Journal of Mechanical Engineering and Automation 2016, 6(1): 18-24

DOI: 10.5923/j.jmea.20160601.03

Continuous Wavelet Transform on Diesel Engine

Vibration Condition Monitoring
Yousef Alhouli*, Abdullah Alkhaledi, Abdulaziz Alzayedi, Mohsen Alardhi, Nawaf Alhaifi

Automotive and Marine Engineering Technology Dept. / Collage of Technological Studies, Public Authority of Applied Education and
Training (PAAET), Kuwait

Abstract In this paper a time-frequency analysis method will be applied with respect to engine vibration signal
processing; this is the Continuous Wavelet Transform (CWT), CWT is used to divide a continuous-time function into
wavelets, the continuous wavelet transform possesses the ability to construct a time-frequency representation of a signal that
offers very good time and frequency localization. The mathematical characteristics of the (CWT) techniques is explained and
the (CWT) method is applied to engine vibration signals under different loads, also the cylinder#1 exhaust valve clearance
variations is considered in this paper. The analysis and results of the (CWT) are presented based on faults and engine
vibration signals.
Keywords Diesel Engine, Vibration, Continuous Wavelet Transform, Condition Monitoring, Combustion

transform was in analyzing seismic signals for modeling by

1. Introduction using a combination of translation and dilations of simple,
oscillatory function of finite duration called a wavelet. The
Time-frequency analysis is a novel signal processing early results were related to what is now known as the
method making it possible to see both the time and frequency continuous wavelets transform (CWT). However, wavelets
information in the same time. It displays the combined expressions can be found in earlier work done in several
results from the time and frequency analysis in a fields such as function representation, quantum mechanics
three-dimensional way which plots the amplitudes against and signal processing [1].
the time and frequency axes.
For the applications of engine vibration analysis the
time-frequency method is very useful in that most engine 2. Literature Review
vibration related events such as combustion and valve The development of microprocessors and associated
operations have fixed occurring times determined by the software has totally changed diesel engine condition
crank mechanism. By performing time-frequency analysis monitoring. Condition monitoring of diesel engines used in
these events can be identified according to their occurrence ships has high priority because failure of the engine might
time and frequency locations [1]. result in loss of the vessel. Different faults within the engine
The most commonly used techniques are the Wavelet and associated driven machinery are normally associated
transform and the Winger-Ville distribution. They are with the increase in the level of vibration. Such faults include
superior to the conventional Fourier based frequency abnormal combustion which is due to faulty valve or
analysis by giving better time and frequency resolutions and abnormal injection, wear or partial seizure of plain bearings,
locations. In the frequency analysis, the temporal defects in engine mountings or drive line, damage to gear
information is obscured. To overcome this problem short teeth, wear and pitting of rolling bearings, damage or fouling
time windows for the frequency analysis are used; but the of turbocharger, etc. This paper concentrates only on
main drawback here is that it is not possible to obtain high combustion related faults such as change exhaust valve
resolution in time and frequency simultaneously. Wavelet clearance.
analysis overcomes this limitation and is very useful for Vibration accelerometer may be mounted at strategic
analyzing non stationary signals such as those from locations on the engine. In this paper, the accelerometer is
reciprocating machinery. The beginning of the wavelet located between cylinder #1 and cylinder #2 to sense
abnormal combustion. However the analysis of vibration
* Corresponding author:
[email protected] (Yousef Alhouli)
signals is not an easy task particularly with machines such as
Published online at http://journal.sapub.org/jmea diesel engines which may have high level of background
Copyright © 2016 Scientific & Academic Publishing. All Rights Reserved vibration [2].
 Journal of Mechanical Engineering and Automation 2016, 6(1): 18-24 19

In reference [3] Haddad et al. analyzed and calculated the using Fast Fourier Transform under different operating
piston slap considering all the possible piston position using conditions (loads and speeds). Also using some advanced
engine block vibration. Nurhadi et al. [4] have studied the techniques such as: time-frequency domain analysis using
correlation between the vibrations measured by wavelets.
accelerometers mounted on an engine block with engine
components as the sources for the excitations. In their
experiment the engine was motored by an electric motor 3. Definition of the Continuous Wavelet
through a v-belt, while compression and combustion were Transform (CWT)
alleviated by moving the spark plug. They concluded that
sources of vibrations such as gear pair, chain-sprocket, If the function f (t) is square integrable, its CWT with
gudgeon pin and abnormal application could be identified by respect to a wavelet Ψ (t) is defined as:
∞ 1 𝑡−𝑏
the application of vibration signature analysis. Gu et al. [5] 𝑊(𝑎, 𝑏) = ∫−∞ 𝑓(𝑡) 𝛹∗〔 〕𝑑𝑡 (1)
�|𝑎| 𝑎
demonstrated that common injector faults change the
vibration energy of the injection pulses and on this basis the Where (a) is dilation factor, (b) is the translation factor and
monitoring of the injector via the comparison of monitored Ψ (t) is the mother wavelet or analyzing wavelet. The factor
pulses with a baseline is described. Thomas et al. [6] have 1/ √𝑎 is used for energy normalisation.
also used a pattern recognition technique to detect engine The wavelet transform is a function of two variables.
knock based on the vibration signal. Debotton et al. [7] have Observe that both f (t) and Ψ (t) belong to 𝐿2 (𝑅) the set of
applied the vibration signature analysis method to determine square integrable functions, also called the set of energy
the condition status of an internal combustion engine. They signals. Equation (1) can be written in a more compact form
used accelerometer which is mounted on an operating spark by defining 𝛹 𝑎,𝑏 (𝑡).
ignition engine. Also they used FFT analyzer to transform 𝛹 𝑎,𝑏 (𝑡) =
1
Ψ〔
𝑡−𝑏
〕 (2)
�|𝑎| 𝑎
the measurements from the time domain to the frequency
domain, and the characteristics are analyzed during normal By combining the equations (1) and (2) we get:
and abnormal operation conditions. Grimmelius and Meiler ∞
𝑊(𝑎, 𝑏) = ∫−∞ 𝑓(𝑡) 𝛹 𝑎,𝑏 ∗ (𝑡) 𝑑𝑡 (3)
[8] have developed a feature extraction and pattern
recognition algorithm to detect cylinder miss-fire in a diesel Where
engine by applying a base-level signal analysis and torsion 𝛹 1,0 (𝑡) = 𝛹 (𝑡)
peak value analysis on crankshaft torsional signal. Teraguchi
et al. [9] investigated the effect of oil film between piston The normalising factor of 1/�|𝑎| ensures that the energy
rings and cylinder walls on the induced vibrations. Alhussain stays the same for all a and b, that is:
et al. [10] extracted useful information about the diesel ∞ ∞
engine lubricating oil quality and condition by analyzing the ∫−∞ ⃓⃓𝛹 𝑎,𝑏 (𝑡) ⃓⃓ 2 𝑑(𝑡) = ∫−∞ ⃒Ψ(t)2 ⃒ 𝑑𝑡 (4)
measured vibration and airborne acoustic signals caused by For all a and b. For any given value of a, the function
piston slap and investigated the effects of load, speed 𝛹 𝑎,𝑏 (𝑡) is a shift of 𝛹 𝑎,0 (𝑡) by an amount b along the time
variation and temperatures. They found that the peaks on the axis. Thus, the variable b represents time shift or translation.
frequency range between 1 kHz and 3 kHz associated with From equation (5).
the oil level and quality, in other words the piston slap 1 𝑡
𝛹 𝑎,0 (𝑡) = Ψ〔 〕 (5)
excitation. Alhussain et al. [11] outlined the diesel engine �|𝑎| 𝑎
acoustic sources and their fundamental characteristics are It follows that 𝛹 𝑎,0 (𝑡) is a time-scaled and
studied in time domain, frequency domain, mean value, amplitude-scaled version of 𝛹 (𝑡) . Since (a) determines
kurtosis, and RMS values. Also they detected and diagnosed the amount of time scaling of dilation, it is referred to as the
exhaust valve clearance fault by monitoring the spikeness scale or dilation variable. If a >1, there is a contraction of
level using the RMS values of the acoustic signals kurtosis 𝛹 (𝑡) along the time axis, whereas if 0< a <1, there is a
for each cylinder. This research study focuses on vibration contraction of 𝛹 (𝑡). Since the CWT is generated using
monitoring techniques and characteristics of the diesel dilates and translates of the signal function 𝛹 (𝑡) , the
engine to detect and diagnose some combustion related faults wavelet for the transform is referred to as the mother wavelet
in diesel engine. In this research vibration data collected [1].
from a Ford FSD diesel engine are used for the detection and Continuous wavelet transform has got some properties
diagnosis of specific incipient engine related faults. The real which make it better than other time-frequency techniques in
data will be collected from the engine test rig operating under analysing such impulsive signals like diesel engine vibration
normal and abnormal conditions. In the abnormal condition signals. These properties are higher resolution and better
the fault in the engine will be on cylinder #1 change exhaust localization characteristics. Also compared with the discrete
valve clearance. The vibration signals will be processed wavelet transform (DWT), (CWT) is better for signal
using conventional vibration signal processing techniques detection and feature extraction [12]. Also (CWT) has been
namely, time domain (waveforms comparison, RMS value, reported to give a non-linear representation of the frequency
kurtosis and skweness) and spectral analysis techniques content of a signal, this is why it is a good tool in detecting
20 Yousef Alhouli et al.: Continuous Wavelet Transform on Diesel Engine Vibration Condition Monitoring

small transient events. It can also be implemented over based on the vibration of engine block using conventional
several frequency scales [13]. and advanced analysis process techniques.
1. Setup a modern vibration monitoring system.
2. Use time domain to represent the vibration signals and
4. Experiment Setup relate it to the engine events.
A four stroke four cylinders is tested, in-line OHV, direct 3. Study the effects of load and speed variation on these
injection, FSD 425 type Ford production diesel engine is signals; loads of 0,20,40,60 Nm and speeds of 1000,
used in this study. This engine is widely used in generators 1500 rpm.
and commercial vehicles. For the sake of exhaust noise 4. Seed specific and quantified fault such as exhaust valve
reduction, two exhaust mufflers are attached to the exhaust clearance (from 0.4 to 0.0 mm).
pipe. Table (1) lists its technical specifications and Figure (1)
shows the test rig.
Table 1. Specifications of the test rig

Bore 93.67 mm
Stroke 90.54mm
Cubic capacity 2496 mm3
Power output 52kW @ 2700 rpm (BS)
Torque 145Nm @ 2700 rpm (BS)
Injection sequence 1,2,4,3
Compression ratio 19:1
Inlet valve timing Opens 130 before TDC
Figure 1. Diesel engine test rig
Exhaust valve timing Opens 51o before BDC, Closes 13o after TDC
Compression pressure 3.38 MPa @ starter motor speed The data acquisition system used to capture the data from
the engine featured the following components; In Figure (2)
To detect and diagnose some combustion related faults below the connection between the components can be seen.

Figure 2. Monitoring system wiring diagram

 Journal of Mechanical Engineering and Automation 2016, 6(1): 18-24 21

A top mount type 4368 accelerometer is located between

cylinder #1 and #2, figure (3) shows the accelerometer
position between the first and the second cylinders.

Figure 4. Exhaust valve clearance alteration

5. CWT of Engine Vibration Analysis

The spectrum analysis was conducted on the vibration
data collected under four operational load conditions of 0Nm,
20 Nm, 40 Nm, and 60 Nm. The engine was operated at
constant speed of 1000 rpm and the exhaust valve clearance
was changed from 0.4mm (Healthy condition) to 0.0 mm
Figure 3. Vibration accelerator type 4368 location
(Faulty condition). Then the engine was operation on the
The main target of this study is to acquire a real data from same conditions at speed of 1500 rpm.
the engine test rig operating under normal and abnormal Figure 5 shows the Continuous Wavelet Transform (CWT)
conditions and apply signal processing techniques for the of the diesel engine vibration. From the (CWT)
purpose of condition monitoring. For this reason, the engine representation it can be seen clearly four peaks representing
was tested under different loading conditions of 0, 20, 40 and the combustion events of the engine cylinders in the firing
60 Nm at different speeds of 1000 and 1500 rpm. Only frequency from left to right (1, 2, 4, and 3).
exhaust valve clearance faults were investigated. As shown The spectral analysis shows that the major part of the
in Figure (4); cylinder #1 exhaust valve clearance was energy is located in the lower frequencies (below 5 kHz), this
altered from 0.4mm (healthy clearance) to 0 mm, 0.25mm, can be seen more clearly in the CWT representation, and also
and 0.6mm. This simulates leakage to some extent, exhaust it can be seen that the peak of the CWT extends to around 35
valve timing and in-cylinder pressure. kHz.

Figure 5. CWT at 0Nm load, 1000 rpm and 0.4 mm clearance

22 Yousef Alhouli et al.: Continuous Wavelet Transform on Diesel Engine Vibration Condition Monitoring

Figure 6. CWT at 0, 20, 40, 60 Nm load, 1000 rpm and 0.4 mm clearance

Figure 7. CWT at 0, 20, 40, 60 Nm load, 1500 rpm and 0.4 mm clearance
 Journal of Mechanical Engineering and Automation 2016, 6(1): 18-24 23

Figure 8. CWT contour plots at 1000 rpm and 20 Nm load with the engine running (a) healthy Exhaust valve clearance 0.4mm, (b) Exhaust valve clearance
0.0mm, (c) Exhaust valve clearance 0.25mm, (d) Exhaust valve clearance 0.6mm

Figures 6 and 7 represent CWT of the engine vibration and to 0.6mm (Figure 8 (d)).
signals for different loads and 1000rpm engine speed, the
figures show the frequency with crank angle at healthy case
and vibration amplitude in meter, as load increases vibration 7. Conclusions
amplitude increases.
By changing the operation loads from 0 Nm to 20 Nm to The study of diesel engine vibration signals can be
40 Nm to 60 Nm at two different speeds 1000 rpm and 1500 investigated in the time-frequency domain by different
rpm with 0.4 clearance (healthy condition), the heights of the methods such as Winger-Ville distribution, conventional
combustion peaks are proportional to the engine load and Fourier based frequency analysis, and Wavelet transform.
speed confirming that the engine vibration signals are load Continuous wavelet transform was chosen in this study to
and speed dependent. because it has got some properties which make it better than
other time-frequency techniques in analyzing such impulsive
signals like diesel engine vibration signals. These properties
6. Fault Detection and Diagnosis are higher resolution and better localization characteristics.
CWT time-frequency analysis technique is capable of
According to the literature review the CWT is able to revealing conditions indicating information embedded in the
represent both low and high frequency bands energy levels vibration signals in which the time, frequency and amplitude
with good resolution. Figure 8 shows the CWT of the engine information can be observed. The CWT determines the
vibration signals based on four exhaust valves clearance sets; signal's energy bands, and the results of the analysis of the
the healthy 0.4mm contour shapes increase in the vibration engine vibration signals confirm that the CWT is most
signal around zero degree (TDC) at the combustion onset as sensitive to low frequency information. CWT was successful
shown in Figure 8 (a). Contour shapes, higher amplitude in detecting the exhaust valve clearance faults in diesel
combustion excitation and a three degree delay are observed engine.
when the exhaust valve changing from 0.4mm to 0.0mm Continuous Wavelet Transform (CWT) was the most
(Figure 8 (b)) which indicated that there is fault in system. powerful vibration technique for fault detection and
The symptoms became clearer when the exhaust valve diagnosis, and showed distinctive capability over time and
clearance increasing from 0.0mm to 0.25mm (Figure 8 (c)) frequency domain.
24 Yousef Alhouli et al.: Continuous Wavelet Transform on Diesel Engine Vibration Condition Monitoring

[7] G. Debotton, J. Ben-Ari, R. Itzhaki and E. Sher, Vibration

Signature Analysis as a fault Detection Method for SI Engine,
1998, SAE publication 980115.
REFERENCES [8] H.T. Grimmelius, Meiler, P.P, H.L.M.M., H.L. Mass,
Bonnier, J.S. Grevink and R.F. Van Kuilenburg, Three
[1] Reghuveer M. Rao. and Ajit S. Bopardikar. Wavelet Stat-of-the-Art Methods for Condition Monitoring, IEEE
Transform, Add Longman, Inc. 1998, ISBN 0-201-63463-5. Transaction on Industrial Electronics, Issue2, 1999.
[2] Bernard Challen and Rodica Baranescu, Diesel Engine [9] S. Teraguchi, W. Suzuki and M. Takiguchi, Effects of
reference Book, 3rd Edition, Society of Automotive Engineer, lubricating oil supply on reductions of piston slap vibration
1999, ISBN 0 7506 2176 1, pp.667-670. and piston friction, 2001, SAE Technical Paper Series
2001-01-0566.
[3] S. D. Haddad and D. A. Howard, ‘Analysis of piston
slap-induced noise and assessment of some methods of [10] A. Alhussain, F. GU, A. Ball and A. Starr, Internal
control in diesel engine’, SAE Paper 800517,1980. combustion engine lubricating oil condition monitoring based
on vibro-acoustic measurements, University of Manchester,
[4] Nurhadi, Bagiasna and Wedianto, Signature Analysis of
July 2007, Dol:10.1784/insi.2007.49.12.715.
4-stroke 1-cylinder engine, Society of Automobile Engineers,
Warrendale, PA, 1993, SAE technical Paper 932011. [11] A. Alhussain, A. Ball and A. Starr, On acoustic
measurement-based condition monitoring of internal
[5] F. Gu, A. D. Ball and K. K. Rao. Diesel injection dynamic
combustion engines, Junury 2008, Dol:10.1784/insi.2008.
modeling and estimation of injection parameters from impact
50.30.
response. Part 2: Prediction of injection parameters from
monitored vibration. Proceeding Of Mechanical Engineers [12] W, Li PhD, Study of Noise for Diesel Engine. University of
Part D-Journal Of Automobile Engineering, University of Manchester 2000.
Manchester, 1996.
[13] F. Gu, W. Li, A.D. Ball and A.Y.T. Leung, The condition
[6] J.H Thomas and B. A Dubuission, diagnostic method using monitoring of diesel engine using Acoustic measurement,
wavelets networks application to engine knock detection. In Part 1, Society of Automotive Engineers, 2000.
IEEE second International Conference on System, Man and
Cybernetics, 1996.