Acoustic Emission Measurement by Fiber Bragg Grating Glued to Cylindrical Sensor

Fiber Bragg grating (FBG) is an optical fiber sensor and is suitable for acoustic emission (AE) measurements of structures such as chemical plants that require water resistance, explosion proofing, and so forth, However, directionality has been a problem in conventional usage which glues the grating range of an FBG to the inspection surface. This study proposes a cylinder-FBG (C-FBG) sensor with the FBG glued to a cylindrical sensor holder. Effectiveness of the sensor was evaluated by characterizing directionality and frequency sensitivity of the sensor and by comparing the detected AE signals in a 3-point bending test of a carbon fiber reinforced plastic (CFRP) laminate against a conventional wide-band piezoelectric AE sensor. The features of the detected signals were similar to the signals detected by the wide-band AE sensor. These results show that it is possible to produce a wide-band cylindrical sensor without the directionality limitations of the conventional usage of FBG.

In particular, a group at Aoyama Gakuin University has developed a Mach-Zehnder measurement system with a phase-compensation feedback control circuit and reported that measurements with high sensitivity are possible [11].
They also proposed resonance type sensors where the sensing fiber is wound on a cylindrical holder.Applications of the above-mentioned system have been reported [12][13][14].However, the shape and the dimensions of the holder are constrained because it is difficult to wind the fiber around holders with small diameters, restricting the manufacture of high-frequency or wide-band sensors.
This study uses fiber Bragg grating (FBG) that can simultaneously measure AE and strain by gluing the detector onto the inspection object [15][16][17][18][19].However, directionality has been a problem in this method because FBG can only detect vibration along the fiber length and cannot detect vibration perpendicular to the grating length.This is improved by using a cylindrical design, resulting in cylinder-FBG (C-FBG) sensors, in which the sensing part (grating length) of the FBG is glued to a cylindrical holder.By using this method, the holder can vibrate with elastic waves from every direction.As a result, the C-FBG sensor can detect waves from all directions because the grating expands and contracts.

Experimental Equipment and Methods
Several aspects of the proposed sensor first must be validated including directionality and frequency sensitivity.The proposed sensor is then compared to a conventional piezoelectric AE sensor using a 3-point bending test of carbon fiber reinforced plastic (CFRP) laminate.in central Bragg wavelength is positive or negative when the FBG is extended or compressed, respectively, by the ultrasonic vibration such as AE.The intensity of the reflected light changes according to the magnitude and direction of the wavelength shift.The transmitted and reflected light reaches a photodetector, and the Bragg shift is converted into a voltage signal [15][16][17][18][19].The signal is stored in a personal computer via an amplifier and a band-pass filter.

Experiment on Directionality and Frequency Sensitivity
Characteristics.The C-FBG sensor is glued to an aluminum plate.Using pencil lead breaks [21] at angle  as shown in Figure 3, the ability of the sensor to detect these AE signals is evaluated.The diameter of lead is 0.5 mm.Directionality is evaluated by the S/N ratio where consistent S/N ratios at every angle  indicate no influence of direction.Figure 4 shows the scheme for evaluating the frequency sensitivity characteristics of the C-FBG sensor.The sensor and the vibrator contact each other as shown in the figure by vacuum grease.As the vibrator source, a wide-band piezoelectric AE sensor (AE-900S-WB: NF Corp.) with a relatively flat frequency characteristic was adopted, and the vibration face uniformly vibrates.Also, the frequency characteristics of the C-FBG sensor are evaluated by comparing the input and the detected frequencies.Here, the input frequencies to the vibrator range from 100 kHz to 1 MHz via a function generator, and the interval is 10 kHz.The C-FBG sensor detects the vibration of the transmitted sine wave from the vibration face.The frequency range was determined based on the guaranteed frequency characteristics of the adopted vibrator.The measuring conditions of the signals are shown in Figure 4.
In the evaluation of the C-FBG sensor sensitivity, the S/N ratio was used.The sensor is also evaluated by comparison of the input frequency and the detected frequency.Therefore, it is necessary to confirm whether the frequency   of the function generator corresponds with the frequency  of the vibration face of the vibrator when the S/N ratio of the C-FBG sensor is evaluated with the method shown in Figure 4.Then, the frequency  was measured by the method shown in Figure 5.The technique uses optical acoustic imaging methods and is the visualization and the measurement method of the sound pressure distribution of ultrasonic waves by holographic interferometry where water is utilized as the pressure sensitive medium [22].The simple explanation of this measurement theory is as follows.
The refraction index of water in the sound field changes by the sound pressure Δp of the ultrasonic wave when the wave oscillates in water.The optical path length of the laser beam (i.e., wavelength   ) changes by the refractive index variation when the laser beam is transmitted through the sound field.The optical path length change of the laser beam that is transmitted through the water is shown in the following equation where the and -axes are the directions of the ultrasonic wave emission and the laser irradiation, respectively, where Δ(, , ) is the amount of phase shift at time ,  is the distance which the laser beam crosses across the sound field, and / is the refractive index variation of the water due to the pressure.
By the interference of the light wave transmitted through this field and the regeneration light wave of the hologram that recorded the condition without ultrasonic wave, a holographic interference fringe appears.The fringe is photographed by CCD camera.By fringe analysis, the distribution of the phase shift Δ(, , ) is determined from the obtained interferogram as shown in Figure 6.It is possible to calculate the average sound pressure distribution Δ(, , , ) from the obtained Δ(, , ) by (2) when the sound pressure distribution of the direction of the optical path  is uniform.Figure 6 is an example of the abovementioned procedure.In the figure, the result of the frequency distribution with FFT results which processed the data is shown: Figure 7 shows the relationship between the frequency of vibration face (i.e., the calculation results) and that of function generator (input).In this figure, the 45 degree line denotes when the input frequency of the function generator and the output frequency of vibrating surface of the vibrator agree perfectly.From this result, the frequencies of the generator correspond approximately to those of the vibration face as the correlation coefficient shows 0.9998.The maximum negative and positive errors are −7.837 and +19.959 kHz, respectively, at generator frequencies of 500 and 840 kHz, respectively.Therefore, the measured vibration face's frequencies were used as input frequencies  to the C-FBG sensors.FFT analysis result of the information that converted the phase data into the sound pressure data (e.g., 500 kHz)  The emitted AE signals during the test are detected by the C-FBG sensor and the wide-band piezo-electric AE sensor (AE-900S-WB: NF Corp.).The measurement conditions are shown in Figure 8.The amplifiers of both sensors are different because the sensitivity of piezo-type sensor is higher than the CFB sensor.Here, the indentation speed is 1 mm/min, and both sensors are glued by vacuum grease at the position of 50 mm from the indenter.The analysis results of each sensor are compared.Evaluation parameters and methods used are AE events, FFT, WMFD (weighted mean frequency distribution [23]), and WT (wavelet transform [24][25][26]).WMFD assembles many FFT analysis results into one result and is useful for extracting the features of the detection results.On the other hand, WT is a time-frequency analysis, enabling the capture of the generation time and the transition conditions of each frequency component.

Experimental Results and Discussion
3.1.Direction and Frequency Sensitivity Characteristics.Figure 9 shows an example of the S/N ratio of the detected signals at each incidence angle  as defined in Figure 3.
Here, twenty measurements were obtained at each angle.Figure 9(a) shows the detected waveforms, and the amplitude of the waveforms fluctuates minimally with changing angle.Therefore, it is estimated that the C-FBG sensor shows low directionality.Figure 9(b) shows the average of the S/N ratio of the detected waveforms.Therefore, it is also clear that the directionality of the FBG has been removed by the holder because the ratios are constant at every angle .show the center of gravity on the frequency axis.The values show the relationship between input frequency and detected frequency; the detected frequencies are almost equal to the three input frequencies.Figure 12: Relationship between load and AE event counts by each sensor in 3-point bending test of CFRP laminates.The continuous and broken lines show the load and the cumulative AE event count, respectively.The cumulative events detected by both sensors are similar.The cumulative event count of the PZT sensor (i.e., AE sensor) is higher than that of C-FBG sensor because the sensitivity of the former is higher than that of the latter.
many points along the 45 ∘ line at frequencies below 700 kHz (correlation coefficient is 0.9965).The maximum negative and positive errors are −82.562and +33.238 kHz, respectively, at generator frequencies of 1017.137 and 383.887 kHz, respectively.
Figure 11 shows the frequency sensitivity of the C-FBG sensor using S/N ratios.From these results, the following things became clear.The S/N ratio does not fluctuate much from the mean, and the Sd is small (2.847).
From the results of Figures 10 and 11, the sensor shows features typical of wide-band sensors because the S/N ratios do not fluctuate at each input frequency, and the detected frequencies almost equal the input frequencies.

Results of Three-Point Bending Test of CFRP Laminate.
In order to verify the accuracy of the AE measurements of the C-FBG sensor, the features of the detected signals by both sensors (C-FBG and the wide band piezo-type AE sensor) were compared.
Figure 12 shows the transitions of the cumulative AE event count and load in the testing, and they were classified into 4 regions.Here, the continuous and the dotted lines show the load and the count, respectively.Region I: few AE events are detected.
Region II: the slope of the AE events line is greater than that in Region I, and the slope is similar to that of the load.This region continues, until the slope of the events line begins to increase further.
Region III: this region extends until 95% of the maximum load is achieved.
Region IV: the load begins to increase further and reaches the maximum value.This region continues, until the specimen fractures.
From the results, the events' counts by the C-FBG sensor are similar to those by the AE sensor.
Figures 13 and 14 show examples of the detected signal of each region shown in Figure 12 and its FFT, its WT result and WMFD in order to compare the frequency components of the signals for each sensor.The signals were detected at the same time by both sensors.The features of the components in the WMFD of the detected signals by the C-FBG sensor within each region are similar to those of the AE sensor.Therefore, these results show that the detected components in each region of Figure 13 resemble those in each region of Figure 14.
From the WT results of each region shown in Figures 13  and 14, it is considered that the character of both sensors is similar.The duration of the wavelets in the high frequency component (i.e., 200-700 kHz) is short, and that of the wavelets in the low frequency (i.e., 60-80 kHz) is long.The C-FBG sensor is able to grasp the time information of the emitted frequency components as well as the wide band

Figure 1 :Figure 2 :
Figure 1: Dimension of C-FBG sensor.FBG is glued to a hollow cylinder with one closed end.

Figure 1
shows the design of the proposed C-FBG sensor.Optical fiber (i.e., FBG) is glued to a hollow cylinder with one closed end.The closed end of the cylinder contacts the inspection surface by vacuum grease.The central Bragg wavelength of the FBG (Broptics Technology Corp.) is 1550 nm ± 0.2 nm with grating length of 10 mm.The entire grating length of the FBG is glued as shown in Figure1along with the dimensions of the cylindrical holder.The glued length is 15 mm.Because the end of the cylindrical holder minimizes the effect of large deformations of the inspection surface, the central Bragg wavelength does not undergo dramatic shifts.Consequently, the C-FBG can detect vibration with uniform sensitivity.

Figure 2
shows the measurement equipment for detecting the AE signals with FBG.The light from a laser source travels to an FBG via an isolator and a circulator.The shift Advances in Materials Science and Engineering 3

Figure 3 : 5 Figure 4 :
Figure 3: Evaluation for directionality of C-FBG sensor.The emitted elastic wave by pencil lead break at each position of circle mark is detected by a C-FBG sensor that was set at the position of the square.The directionality of the sensor is removed if the features of the detected waveforms at each position are similar.

Figure 5 :
Figure 5: Basic configuration of holographic measurement of ultrasonic sound pressure.

Figure 6 :
Figure 6: An example of fringe pattern analysis, phase shift, and its fast Fourier transform result.

Figure 7 :
Figure 7: Comparison between frequencies of vibration face of vibrator and those of function generator.The data points fall on the 45 degree line when the input frequency of function generator and the output frequency of oscillating surface of the vibrator agree perfectly.The black circles denote the maximum negative and positive error.

Figure 8 : 3 0
Figure 8: Conceptual scheme of 3-point bending test.The amplifiers are different because the sensitivity of the PZT is higher than the C-FBG.

Figure 9 :
Figure 9: Evaluation of directionality of C-FBG sensor.The S/N ratios of the emitted waveform from each angle are shown.The directionality of the FBG was removed by using a cylindrical holder because the ratios are constant at each angle.

Figure 10 (
a) shows examples of the detected waveforms and their frequency analysis results with three input frequencies: 98.433, 492.163, and 918.900 kHz.Ten FFT analysis results were used for WMFD.The dots in WMFD results Examples of detected AE waveform by C-FBG sensor at each input frequency, its FFT, and WMFD (large dot is the center of gravity position for frequency component in WMFD) Detected frequency by C-FBG sensor (kHz) Measuring conditions Sampling frequency: 5 MHz Band-pass filter: 100∼1000 kHz Amplifier: 60 dB Hit length: 10 kword Comparison results Average error: 20.324 kHz Correlation coefficient: 0.9965 Maximum positive error: 33.238 kHz at input frequency = 383.887kHz Maximum negative error: 82.562 kHz at input frequency = 1017.137kHz Input frequency from vibrator (kHz) (b) Comparison of detected AE frequencies and input frequencies

Figure 10 :
Figure 10: Example of detected AE waveforms by C-FBG sensor and comparison of detected and input frequencies.The 45 degree line shown in (b) occurs when the C-FBG sensor accurately detects the input frequency.The black circles denote the maximum negative and positive error.

Figure 10 (FrequencyFigure 11 :
Figure 10(b)  shows the relationship between the input frequencies to the C-FBG sensor from the vibrator and the frequencies detected by the C-FBG sensor.The plot indicates that the sensor detected the correct frequency as there are