.
Some systems which are usable for vibration study are oscilloscope,
advanced analyzers, total vibration recorder, data savers, and FFT analyzers.
In this research, during engine function and fault creation, vibration features
are calculated by ADASH 4400 analyzer shown in fig
Harmony ratio of mass-spring is very high in accelerometers (fig ). It means
that this sensor is working under its natural frequency. We used quartz
piezoelectricity feature for transforming mechanical moves. Piezoelectricity
feature results in electric charge at the end of a crystal which under pressure
or strain. Electrical charge is the result of polar molecules movements in
crystal. High frequency resonances in sensors are happened because of their
structural composition. This frequency is about 35 KHz in Accelerometers
which have applications in industries. Allowed maximum thresholds
frequency is about 20 KHz and allowed minimum one is 1Hz. In this study
we used 4 one directional piezoelectric CTC Accelerometer sensors. This
Accelerometer sensor is product of Sweden and can receive vibration in
temperature zone of -50 to +121. This sensor is 90 grams. Using magnets, all
11
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

sensors are fixed on engine body. Therefore sensor 1 was fixed under
cylinder plug 1and sensor number 2 was fixed under cylinder plug 2 and
other two sensors were fixed on engine block.

9. RESULTS
Sample of measured time wave and frequency range calculated by analyzer which is
related to healthy engine are shown in Fig. 5 and Fig. 6 respectively.

12
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig

Fig

As you can see in last figure there are four peaks for two engine rounds in time domain a
healthy engine which is related to four combustions in two engine rounds and the four
peaks are repeated during measurement time. Results of time and frequency domain
measured by analyzer when the wire was taken out are shown in Fig. 7 and Fig. 8.

13
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig

fig

Studying this signal and its related frequency spectrum and comparing it with healthy
engine, it is cleared that peaks extent are reduced considerably and among them one of
reductions are more than others which is related to incomplete combustion in cylinder 1.
In Fig. 9 and Fig. 10 the measured sample of time signal and frequency domain can be
found when wire #1 and #2 are taken out simultaneously

14
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig

In this signal, reduction of peaks amounts is more obvious so two of them are dropped
considerably which is related to cylinder 1 and 2. Sample of signal and frequency range
measured by analyzer are shown in figures 11 and 12 which is related to redundant poppet
valve clearance. In case of poppet valve clearance it is found that there exist much
similarity between this signal and healthy signal. However, we see that one of the peaks in
a twin engine round has a tangible drop which is related to combustion of cylinder 3. Its
reason is that poppet smoke valve clearance results to incomplete evacuation and
combustion does not have enough suitable power efficiency.

10. RESULTS RELATED TO USING OF PCA

15
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Results related to use of PCA are shown in figures 11 to 15. For performing this test,
engine worked in a normal and faultless way about 30 seconds. Then we programmed for
every 15 seconds respectively fault of taking out wire #1 due to incomplete combustion of
cylinder 1, return to normal mood subsequently, fault of taking out wires #1 and #2
simultaneously and returning to normal mood. We calculated principal components using
results of 4 sensors in first 20 seconds. In the next 10 seconds of correct function of
engine, we use obtained principle components to validate automatic fault detection system.
As you see in Fig. 11, in time interval between 20th second to 30th second, interval of
t=45 to t=60 and interval of t=75 to t=90, the healthy periods, PCA technique works
correctly due to components calculated from first 20 seconds of signal. In time intervals of
30 to 45 and 60 to 75 with faulty engine, PCA output is completely different from thats
for healthy engine and these differences can be used for fault detection. Considering facts
like calculation reduction, more realistic results and enough perfection in this research it is
decided to use only two first principal components. Shown results related to two different
engine rounds. Horizontal line shown in figures can be introduced as a border which
indicates healthy behavior from faulty function of engine.

16
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Fig

Fig

17
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

Using Fig. 14, it can be seen that if standard deviation and average of data driven from
PCA were considered as two extracted features, four categories will be separated
efficiently. However, as you see in Fig. 15, separation of poppet valve fault and healthy
engine is not simple as Fig. 14 because of small difference between features.

Fig

fig

18
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

11. CONCLUSION
In this paper we used PCA to diagnose possible faults in combustion procedure of a
special internal combustion engine called OPV. Comparison of performed method and its
findings with previous literatures shows that vibration analysis of engine by PCA analysis
is done by this research for the first time in Iran. In only research done in subject of using
vibrational analysis for fault detection of internal combustion engine, Fourier transform
and discrete wavelet transformation were used.
In current research we used a method based on compression and of input data set or
dimensionality reduction via PCA analysis. We transformed current data in a space with
fewer dimensions which included all information of input data after calculation of values
and special vector using a special registry and in fact composing their values.
Specifications of this new subset were reconstructed with controllable parameters in order
to categorize features with high accuracy and fewer calculations. Results show that in case
of suitable implementation of PCA, we can perform this method in form of online too.
High accuracy of results due to PCA and high speed of calculations are two mandatory
conditions for using a technique in online situation.

19
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

REFERENCE
[1] M. Zhao, X. Jin, Z. Zhang, B. Li, Fault diagnosis of rolling element bearings via
discriminative subspace learning: Visualization and classification, Expert Systems with
Applications, 2014.
[2] L. Liao, J. Lee, J. A novel method for machine performance degradation assessment
based on fixed cycle features test. Journal of Sound Vibration, 326, 894908, 2009.
[3] K. Elbhbah, J. K. Sinha, Vibration-based condition monitoring of rotating machines
using a machine composite spectrum, Journal of Sound and Vibration 332, 2013.
[4] Y. Guo, T.W. Liu, J. Na, R.F. Fung, Envelope order tracking for fault detection in
rolling element bearings, Journal of Sound and Vibration 331, 2012.
[5] Y. Guo, N. Jing, L. Bin, F. Rong-Fong, Envelope extraction based dimension reduction
for independent component analysis in fault diagnosis of rolling element bearing, Journal
of Sound and Vibration 333, 2014.
[6] A. Alvarez, Fault detection using Principal Component Analysis (PCA) in a
Wastewater Treatment Plant (WWTP), IEEE Transactions on Control & Automation,
2010.

20
College of Engineering, Chengannur

FAULT DETECTION OF AUTOMOBILE ENGINE

[7] S. Mohammadpour, A Survey on Diagnostics Methods for Automotive Engines,

American Control Conference, p985-990, 2011.

21
College of Engineering, Chengannur