Journal of Sound and <ibration (2001) 246(5), 777}793

doi:10.1006/jsvi.2000.3611, available online at http://www.idealibrary.com on

WAVELET ANALYSIS OF VIBRATION SIGNALS OF

AN OVERHANG ROTOR WITH A PROPAGATING
TRANSVERSE CRACK
S. A. ADEWUSI
Mechanical Engineering Department, King Fahd ;niversity of Petroleum and Minerals, P.O. Box 1069,
Dhahran 31261, Saudi Arabia. E-mail: sadewusi@kfupm.edu.sa

AND

B. O. AL-BEDOOR
Mechanical Engineering Department, King Fahd ;niversity of Petroleum and Minerals, P.O. Box 841,
Dhahran 31261, Saudi Arabia, E-mail: bobedoor@kfupm.edu.sa

(Received 28 June 2000, and in ,nal form 1 November 2000)

This paper presents an experimental study of the dynamic response of an overhang rotor
with a propagating transverse crack using the discrete wavelet transform (DWT)*a joint
time frequency analysis technique. Start-up and steady state vibration signatures are
analyzed using Daubechies (Db6) mother wavelet and the results are presented in the form of
scalograms and space-scale energy distribution graphs. The start-up results showed that
crack reduces the critical speed of the rotor system. The steady state results showed that
propagating crack produces changes in vibration amplitudes of frequency scale levels
corresponding to 1X, 2X and 4X harmonics. The vibration amplitude of frequency scale
level corresponding to 1X may increase or decrease depending on the location of the crack
and side load. However, the amplitude of frequency scale level corresponding to 2X
increases continuously as the crack propagates.
 2001 Academic Press

1. INTRODUCTION

Overhang rotors are found in many industrial applications. Although rotors are carefully
designed for fatigue loading and high level of safety by using high-quality materials and
precise manufacturing techniques, catastrophic failures of rotors as a result of cracks can
occur particularly in high-speed rotating machines, in which the rotor is carrying discs,
blades, gears, etc. of considerable weight, which may induce fatigue crack propagation.
Despite the extensive studies of the vibration analysis of cracked rotating shafts for
diagnostic purposes, the problem is still not fully understood and no unique identi"cation
technique is found.
Due to the importance of detecting cracks in rotors, research was directed towards the
use of vibration signals as they provide a non-intrusive detection technique. Many
researchers [1}16] modelled and studied the dynamic response of cracked rotors using
di!erent approaches. The conventional fast Fourier transform (FFT)-based spectral
analysis method was used to analyze steady state vibration signals of cracked rotors. Imam
et al. [4] used a three-dimensional (3-D) "nite element method and a non-linear rotor

0022-460X/01/400777#17 $35.00/0  2001 Academic Press

778 S. A. ADEWUSI AND B. O. AL-BEDOOR

dynamics model to study cracked rotor system vibration and developed an on-line rotor
crack detection and monitoring system. Histogram signature analysis, which is the FFT of
the di!erence between the averaged vibration signals from a cracked and an uncracked
shaft was used. Experimental results showed 1X and 2X vibration frequency harmonics.
Dirr and Schmalhorst [6] used a vibration measurement method and a potential di!erence
method of fatigue crack measurement to study the shape of cracks at di!erent depth during
crack propagation in a rotating shaft. The FFT of experimental vibration signals showed
1X and 2X harmonics. A three-dimensional (3-D) "nite element crack model was also used
to study the bending stress distribution near the crack tip. Wauer [9] carried out
a comprehensive literature survey of the state-of-the art of the dynamics of cracked rotors.
None of the 162 papers cited in reference [9] used wavelet analysis to study vibrations of
cracked rotors. Collins et al. [11] studied a cracked Timoshenko rotor by solving the
six-coupled equations obtained by Wauer [9]. The frequency spectrum of the rotor
response to a periodic axial impulse was also studied. The results showed an increase in the
coupling between axial, torsional and transverse vibration. Diana et al. [12] studied
theoretically and experimentally on-line crack detection for turbo-generator rotors with
a transverse crack. The di!erence between the current vibration and the previous vibration
is used to calculate the vibration forces. The FFT of vibration signals showed 1X and 2X
harmonic components. Dimarogonas and Papadopoulas [13] studied the stability of
cracked rotors in the coupled vibration mode. The frequency spectra of the vibration signal
of a 300 MW steam turbine showed high 2X, X and X vibration components that
 
suggested the existence of deep crack. Wu and Huang [15] studied the dynamic response of
a rotor with a transverse crack by numerically solving the dynamic equations of a cracked
rotor model. FFT of the response at various speeds, crack depths and crack locations
showed 1X and 2X harmonics. Zheng [16] studied numerically the vibration of a rotor
system with a switching crack. He suggested the use of features other than 1X and 2X
harmonics and signal analysis technique other than FFT-based spectral analysis for crack
detection. He used Gabor's analysis on the vibration signals after the 1X and 2X harmonics
were removed. His results showed the presence of transient signals.
Other papers reveal that attention was focused on the study of simply supported cracked
rotors, the use of FFT-based spectral analysis and orbits in the vibration signals analysis of
rotors with transverse crack. It is known that FFT is highly localized only in the frequency
domain and that it cannot handle transient or non-stationary signals. The limitation of the
FFT led to the development of the joint time frequency analysis technique, of which wavelet
transform is a family, giving information in both time and frequency domains.
A propagating crack will change continuously the dynamic properties of rotors and the
resulting vibration signals. Therefore, wavelet analysis, a joint time frequency analysis
technique, will represent the variations in features of vibration signals with time. Spectral
analysis gives an averaged result without giving information on time localization. Wavelet
transform (WT) is increasingly becoming useful in vibration signals analysis. Some of its
unique characteristics are that it can handle both stationary and non-stationary signals, and
can transform any signal directly into time/space and frequency/scale domains, which can
provide detailed information about signal evolution. Mallat [17], Newland [18] and Burrus
et al. [19] discussed the theory of wavelets and their applications in signal analysis.
Newland [20, 21] presented wavelet transform and wavelet maps in vibration analysis.
Onsay and Haddow [22] used wavelet transform to analyze experimental vibration data of
the transient #exural vibrations of an impact excited uniform beam. The results showed the
e$cacy of WT in detecting transient waves in a dispersive medium. Kishimoto et al. [23]
used WT to analyze dispersive waves in a beam. The results showed that wavelet could
decompose strain response into time and frequency components. Hamdan et al. [24] carried
 WAVELET ANALYSIS OF A PROPAGATING CRACK 779
out the comparison of various basic wavelets for the analysis of #ow-induced vibration of
a cylinder in a cross-#ow. Newland [25] demonstrated the application of harmonic wavelets
in time}frequency mapping of transient signals. The advantage of WT over short-time
Fourier transform and Wigner}Ville frequency decomposition methods is that its bandwidth
can be chosen arbitrarily and hence it o!ers a variable Q transform. Aretakis and
Mathioudakis [26] applied wavelet analysis to gas turbine fault diagnosis and compared the
results with Fourier analysis.
This work presents an experimental study on the dynamic response of an overhang rotor
with a propagating transverse crack using the discrete wavelet transform (DWT), a joint time
frequency analysis technique. Start-up and steady state vibration signatures are analyzed
using Daubechies (DbN) mother wavelet and the results are presented in the form of 2-D and
3-D graphs. The results are presented and discussed and some conclusions are extracted.

2. EXPERIMENTAL SET-UP
The equipment used for the experiment include a rotor kit, rotor kit motor speed control,
Data Acquisition Interface Unit (DAIU-208P) with accessories, oscilloscopes, personal
computer (PC), eddy current displacement/proximity probes, automated diagnostics for
rotating equipment (ADRE) for Windows Software and shafts. The experiment is set-up as
shown in Figure 1(a). The proximity probes are connected via auxiliary components to

Figure 1. Experiment rig. (a) Equipment arrangement. (b) Schematic diagram for overhang rotor arrangement.
(c) Crack geometry.
780 S. A. ADEWUSI AND B. O. AL-BEDOOR

oscilloscopes to observe the amplitude}time waveforms and orbits of the vibration signals.
The probes are connected to the DAIU-208P, which is in turn connected to the PC. The
DAIU operation is controlled by the ADRE for Window Software to collect and store
vibration data. The vibration signatures are processed using the Daubachies (DbN) wavelet
transform.
Figure 1(b) is the schematic diagram of the overhang cracked rotor arrangement. The
shaft material is ductile steel bar AISI 4140. The shaft has the following dimensions and
material properties: 10 mm diameter, 540 mm length, mass per unit length of 0)72 kg/m,
Young's modulus of about 200 GPa. Two discs each of mass 0)8 kg and a hanging side load
are "tted on the shaft as shown in Figure 1(b). The hanging side load was applied to aid
crack propagation. The shafts are supported in self-lubricating sleeve bearings. Three shafts
were machined: one without crack and the other two with 4 mm deep v-notch to induce
stress concentration to start propagating the crack. The crack angle is 303 and the crack tip
open displacement (CTOD) was measured with the optical microscope and was found to be
0)117 mm. Figure 1(c) shows the crack geometry.
Three experiments were carried out. The "rst experiment was for the uncracked shaft
with the discs and 8 kg hanging side load located at 7 and 13 cm from the bearing support
number 1 (Figure 1(b)) respectively. The second experiment consisted of the "rst cracked
shaft with 4 mm depth v-notch surface crack located at 1)5 cm from the bearing support
number 1, Figure 1(b). The locations of the discs, 8 kg hanging side load, and bearing
supports are the same as those for the "rst experiment for the uncracked shaft. The third
experiment was for the second cracked shaft with 4 mm deep v-notch surface crack located
at 3)5 cm from the bearing support number 1. The discs and 8 kg hanging side load are
located at 10 and 16 cm from the bearing support number 1 respectively.
The start-up and steady state vibration signals for each experiment are collected and
analyzed with the Daubechies wavelet and the results are presented and discussed in the
following section. The vibration signals in both the vertical and horizontal directions
measured by the proximity probes located close to bearing support number 1 are
considered. The steady state running speed for all experiments was 3500 r.p.m., which is
higher than the "rst critical speed of the rotor system.

3. RESULTS AND DISCUSSIONS

Unlike FFT, which gives frequency information directly, wavelet transform (WT) does
not give frequency information directly. WT gives a dimensionless scale, which can be
related to frequency and dimensionless space, which can be related to time. For this reason,
attention and discussion of results will focus on comparison, rather than absolute values, of
di!erent features of wavelet graphs; namely the 2-D and 3-D plots. Dyadic discrete wavelet
transform (DWT) using Daubechies mother wavelet (Db6) in MATLAB Wavelet Toolbox
[27] is used for the analyses. The relationship between frequency f and wavelet scale a is
fa1/a. For the dyadic discrete wavelet transform, Scale a is de"ned as a"2H, where
j"1, 2, 32 is the frequency level. Each level represents a unique frequency resolution. The
dyadic scale a increases in multiples of 2, e.g., the lowest frequency resolution scale is
2 followed by 4, 8, 16, etc.
Wavelet analysis of start-up and steady state vibration signals are presented in the form
of 2-D and 3-D graphs. Since dyadic scale a is de"ned as 2H, changes in critical speed of the
rotor system that are not multiples of 2 cannot be clearly identi"ed along the scale-axis
(y-axis) of the 2-D graphs of start-up signals. Therefore, small changes in critical speed can
be identi"ed on the start-up 2-D graphs (scalograms) by considering the time or position of
 WAVELET ANALYSIS OF A PROPAGATING CRACK 781

Figure 2. Scalogram of start-up signal for uncracked shaft with 8 kg hanging side load. (a) Vertical.
(b) Horizontal.

resonance along the space (or dimensionless time) axis. During start-up, the rotor speed
increases with time, which is represented as space on wavelet 2-D graphs; hence the position
of the resonance is proportional to the critical speed. The centre point of the position of
resonance for each start-up scalogram is indicated below in each graph.
The steady state vibration response of any system usually contains harmonics and
sub-harmonics of the excitation frequency. Therefore, 1X, 2X, 4X and 8X vibration
harmonics can easily be identi"ed on steady state wavelet scalograms with the highest scale
level corresponding to 1X. The vertical overhang 8 kg side load was attached to the rotor
(Figure 1(b)) to induce stress on the v-notch to aid crack propagation.
782 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 3. 3-D plot of start-up signals for uncracked shaft with 8 kg hanging side load. (a) Vertical.
(b) Horizontal.

In the scalogram (the 2-D plot), the x- and y-axis represent dimensionless time and
frequency levels respectively. The intensity indicates the vibration amplitude as shown by
the intensity scale. The 3-D plots show dimensionless time and frequency level on the plane,
the energy axis indicates the vibration amplitude.

3.1. UNCRACKED SHAFT

Figures 2 and 3 represent the 2-D (scalogram) and the 3-D plot of start-up signals for the
uncracked shaft. Resonance occurred at frequency level 4 and position 18542 in the vertical
direction with wide resonance bandwidth (Figure 2(a)). In the horizontal direction, it
occurred at level 5, position 5833 and level 4, position 17600 (Figure 2(b)). The regions
without any trace in Figure 2(b) correspond to the vibration region with amplitude less than
 WAVELET ANALYSIS OF A PROPAGATING CRACK 783

Figure 4. Scalogram of steady state signals for uncracked shaft with 8 kg hanging side load. (a) Vertical.
(b) Horizontal.

0)5 mil. The contour plot that is used to get the scalograms chooses the intensity range by
default neglecting small amplitude values compared with the maximum value. In the 3-D
plot, (Figure 3) the minimum amplitude is zero, therefore all amplitudes are well
represented.
Figures 4 and 5 present the wavelet analysis results for the steady state vibration of the
uncracked shaft with an 8 kg hanging load and at a speed of 3500 r.p.m. Two prominent
frequency levels, levels 7 and 6 corresponding to 1X and 2X harmonics, respectively, are
present in Figure 4(a) and 5(a). The vibration amplitude at level 7 (1X) is higher than that at
level 6 (2X). In Figures 4(b) and 5(b), only level 6 is prominent. Since this is the highest level,
784 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 5. 3-D plot of steady state signals for uncracked shaft with 8 kg hanging side load. (a) Vertical.
(b) Horizontal.

it corresponds to 1X vibration. The amplitude of vibration in the horizontal direction

(Figure 5(b)) is unsteady. Also, some scattered spots appeared between levels 8 and 10 in
Figure 4(b). These features in the horizontal direction suggest the presence of transient
vibration, which may be due to non-linearity in the system and the gyroscopic e!ect of the
overhang masses.

3.2. CRACKED SHAFT

Scalograms and 3-D graphs of start-up signals for the "rst cracked shaft are presented in
Figures 6 and 7. The resonance position in the vertical direction (Figure 6(a)) occurred at
15417 with a bandwidth smaller than the bandwidth in Figure 3(a). The resonance positions
in the horizontal direction (Figure 6(b)) occurred at 4167 and 15100 for the "rst and second
 WAVELET ANALYSIS OF A PROPAGATING CRACK 785

Figure 6. Scalogram of start-up signals for the "rst shaft with 4 mm notch crack and 8 kg hanging side load.
(a) Vertical. (b) Horizontal.

critical speeds respectively. The resonance bandwidths in the horizontal direction are also
smaller than the bandwidths for the uncracked shaft, Figure 3(b). Comparison of Figures
6 and 7 for the "rst cracked shaft and Figures 3 and 4 for the uncracked shaft show that the
crack decreases the critical speed in both directions, as shown by the positions of resonance.
The results are consistent with the results of Dimaroganas and Paipetis [3]; they reported
that a crack reduces the sti!ness of shaft, which is re#ected in the decrease in critical speed.
The start-up experiment for the "rst cracked shaft was repeated three times, which caused
crack propagation and the shaft fractured after few minutes when rotated at a steady state
786 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 7. 3-D plot of start-up signals for the "rst shaft with 4 mm notch crack and 8 kg hanging side load.
(a) Vertical. (b) Horizontal.

speed of 3500 r.p.m. as shown in Figures 8 and 9. Figures 8 and 9 show scalograms and
3-D graphs of steady state signals for the "rst cracked shaft. Two levels (6 and 5)
corresponding to 1X and 2X harmonics, respectively, are present. Figures 8 and 9 show
that crack propagation produces a tremendous increase in the vibration amplitude of
level 5 (2X harmonic), far greater than level 6 (1X) until the shaft fractured. This observation
agrees with the results reported by Gasch [1], Grabowski [2], Imam et al. [4], Inagaki et al.
[5]. Dirr and Shmalhorst [6], Davies and Mayes [7] and Diana et al. [12]. After the
fracture of the rotor, level 5 (2X) amplitude became very small compared with level 6 (1X)
amplitude.
Figures 10 and 11 represent the scalogram and the 3-D plot of start-up signals for the
second cracked shaft experiment. Resonance occurred at level 4 and point 15625 in the
vertical direction with a small resonance bandwidth (Figure 10(a)) while it occurred at level
5, position 4670 and level 4, position 15330 in the horizontal direction (Figure 10(b)). The
 WAVELET ANALYSIS OF A PROPAGATING CRACK 787

Figure 8. Scalogram of the steady state signals for "rst shaft with 4 mm notch crack and 8 kg hanging side load.
(a) Vertical. (b) Horizontal.

results for the uncracked and the two cracked shafts show that the in#uence of a crack on
the sti!ness and dynamic response of a rotor system depends on the crack location. Gasch
[1] and Meng and Hahn [14] reported the same observation. Figures 12 and 13 show
steady state results for the second cracked shaft. The results show that vibration amplitudes
of levels 6 and 5 corresponding to 2X and 4X harmonics increase continuously as the crack
propagates; the increase in 2X amplitude is, however, higher than 4X amplitude. The
change in amplitude of level 7 (1X) is small compared with levels 6 and 5 corresponding to
2X and 4X harmonics respectively. The wavelet transformation was used in the monitoring
of the vibration of cracked rotors. The results of this study, particularly for the steady state
response of the cracked shaft, have very clearly captured the propagation of the crack in the
rotating shaft in the form of exponential increase in 2X vibration amplitudes in 3-D plots.
788 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 9. 3-D plot of the steady state signals for the "rst shaft with 4 mm notch crack and 8 kg hanging side
load. (a) Vertical. (b) Horizontal.

The rate at which this component (2X) grows can be an excellent way of detecting
propagating cracks.

4. CONCLUSIONS

Wavelet analysis of vibration signals of an overhang rotor with a 4 mm v-notch

propagating transverse crack is presented. The dyadic discrete wavelet transform using
Daubechies (Db6) mother wavelet of start-up and steady-state vibration signals for
uncracked and cracked rotors are presented in the form of 2-D graphs known as scalograms
and 3-D graphs called space-scale energy distribution graphs. Wavelet transform, a joint
time frequency analysis technique, gave a good pictorial representation of the changing
features in vibration signals of a rotor with a propagating crack. The start-up results
 WAVELET ANALYSIS OF A PROPAGATING CRACK 789

Figure 10. Scalogram of start-up signals for the second shaft with 4 mm notch crack and 8 kg hanging side load.
(a) Vertical. (b) Horizontal.

showed that the crack reduces the critical speed of the rotor system. The steady state results
showed that the propagating crack produces continuous changes in vibration amplitudes of
frequency scale levels corresponding to 1X, 2X and 4X. During crack propagation, the
vibration amplitude of the frequency scale level corresponding to 1X may increase or
decrease depending on the location of the crack, while the amplitude of the frequency scale
corresponding to 2X always increases continuously as the crack propagates; this can
790 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 11. 3-D plot of start-up signals for the second shaft with 4 mm notch crack and 8 kg hanging side load.
(a) Vertical. (b) Horizontal.

provide an excellent tool for predicting the existence of a propagating crack. Changes
in amplitudes of 1X and 2X vibration harmonics at a constant running speed are
an important feature that distinguishes a propagating crack from imbalance
and misalignment, which usually show constant amplitude. Wavelet analysis presents
these relative changes in amplitude very clearly; this feature is not present in other
signal-processing techniques that are not localized in time and frequency. Further
studies using mathematical modelling to quantify the rate of change in 2X
vibration amplitude and how this can be related to the changing crack depth are
recommended.
 WAVELET ANALYSIS OF A PROPAGATING CRACK 791

Figure 12. Scalogram of steady state signals for the second shaft with 4 mm notch crack and 8 kg hanging side
load. (a) Vertical. (b) Horizontal.

ACKNOWLEDGMENTS

The authors acknowledge the support of King Fahd University of Petroleum and
Minerals, Dhahran, Saudi Arabia. The support of Mr. Don Bently of Bently Nevada
Corporation, Minden-Nevada, U.S.A. is also acknowledged and appreciated.

REFERENCES

1. R. GASCH 1993 Journal of Sound and <ibration 160, 313}332. A survey of the dynamic behavior of
a simple rotating shaft with a transverse crack.
792 S. A. ADEWUSI AND B. O. AL-BEDOOR

Figure 13. 3-D plot of steady state signals for the second shaft with 4 mm notch crack and 8 kg hanging side
load. (a) Vertical. (b) Horizontal.

2. B. GRABOWSKI 1980 ransactions of the American Society of Mechanical Engineers 102, 141}146.
The vibrational behavior of a turbine rotor containing a transverse crack.
3. A. D. DIMAROGONAS and S. PAIPETIS 1983 Analytical Methods in Rotor Dynamics. London:
Applied Science Publishers.
4. S. H. IMAM, R. J. AZZARO BANKERT and J. SCHEIBEL 1989 Journal of <ibration Acoustics Stress
and Reliability in Design 111, 241}250. Development of an on-line rotor crack detection and
monitoring system.
5. T. INAGAKI, H. KANKI and S. SHIRAKI 1982 Journal of Mechanical Design 104, 345}355.
Transverse vibrations of a general cracked rotor bearing system.
6. B. O. DIRR and B. K. SCHMALHORST 1987 11th ASME Conference on Mechanical <ibration and
Noise, Boston, 607}614. Crack depth analysis of a rotating shaft by vibration measurements.
 WAVELET ANALYSIS OF A PROPAGATING CRACK 793
7. I. W. MAYES and W. G. R. DAVIES 1984 Journal of <ibration Acoustics Stress and Reliability in
Design 106, 139}145. Analysis of the response of a multi-rotor-bearing system containing
a transverse crack in a rotor.
8. D. E. BENTLY and R. F. BOSMANS 1989 Orbit. 10, 8}12. Case study of shaft crack failure.
9. J. WAUER 1990 Applied Mechanics Review 43, 13}17. On the dynamics of cracked rotors:
a literature survey.
10. J. WAUER 1990 International Journal of Solids Structures 26, 901}914. Modeling and formulation
of equations of motion for cracked rotating shafts.
11. K. R. COLLINS and R. H. PLAUT, E. VIA and J. WAUER 1991 ransactions of the American Society
of Mechanical Engineers 113, 74}78. Detection of cracks in rotating Timoshenko shafts using axial
impulses.
12. G. DIANA, N. BACHSCHMID and F. ANGELI 1986 Proceedings of International Conference on
Rotordynamics, okyo, 385}390. An on-line crack detection method for turbogenerator rotors.
13. A. D. DAMAROGONAS and C. A. PAPADOPOULAS 1983 Journal of Sound and <ibration 91, 583}593.
Vibration of cracked shaft in bending.
14. G. MENG and E. J. HAHN 1997 Journal of Engineering for Gas urbine 119, 447}455. Dynamic
response of a cracked rotor with some comments on crack detection.
15. M-C. WU and S.-S. HUANG 1998 Journal of <ibration and Acoustics 120, 551}555. In-plane
vibration and crack detection of a rotating shaft-disk containing a transverse crack.
16. G. T. ZHENG 1998 Journal of Engineering for Gas urbine 120, 149}154. Vibration of a rotor
system with a switching crack and detection of the crack.
17. S. G. MALLAT 1989 IEEE ransactions on Pattern Analysis and Machine Intelligence 11, 674}693.
A theory for multiresolution signal decomposition: the wavelet representation.
18. D. E. NEWLAND 1993 An Introduction to Random <ibrations Spectral and =avelet Analysis.
Singapore: Longman: third edition.
19. C. BURRUS, A. SIDNEY RAMESH and H. GUO 1998 Introduction to =avelets and =avelet
ransforms: A primer. NJ, U.S.A.: Prentice-Hall PTR.
20. D. E. NEWLAND 1994 Journal of <ibration and Acoustics 116, 409}416. Wavelet analysis of
vibration, Part 1: Theory.
21. D. E. NEWLAND 1994 Journal of <ibration and Acoustics 116, 417}425. Wavelet analysis of
vibration. Part 2: Wavelet Maps.
22. T. ONSAY and A. G. HADDOW 1994 Journal of the Acoustical Society of America 95, 1441}1449.
Wavelet transform analysis of transient wave propagation in a dispersive medium.
23. K. KISHIMOTO, H. INOUE, M. HAMADA and T. SHIBUYA 1995 Journal of Applied Mechanics 62,
841}846. Time-frequency analysis of dispersive waves by means of wavelet transform.
24. M. N. HAMDAN, B. A. JUBRAN, N. H. SHABANEH and M. ABU-SAMAK 1996 Journal of Fluids and
Structures 10, 1}19. Comparison of various basic wavelets for the analysis of #ow induced
vibration of a cylinder in cross #ow.
25. D. E. NEWLAND. 1997 Fifth International Congress on Sound and <ibration. Application of
harmonic wavelets to time-frequency mapping. Adelaide, South Australia.
26. N. ARETAKIS and K. MATHIOUDAKIS 1997 ransaction of the American Society of Mechanical
Engineers 119, 870}876. Wavelet analysis for gas turbine fault diagnostics.
27. M. MISITI, Y. MISITI, G. OPPENHEIM and J.-M. POGGI 1996 =avelet oolbox for ;se with
MAAB: the Mathworks Inc. MA, U.S.A.