JP2008008815A - Signal detecting device, signal detecting method, and signal detecting program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To determine whether an AE signal is included in a signal detected by an optical fiber AE sensor. <P>SOLUTION: A FFT part 130 carries out FFT with respect to signal data acquired by an optical fiber AE sensor 10. A power spectrum intensity calculating part 150 calculates the intensity of the power spectrum of each frequency width by calculating the sum of the power spectrum with respect to a prescribed frequency width. A feature quantity deciding part 170 decides a frequency corresponding to the peak power spectrum intensity of top n as feature quantity. A signal recognizing part 180 is configured so as to determine whether the AE signal exists by using the feature quantity decided by the feature quantity deciding part 170. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a signal detection apparatus, a signal detection method, and a signal detection program for detecting whether or not a signal detected by a sensor includes a signal other than noise, and in particular, even when the S / N ratio is low, other than noise. The present invention relates to a signal detection apparatus, a signal detection method, and a signal detection program that can accurately identify the presence or absence of a signal.

  In thermal power plants, it is desired to introduce technology to detect abnormal signs of operating high-temperature equipment early and monitor the condition. As part of this, abnormal signs and diagnostic techniques using the Acoustic Emission (AE) method are being studied. AE is a phenomenon in which elastic energy is released with the destruction and deformation of a solid material and a sound wave (AE wave) is generated (for example, see Non-Patent Document 1).

  Conventionally, a flaw detection method using ultrasonic waves or X-rays has been used for defect inspection of large structures. However, a wide range of inspections requires a large-scale apparatus, inspection time and labor. On the other hand, the AE method can monitor the occurrence and growth of defects in time series, and by arranging multiple sensors, it is possible to quantitatively evaluate when, where, how and how much damage has occurred. . However, since AE sensors using piezoelectric elements (hereinafter referred to as piezoelectric element AE sensors) that are generally used have a low service temperature of about 200 ° C., AE is observed by directly contacting the sensor with high-temperature equipment. I can't.

  In recent years, strain sensors using optical fibers have been developed, and their application to AE measurement is being studied. Since the optical fiber is made of quartz glass, it can be expected to be used in a high temperature environment. In addition, there is an advantage that it is not affected by environmental electrical noise which is a problem in the piezoelectric element AE sensor. For example, Non-Patent Documents 2 and 3 describe high-temperature AE sensors that can be used up to 600 ° C.

AE Special Research Committee Education WG, “Acoustic Emission I”, Japan Nondestructive Inspection Association, 1995 P. Chivavibul, Hiroyuki Fukutomi, Takashi Ogata, Makoto Takahashi, Fumio Matsumura, Yuichi Machishima, “Development of condition monitoring technology based on AE method using optical fiber sensor”, Research Report of Central Research Institute of Electric Power, Q04014, 2005 P. Chivavibul, H. Fukutomi, S. Takahashi and Y. Machijima, “Development of Heat-resistant Optical Fiber AE Sensor”, Progress in Acoustic Emission XV, 2004

  However, since the optical fiber AE sensor has a low S / N ratio, there is a problem that noise is very large compared to the piezoelectric element AE sensor. FIG. 15 shows signal waveforms obtained by the optical fiber AE sensor and the piezoelectric element AE sensor. The resonance frequency of the piezoelectric element AE sensor is determined by its thickness and material. For example, for a metal (100 kHz or more) or concrete (100 kHz or less), a sensor having a resonance frequency that matches the frequency band of the target AE is selected. Thus, a signal with a small amplitude can be detected. On the other hand, the signal is greatly modulated in frequency, and in order to detect an AE signal due to a composite factor, it is necessary to arrange a plurality of sensors at the same position.

  On the other hand, the optical fiber AE sensor has a flat frequency characteristic and can take AE signals of various frequencies, but it is very noisy. Therefore, it is difficult to determine whether the measured waveform is only noise or includes an AE signal.

  The present invention has been made to solve the above-described problems caused by the prior art, and accurately identifies the presence / absence of signals other than noise even when the S / N ratio is low as in an optical fiber AE sensor. An object of the present invention is to provide a signal detection device, a signal detection method, and a signal detection program capable of performing the above.

  In order to solve the above-described problems and achieve the object, a signal detection device according to the invention of claim 1 includes discrete Fourier transform means for performing discrete Fourier transform on a signal detected by a sensor, and the discrete Fourier transform means. The power spectrum intensity calculating means for calculating the power spectrum intensity by taking the sum of the power spectrum for each predetermined frequency width with respect to the power spectrum obtained by the above, and calculated for each predetermined frequency width by the power spectrum intensity calculating means. A signal presence / absence identification for identifying whether or not a signal other than noise is included in the signal, using a predetermined number of power spectrum intensities of the power spectrum intensities and frequencies corresponding to the power spectrum intensities as features. Means.

  According to a second aspect of the present invention, there is provided the signal detection device according to the first aspect, wherein the signal presence / absence identifying means identifies whether the signal includes a signal other than noise by pattern identification. Be

  According to a third aspect of the present invention, in the signal detection device according to the second aspect, the signal presence / absence identifying means identifies whether the signal includes a signal other than noise by a support vector machine. Features.

  According to a fourth aspect of the present invention, in the signal detection device according to the first, second, or third aspect, the signal is a signal detected by an optical fiber AE sensor.

  According to a fifth aspect of the present invention, there is provided a signal detection method comprising: a discrete Fourier transform step for performing a discrete Fourier transform on a signal detected by a sensor; and a predetermined spectrum for a power spectrum obtained by the discrete Fourier transform step. A power spectrum intensity calculating step for calculating the power spectrum intensity by taking the sum of the power spectrum for each frequency width, and a power spectrum intensity calculated for each predetermined frequency width by the power spectrum intensity calculating step from a larger one to a predetermined one A signal presence / absence identifying step for identifying whether or not a signal other than noise is included in the signal, using a number of power spectrum intensities and a frequency corresponding to each of the power spectrum intensities as feature amounts. .

  A signal detection program according to a sixth aspect of the invention includes a discrete Fourier transform procedure for performing a discrete Fourier transform on a signal detected by a sensor, and a predetermined spectrum for a power spectrum obtained by the discrete Fourier transform procedure. A power spectrum intensity calculation procedure for calculating the power spectrum intensity by taking the sum of the power spectrum for each frequency width, and a power spectrum intensity calculated for each predetermined frequency width by the power spectrum intensity calculation procedure from a larger one to a predetermined one A signal presence / absence identifying procedure for identifying whether or not a signal other than noise is included in the signal, using a number of power spectrum intensities and frequencies corresponding to the power spectrum intensities as feature quantities, And

  According to the first, fifth and sixth aspects of the invention, the discrete Fourier transform is performed on the signal detected by the sensor, and the power spectrum is summed for each predetermined frequency width with respect to the obtained power spectrum. Spectral intensity is calculated, and a signal other than noise is included in the signal, with the frequency corresponding to a predetermined number of power spectral intensities and the frequency corresponding to each of the power spectral intensities calculated from each of the power spectral intensities calculated for each predetermined frequency width as a feature quantity. Since it is configured to identify whether or not it is included, even if the S / N ratio is low, there is an effect that the presence or absence of signals other than noise can be accurately identified.

  Further, according to the invention of claim 2, since it is configured to identify whether or not a signal other than noise is included in the signal by pattern identification, it is possible to accurately identify the presence or absence of a signal other than noise. Play.

  According to the invention of claim 3, since it is configured to identify whether or not a signal other than noise is included in the signal by the support vector machine, the presence or absence of a signal other than noise can be accurately identified. There is an effect.

  According to the invention of claim 4, since the signal is configured to be a signal detected by the optical fiber AE sensor, there is an effect that the presence / absence of an AE signal other than noise can be accurately identified.

  Exemplary embodiments of a signal detection device, a signal detection method, and a signal detection program according to the present invention will be explained below in detail with reference to the accompanying drawings.

  First, feature amounts extracted from a signal by the signal detection apparatus according to the present embodiment will be described with reference to FIGS. FIG. 1 is a diagram showing parameters characterizing the AE signal. As shown in the figure, there are the following three main parameters characterizing the AE signal.

The maximum amplitude level in the maximum amplitude signal. In the case of AE caused by a crack, it reflects the size of the crack and changes depending on the sensor mounting position.
A frequency that shows the maximum power spectrum when the peak frequency signal is analyzed. Reflects the type of AE and varies depending on the material and shape of the measurement target.
The time from when the duration signal exceeds the detection threshold until it decays below a certain level. Reflects the type of AE and varies depending on the size of the crack.

  Here, the detection threshold setting method used as a parameter extraction trigger is to record noise while gradually reducing the threshold value in the measurement environment to avoid noise. It is common to set to. The piezoelectric element AE sensor has no problem because the noise is small, but in the case of the optical fiber AE sensor, the threshold value is set high because the noise is large, and it is difficult to apply the conventional definition of duration. .

  Also, the ideal AE signal waveform has different maximum amplitudes before and after the trigger (in the case of background noise, the amplitude before and after the trigger does not change much, but greatly varies if an AE signal is generated), and the peak frequency is different. (When the background noise and the AE signal are compared, there is a difference in the peak frequency), but the actual waveform obtained from the optical fiber AE sensor has a low S / N ratio, so the amplitude level of the signal and the noise However, there is a problem that the initial position of the signal waveform is not always the trigger part when the threshold value is set high because the detection threshold value is high due to large noise and difficult to set. Because of these problems, the maximum amplitude is not an effective feature. Also, it is difficult to take a method of comparing before and after the trigger.

  Therefore, in this embodiment, attention is paid to the peak frequency of the signal waveform after the trigger in which the signal always exists if AE occurs. However, since it is difficult to represent the difference in signal waveform only with the frequency of one peak at which the power spectrum is maximum, the frequency and power spectrum values are taken for a plurality of peaks in the power spectrum distribution. FIG. 2 shows the problem of the characteristic parameter.

  FIG. 3 is an explanatory diagram for explaining the feature amount extracted from the signal by the signal detection apparatus according to the present embodiment. For the power spectrum distribution obtained by frequency analysis of the signal waveform, if the values are simply taken in descending order of the power spectrum, the peaks are usually broad, so many values around a single peak frequency are taken. . In this case, the result may be the same as when a single peak frequency is taken. Therefore, it is considered that a peak at a distant position is desirable as a feature for representing a difference in signal waveform. Therefore, in the signal detection apparatus according to the present embodiment, the frequency band is defined by dividing the frequency at regular intervals, the power spectrum intensity for the frequency band is taken, and a plurality of frequency band positions and their strengths are taken in descending order. The amount was decided.

  That is, first, FFT is performed on the waveform portion after the trigger to calculate a power spectrum. According to Parseval's theorem, since the sum of the power spectrum is the integral of the square of the time function of the frequency band, the power spectrum is summed for each fixed frequency band to obtain the intensity for each frequency band. Then, a plurality of frequency band positions and the sum of the power spectra are taken as the feature amount in descending order of the sum of the power spectra.

  Thus, in this embodiment, the power spectrum is summed for each fixed frequency band to obtain the intensity for each frequency band, and the frequency band position and the sum of the power spectrum are calculated in descending order of the sum of the power spectrum. The presence or absence of the AE signal can be accurately identified by using a plurality of features as feature amounts and identifying the presence or absence of the AE signal using these feature amounts.

  Next, the configuration of the AE signal detection apparatus according to the present embodiment will be described. FIG. 4 is a functional block diagram illustrating the configuration of the AE signal detection apparatus according to the present embodiment. As shown in the figure, the AE signal detection apparatus 100 includes a signal reading unit 110, a signal data storage unit 120, an FFT unit 130, a power spectrum storage unit 140, a power spectrum intensity calculation unit 150, and a power spectrum. An intensity storage unit 160, a feature amount determination unit 170, and a signal identification unit 180 are included.

  The signal reading unit 110 reads the data sampled from the output signal of the Mach-Zehnder optical fiber laser interferometer 20 that measures the AE signal using the optical fiber AE sensor 10 and recorded in the data recording device 30. The signal data storage unit 120 is a storage unit that stores sampling data of signals detected by the optical fiber AE sensor 10. The data recording device 30 records signal data triggered by exceeding a predetermined threshold value.

Here, the principle of the optical fiber AE sensor 10, the optical fiber type laser interferometer 20, and noise during AE signal measurement will be described. In a medium with a refractive index n, the frequency change Δf due to the Doppler effect of light when the observation point moves as viewed from the light source is described by Equation (1).

Here, C ₀ is the speed of light, f ₀ is the frequency of light, L is the distance between the light source and the observation point, and t is time. When the optical fiber is wound in a loop shape, dL / dt in the equation (1) is expressed as follows.

Here, ε _x and ε _y are strain rates in the x and y directions, respectively. N is a coefficient determined from the length of the optical fiber constituting the loop, the wavelength of light propagating through the optical fiber, and the state of adhesion between the optical fibers. π is the circumference, and a is the loop diameter.

From Equation (1) and Equation (2), a theoretical formula for the frequency change Δf in a plane multiple circular waveguide with a uniform strain field such as the loop sensor shown in FIG. 5 is derived.

Here, a ₁ and a ₂ are the inner diameter and outer diameter of the loop, respectively. A frequency shift occurs in the light propagating in the fiber due to the strain rate change received by the loop from Equation (3), and the strain rate change can be detected by detecting Δf. By increasing the number of turns and increasing N, Δf can be increased and sensitivity can be increased.

  FIG. 6 is a diagram showing an optical circuit of the optical fiber type laser interferometer 20. Light from the light source is divided into signal light and reference light. When the optical fiber is distorted by the elastic wave, a frequency shift occurs in the signal light passing therethrough. The optical fiber type laser interferometer 20 superimposes the reference light and the signal light whose frequency is shifted by 80 MHz with an acousto-optic device (AOM), converts the intensity change of the interference light into an electric signal with a photodiode, and then detects the frequency. The circuit detects a frequency change as a voltage change. And a frequency component of 10 Hz-250 kHz is detected by the built-in frequency filter.

  The loop sensor is fixed to the surface to be measured using vacuum grease. Further, a 50-volume sensor (inner diameter: 5 mm, outer diameter: 21 mm) having a structure in which a polyimide-coated fiber is formed in a loop shape and sandwiched between Kapton films is used.

  The AE signal measured by the optical fiber type laser interferometer 20 is usually an ultrasonic wave in a frequency band of 10 kHz to several MHz. Since human audible sound is about 20 Hz to 20 kHz, the AE signal has a higher frequency component.

  A well-known AE signal waveform is a sudden waveform with a sharp rise and gradually decaying, and is detected as cracks are generated and grown. Apart from this, a continuous signal is detected in accordance with friction or plastic deformation of the material. This signal cannot be separated as a result of the continuous occurrence of many sudden waveforms. FIG. 7 shows sudden and continuous AE signal waveforms.

  In addition, there are roughly three types of noise generated during measurement of the optical fiber AE. The first is noise inside the apparatus, which is inherent to the optical fiber type laser interferometer 20 and the AE waveform recording apparatus constituting the measurement system.

  While the piezoelectric element sensor generates a voltage due to the piezoelectric effect only when vibration is detected and amplifies it, the optical fiber AE sensor 10 causes the optical fiber type laser interferometer 20 to constantly interfere with light. The change in the amount of light is detected by a photodiode. At this time, the output does not become zero even if there is no vibration due to fluctuations in the laser light source, return light at the connector used for splitting / coupling light, slight difference in optical path length, and the like. For this reason, the optical fiber AE sensor 10 has a larger noise than the piezoelectric element sensor.

  The second is acoustic noise, which indicates acoustic waves propagating through the object to be measured other than AE. The most typical mechanical vibration is removed by a high-pass filter because of its low frequency. On the other hand, what is caused by friction or fluid may be close to the focused AE and frequency components, and is difficult to remove.

  The third is electrical noise, which is mixed in from power lines and cables. In the optical fiber AE sensor 10, although the sensor unit does not directly receive electromagnetic noise, a coaxial cable or the like used for connection between the connector portion and the device causes electrical noise. In addition, when there is an instantaneous voltage change such as discharge or ON / OFF of a high voltage device in the surroundings, the cable or connector directly receives the signal on the spike. If a noise cut transformer is used, noise mixed from the power supply line can be significantly cut.

  Returning to FIG. 4, the FFT unit 130 is a processing unit that reads out the signal data stored in the signal data storage unit 120 and performs FFT (Fast Fourier Transform), and converts the power spectrum obtained by the FFT into a power spectrum storage unit. Stored in 140. The power spectrum storage unit 140 is a storage unit that stores a power spectrum.

  The power spectrum intensity calculation unit 150 is a processing unit that reads the power spectrum from the power spectrum storage unit 140, calculates the sum of the power spectrum for a predetermined frequency width, and calculates the power spectrum intensity for each frequency width. The power spectrum intensity is stored in the power spectrum intensity storage unit 160. The power spectrum intensity storage unit 160 is a storage unit that stores the power spectrum intensity for each frequency width.

  The feature quantity determination unit 170 is a processing unit that reads the power spectrum intensity from the power spectrum intensity storage unit 160 and determines the power spectrum intensity of the top n peaks and the corresponding frequency as the signal feature quantity.

  The signal identification unit 180 determines whether or not an AE signal is included in a signal detected using the optical fiber AE sensor 10 and the feature amount determined by the feature amount determination unit 170 and a support vector machine (hereinafter referred to as SVM). It is a processing part identified using.

  FIG. 8 is an explanatory diagram for explaining the concept of SVM. Circles in the figure represent one data pattern represented on the feature space, and white circles and black circles represent an AE signal pattern and a noise pattern, respectively. The SVM finds a difficult-to-classify data pattern (a circle on the dotted line in FIG. 8) called a “support vector” and determines the discriminant function at the center. Then, using this discrimination function, it is determined whether the data pattern is an AE signal or noise. The details of SVM are described in the literature (V. N. Vapnik: The Nature of Statistical Learning Theory, Springer, 1995.).

  Next, a processing procedure of the AE signal detection apparatus 100 according to the present embodiment will be described. FIG. 9 is a flowchart illustrating the processing procedure of the AE signal detection apparatus 100 according to the present embodiment. As shown in the figure, in the AE signal detection device 100, the signal reading unit 110 reads the signal data detected by the optical fiber type laser interferometer 20 and recorded in the data recording device 30 (step S1). Store in the data storage unit 120.

  Then, the FFT unit 130 reads the signal data from the signal data storage unit 120, executes the FFT, and stores the power spectrum in the power spectrum storage unit 140 (step S2).

  Then, the power spectrum intensity calculation unit 150 reads the power spectrum from the power spectrum storage unit 140, calculates the power spectrum intensity for each predetermined frequency width (step S3), and stores the calculated power spectrum intensity in the power spectrum intensity storage unit 160. Store.

  Then, the feature amount determination unit 170 reads the power spectrum intensity from the power spectrum intensity storage unit 160, and determines n power spectrum intensities and corresponding frequencies from the larger one as the feature amount (step S4). Then, the signal identification unit 180 identifies whether or not the AE signal is included in the signal detected by the optical fiber type laser interferometer 20 using n feature values (step S5).

  As described above, the presence / absence of the AE signal can be accurately detected by using the n largest power spectrum intensities and corresponding frequencies as the feature amounts.

  Next, the evaluation result of the AE signal detection apparatus 100 according to the present embodiment will be described. In order to obtain the data of the signal waveform with mixed AE signal and noise and the signal waveform of only noise for evaluation, input conditions such as size, speed, generation position and timing can be controlled, and the AE source has reproducibility. On the other hand, the detection waveform by the optical fiber AE sensor 10 was acquired. FIG. 10 shows a schematic diagram of the evaluation experiment.

  Measurement was performed by fixing the optical fiber AE sensor 10 at the center of a SUS flat plate (100 × 100 × 1 mm), and an AE source was introduced near the end. In the evaluation experiment, there was no disturbance (noise) that was intentionally entered, and background noise due to the influence of the optical fiber type laser interferometer 20, which is mainly noise inside the apparatus, is included in all waveforms. Went in state.

The following was used for the AE source, and the detection threshold was set to 60 mV.
(1) AE by scratch
Generated by introducing scratch marks with a diamond indenter near the edge
(2) AE due to metal needle collision
In the vicinity of the end, a metal rod with a radius of curvature of 2 mm at the tip is dropped from a height of about 50 mm and collided.

  In addition, two types of waveforms with only the background noise were recorded, with the detection threshold being 80 mV and 40 mV, in order to obtain noise waveforms in various states. All the data of 50 examples of AE signal waveforms and 15 examples of noise waveforms were obtained. The breakdown of the AE signal data is (1) 40 examples and (2) 10 examples. All waveforms are recorded with the following observation parameters.

・ Sampling rate: 1e-7sec
-Number of samples: 10,000
-Pre-trigger: 1/2 (trigger position is the center of the entire waveform)
・ Signal unit: Volts

  In addition, LIBSVM (http://www.csie.ntu.edu.tw/˜cjlin/libsvm/) was used for calculation of SVM. The SVM parameter C is set to 1000 in order to minimize learning data classification errors.

Moreover, as an evaluation, accuracy comparison was performed by a single cross-validation method (hereinafter referred to as LOO). LOO removes one pattern x _i from the pattern set T of the number of data n, performs learning with (T−x _i ), and tests with the removed x _i i = 1. . . This is performed for n, and the average value of n times is used as the estimation accuracy.

Moreover, the calculation of FFT was performed on the following conditions using the function of MATLAB.
-Number of samples: 5000 (for 500 μsec, frequency resolution: 2 kHz)
Window function: rectangular window Note that each feature is normalized so that it has an average of 0 and a variance of 1.

The experiment was performed by changing the following three conditions.
(1) A width for collecting spectral data obtained by the frequency width FFT. Here, the width is 2 kHz, 4 kHz, 10 kHz, and 20 kHz.
(2) Number of peaks The number of spectral data used. Here, the top one is set to the top ten.
(3) Removal of fundamental frequency component In order to reduce the influence of low frequency noise, the case where the fundamental frequency component (2 kHz) was removed was separated from the case where it was not removed. When removed, the spectral data other than the fundamental frequency components were collected for each frequency width.

  FIG. 11 and FIG. 12 show the discrimination accuracy when the fundamental frequency component is removed and when it is not removed. The vertical axis represents the discrimination accuracy, the horizontal axis represents the change in the number of peaks, and each graph corresponds to each frequency width.

  From FIG. 11, it can be seen that 100% discrimination accuracy can be obtained by using the upper 3, 4, 5, and 10 peaks when the fundamental frequency is removed and the frequency width is 4 kHz. The number of features is the smallest when the top three peaks are used, and the number of features is six, with the top frequency width and three spectral intensities in total. In LOO, the number of patterns (support vectors) used for discrimination is 4 to 5, and in any case, complete linear discrimination is possible. Under this condition, the number of support vectors when learning all waveform patterns was four. At this time, the waveform that became the support vector was AE by scratch: 3 examples, and background noise: 1 example.

  FIG. 13 is a diagram illustrating an example of a waveform obtained by the optical fiber AE sensor 10 and a feature amount thereof. FIG. 4A shows a case where no AE signal is included, and FIG. 4B shows a case where an AE signal is included.

  Comparing FIG. 6A and FIG. 6B, it is difficult to identify whether or not an AE signal is included from the waveform, but whether or not an AE signal is included from the feature amount. It turns out that it is easy to do.

  As described above, in this embodiment, the FFT unit 130 performs FFT on the signal data obtained by the optical fiber AE sensor 10, and the power spectrum intensity calculation unit 150 performs power spectrum analysis for a predetermined frequency width. The sum is calculated to calculate the power spectrum intensity for each frequency width, the feature quantity determining unit 170 determines the frequency corresponding to the power spectrum intensity of the top n peaks as the feature quantity, and the signal identifying unit 180 is the feature quantity determining unit. Since the presence / absence of the AE signal is identified using the feature amount determined in 170, the presence / absence of the AE signal can be accurately identified.

  In this embodiment, the AE signal detection apparatus has been described. However, an AE signal detection program having the same function can be obtained by realizing the configuration of the AE signal detection apparatus with software. A computer that executes this AE signal detection program will be described.

  FIG. 14 is a functional block diagram illustrating the configuration of a computer that executes an AE signal detection program according to the present embodiment. As shown in the figure, the computer 200 includes a RAM 210, a CPU 220, an HDD 230, a LAN interface 240, an input / output interface 250, and a DVD drive 260.

  The RAM 210 is a memory that stores a program and a program execution result, and the CPU 220 is a central processing unit that reads the program from the RAM 210 and executes the program.

  The HDD 230 is a disk device that stores programs and data, and the LAN interface 240 is an interface for connecting the computer 200 to other computers via the LAN.

  The input / output interface 250 is an interface for connecting an input device such as a mouse or a keyboard and a display device, and the DVD drive 260 is a device for reading / writing a DVD.

  The AE signal detection program 211 executed in the computer 200 is stored in the DVD, read from the DVD by the DVD drive 260, and installed in the computer 200.

  Alternatively, the AE signal detection program 211 is stored in a database or the like of another computer system connected via the LAN interface 240, read from these databases, and installed in the computer 200.

  The installed AE signal detection program 211 is stored in the HDD 230, read into the RAM 210, and executed by the CPU 220 as the AE signal detection process 221.

  In the present embodiment, the case where the signal detected by the optical fiber AE sensor identifies whether or not the AE signal is included has been described. However, the present invention is not limited to this and is buried in noise. The present invention can be similarly applied when the presence or absence of an arbitrary signal is identified.

  As described above, the signal detection device, the signal detection method, and the signal detection program according to the present invention are useful for monitoring abnormalities such as plant equipment, and are particularly suitable when the contained noise is large.

It is a figure which shows the parameter which characterizes AE signal. It is a figure which shows the problem of a characteristic parameter. It is explanatory drawing for demonstrating the feature-value extracted from the signal by the signal detection apparatus which concerns on a present Example. It is a functional block diagram which shows the structure of the AE signal detection apparatus which concerns on a present Example. It is a figure which shows a loop sensor. It is a figure which shows the optical circuit of an optical fiber type laser interferometer. It is a figure which shows a sudden type and a continuous type AE signal waveform. It is explanatory drawing for demonstrating the concept of SVM. It is a flowchart which shows the process sequence of the AE signal detection apparatus which concerns on a present Example. It is a schematic diagram of an evaluation experiment. It is a figure which shows the discrimination | determination precision at the time of removing a fundamental frequency component. It is a figure which shows the discrimination | determination precision when not removing a fundamental frequency component. It is a figure which shows the example of the waveform obtained with the optical fiber AE sensor, and its feature-value. It is a functional block diagram which shows the structure of the computer which performs the AE signal detection program which concerns on a present Example. It is a figure which shows the signal waveform by an optical fiber AE sensor and a piezoelectric element AE sensor.

Explanation of symbols

DESCRIPTION OF SYMBOLS 10 Optical fiber AE sensor 20 Optical fiber type laser interferometer 30 Data recording device 100 AE signal detection device 110 Signal reading part 120 Signal data storage part 130 FFT part 140 Power spectrum storage part 150 Power spectrum intensity calculation part 160 Power spectrum intensity storage part 170 Feature Determining Unit 180 Signal Identification Unit 200 Computer 210 RAM
211 AE signal detection program 220 CPU
221 AE signal detection process 230 HDD
240 LAN interface 250 I / O interface 260 DVD drive

Claims (6)

Discrete Fourier transform means for performing discrete Fourier transform on the signal detected by the sensor;
Power spectrum intensity calculating means for calculating the power spectrum intensity by taking the sum of the power spectrum for each predetermined frequency width with respect to the power spectrum obtained by the discrete Fourier transform means;
The power spectrum intensity calculating means calculates a predetermined number of power spectrum intensities out of the power spectrum intensities calculated for each predetermined frequency width and the frequency corresponding to each of the power spectrum intensities as feature quantities, and adds noise to the signal. Signal presence / absence identifying means for identifying whether or not a signal other than
A signal detection apparatus comprising:   The signal detection apparatus according to claim 1, wherein the signal presence / absence identifying unit identifies whether or not a signal other than noise is included in the signal by pattern identification.   The signal detection apparatus according to claim 2, wherein the signal presence / absence identifying unit identifies whether a signal other than noise is included in the signal by a support vector machine.   The signal detection apparatus according to claim 1, wherein the signal is a signal detected by an optical fiber AE sensor. A discrete Fourier transform step for performing a discrete Fourier transform on the signal detected by the sensor;
A power spectrum intensity calculating step for calculating the power spectrum intensity by taking the sum of the power spectrum for each predetermined frequency width with respect to the power spectrum obtained by the discrete Fourier transform step;
A noise is added to the signal by using a predetermined number of power spectrum intensities and frequencies corresponding to the power spectrum intensities among the power spectrum intensities calculated for each predetermined frequency width in the power spectrum intensity calculating step as features. A signal presence / absence identifying step for identifying whether or not a signal other than
The signal detection method characterized by including. A discrete Fourier transform procedure for performing a discrete Fourier transform on the signal detected by the sensor;
A power spectrum intensity calculation procedure for calculating the power spectrum intensity by taking the sum of the power spectrum for each predetermined frequency width with respect to the power spectrum obtained by the discrete Fourier transform procedure;
Among the power spectrum intensities calculated for each predetermined frequency width by the power spectrum intensity calculation procedure, a predetermined number of power spectrum intensities and frequencies corresponding to the power spectrum intensities are used as feature quantities, and noise is included in the signal. A signal presence / absence identification procedure for identifying whether or not a signal other than
A signal detection program for causing a computer to execute.

Images