U.S. patent application number 13/047517 was filed with the patent office on 2011-09-22 for system and method for damage diagnosis.
This patent application is currently assigned to FUJI JUKOGYO KABUSHIKI KAISHA. Invention is credited to Yoji Okabe, Hideki Soejima.
Application Number | 20110231112 13/047517 |
Document ID | / |
Family ID | 44010056 |
Filed Date | 2011-09-22 |
United States Patent
Application |
20110231112 |
Kind Code |
A1 |
Soejima; Hideki ; et
al. |
September 22, 2011 |
System and Method for Damage Diagnosis
Abstract
The object of the invention is to provide a damage diagnostic
system that uses a damage detection system that obtains propagation
intensity distribution data, which is expanded in the two
dimensions frequency and propagation time, by converting the output
value from an oscillation detection sensor that was obtained when
oscillation is performed by an oscillator, and for one mode or two
or more modes that are selected from the fundamental mode and
higher mode of Lamb waves, obtains certain characteristic values
from the data, for example three indices, which are the slope of
the mode dispersion of the A1 mode (rate of change of the
propagation time with respect to the frequency), the amount of
decrease in the propagation time of the A1 mode, and the amount of
increase in the propagation time of the S0 and S1 modes, and
outputs the measurement results. The measurement results are
displayed on a display device.
Inventors: |
Soejima; Hideki; (Tokyo,
JP) ; Okabe; Yoji; (Tokyo, JP) |
Assignee: |
FUJI JUKOGYO KABUSHIKI
KAISHA
Tokyo
JP
THE UNIVERSITY OF TOKYO
Tokyo
JP
|
Family ID: |
44010056 |
Appl. No.: |
13/047517 |
Filed: |
March 14, 2011 |
Current U.S.
Class: |
702/35 |
Current CPC
Class: |
G01N 29/4436 20130101;
G01N 29/46 20130101; G01N 2291/0289 20130101; G01N 2291/0427
20130101; G01N 29/07 20130101; G01N 2291/011 20130101 |
Class at
Publication: |
702/35 |
International
Class: |
G06F 19/00 20110101
G06F019/00; G01B 17/08 20060101 G01B017/08 |
Foreign Application Data
Date |
Code |
Application Number |
Mar 16, 2010 |
JP |
2010-058784 |
Claims
1. A system for damage diagnosis for diagnosing a damage that
occurred on or within an object, the system comprising: an
oscillator for applying a broadband ultrasonic oscillation to the
object to generate a broadband Lamb wave within the object; an
oscillation detection sensor for detecting the broadband Lamb wave
from the object, the detected broadband Lamb wave including at
least one mode of Lamb wave; and a processing unit, being connected
to the oscillator and the oscillation detection sensor, for (1)
obtaining time-frequency transformation data by performing a
time-frequency transformation to the broadband Lamb wave detected
by the oscillation detection sensor, wherein the time-frequency
transformation data indicates a propagation time of the at least
one mode of Lamb wave, and the propagation time is the time for
Lamb wave to propagate from the oscillator through the oscillation
detection sensor, and (2) identifying, based on the propagation
time of the at least one mode of Lamb wave in the time-frequency
transformation data, whether or not the damage has occurred on or
within the object, and/or identifying the size or length of the
damage that occurred on or within the object.
2. The system for damage diagnosis according to claim 1, wherein
the processing unit, in the identifying process (2), based on
whether or not the propagation time of the at least one mode of
Lamb wave matches a reference value, identifies whether or not the
damage has occurred on or within the object, and/or identifies the
size or length of the damage that occurred on or within the
object.
3. The system for damage diagnosis according to claim 1, wherein
the broadband Lamb wave includes two or more modes of Lamb waves;
and the processing unit, in the identifying process (2), based on
whether or not the propagation time of at least one mode of Lamb
wave of the two or more modes of Lamb waves matches a reference
value, identifies whether or not damage has occurred in the object,
and/or identifies the size or length of the damage that occurred on
or within the object.
4. The system for damage diagnosis according to claim 3, wherein
the identifying process (2) comprises a selection step for
selecting the at least one mode Lamb wave from the two or more
modes of Lamb waves to be compared with the reference value.
5. The system for damage diagnosis according to claim 1, wherein
the time-frequency transformation data obtained by the processing
unit is a two-dimensional propagation intensity distribution data
in which frequency is one of the two dimension and propagation time
is the other.
6. The system for damage diagnosis according to claim 1, wherein
the at least one mode Lamb wave includes a plurality of waves
having mutually different frequencies; and the propagation time of
the at least one mode of Lamb wave is a propagation time of a
maximum intensity portion of at least one of the plurality of
waves.
7. The system for damage diagnosis according to claim 1, wherein
said at least one mode is A1 mode.
8. The system for damage diagnosis according to claim 1, wherein
said at least one mode is S0 mode and S1 mode.
9. The system for damage diagnosis according to claim 1, wherein
said at least one mode is A1 mode, S0 mode, and S0 mode.
10. The system for damage diagnosis according to claim 1, wherein
the at least one mode Lamb wave includes a plurality of waves
having mutually different frequencies; and the processing unit, in
the identifying process (2), calculates propagation times of two of
the plurality of waves, calculates a change ratio of the
propagation times by means of dividing a difference of the two
propagation times by a difference of the frequencies of the two
waves, and based on whether or not the change ratio matches a
reference value, identifies whether or not the damage has occurred
on or within the object, and/or identifies the size or length of
the damage that occurred on or within the object.
11. The system for damage diagnosis according to claim 10, wherein
said at least one mode is A1 mode.
12. The system for damage diagnosis according to claim 4, wherein
the system comprises two oscillators, with one oscillator being
attached to one surface in the thickness direction of the object,
and the other oscillator being attached to the other surface in the
thickness direction of the object; and the processing unit executes
an oscillation control process to control the oscillators, and
executes the oscillation control process and the selection step
under any of the conditions (a) to (c) below, where condition (a)
is such that the processing unit, in the oscillation control
process, controls the two oscillators so that a symmetrical mode
Lamb wave is generated in the object, and selects the symmetric
mode Lamb wave in the selection process; condition (b) is such that
the processing unit, in the oscillation control process, controls
the two oscillators so that an asymmetric mode Lamb wave is
generated in the object, and selects the asymmetric mode Lamb wave
in the selection process; and condition (c) is such that the
processing unit executes the processes under conditions (a) and the
processes under condition (b) at different times.
13. The system for damage diagnosis according to claim 4, wherein
this system comprises two oscillation detection sensors, with one
oscillation detection sensor being attached to one surface in the
thickness direction of the object, and the other oscillation
detection sensor being attached to the other surface in the
thickness direction of the object; and the processing unit executes
the processes (1) and (2) under any one of the conditions (a) to
(c) below; where condition (a) is such that the processing unit, in
the obtaining process (1), creates data in which an asymmetric mode
is canceled out and a symmetric mode is emphasized by adding the
broadband Lamb waves detected by the two oscillation detection
sensors, and then the processing unit obtains time-frequency
transformation data by performing the time-frequency transformation
to the created data, and in the identifying process (2), selects a
symmetric mode Lamb wave; condition (b) is such that the processing
unit, in the obtaining process (1), creates data in which the
symmetric mode is canceled out and the asymmetric mode is
emphasized by subtracting the broadband Lamb waves detected by the
two oscillation detection sensors, and then the processing unit
obtains time-frequency transformation data by performing the
time-frequency transformation to the created data, and in the
identifying process (2), selects an asymmetric mode Lamb wave; and
condition (c) is such that the processing unit executes the
processes under conditions (a) and the processes under condition
(b).
14. The system for damage diagnosis according to claim 1, wherein
the time-frequency transformation is any one of the wavelet
transformation, short-time Fourier transformation, chirplet
transformation, Wigner transformation, and Stockwell
transformation, or a combination of any two or more of said
transformations.
15. The system for damage diagnosis according to claim 1, wherein
the oscillator is attached to the object.
16. The system for damage diagnosis according to claim 1, wherein
the oscillation detection sensor is attached to the object.
17. A method for damage diagnosis for diagnosing damage that
occurred on or within an object, the method using an oscillator for
applying a broadband ultrasonic oscillation to the object to
generate a broadband Lamb wave within the object, an oscillation
detection sensor for detecting the broadband Lamb wave from the
object, the detected broadband Lamb wave including at least one
mode of Lamb wave, and a processing unit being connected to the
oscillator and the oscillation detection sensor, and the method
comprises the steps of: (1) obtaining time-frequency transformation
data by performing a time-frequency transformation to the broadband
Lamb wave detected by the oscillation detection sensor, wherein the
time-frequency transformation data indicates a propagation time of
the at least one mode of Lamp wave, and the propagation time is the
time for Lamb wave to propagate from the oscillator through the
oscillation detection sensor; and (2) identifying, based on the
propagation time of the at least one mode of Lamb wave in the
time-frequency transformation data, whether or not the damage has
occurred on or within the object, and/or identifying the size or
length of damage that occurred on or within the object.
18. The method for damage diagnosis according to claim 17, wherein
the time-frequency transformation is any one of the wavelet
transformation, short-time Fourier transformation, chirplet
transformation, Wigner transformation, and Stockwell
transformation, or a combination of any two or more of said
transformations.
Description
CROSS REFERENCE TO RELATED APPLICATIONS
[0001] This application claims priority under 35 U.S.C. 119 based
upon Japanese Patent Application Serial No. 2010-058784, filed on
Mar. 16, 2010. The entire disclosure of the aforesaid application
is incorporated herein by reference.
FIELD OF THE INVENTION
[0002] The present invention relates to a system and method for
damage diagnosis that uses Lamb waves.
BACKGROUND OF THE INVENTION
[0003] In fields where strength and weight reduction of materials
are required, for example in the field of fuselage of an aircraft,
in order to meet such demands, the use of many composite materials
such as CFRP (Carbon Fiber Reinforced Plastics) is essential. In
order to maintain a high level of reliability of structures made
from such composite materials and to perform more efficient design
work, damage detection technology (health monitoring technology) is
attracting much attention. As devices for performing this kind of
detection of damage and defects in composite materials, there are
damage detection devices as disclosed in patent publication 1 and 2
that use a FBG (Fiber Bragg Grating) optical fiber sensor.
Recently, optical fibers are becoming very thin (for example, 52
.mu.m diameter). As a result, even when embedded in a structure,
there is not much of a decrease in strength of the structure.
Therefore, optical fibers have an advantage that they have a high
degree of freedom regarding placement.
[0004] The inventions disclosed in Japanese patent publication) and
2 below use an optical fiber sensor having a grating section
wherein piezo elements that are fixed and arranged at specified
locations in a structural composite material, lead wires that
transmit signals to the piezo elements, an optical fiber sensor
attached to the structural composite material so that the composite
material of the structural composite material is located between
the optical fiber sensor and the piezo elements, the optical fiber
sensor having a grating section to reflect light of a specified
wavelength to a core section, a light source that shines light on
the core section, and a characteristic detection unit that detects
a characteristic of the reflected light from the grating section,
and vibrating the material by the piezo elements, detects damage
from the change in output from the characteristics detection unit.
A spectrum analyzer that detects the frequency characteristic of
reflected light from the grating section is used as the
characteristic detection unit.
[0005] Furthermore, in the invention disclosed in Japanese Patent
Publication No. 2005-098921, a comparison is performed with
detected data of a normal structural composite material that was
acquired beforehand. Alternatively, another method is disclosed in
which in the frequency distribution that is detected by the
spectrum analyzer, a threshold value is set for the fluctuation
value from when there is no oscillation of a specified frequency,
and when the detected value is equal to or less than that threshold
value, it may be determined that there is damage (paragraph
0032).
[0006] In the invention disclosed in Japanese Patent Publication
No. 2007-232371, two optical filters are provided in a spectrum
analyzer. It is proposed that by outputting reflected light to an
arithmetic processing unit via the two optical filters, the
spectrum analyzer will detect a wavelength oscillation signal of
the reflected light with high sensitivity. It is also proposed that
the arithmetic processing unit will calculate a value (DI value)
that corresponds to the scale of the damage of the test object
based on the obtained wavelength oscillation signal.
[0007] As a method of damage detection technology, research is
being performed regarding a method in which ultrasonic waves called
Lamb waves are generated and detected, and the occurrence of
damages is diagnosed based on the change in the detected waves. The
Lamb wave is an ultrasonic wave that propagates through a thin
plate, and propagates over a long distance with a relatively small
amount of damping. Therefore, it is a form of ultrasonic wave
propagation that is suitable to damage detection. Moreover, Lamb
waves have two characteristics; a multi-mode characteristic and
velocity dispersion characteristic (frequency dependence), and
depending on the plate thickness and frequency, there are plurality
of modes having different speeds. Due to these complex
characteristics, conventionally, damage detection was performed by
using only information about a specific frequency of the Lamb
waves.
SUMMARY OF THE INVENTION
[0008] Considering the above situation, the purpose of the present
invention is to provide a system and method for damage diagnosis
that use the dispersion characteristic of Lamb waves in order to
make it possible to measure the mode dispersion over a broad band
frequencies, make it possible to perform quantitative evaluation of
the peeling length by obtaining more useful information for damage
detection than in conventional technology, and make it possible to
detect and diagnose damages with high precision and high
reliability
[0009] According to a first embodiment of the present invention to
achieve the purpose described above, there is provided
[0010] a system for damage diagnosis for diagnosing a damage that
occurred on or within an object, the system comprising:
[0011] an oscillator for applying a broadband ultrasonic
oscillation to the object to generate a broadband Lamb wave within
the object;
[0012] an oscillation detection sensor for detecting the broadband
Lamb wave from the object, the detected broadband Lamb wave having
at least one mode of Lamb wave; and
[0013] a processing unit, being connected to the oscillator and the
oscillation detection sensor, for
[0014] (1) obtaining a time-frequency transformation data by
performing a time-frequency transformation to the broadband Lamb
wave detected by the oscillation detection sensor, wherein the
time-frequency transformation data indicates a propagation time of
the at least one mode of Lamb wave, and the propagation time is the
time for Lamb wave to propagate from the oscillator through the
oscillation detection sensor, and
[0015] (2) identifying, based on the propagation time of the at
least one mode of Lamb wave in the time-frequency transformation
data, whether or not the damage has occurred on or within the
object, and/or identifying the size or length of the damage that
occurred on or within the object.
[0016] According to a second embodiment of the present invention to
achieve the purpose described above, there is provided
[0017] the system for damage diagnosis according to the first
embodiment, wherein
[0018] the time-frequency transformation data obtained by the
processing unit is a two-dimensional propagation intensity
distribution data in which frequency is one of the two dimension
and propagation time is the other.
[0019] According to a third embodiment of the present invention to
achieve the purpose described above, there is provided
[0020] the system for damage diagnosis according to the first
embodiment, wherein
[0021] the at least one mode Lamb wave includes a plurality of
waves having mutually different frequencies; and
[0022] the propagation time of the at least one mode of Lamb wave
is a propagation time of the maximum intensity portion of at least
one of the plurality of waves.
[0023] According to a fourth embodiment of the present invention to
achieve the purpose described above, there is provided
[0024] the system for damage diagnosis according to the first
embodiment, wherein
[0025] the identifying process (2) comprises a selection step for
selecting the at least one mode Lamb wave among the two or more
modes of Lamb waves to be compared with the reference value.
[0026] According to a fifth embodiment of the present invention to
achieve the purpose described above, there is provided
[0027] the damage diagnostic system according to the first
embodiment, wherein the at least one mode is the S0 mode and S1
mode.
[0028] According to a sixth embodiment of the present invention to
achieve the purpose described above, there is provided
[0029] the damage diagnostic system according to the first
embodiment, wherein the at least one mode is the A1 mode, S0 mode,
and S0 mode.
[0030] According to a seventh embodiment of the present invention
to achieve the purpose described above, there is provided
[0031] the system for damage diagnosis according to the first
embodiment, wherein
[0032] the at least one mode Lamb wave includes a plurality of
waves having mutually different frequencies; and
[0033] the processing unit, in the identifying process (2),
calculates propagation times of two of the plurality of waves,
calculates a change ratio of the propagation times by means of
dividing a difference of the two propagation times by a difference
of the frequencies of the two waves, and based on whether or not
the change ratio matches the reference value, identifies whether or
not damage has occurred on or within the object, and/or identifies
the size or length of the damage that occurred on or within the
object.
[0034] According to an eighth embodiment of the present invention
to achieve the purpose described above, there is provided
[0035] the damage diagnostic system according to the seventh
embodiment, wherein the at least one mode is the A1 mode.
[0036] According to a ninth embodiment of the present invention to
achieve the purpose described above, there is provided
[0037] the system for damage diagnosis according to the fourth
embodiment, wherein
[0038] the system comprises two oscillators, with one oscillator
being attached to one surface in the thickness direction of the
object, and the other oscillator being attached to the other
surface in the thickness direction of the object; and
[0039] the processing unit executes an oscillation control process
to control the oscillators, and executes the oscillation control
process and the selection step under any of the conditions (a) to
(c) below, where
[0040] condition (a) is such that the processing unit, in the
oscillation control process, controls the two oscillators so that a
symmetrical mode Lamb wave is generated in the object, and selects
the symmetric mode Lamb wave in the selection process;
[0041] condition (b) is such that the processing unit, in the
oscillation control process, controls the two oscillators so that a
asymmetric mode Lamb wave is generated in the object, and selects
the asymmetric mode Lamb wave in the selection process; and
[0042] condition (c) is such that the processing unit executes the
processes under conditions (a) and the processes under condition
(b) at different times.
[0043] According to a tenth embodiment of the present invention to
achieve the purpose described above, there is provided
[0044] the system for damage diagnosis according to the fourth
embodiment, wherein
[0045] this system comprises two oscillation detection sensors,
with one oscillation detection sensor being attached to one surface
in the thickness direction of the object, and the other oscillation
detection sensor being attached to the other surface in the
thickness direction of the object; and
[0046] the processing unit executes the processes (1) and (2) under
any one of the conditions (a) to (c) below; where
[0047] condition (a) is such that the processing unit, in the
obtaining process (1), creates data in which the asymmetric mode is
canceled out and the symmetric mode is emphasized by adding the
broadband Lamb waves detected by the two oscillation detection
sensors, and then the processing unit obtains time-frequency
transformation data by performing the time-frequency transformation
to the created data, and in the identifying process (2), selects a
symmetric mode Lamb wave;
[0048] condition (b) is such that the processing unit, in the
obtaining process (1), creates data in which the symmetric mode is
canceled out and the asymmetric mode is emphasized by subtracting
the broadband Lamb waves detected by the two oscillation detection
sensors, and then the processing unit obtains time-frequency
transformation data by performing the time-frequency transformation
to the created data, and in the identifying process (2), selects a
asymmetric mode Lamb wave; and
[0049] condition (c) is such that the processing unit executes the
processes under conditions (a) and the processes under condition
(b).
[0050] According to an eleventh embodiment of the present invention
to achieve the purpose described above, there is provided
[0051] a method for damage diagnosis for diagnosing a damage that
occurred on or within an object, the method using
[0052] an oscillator for applying a broadband ultrasonic
oscillation to the object to generate a broadband Lamb wave within
the object,
[0053] an oscillation detection sensor for detecting the broadband
Lamb wave from the object, the detected broadband Lamb wave having
at least one mode of Lamb wave, and
[0054] a processing unit being connected to the oscillator and the
oscillation detection sensor, and
[0055] the method comprises the steps of:
[0056] (1) obtaining a time-frequency transformation data by
performing a time-frequency transformation to the broadband Lamb
wave detected by the oscillation detection sensor, wherein the
time-frequency transformation data indicates a propagation time of
the at least one mode of Lamp wave, and the propagation time is the
time for Lamb wave to propagate from the oscillator through the
oscillation detection sensor; and
[0057] (2) identifying, based on the propagation time of the at
least one mode of Lamb wave in the time-frequency transformation
data, whether or not the damage has occurred on or within the
object, and/or identifying the size or length of damage that
occurred on or within the object.
[0058] As Lamb waves, there is a Lamb wave of Symmetric mode (S
mode) which has symmetric amplitude relative to the center in the
thickness direction of the oscillation propagation object having a
plate-like shape, and a Lamb wave of Asymmetric mode (A mode) which
has asymmetric amplitude relative to the center of the thickness
direction of the oscillation propagation object. Also, there are
plural n dimension modes (Sn, An) which are respectively higher
dimension modes of the fundamental symmetric mode (S0) and the
fundamental asymmetric mode (A0). Therefore, the waveform of the
Lamb wave becomes complicated.
[0059] In the research conducted by the inventors, a method to
divide the symmetric and asymmetric modes by means of generating
and detecting a broadband Lamb wave was established. As a result of
analyzing each mode using this method, it is found that S1 mode is
transformed to S0 and A1 modes at a peeling portion occurred
between layers, those modes propagate through the peeling portion,
those modes go back to S1 mode again after having passed the
peeling portion, and that S1 mode propagates through the
object.
[0060] Also, it is found that A1 mode is transformed at the peeling
portion to S0 mode which has a propagation speed faster than that
of A1 mode, the S0 mode propagates through the peeling portion, the
S0 mode goes back to A1 mode again after having passed the peeling
portion, and the A1 mode propagates through the object
[0061] Thus, the change of velocity leads to the change of arrival
time. Also, it is found that the arrival time of each mode shows
its particular change in accordance with the length of the peeling
portion.
[0062] Therefore, by obtaining a two-dimensional propagation
intensity distribution data that is expanded 2-dimensinally
according to frequency and propagation time, and by obtaining from
the data, as to a specified mode, a predetermined characteristic
value (index which represents size of the damage) which shows the
change of the arrival time of the objective mode that occurs by the
damage, it becomes possible to determine as to whether or not the
damage has occurred, and as to the size of the damage.
[0063] Other features and advantages of the present invention will
become apparent from the following detailed description, taken in
conjunction with the accompanying drawings, which illustrate, by
way of example, the principles of the invention.
BRIEF DESCRIPTION OF THE DRAWINGS
[0064] FIG. 1 is a diagram illustrating the configuration of a
damage detection system of an embodiment of the present
invention.
[0065] FIG. 2A is a diagram illustrating the construction of an
optical fiber sensor, and FIG. 2B is a line diagram illustrating
the change in the index of refraction of the grating section in the
direction that light advances.
[0066] FIG. 3A is a diagram illustrating the construction of an
optical fiber sensor and the spectrum analyzer that is connected to
the optical fiber sensor of an embodiment of the present invention,
and FIG. 3B is diagram illustrating a spectrum of the passband of
eight optical filters.
[0067] FIG. 4A is a diagram of an input waveform that is inputted
to an optical filter of an embodiment of the present invention,
FIG. 4B is a spectrum of the passbands of two optical filters, and
FIG. 4C is an output waveform of the optical filter.
[0068] FIG. 5 is a block diagram illustrating the control system of
a damage detection system of an embodiment of the present
invention.
[0069] FIG. 6A is an input waveform of an FC actuator related to
testing, and FIG. 6B is a Fourier spectrum of that FC actuator.
[0070] FIG. 7A is a detected wave that was detected by an FBG
sensor related to testing, FIG. 7B is a Fourier spectrum of that
FBG sensor, and FIG. 7C is the wavelet transformation result.
[0071] FIG. 8 is a theoretical dispersion curve of a Lamb wave
under the same conditions as testing.
[0072] FIG. 9 is a concept diagram illustrating a mode separation
method that uses an MFC actuator.
[0073] FIG. 10 is a concept diagram illustrating a mode separation
method that uses an FBG sensor.
[0074] FIGS. 11A and 11B are diagrams illustrating mode
identification results related to testing.
[0075] FIG. 12A is a concept diagram of the mode conversion
behavior of a Lamb wave, FIG. 12B is a theoretical dispersion curve
of a Lamb wave that propagates inside a 2.4 mm thick plate, and
FIG. 12C is a theoretical dispersion curve of a Lamb wave that
propagates inside a 1.7 mm thick plate.
[0076] FIG. 13 is a cross-sectional diagram of a test specimen
related to testing.
[0077] FIG. 14 is a diagram illustrating the mode conversion
behavior of the S mode found from testing.
[0078] FIG. 15 is a diagram illustrating the mode conversion
behavior of the A mode found from testing.
[0079] FIG. 16 is a cross-sectional diagram of a finite-element
analysis model.
[0080] FIG. 17 is a diagram illustrating the mode conversion
behavior of the S mode found from finite-element analysis.
[0081] FIG. 18 is a diagram illustrating the mode conversion
behavior of the A mode found from finite-element analysis.
[0082] FIGS. 19A and 19B are diagrams illustrating the difference
in propagation states, where FIG. 19A is for when the overall
structure is healthy, and FIG. 19B is for when peeling occurs.
[0083] FIG. 20 is a diagram illustrating the difference between the
speed in the A1 mode in a healthy section and in the S0 mode in a
peeling section.
[0084] FIG. 21 is a perspective diagram of a section of a measured
object that is related to detection testing of peeling between
artificial layers.
[0085] FIG. 22 is a diagram illustrating a plot of the time when
the maximum wavelet coefficient occurred in the A1 mode at each
frequency for each of the test specimens of different peeling
lengths.
[0086] FIG. 23 is a diagram illustrating a plot of the time when
the maximum wavelet coefficient occurred in the S0 and S1 modes at
each frequency for each of the test specimens of different peeling
lengths.
[0087] FIG. 24 is a graph illustrating the change in the slope of
the approximation straight line of the measurement data group for
each test specimen in the 250 to 400 kHz range in FIG. 22 with
respect to the peeling length.
[0088] FIG. 25 is a graph illustrating the change in the amount of
decrease in propagation time in the A1 mode at 300 kHz with respect
to the peeling length.
[0089] FIG. 26 is a graph illustrating the change in the amount of
increase in propagation time in the S0 an S1 modes at 350 kHz
(finite-element analysis) and at 400 kHz (testing) with respect to
the peeling length.
DETAILED DESCRIPTION OF THE INVENTION
[0090] In the following, a preferred embodiment of the present
invention will be described in detail with reference to the
accompanying, exemplary diagrams. The following is only an
embodiment and does not limit the present invention.
[0091] [Basic Configuration]
[0092] First, the basic configuration of the damage detection
system of this embodiment is explained below.
[0093] FIG. 1 is a diagram illustrating the configuration of a
damage detection system 10 that detects damage in a structural
composite material Z. In this embodiment, the structural composite
material is the test object for which detection is performed.
[0094] In this embodiment, an MFC (Macro Fiber Composite) actuator
is used as an oscillator for applying Lamb wave type ultrasonic
wave oscillation to a test object. The MFC actuator has ultra thin
rectangular column shaped piezoelectric ceramic lined up in one
direction and embedded in an epoxy resin, with electrodes being
adhered to the upper and lower surfaces, and is capable of causing
a relatively large in-plane strain to occur in one direction.
Because of that characteristic, it is known that an MFC actuator
can also be used as an ultrasonic oscillation element. It is also
possible to apply another kind of oscillation actuator, such as
piezoelectric elements as the oscillator.
[0095] As illustrated in FIG. 1, the damage detection system 10 of
this embodiment comprises: an MFC actuator 21 that is adhered to
the surface section of a structural composite material Z near the
location where damage detection of the structural composite
material Z is to be performed; an optical fiber sensor 30 as an
oscillation detection sensor that is placed near the location where
damage detection of the structural composite material Z is to be
performed; a controller 41 for controlling the MFC actuator 21; a
spectrum analyzer 42 that detects the wavelength characteristic of
reflected light that is obtained from the optical fiber sensor 30;
and an arithmetic processing unit 50 that performs arithmetic
processing of output values from the spectrum analyzer 42. The
power source 43 for the spectrum analyzer 42 is also illustrated in
the figure.
[0096] When a driving voltage is applied from the outside, the MFC
actuator 21 causes a relatively large in-plane strain to occur in
one direction in that plane. Using this, the controller 41 applies
a driving voltage to the MFC actuator 21 in order to apply an
instantaneous oscillation to the structural composite material
Z.
[0097] The optical fiber sensor 30 is an FBG (Fiber Bragg Grating)
optical fiber sensor, and as illustrated in FIG. 2A is made from an
optical fiber 34 having a grating section 33 located inside the
core section 32 that reflects light of a certain wavelength.
[0098] One of the end sections of the optical fiber 34 is connected
to the spectrum analyzer 42, and light covering a specified range
of wavelength bands is irradiated from the light source of that
spectrum analyzer 42 and enters into the core section 32. The light
that enters from the spectrum analyzer 42 propagates through the
core section 32 and only some of the light wavelengths are
reflected by the grating section 33.
[0099] FIG. 2B is a line diagram illustrating change in the
refractive index in the advancement direction of the light through
the core section 32, and in the figure, the range L illustrates the
refractive index in the grating section 33.
[0100] As illustrated in the figure, the grating section 33 is
constructed such that the refractive index of the core section 32
changes at a fixed cycle. The grating section 33 selectively
reflects only light of certain wavelength at the boundary sections
where the refractive index changes.
[0101] Here, the change in the wavelength .DELTA..lamda.B of the
reflected light from the FBG optical fiber sensor is represented by
equation (1) below where n is the effective refractive index of the
core, .LAMBDA. is the grating interval, P11 and P12 are Pockels
coefficients, .nu. is the Poisson's ratio, .epsilon. is the applied
strain, .alpha. is the temperature coefficient of the fiber
material and .DELTA.T is the change in temperature (Alan D. Kersey,
"Fiber Grating Sensors", JOURNAL OF LIGHTWAVE TECHNOLOGY, Vol. 15,
No. 8, 1997).
[ Formula 1 ] .DELTA..lamda. B = 2 n .LAMBDA. ( { 1 - ( n 2 2 ) [ P
12 - v ( P 11 + P 12 ) ] } + [ .alpha. + ( n T ) n ] .DELTA. t ) (
1 ) ##EQU00001##
[0102] Therefore, when oscillation occurs in the grating section
33, the amount of strain .epsilon. in the grating section 33
changes, and as a result, the wavelength of the reflected light
fluctuates according to the amount of strain .epsilon.. As long at
the oscillation is transmitted in a good manner from the
oscillation source, the grating section 33 generates a large
strain, and the amount of change in the wavelength .DELTA..lamda.B
fluctuates a lot, however when the oscillation is not transmitted
in a good manner from the oscillation source, the grating section
33 generates small strain, and the amount of change in the
wavelength .DELTA..lamda.B fluctuates only a little.
[0103] The MFC causes strain orthogonal to the axial direction of
the fibrous piezoelectric element to occur, and the FBG detects
strain in the axial direction that occurred in the fibrous optical
fiber. These elements have a wide frequency characteristic without
having a resonant frequency, and because these elements have strong
directivity, the propagation path is distinct. Using these two
characteristics, the measurement system of this embodiment is able
to allow propagation of broadband Lamb waves having directivity.
The FBG and MFC are both compact and lightweight, are flexible and
have a high failure strain, so can be integrated with a laminated
plate, they will not fail even under large strain, so have high
reliability, and having such characteristics are suitable for use
in structural health monitoring.
[0104] FIG. 3A illustrates an example of the construction of the
optical fiber sensors and the spectrum analyzer that is connected
to the sensors. As illustrated in FIG. 3A, the spectrum analyzer 42
comprises a light source 61, an optical circulator 62, an AWG
module 63 and a photoelectric converter 60. In this embodiment, an
optical fiber 34, in which four optical fiber sensors 30a to 30d
having different reflected wavelengths are arranged in series, is
connected to the spectrum analyzer 42. The minimum structure is the
optical sensors 30.
[0105] The light source 61 is a broadband light source that
includes all of the oscillation areas of the reflected wavelengths
of the optical fiber sensors 30a to 30d. This is so that even when
there is oscillation at the reflected wavelengths of the optical
sensors 30a to 30d due to a Lamb wave, it is always possible to
obtain the fully reflected light.
[0106] The optical circulator 62 causes light from the light source
62 to advance toward the optical fiber sensors 30a to 30d, and
directs the reflected light from the optical fiber sensors 30a to
30d to the optical fiber 69. The reflected light that is guided to
the optical fiber 69 is led to the input port P0 of the AWG module
63.
[0107] The AWG module 63 has an AWG substrate 64. A monolithic
integrated lightwave circuit is formed on the AWG substrate 64. The
lightwave circuit on the AWG substrate 64 has input/output slab
waveguides 65, 66, an array waveguide 67 and an output waveguide
68, and forms eight optical filters having different passbands that
are connected in parallel to the input port P0. The lightwave
circuit on the AWG substrate 64 divides the wavelength multiplexed
input light into different wavelengths by passing the light through
the eight optical filters, and outputs that light in parallel to
eight output ports P1 to P8. However, the actual number of output
ports is not limited to eight.
[0108] Each of the passbands of the optical filters that
corresponds to the eight output ports P1 to P8 are illustrated in
the spectrum in FIG. 3B. For example, in FIG. 3B, the reflected
light that corresponds to the portion where the reflected light
input distribution 70 of the optical fiber sensor 3b that has a
center reflected wavelength of .lamda.2 overlaps passband 71 is
allowed to pass through one of the optical filters and is outputted
to the output port P3, and at the same time, the reflected light
that corresponds to the portion that overlaps passband 72 is
allowed to pass through another optical filter and is outputted to
the output port P4. Similarly, output ports P1 and P2 correspond to
the optical fiber sensor 30a having a center reflected wavelength
of .lamda.1, output ports P5 and P6 correspond to the optical fiber
sensor 30c having a center reflected wavelength of .lamda.3, and
output ports P7 and P8 correspond to the optical fiber sensor 30d
having a center reflected wavelength of .lamda.4, and dividing the
wavelengths is possible in the same way. As described above the
minimum structure is one optical fiber sensor, and in this case two
optical filters are sufficient.
[0109] On behalf of all, the processing that is performed on the
reflected light from one optical fiber sensor 30 will be explained
with reference to FIGA. 4A and 4B.
[0110] As illustrated in FIG. 4B, an input distribution 73T of the
reflected light from the optical fiber sensor 30 appears. When
oscillation is applied by the MFC actuator 21, a Lamb wave, having
the MFC actuator as the oscillation source, propagates through the
structural composite material Z, and the optical fiber sensor 30
causes oscillation at the wavelength of the outputted reflected
light according to the Lamb wave that is transmitted from the
structural composite material Z. The oscillation at this wavelength
is graphically illustrated by input wave 73W in FIG. 4A.
[0111] Due to the oscillation at this wavelength, the reflected
light input distribution 73T illustrated in FIG. 4B alternately
shifts upward and downward a little, so there is a repeated
increase and decrease in the wavelength value.
[0112] At such a wavelength oscillation, 73C in the figure is the
oscillation center having the center wavelength of the reflected
light input distribution 73T. On the other hand, the center
wavelength 75C of the passband 75T of the optical filter is fixed
in the area above the oscillation center 73C.
[0113] Moreover, the center wavelength 75C and center wavelength
74C are fixed at positions that are separated by at least the
amplitude of the wavelength oscillation of the reflected light from
the oscillation center 73C.
[0114] Furthermore, when the reflected light input distribution 73T
is still, the slope 75T-1 on the lower side of the upper passband
75T crosses the slope 73-T on the upper side of the reflected light
input distribution 73T, and the upper passband 75T and the
reflected light input distribution 73T overlap with a width that is
equal to or greater than the amplitude of the wavelength
oscillation.
[0115] Similarly, when the reflected light input distribution is
still, the slope 74T-1 on the upper side of the lower passband 74T
crosses the slope 73T-2 on the lower side of the reflected input
light distribution 73T, and the lower passband 74T and reflected
light input distribution 73T overlap with a width that is equal to
or greater than the amplitude of the wavelength oscillation.
[0116] By fixing the passband 75T and passband 74T with a position
relationship with respect to the reflected light input distribution
73T as described above, it is possible to detect wavelength
oscillation of the reflected light with high sensitivity.
[0117] The upper optical filter allows the reflected light
corresponding to the portion where the reflected light input
distribution 73T overlaps the passband 75T to pass, and outputs the
reflected light. Similarly, the lower optical filter allows the
reflected light corresponding to the portion where the reflected
light input distribution 73T overlaps the passband 74T to pass, and
outputs the reflected light.
[0118] Therefore, when the value of the wavelength of the reflected
light increases and the reflected light input distribution 73T
shifts upward, the output value of the upper optical filter having
passband 75T increases, and the output value of the lower optical
filter having passband 74T decreases. However, when the value of
the wavelength of the reflected light decreases and the reflected
light input distribution 73T shifts downward, the output value of
the upper optical filter having passband 75T decreases, and the
output value of the lower optical filter having passband 74T
increases.
[0119] Consequently, when the change in the center wavelength of
the reflected light oscillates due to the input wave 73W
illustrated in FIG. 4A, the output value of the upper filter having
passband 75T generates the output wave illustrated in FIG. 4C, and
the lower optical filter having passband 74T generates the output
wave 74W illustrated in FIG. 4C. As illustrated in FIG. 4C, the
output wave 74W and output wave 75W have opposite phase wave
motion.
[0120] According to the theory above, the spectrum analyzer 42
illustrated in FIG. 3 outputs lightwaves to the eight output ports
P1 to P8 when oscillated, and these lightwaves are changed to
electrical signals by the photoelectric converter 60 and outputted
to the outside. The output from the spectrum analyzer 42 undergoes
A/D conversion by an interface (not illustrated in the figure) and
inputted to the arithmetic processing unit 50.
[0121] As illustrated in FIG. 5, the arithmetic processing unit 50
comprises a CPU 51 that performs arithmetic processing according to
a program, ROM 52 that stores programs for performing various
processing and control, RAM 53 that becomes a work area that
temporarily stores data and the like for various processing, an
interface 54 that makes it possible to transmit data to or receive
data from the control unit 41, an interface 55 that inputs data
from the spectrum analyzer 42, an image output interface 57 that
converts the display data of the processing results to an image
signal having a format that is suitable to the display 56, and
outputs that signal to the display 56, and a data bus 58 that is
used for transmitting various instructions or data between all of
the components above.
[0122] The damage detection system 10, together with applying
oscillation to a structural composite material Z, which is the
object of damage detection, by way of the MFC actuator that is
placed on the structural composite material, detects whether or not
damage has occurred near the optical fiber sensors 30 according to
the propagation state of the oscillation wave that is detected by
the optical fiber sensors 30. In order to accomplish that, the
arithmetic processing unit 50 executes various functions explained
below by the CPU 51 using the RAM 53 to perform the processing of
the various programs stored in the ROM 52.
[0123] The CPU 51, according to the programs stored in the ROM 52,
controls the operation of the control unit 41 so that a driving
voltage is applied to the MFC actuator 21. When there is a
plurality of MFC actuators 21, any one of the actuators can be
selected as the MFC actuator 21, however, when used as an
oscillation source, for example, it is preferred that an MFC
actuator be selected such that there is a portion between the
optical fiber sensors 30 and the grating section 33 where damage to
the structural composite material Z occurs easily.
[0124] The CPU 51, according to a program stored in the ROM 52,
performs processing of applying a driving voltage, acquiring output
wave data that is outputted in parallel from the spectrum analyzer
42 during the fixed period of oscillation caused by the MFC
actuator 21, and storing the acquired data in the RAM 53.
[0125] The CPU 51, issues control instructions, and by way of the
MFC actuator 21 applies the ultrasonic oscillation of a Lamb wave
to the structural composite material Z, and quantifies and obtains
the difference signal of the output wave 74W and the output wave
75W from the optical filter that is obtained during oscillation.
For example, the difference signal f(t) illustrated in FIG. 7A is
obtained.
[0126] The CPU 51 also performs wavelet conversion of the f(t) data
according to Equation (2). As a result, the f(t) data is converted
to propagation intensity distribution data that is expanded
2-dimensionally according to frequency and propagation time. This
data corresponds to the propagation intensity distribution of the
Lamb wave to the optical fiber sensors 30, and when represented
graphically becomes as illustrated in FIG. 7C.
[Formula 2]
F(a,b)=.intg..sub.-.infin..sup..infin.f(t).omega.*.sub.a,b(t)dt
(2)
[0127] [Damage Detection Operation]
[0128] Using the basic configuration explained above, and further
as illustrated in FIG. 19 or FIG. 21, MFC actuators 21, 21 and
optical fiber sensors 30, 30 are placed at the same positions on
the top and bottom of the structural composite material Z, and the
damage detection operation described below is executed.
[0129] The CPU 51, by causing the top and bottom MFC actuators 21,
21 to generate an oscillating wave in the same phase, applies
oscillation in just the symmetrical mode to the structural
composite material Z, then performs wavelet conversion of the f(t)
data as described above and obtains 2-dimensional expanded data
according to the frequency and propagation time of just the S mode
as illustrated in FIG. 11A. After that, the CPU 15, based on the
theoretical dispersion curve illustrated in FIG. 8, specifies a
mode, such as the S0 mode, S1 mode, S2 mode and the like, and
calculates the propagation time where the maximum wavelet
coefficient value occurs for each frequency in the specified mode.
For example, when specified modes are the S0 mode and S1 mode, as
illustrated in FIG. 23, the relationship between frequency and time
at which the maximum wavelet coefficient value occurs is specified.
This is one characteristic value and one measurement result that is
extracted from the 2-dimensional expanded data. The CPU 51 displays
this on the display 56 as illustrated in FIG. 23.
[0130] The CPU 51 displays the measurement results for a test
object for which the damage state is unknown in the same way as and
together with the measurement results for a structure for which the
damage state is known. The tester references this, and through
comparison, can estimate whether or not there is damage, and to
what extent damage has occurred.
[0131] Alternatively, the propagation time in the S0 mode and the
S1 mode increases as the length of the lamination peeling
increases, so, as illustrated in FIG. 26, the CPU 51 displays the
amount of increased propagation time with respect to the case when
there is no damage. The tester references this and can estimate
whether or not there is damage, and to what extent damage has
occurred.
[0132] The CPU 51 further advances, and based on the multiple
measurement results stored in the ROM 52 for a structure for which
the damage state is known, and the measurement results for a test
object for which the damage state is not known, performs estimation
of the extent of the damage that has occurred in the test object,
and can display those results on the display 56.
[0133] In order to acquire data for the symmetrical mode (S mode),
instead of the method above, by adding the output values of the top
and bottom optical fiber sensors 30, 30, it is possible to obtain
2-dimensional data (propagation intensity distribution data) of
which the asymmetrical mode is cancelled and the symmetrical mode
is emphasized.
[0134] The CPU 51, by causing the top and bottom MFC actuators 21,
21 to generate an oscillating wave having opposite phase, applies
just the asymmetrical mode to the structural composite material Z,
performs wavelet conversion of the f(t) data as described above,
and obtains 2-dimensional expanded data according to the frequency
and propagation time of just the A mode as illustrated in FIG. 11B.
After that, the CPU 51 specifies a mode such as the A0 mode and A1
mode, and calculates the propagation time at which the maximum
wavelet coefficient value occurs for each frequency of the
specified mode. For example, when the specified mode is the A1
mode, the relationship between the frequency and the propagation
time at which the maximum wavelet coefficient value occurs as
illustrated in FIG. 22 is specified. This is one characteristic
value and one measurement result that is extracted from the
2-dimensional expanded data.
[0135] The CPU 51 displays this on the display 56 as illustrated in
FIG. 22. The CPU 51 displays the measurement results for a test
object for which the damage state is not know in the same way as
and together with the measurement results for a structure for which
the damage state is known. The tester references this, and by
making a comparison, is able to estimate whether or not there is
damage, and to what extent damage has occurred.
[0136] Alternatively, the propagation time in the A1 mode is
reduced due to conversion to the S0 mode, which has a faster
propagation time than the A1 mode in the damaged area, so the CPU
51 displays the amount of the reduction in propagation time with
respect to the case in which there is not damage. The tester
references this, and is able to estimate whether or not there is
damage, and to what extent damage has occurred.
[0137] Moreover, the CPU 51 calculates the rate of change in the
frequency with respect to the propagation time for the A1 mode. The
approximation straight line of the measurement data sets of each
test specimen in the 250 to 450 kHz range is calculated, and the
rate of change corresponds to the slope of the approximation
straight line. This also is one characteristic value and one
measurement value that is extracted from 2-dimensional expanded
data. The CPU 51 displays this rate of change (slope) numerically
and in a graph as illustrated in FIG. 4. Here also, the CPU 51
displays measurement results for the test object for which the
state of damage is unknown in the same way as and together with the
measurement results for a structure for which the state of damage
is known. The tester references this, and by making a comparison,
is able to estimate whether or not there is damage, and to what
extent damage has occurred.
[0138] Advancing further, based on the multiple measurement results
stored in ROM 52 for a structure for which the damage state is
known and measurement results for a test object for which the
damage state is unknown, the CPU 51 performs estimation of the
extent of data to the test object, and can display the results on
the display 56. The basic data for estimation calculation is the
amount of increase in propagation time in the S0 mode and S1 mode
described above, and the amount of decrease and rate of change
(slope) in the propagation time in the A1 mode.
[0139] In order to acquire data in the asymmetrical mode (A mode),
instead of the method above, it is possible to obtain 2-dimensional
expanded data (propagation intensity distribution data), in which
the symmetrical mode is canceled out and the asymmetrical mode is
emphasized, by subtracting the output values from the top and
bottom optical fiber sensors 30, 30.
[0140] In the embodiment described above, the difference signal
between the output values of two optical filters was taken to be
the basic data for wavelet conversion, however, the invention is
not limited to this, and the output value of one optical filter
could be taken to be the basic data for wavelet conversion.
[0141] Moreover, in the embodiment above, the maximum peak value of
the wavelet coefficient for a specified mode was calculated,
however, the value of any parameter may be used as long as the
parameter is suitable for use in comparing the acquired Lamb wave
in specified modes.
[0142] Furthermore, in the embodiment described above, wavelet
conversion was applied as the method of conversion for
2-dimensionally expanding the detected values from the optical
sensors according to frequency and propagation time, however, the
present invention is not limited to this, and it is also possible
to apply other conversion methods such as short-time Fourier
transformation, chirplet transformation, Wigner transformation,
Stockwell transformation, or a combination of any two or more of
said transformations.
[0143] [Verification Testing and Analysis]
[0144] Next, as a reference when explaining the theory of the
present invention and when embodying the present invention, a
description of verification performed through testing and analysis
is given below.
1. MODE IDENTIFICATION METHOD (MODE SEPARATION METHOD)
[0145] First, measurement was performed for a quasi-isotropic CFRP
laminated plate (T700S/2500, Toray Industries Inc., [45/0/-45]3s,
thickness: 3.4 mm). The MFC (M-2814-P2, Smart Material Co., Ltd.)
had a length of 6 mm, width of 14 mm and thickness of 0.3 mm, and
the FBG sensor (Fujikura, Ltd.) had a sensor length of 1.5 mm, and
diameter with polyimide coating of 150 .mu.m. Both were adhered to
the surface of the CFRP laminated plate, being separated by 100 mm,
and measurement was performed. Both were adhered to the surface
using Aron Alpha (Konishi Co., Ltd.), which is a Cyanoacrylate type
adhesive. A broadband signal with a hamming window in the first
cycle of an fc=400 kHz sine wave as illustrated in FIG. 6 was used
as the input signal to the MFC. In order to remove the noise from
the received oscillation waveform of the Lamb wave, which is
generated by the MFC, propagates through the laminated plate and is
received by an FBG sensor, averaging was performed by waveforms
measured 32,768 times. After that, signal analysis of the obtained
received waveform was performed, and the mode dispersion
characteristics that were included in the received oscillation wave
were expressed in the time-frequency domain. A complex Morlet
function was used as the window function in signal analysis, and 1D
complex continuous wavelet analysis was performed. The waveform of
the wave received by the FBG sensor, the Fourier spectrum of that
waveform, and the wavelet conversion results are illustrated in
FIG. 7. As a result, it could be confirmed that a wave component
covering a broadband was received without a large peak appearing at
the specified frequency. Moreover, from the wavelet conversion
results, a plurality of modes having different speeds and
frequencies were observed and found to have mode dispersion. Next,
the theoretical dispersion curve for identifying each mode that
appears in the received oscillation wave is derived.
[0146] FIG. 8 illustrates the theoretical dispersion curve that was
derived at the arrival time at a propagation distance of 100 mm in
a 3.4 mm CFRP laminated plate that is the same as used in the
testing above. In this dispersion curve, the arrival time of
high-dimension modes suddenly becomes late, and where the frequency
becomes infinitely large is called the cutoff frequency. By
comparing this theoretical dispersion curve with the wavelet
conversion results of the received oscillation waveform above, it
can be seen that the mode dispersion coincides well between both.
However, in the frequency domain of 300 kHz and greater where a
plurality of modes overlap, it is difficult to identify the mode,
so in order to perform accurate mode identification, it is
necessary to separate these overlapping modes.
[0147] Therefore, a method of adhering both an MFC and FBG sensor
at the same locations on the top and bottom surface of the
laminated plate was used as a method for separating these modes. As
illustrated in FIG. 9, an MFC is adhered at the same location on
the top and bottom surface, and when the MFC generate waves that
are in phase, it is possible to perform oscillation in just the
symmetrical mode. On the other hand, when waves are generated
having opposite phase, it is possible to perform oscillation in
just the asymmetrical mode.
[0148] As illustrated in FIG. 10, an FBG sensor is adhered to the
same position on both the top and bottom surface, and by taking the
sum of the received oscillation waveform of the top and bottom of
the plate, it is possible to separate the symmetrical modes, and by
taking the difference, it is possible to separate the asymmetrical
modes.
[0149] The result of using these two methods to separate the S
(symmetrical) modes and A (asymmetrical) modes, perform wavelet
conversion, and then perform comparison with the theoretical
dispersion curve above is illustrated in FIG. 11. As a result, by
separating the modes, overlapping of a plurality of modes is
eliminated, and the modes could be identified accurately. It was
also confirmed that in the received oscillation waveform there were
the A0, S0, A1, S1 and S2 modes.
[0150] From the results above, it is possible to identify each mode
included in a received Lamb wave by using the mode separation
method above.
2. CAUSE OF CHANGE IN THE PROPAGATION TIME IN A SPECIFIED MODE
(MODE CONVERSION BEHAVIOR IN A SECTION OF PEELING BETWEEN
LAYERS)
[0151] In the previous section, identification of each mode was
performed, and it became possible to understand the mode dispersion
included in the measurement results. Next, the mode conversion
behavior that occurs due to changes in the mode dispersion is
clarified through testing and analysis.
[0152] (1) Mode Conversion Due to Changes in Sheet Thickness at
Peeling Areas
[0153] The propagation speed of a Lamb wave depends on the product
of the frequency and plate thickness, so as the plate thickness
changes, the mode dispersion of a Lamb wave also changes.
Therefore, as illustrated in FIGS. 12A to 12C, when peeling occurs
between layers inside a laminated plate, the plate thickness of the
propagation path at the area of peeling is less than in a healthy
section, so the mode dispersion is different in healthy sections
and peeling sections. Due to this change in mode dispersion, it is
thought that mode conversion occurs in the Lamb wave that
propagated through a healthy section, and propagates through a
peeling section in a different mode form than in a healthy
section.
[0154] For example, in a laminated plate having a plate thickness
of 3.4 mm, there are three modes, A0, S0 and A1 modes, as the
propagation form of a Lamb wave having a frequency of 300 kHz,
however, when peeling occurs between layers in the center of the
laminated plate, the plate thickness in the peeling section changes
to 1.7 mm and there are only two propagation forms, the A0 mode and
S0 mode.
[0155] Therefore, a Lamb wave that propagated though a healthy
section as the A1 mode, undergoes mode conversion in the peeling
section, and propagates as the A0 and S0 modes. However, which mode
the wave will propagate by through the peeling section cannot be
found from the theoretical dispersion curve. Therefore, the actual
mode conversion behavior that occurs in peeling sections between
layers is made clear by performing testing and finite-element
analysis.
[0156] (2) Experiment
[0157] In order to make clear the actual mode conversion behavior
that occurs at the beginning and ending of a peeling section
between layers, a quasi-isotropic CFRP laminated plate (T700S/2500,
Toray Industries Inc., [45/0/-45]3s, thickness: 3.4 mm) was used to
simulate the case in which peeling between layers occurs in the
center in the thickness direction of the plate. Mode identification
of a received Lamb wave is performed by mode separation, so in
order to measure the mode dispersion in a peeling section it is
necessary to adhere an FBG sensor to the interior surface of the
simulated peeling between layers. Therefore, two 1.7 mm thick CFRP
laminated plates were prepared, and after an FBG sensor was adhered
to one, and in order that the surface to which the FBG sensor was
adhered was inside, an epoxy type adhesive, Araldite Standard
(Huntsman Advanced Materials, Inc.) was applied in a 60 mm range
from one end of the plate. The two CFRP laminated plates, in order
to simulate a laminated structure [45/0/-5/90]3s, were made with a
laminated structure [45/0/-45/90]3, and were symmetrically adhered
to the mounting surface. The dimensions of the test specimen are
illustrated in FIG. 13. The width of the plate is 90 mm.
[0158] The MFC (M-2814-P2) that was used had a length of 6 mm, a
width of 14 mm and a thickness of 0.3 mm, and one was adhered to
both the top and bottom surface of the laminated plate. FBG sensors
were adhered to the top and bottom surface of the laminated plate
at two points; a distance 30 mm from the tip end of the MFC where
the plate thickness was 3.4 mm (healthy section), and at a distance
70 mm where the plate thickness was 1.7 mm (peeling section), and
these sensors received the Lamb wave. The FBG sensors (Fujikura
Ltd.) that were used in testing had a sensor length 1.5 mm, and
diameter with polyimide coating of 150 .mu.m. Aron Alpha (Konishi
Co., Ltd.) was used for adhering the elements. The input signal was
a fc=400 kHz sine wave with a hamming window in one cycle, and in
order to remove noise from the received oscillation waveform,
averaging was performed by waveforms measured 32,768 times. The
mode conversion behavior of the S mode that was found by performing
oscillation in just the S (symmetrical) mode using the MFC on both
the top and bottom surfaces is illustrated in FIG. 14, and the mode
conversion behavior of the A mode that was found by performing
oscillation in just the A (asymmetrical) mode is illustrated in
FIG. 15.
[0159] From the results in FIG. 14, in the case of generating
oscillation in only the S mode, only the S0 mode and S1 mode were
observed in the healthy section. From the theoretical dispersion
curve in the case of the 1.7 mm plate thickness illustrated in FIG.
12C, it is seen that in the peeling section, the S1 mode only
existed at 800 kHz or greater, so it is thought that in the healthy
section, the S1 mode undergoes mode conversion in the peeling
section and the wave propagates as another mode. Therefore,
observing the modes that exist in the peeling section, two modes,
the S0 mode and A1 mode, were observed. From this result, at the
start of the peeling section, it was confirmed that "S1
mode.fwdarw.S0 mode.fwdarw.A1 mode" mode conversion occurred.
[0160] From the results in FIG. 15, when oscillation was generated
in only the A mode, only the A0 mode and A1 mode were observed in
the healthy section. From the theoretical dispersion curve in the
case of a 1.7 mm plate thickness illustrated in FIG. 12C, it was
seen that in the peeling section, the A1 mode existed at only 500
kHz or greater, so it is thought that the A1 mode that was observed
in the healthy section undergoes mode conversion in the peeling
section and that the wave propagates as another mode. Therefore,
observing the modes that exist in the peeling section, two modes,
the S0 mode and A1 mode are observed (the arrival time of the A0
mode is late, so is not considered to be an object of mode
conversion). In this peeling section, it is thought that the
observed A1 mode does not undergo mode conversion, but that at 500
kHz or greater, the A1 mode propagates as is. Therefore, from this
result, at the start of the peeling section, it was confirmed that
"A1 mode.fwdarw.S0 mode" mode conversion occurred.
[0161] (3) Verification by Finite-Element Analysis
[0162] In order to verify the mode conversion behavior found from
observation in (2) above, 2D finite-element analysis was performed.
The finite-element model and dimensions are illustrated in FIG. 16.
An LS-DYNA 971 was used for the model construction and analysis.
The elements used for the analysis model were 2D shell elements
(plane strain). The mesh size was 0.125 mm, which was sufficiently
small enough to be able to calculate high-frequency waves with a
short wavelength. Node bonding is performed for the contact
sections between the MFC and CFRP laminated plate, and as in
testing, a sine wave having a frequency of 400 kHz and a hamming
winding in 1 cycle was used as the input waveform to the MFC. With
the LS-DYNA it is not possible to calculate the piezoelectric
effect, so the piezoelectric effect was applied as the coefficient
of thermal expansion in the direction of expansion of the MFC and
simulated. Under the conditions described above, the time history
of the strain in the x direction was calculated at the three
oscillation receiving points illustrated in FIG. 16 (healthy
section: 20 mm propagation distance 20, 3.4 mm plate thickness;
peeling section: 60 mm propagation distance, 1.7 mm plate
thickness; and healthy section (after passing the peeling section):
100 mm propagation distance, 3.4 mm plate thickness), and this was
taken to be the received oscillation waveform. The mode conversion
behavior in the S mode that was found by performing oscillation in
only the S (symmetrical) mode using both the top and bottom MFC is
illustrated in FIG. 17, and the mode conversion behavior in the A
mode that was found by performing oscillation in only the A
(asymmetrical) mode is illustrated in FIG. 18.
[0163] From the results in FIG. 17, at the start of peeling between
layers, it was confirmed that the same "S1 mode S0 mode A1 mode"
mode conversion as in the testing occurred. Moreover, in the
healthy section after passing the peeling between layers, the same
dispersion as in the healthy section before passing the peeling was
observed, and the S0 mode and S1 mode were observed. When this S1
mode propagated through the peeling section and returned to the
healthy section, it is thought that the S0 mode and A1 mode of the
peeling were converted again to the S1 mode in the healthy
section.
[0164] Next, from the results in FIG. 18, it was confirmed that at
the start of the peeling between layers, there was the same "A1
mode.fwdarw.S0 mode" mode conversion as in testing. Moreover, in
the healthy section after passing the peeling between layers, the
same dispersion as in the healthy section before passing the
peeling was observed, and the A0 mode and A1 mode were
observed.
[0165] When this A1 mode propagates through the peeling section and
returns to the healthy section, it is thought that the S0 mode in
the peeling section undergoes mode conversion and is converted
again to the A1 mode in the healthy section.
[0166] The above indicates the validity of the mode conversion
behavior found through testing, and further makes clear the mode
conversion behavior that occurs after the peeling between layers
ends. As a result, it was confirmed that when passing through the
peeling section between layers, the following two mode conversions
exist. [0167] "S1 mode.fwdarw.S0 mode, A1 mode.fwdarw.S1 mode"
[0168] "A1 mode.fwdarw.S0 mode A1 mode"
[0169] Due to this kind of mode conversion behavior, the mode
during propagation through a peeling section and the mode during
propagation though a healthy section differ. For example, as
illustrated in FIG. 19, in the "A1 mode S0 mode A1 mode" mode
conversion, in the healthy sections, propagation in all propagation
paths is in the A1 mode, however, when peeling occurs, propagation
through that area is in the S0 mode. Here, the propagation speeds
in the A1 mode in healthy sections (3.4 mm plate thickness) and in
the S0 mode in peeling section (1.7 mm plate thickness) differ,
with the speed of the S0 mode being faster than that of the A1
mode. Therefore, the arrival time of oscillation in the A1 mode
that is received by the FBG sensor is earlier when peeling has
occurred than when healthy. This change in the arrival time occurs
due to the difference in propagation speeds of a mode propagating
though healthy sections and a mode propagating through peeling
sections, and the length of the peeling. Therefore, by taking this
difference as an index, it is possible to detect peeling between
layers, and to quantitatively evaluate the peeling length.
3. VERIFICATION TESTING AND ANALYSIS OF DETECTION OF ARTIFICIAL
PEELING BETWEEN LAYERS
[0170] (1) Verification Testing
[0171] The effectiveness of the present invention is illustrated by
verifying through testing whether or not there is actually a change
in arrival time of a wave after passing through a peeling
section.
[0172] Therefore, three kinds of laminated plates were made in
which, while forming an isotropic laminated plate such as a
quasi-isotropic CFRP laminated plate, peeling, having peeling
lengths L=20, 40 and 60 mm, was artificially introduced between
layers in the center in the plate thickness direction by embedding
two layers of 50 .mu.m thick Teflon (registered trademark) film
between adjacent 90.degree. layers in the center in the plate
thickness direction. A broadband Lamb wave was caused to propagate
such that it passed through these peeling areas, and the received
oscillation waveform was measured. The testing configuration is
illustrated in FIG. 21. An MFC and FBG sensor were adhered to the
same locations on the top and bottom surfaces of the plate, and
mode separation was performed in the same way as was performed in
the case of clarifying the mode conversion behavior in the previous
section.
[0173] Using this testing configuration, detection of peeling
between artificial layers was tested using a laminated plate in
which artificial peeling, having lengths L=20, 40 and 60 mm, was
introduced. The results of this testing were compared with the
results when a healthy laminated plate (L=0) was measured, and the
change in the arrival time was evaluated.
[0174] In order to do this, after the received oscillation waveform
underwent wavelet conversion, the maximum value of the wavelet
coefficients for each frequency was extracted. When there was
peeling between layers, the amount of change in the time of this
maximum wavelet coefficient from the healthy state corresponds to
the change in arrival time.
[0175] When oscillation was generated in the A mode using the MFC
on both the top and bottom, the "time at which the maximum wavelet
coefficient appeared for each frequency" in the A mode that was
measured by the FBG sensors is illustrated in FIG. 22 for the 200
to 700 kHz A1 mode in which relatively large change occurred in the
arrival time for the cases of L=0, 20, 40 and 60 mm.
[0176] Next, when oscillation was generated in the S mode using the
MFC on both the top and bottom, the "time at which the maximum
wavelet coefficient appeared for each frequency" in the S mode that
was measured by the FBG sensors is illustrated in FIG. 23 for the
400 to 600 kHz S0 and S1 modes in which relatively large change
occurred in the arrival time for the cases of L=0, 20, 40 and 60
mm.
[0177] From FIG. 22 it is observed that as the peeling length
becomes longer, the arrival time of the A1 mode becomes earlier. It
is also observed that there is change in the slope of the mode
dispersion in the 200 to 500 kHz frequency range. As illustrated in
FIG. 20, this means that the closer the frequency is to the cutoff
frequency of the A1 mode (frequency become low), the greater the
difference between the propagation speed of the S0 mode and A1 mode
becomes.
[0178] Moreover, from FIG. 23 it is observed that as the peeling
length becomes long, the arrival time of the S0 and S1 modes
becomes later.
[0179] The results above, show that there is indeed change in the
arrival time when peeling between layers occurs, and that the
present invention is effective.
[0180] (2) Verification Through Finite-Element Analysis
[0181] In section 2.(3) above, the peeling length of a 2D
finite-element analysis model was changed as L=20, 40 and 60 mm,
and analysis was performed using the same testing configuration as
in the testing above. After that, as in the testing, the maximum
amplitude was found for the A1 mode and S0 and S1 modes, and in
observing the change in the arrival time, the results of the
testing (FIG. 22, FIG. 23) coincide and the same change was
observed.
[0182] (3) Quantitative Evaluation of the Peeling Length
[0183] Furthermore, using the change in the arrival time that was
observed from test results in (1) above and from the analysis
results in (2) above, or the slope of the mode dispersion as an
index, it is shown that it is possible to quantitatively evaluate
the peeling length.
[0184] A linearly approximated straight line is calculated from a
plot of maximum amplitude values in the frequency ranges where
change occurred in the arrival time, and using that approximated
straight line, the following indices were found. The indices were
found from test results and analysis results. The results of
plotting the indices for each peeling length are illustrated in
FIG. 24, FIG. 25 and FIG. 26.
[0185] FIG. 24 illustrates the slope of the mode dispersion of the
250 to 400 kHz A1 mode, FIG. 25 illustrates the amount of decrease
in the arrival time of the A1 mode at 300 kHz, and FIG. 26
illustrates the amount of increase in the arrival time of the S0
and S1 modes at 400 kHz (analysis was at 350 kHz).
[0186] From the results in FIG. 24, it is observed that as the
peeling length becomes longer, the slope of the mode dispersion
becomes smaller. Also, from the results of FIG. 25 and FIG. 26, it
is observed that as the peeling length becomes larger, the amount
of decrease in the arrival time of the A1 mode, and the amount of
increase in the arrival time of the S0 and S1 modes becomes
greater. These indices change nearly proportional to the peeling
length. Therefore, it is possible to use these indices to
quantitatively evaluate the peeling length.
4. CONCLUSION
[0187] As described above, first, identification was performed of
each mode of a broadband Lamb wave that is measured in a broadband
ultrasonic transmission system. A method of separating
symmetrical/asymmetrical modes is proposed as a method for doing
this, and it was shown that mode identification is possible by
using this separation method.
[0188] Next, the mode conversion behavior in peeling sections
between layers was clarified through testing and analysis, and it
was confirmed that there are two types of mode conversion behavior,
"S1 mode.fwdarw.S0 mode, A1 mode.fwdarw.S1 mode" and "A1
mode.fwdarw.S0 mode.fwdarw.A1 mode".
[0189] After that, the validity of the peeling detection method of
the present invention, which uses the change in speed of a Lamb
wave due to mode conversion, was verified through testing and
analysis. As a result, it was confirmed that the change in speed in
the peeling sections was observed as the change in arrival
time.
[0190] Finally, it was shown that the peeling length could be
quantitatively evaluated using the slope of the mode dispersion in
the A1 mode, the amount of decrease in the arrival time of the A1
mode and the increase in the arrival time of the S0 and S1 modes as
indices.
[0191] It is to be understood that the above-described embodiments
are illustrative of only a few of the many possible specific
embodiments which can represent applications of the principles of
the invention. Numerous and varied other arrangements can be
readily devised by those skilled in the art without departing from
the spirit and scope of the invention.
* * * * *