Non-destructive testing method and non-destructive testing system for concrete utility poles

Abstract

To provide a non-destructive testing method for concrete utility poles that can accurately detect and determine the extent of deterioration that has not yet become apparent. [Solution] This non-destructive testing method includes a correction processing step, an image generation step, and a determination step. In the correction processing step, the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasonic signal for inspection is enhanced and corrected for each frequency. In the image generation step, time-frequency images G1 and G2 are generated based on the corrected time-frequency analysis data. In the determination step, the deterioration status of the concrete utility pole 1 is determined based on the line segment Ln, which shows the signal intensity of the input frequency band, and the line segment Lc, which shows the signal intensity of the natural frequency band of the concrete utility pole 1 that appears at a frequency lower than the input frequency band, in the generated time-frequency images G1 and G2.

Description

Technical Field

[0001] The present invention relates to a non-destructive inspection method and a non-destructive inspection system for concrete utility poles.

Background Art

[0002] Conventionally, as one method for non-destructively inspecting the deterioration status of concrete utility poles, a method using ultrasonic waves has been proposed (see, for example, Patent Documents 1 and 2). In this method, for example, while an ultrasonic sensor is disposed on a concrete utility pole, the concrete utility pole is struck to generate a percussion sound, STFT data is acquired from the received signal of the percussion sound, and a spectrogram image is further generated from this STFT data. Then, by comparing the generated spectrogram image with a spectrogram stored in advance, a determination of whether there is no damage or there is damage is made. Also, a method of determining whether there is no damage or there is damage based on whether the frequency component in the spectrogram image is closer to low frequency or high frequency has also been conventionally proposed.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, with the above-described conventional techniques, although it is possible to some extent determine the presence or absence of deterioration manifested in the form of damage on the surface of the concrete utility pole, it has not been possible to accurately determine the presence or absence of deterioration that has not yet manifested or the degree of deterioration.

[0005] The present invention has been made in view of the above-mentioned problems, and its objective is to provide a non-destructive testing method and a non-destructive testing system for concrete utility poles that can detect and accurately determine the state of deterioration that has not yet become apparent. [Means for solving the problem]

[0006] To solve the above problems, the invention described in claim 1 is a method for non-destructively inspecting a concrete utility pole, comprising: causing an oscillating probe installed on the concrete utility pole to oscillate and input an inspection ultrasonic signal of a predetermined input frequency to the concrete utility pole; obtaining the received waveform of the inspection ultrasonic signal with a receiving probe installed on the concrete utility pole to calculate time-frequency analysis data; and inspecting the concrete utility pole non-destructively based on a time-frequency image generated from the time-frequency analysis data, the method comprising: a correction processing step of performing a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the inspection ultrasonic signal for each frequency; an image generation step of generating a time-frequency image based on the corrected time-frequency analysis data obtained through the correction processing step; and a determination step of determining the state of deterioration of the concrete utility pole based on a line segment indicating the signal intensity of the input frequency band and a line segment indicating the signal intensity of the concrete utility pole's natural frequency band that may appear at a frequency lower than the input frequency band in the generated time-frequency image.

[0007] Therefore, according to the invention described in claim 1, in the correction processing step, the signal intensity of the time-frequency analysis data is enhanced and corrected for each frequency, and then in the image generation step, a time-frequency image is generated. As a result, a time-frequency image that facilitates judgment in the subsequent judgment step can be obtained. Specifically, a time-frequency image can be obtained in which line segments indicating signal intensity appear at positions corresponding to the input frequency band and positions corresponding to the natural frequency band of the concrete utility pole. Then, in the judgment step, by making a judgment based on these line segments, it is possible to detect and accurately determine the state of deterioration that has not yet become apparent.

[0008] The invention described in claim 2 is characterized in that, in the correction processing step of claim 1, the signal intensity of the time-frequency analysis data is raised to the power of each frequency.

[0009] Therefore, according to the invention described in claim 2, in addition to primarily enhancing and correcting the signal in the input frequency band, if a signal in the natural frequency band of the concrete utility pole appears at a frequency lower than the input frequency band, that signal is also enhanced and corrected. As a result, line segments indicating signal strength clearly appear in the input frequency band and natural frequency band in the time-frequency image. Therefore, in the judgment step, it becomes possible to easily compare these line segments, for example, by visual inspection. Thus, it becomes easier to determine the degree of deterioration of the concrete utility pole.

[0010] The invention described in claim 3 is characterized in that, in claim 2, the time-frequency analysis data is STFT data, and the time-frequency image is a spectrogram image generated based on the corrected STFT data obtained through the enhancement processing step.

[0011] The invention described in claim 4 is characterized in that, in claim 2, the time-frequency analysis data is wavelet transform data, and the time-frequency image is a scaler image generated based on the corrected wavelet transform data obtained through the enhancement step.

[0012] The invention described in claim 5 is characterized in that, in any one of claims 1 to 4, the determination step is as follows: in the first case, when only line segments indicating the signal strength of the input frequency band appear in the time-frequency image and line segments indicating the signal strength of the natural frequency band do not appear, it is determined that there is no deterioration of the concrete utility pole or that the degree of deterioration is relatively small; in the second case, when only line segments indicating the signal strength of the natural frequency band appear in the time-frequency image, or in the third case, when both line segments indicating the signal strength of the input frequency band and line segments indicating the signal strength of the natural frequency band appear, it is determined that the degree of deterioration of the concrete utility pole is greater than in the first case.

[0013] The invention described in claim 6 is characterized in that, in any one of claims 1 to 4, the determination step is as follows: in the first case, when both a line segment indicating the signal strength of the input frequency band and a line segment indicating the signal strength of the natural frequency band appear in the time-frequency image and the signal strength of the input frequency band is relatively stronger, it is determined that there is no deterioration of the concrete utility pole or that the degree of deterioration is small; in the second case, when only a line segment indicating the signal strength of the natural frequency band appears in the time-frequency image, or in the third case, when both a line segment indicating the signal strength of the input frequency band and a line segment indicating the signal strength of the natural frequency band appear and the signal strength of the input frequency band is relatively weaker, it is determined that the degree of deterioration of the concrete utility pole is greater than in the first case.

[0014] The invention described in claim 7 is a method for non-destructively inspecting a concrete utility pole by causing an oscillating probe installed on the concrete utility pole to oscillate and input an inspection ultrasonic signal of a predetermined input frequency to the concrete utility pole, and by acquiring the received waveform of the inspection ultrasonic signal with a receiving probe installed on the concrete utility pole to calculate time-frequency analysis data, and based on a time-frequency image generated from the time-frequency analysis data, the method comprising a correction processing step of performing a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the inspection ultrasonic signal for each frequency, and based on the corrected time-frequency analysis data obtained through the correction processing step The gist of this non-destructive testing method for concrete utility poles is characterized by including an image generation step of generating a time-frequency image; a calculation step of calculating the Mahalanobis distance from a first output value obtained by inputting a group of time-frequency images of healthy concrete utility poles without deterioration to a trained autoencoder that has been trained by inputting a group of time-frequency images of healthy concrete utility poles without deterioration, and a second output value obtained by inputting a time-frequency image of the concrete utility pole to be inspected to the trained autoencoder; and a determination step of determining the state of deterioration of the concrete utility pole to be inspected based on the calculated Mahalanobis distance.

[0015] Accordingly, according to the invention described in claim 7, after the signal intensity of the time-frequency analysis data is enhanced and corrected for each frequency in the correction processing step, a time-frequency image is generated in the image generation step. As a result, a time-frequency image that facilitates calculation and determination in the subsequent calculation and determination steps can be obtained. Specifically, a time-frequency image can be obtained in which line segments indicating signal intensity appear at positions corresponding to the input frequency band and positions corresponding to the natural frequency band of the concrete utility pole. Then, in the calculation step, a suitable trained autoencoder can be obtained relatively easily by using the time-frequency image group containing these line segments. Furthermore, by inputting the time-frequency image group containing these line segments into this trained autoencoder and outputting predetermined first and second output values, an appropriate Mahalanobis distance can be calculated relatively easily. In the determination step, by making a determination based on the calculated Mahalanobis distance (score), it is possible to detect and accurately determine the state of deterioration that has not yet become apparent. In addition, the degree of deterioration can be evaluated objectively and quantitatively.

[0016] The invention described in claim 8 is characterized in that, in the correction processing step of claim 7, the signal intensity of the time-frequency analysis data is raised to the power of each frequency.

[0017] Therefore, according to the invention described in claim 8, in addition to primarily enhancing and correcting the signal in the input frequency band, if a signal in the natural frequency band of the concrete utility pole appears at a frequency lower than the input frequency band, that signal is also enhanced and corrected. As a result, line segments indicating signal strength clearly appear in the input frequency band and natural frequency band in the time-frequency image. Therefore, it becomes possible to easily compare these line segments, for example, by visual inspection, during the determination step. Thus, it becomes easier to determine the degree of deterioration of the concrete utility pole.

[0018] The invention described in claim 9 is characterized in that, in claim 8, the time-frequency analysis data is STFT data, and the time-frequency image is a spectrogram image generated based on the corrected STFT data obtained through the enhancement processing step.

[0019] The invention described in claim 10 is characterized in that, in claim 8, the time-frequency analysis data is wavelet transform data, and the time-frequency image is a scaler image generated based on the corrected wavelet transform data obtained through the enhancement step.

[0020] The invention described in claim 11 is characterized in that, in any one of claims 7 to 10, the calculation step involves converting the reorganization error and pixel difference obtained when a group of time-frequency images of a healthy concrete utility pole without deterioration is input to the trained autoencoder into two-dimensional data, which is used as the normal data group; converting the reorganization error and pixel difference obtained when a time-frequency image of the concrete utility pole to be inspected is input to the trained autoencoder into two-dimensional data, which is used as the inspection data; and calculating the Mahalanobis distance from the center point of the normal data group to the point indicated by the inspection data.

[0021] The invention described in claim 12 is characterized in that, in the determination step of claim 11, the calculated Mahalanobis distance is compared with a preset threshold, and if the calculated Mahalanobis distance is less than the threshold, it is determined that there is no deterioration, and if it is equal to or greater than the threshold, it is determined that there is deterioration.

[0022] The invention described in claim 13 is characterized in that, in the determination step of claim 11, if the calculated Mahalanobis distance is greater than or equal to the threshold and the difference from the threshold is small, it is determined that the degree of deterioration is small, and if the calculated Mahalanobis distance is greater than or equal to the threshold and the difference from the threshold is large, it is determined that the degree of deterioration is large.

[0023] The invention according to claim 14 is installed on a concrete pole, and includes an oscillation-side probe for inputting an ultrasonic signal for inspection with a predetermined input frequency into the concrete pole, a reception-side probe installed on the concrete pole for acquiring a reception waveform of the inspection ultrasonic signal, and a determination device for acquiring time-frequency analysis data from the acquired reception waveform and determining the deterioration status of the concrete pole based on a time-frequency image generated from the time-frequency analysis data. The determination device includes a correction processing unit that performs a process of emphasizing and correcting the signal intensity of the time-frequency analysis data calculated from the reception waveform of the inspection ultrasonic signal for each frequency, an image generation unit that generates a time-frequency image based on the corrected time-frequency analysis data obtained by the process of the correction processing unit, and a determination unit that determines the deterioration status of the concrete pole based on a line segment indicating the signal intensity in the input frequency band and a line segment indicating the signal intensity in the natural frequency band of the concrete pole that can appear at a frequency lower than the input frequency band in the generated time-frequency image. The gist of the non-destructive inspection system for a concrete pole is as described above.

[0024] Therefore, according to the invention of claim 14, the correction processing unit emphasizes and corrects the signal intensity of the time-frequency analysis data for each frequency, and the image generation unit generates a time-frequency image based on the corrected time-frequency analysis data after the emphasis correction. For this reason, a time-frequency image that is easy for the determination unit to make a determination can be obtained. Specifically, a time-frequency image in which a line segment indicating the signal intensity appears at a position corresponding to the input frequency band or a position corresponding to the natural frequency band of the concrete pole can be obtained. Then, by the determination unit making a determination based on these line segments, it is possible to detect the deterioration status that has not yet manifested and make an accurate determination.

[0025] The invention described in claim 15 comprises: an oscillator-side probe installed on a concrete utility pole and inputting an inspection ultrasonic signal of a predetermined input frequency to the concrete utility pole; a receiver-side probe installed on the concrete utility pole and acquiring the received waveform of the inspection ultrasonic signal; and a determination device that acquires time-frequency analysis data from the acquired received waveform and determines the state of deterioration of the concrete utility pole based on a time-frequency image generated from the time-frequency analysis data, wherein the determination device comprises a correction processing unit that performs a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the inspection ultrasonic signal for each frequency, and based on the corrected time-frequency analysis data obtained by the processing of the correction processing unit The gist of this non-destructive testing system for concrete utility poles is that it comprises: an image generation unit that generates time-frequency images; a storage unit that stores a trained autoencoder that has been trained by inputting a group of time-frequency images of healthy concrete utility poles that are not degraded; a determination unit that has a function to calculate the Mahalanobis distance from a first output value obtained by inputting a group of time-frequency images of healthy concrete utility poles that are not degraded to the trained autoencoder and a second output value obtained by inputting a time-frequency image of the concrete utility pole to be inspected to the trained autoencoder; and a determination unit that has a function to determine the state of deterioration of the concrete utility pole to be inspected based on the calculated Mahalanobis distance.

[0026] Therefore, according to the invention described in claim 15, the correction processing unit emphasizes and corrects the signal intensity of the time-frequency analysis data for each frequency, and the image generation unit generates a time-frequency image based on the corrected time-frequency analysis data after the emphasis correction. For this reason, it is possible to obtain a time-frequency image in which the calculation unit and the determination unit can easily perform calculations and determinations. Specifically, it is possible to obtain a time-frequency image in which line segments indicating signal intensity appear at positions corresponding to the input frequency band and positions corresponding to the natural frequency band of the concrete pole. Then, by using the group of time-frequency images including these line segments, the calculation unit can relatively easily obtain a suitable pre-trained autoencoder. Furthermore, by inputting the group of time-frequency images including these line segments into this pre-trained autoencoder and outputting a predetermined first output value and a second output value, it is possible to relatively easily calculate an appropriate Mahalanobis distance. Furthermore, by the determination unit performing a determination based on the calculated numerical value (score) of the Mahalanobis distance, it is possible to detect a deterioration situation that has not yet manifested and perform an accurate determination. In addition, the degree of deterioration can be evaluated objectively and quantitatively without human intervention.

Advantages of the Invention

[0027] As described in detail above, according to the inventions described in claims 1 to 15, it is possible to provide a non-destructive inspection method and a non-destructive inspection system for a concrete pole that can detect a deterioration situation that has not yet manifested and perform an accurate determination.

Brief Description of the Drawings

[0028] [Figure 1] Schematic diagram showing the non-destructive inspection system of the first embodiment. [Figure 2] Block diagram showing the non-destructive inspection system of the first embodiment. [Figure 3] (a) to (f) are spectrogram images obtained by the non-destructive inspection method of the first embodiment. [Figure 4] Block diagram showing the non-destructive inspection system of the second embodiment. [Figure 5]A schematic diagram illustrating an autoencoder in a non-destructive testing system according to a second embodiment. [Figure 6] A graph showing the rearrangement error and pixel difference converted into two-dimensional data in the non-destructive testing method of the second embodiment. [Figure 7] A block diagram showing a modified non-destructive testing system of the second embodiment. [Modes for carrying out the invention]

[0029] [First Embodiment]

[0030] The non-destructive testing method and non-destructive testing system 11 for a concrete utility pole 1, which embody the present invention, will be described in detail below with reference to Figures 1 to 3.

[0031] Figure 1 is a schematic diagram showing the non-destructive testing system 11 of this embodiment. Figure 2 is a block diagram of the non-destructive testing system 11 of this embodiment. Figures 3(a) to 3(f) are spectrogram images G1 and G2 obtained by the non-destructive testing method of this embodiment.

[0032] The non-destructive testing system 11 of this embodiment is a system for non-destructively inspecting a concrete utility pole 1 and determining its state of deterioration. The concrete utility pole 1 to be inspected is not particularly limited; for example, it may be a single pole, or it may be a segmented pole made up of multiple poles forming one long pole. Furthermore, the concrete utility pole 1 to be inspected may be in a horizontal position before erection, or it may be in an erected state. In this example, a segmented pole in an erected state is the object of inspection (see Figure 1).

[0033] As shown in Figures 1 and 2, this non-destructive testing system 11 comprises an oscillating probe 21, a receiving probe 31, and a personal computer (PC) 41 having a determination device 51. Both the oscillating probe 21 and the receiving probe 31 are installed on the outer surface of the concrete utility pole 1. Specifically, the oscillating probe 21 is fastened and fixed to the outer surface of the base end of the concrete utility pole 1 using a metal fixing band 22. A sheet-like acoustic matching material 23 is interposed between the front surface, which is the oscillating surface of the oscillating probe 21, and the outer surface of the concrete utility pole 1 to improve the adhesion between the two. The receiving probe 31 is fastened and fixed to the outer surface of the tip end of the concrete utility pole 1 using a metal fixing band 22. A sheet-like acoustic matching material 23 is interposed between the front surface, which is the receiving surface of the receiving probe 31, and the outer surface of the concrete utility pole 1 to improve the adhesion between the two. Furthermore, it is preferable that the portion of the fixed band 22 that contacts the rear surfaces of the oscillator 21 and the receiver 31 is covered with a soft material such as rubber to prevent damage to the oscillator 21 and the receiver 31.

[0034] The oscillating probe 21 is a device that emits ultrasonic waves at a predetermined frequency to input an inspection ultrasonic signal at a predetermined input frequency to the concrete utility pole 1. The receiving probe 31 is a device for acquiring the received waveform of the inspection ultrasonic signal originating from the ultrasonic waves emitted by the oscillating probe 21. The frequency of the ultrasonic waves emitted by the oscillating probe 21 is not particularly limited and can be set arbitrarily, but in this embodiment, it is set to 50 kHz or higher, preferably 50 kHz to 100 kHz. The reason for this is that the natural frequency band of the concrete utility pole 1 to be inspected is at a frequency lower than 25 kHz, so the oscillation frequency (input frequency) of the oscillating probe 21 is set to be sufficiently higher than that in advance. In other words, the oscillation frequency (input frequency) of the oscillating probe 21 is preferably at least twice the frequency of the natural frequency band of the concrete utility pole 1, and more preferably between twice and five times. If this frequency difference is less than twice, it may become difficult to separate or distinguish between the oscillation frequency and the natural frequency.

[0035] As shown in Figure 2, the oscillator probe 21 houses a communication device 25, a control device 26, and a transducer 27 in a housing (not shown). The transducer 27 is made of a piezoelectric element and is configured to emit ultrasonic waves between 50 kHz and 100 kHz by supplying a high-frequency signal, thereby inputting an ultrasonic inspection signal to the concrete utility pole 1. The control device 26 is configured to include a computer having a CPU, ROM, RAM, etc. The control device 26 drives the transducer 27 by generating a predetermined oscillation command signal and transmitting this oscillation command signal to the transducer 27 via an oscillation circuit (not shown). The control device 26 also controls the operation of the communication device 44. The communication device 25 is configured to send and receive signals with the communication device 44 on the PC 41 side via wireless communication such as Bluetooth®. Specifically, the communication device 25 is configured to receive instruction signals transmitted from the PC 41 side, for example. The control device 26 then generates an oscillation command signal upon receiving this instruction signal.

[0036] The receiving probe 31 houses a communication device 35, a control device 36, and a transducer 37 in a housing (not shown). The transducer 37 is made of a piezoelectric element and is configured to receive the reflected wave signal of the inspection ultrasonic signal input to the concrete utility pole 1 by the transducer 37 of the oscillating probe 21. The control device 36 is configured to include a computer having a CPU, ROM, RAM, etc. The control device 36 takes the received waveform signal received by the transducer 37 and creates transmission data. The control device 36 also controls the operation of the communication device 44. The communication device 35 is configured to send and receive with the communication device 44 on the PC 41 side via wireless communication such as Bluetooth (registered trademark). This communication device 35, for example, receives instruction signals from the PC 41 side and transmits transmission data including the received waveform signal of the inspection ultrasonic signal to the PC 41 side.

[0037] As shown in Figure 2, the PC 41 is composed of a CPU and other components, and includes a display device 42, an input device 43, a communication device 44, a determination device 51, and the like.

[0038] The display device 42 is, for example, a monitor display such as a liquid crystal, plasma, or organic EL (electroluminescence) display. The display device 42 can be used for color or monochrome display, but color display is preferable. This display device 42 is used to display time-frequency images or to display input screens for various settings.

[0039] The input device 43 is an input user interface such as a touch panel, mouse, keyboard, or pointing device, and is used for inputting requests, instructions, and parameters from the user.

[0040] The communication device 44 is configured to transmit and receive data with the communication device 25 of the oscillating probe 21 and the communication device 35 of the receiving probe 31 via wireless communication such as Bluetooth®.

[0041] The determination device 51 acquires time-frequency analysis data from the acquired received waveform and determines the deterioration status of the concrete utility pole 1 based on the time-frequency image generated from the time-frequency analysis data. The determination device 51 in this embodiment includes a correction processing unit 52, an image generation unit 53, a determination unit 54, and a storage unit 55. Here, time-frequency analysis data is data obtained after performing time-frequency analysis on the acquired received waveform using a conventionally known method, and in this embodiment, it is STFT (short-time Fourier transform) data. Performing STFT extracts the characteristics of the acquired received waveform.

[0042] The correction processing unit 52 performs a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasound signal for inspection, for each frequency. In this embodiment, the correction processing unit 52 performs a process to square (raise to a power) the signal intensity of the STFT data, which is the time-frequency analysis data, for each frequency. As a result, corrected STFT data (corrected time-frequency analysis data) is generated in which the parts with stronger signal intensity are emphasized.

[0043] The image generation unit 53 generates a time-frequency image based on the corrected time-frequency analysis data obtained by the processing of the correction processing unit 52. In this embodiment, a spectrogram image G1 is generated based on the corrected STFT data obtained by the processing of the correction processing unit 52 (see Figure 3).

[0044] In the spectrogram images G1 and G2, the horizontal axis represents time (msec) and the vertical axis represents frequency (kHz), with brightness and color representing the intensity at a certain frequency at a given time. For example, in this embodiment, as the signal intensity increases, the colors change from blue → light blue → yellow-green → green → yellow. The spectrogram images G1 and G2 generated in this way typically include at least one of the line segments Ln, which represents the signal intensity in the input frequency band, and Lc, which represents the signal intensity in the natural frequency band of the concrete utility pole 1, which may appear at frequencies lower than the input frequency band.

[0045] The determination unit 54 determines the deterioration status of the concrete utility pole 1 based on the generated spectrogram images G1 and G2 (time-frequency images), specifically the line segment Ln which shows the signal strength in the input frequency band and the line segment Lc which shows the signal strength in the natural frequency band of the concrete utility pole 1 that may appear at frequencies lower than the input frequency band. The storage unit 55 stores various programs executed by the CPU, as well as data for determining the deterioration status.

[0046] Next, the procedure for non-destructive testing of the concrete utility pole 1 using this non-destructive testing system 11 will be described.

[0047] First, as shown in Figure 1, the oscillating probe 21 and the receiving probe 31 are pre-installed on the outer surface of the concrete utility pole 1 (preparation step). Next, the operator operates the input device 43 of the PC 41 to instruct the start of non-destructive testing (start step). The control device 26 then transmits an oscillation command signal to the oscillating probe 21, driving the transducer 27. As a result, the oscillating probe 21 starts oscillating, and an ultrasonic inspection signal of a predetermined input frequency is input to the concrete utility pole 1 (oscillation step). The predetermined input frequency can be selected from 50kHz, 80kHz, or 100kHz. Subsequently, the received waveform of the ultrasonic inspection signal input to the concrete utility pole 1 is acquired by the transducer 37 of the receiving probe 31 (reception step). Next, the control device 36 of the receiving probe 31 creates transmission data including the received waveform signal and transmits that transmission data to the PC 41 (data transmission step). Next, the judgment device 51 in the PC 41, which has received the transmission data, performs a short-time Fourier transform on the received waveform of the ultrasound signal for inspection to calculate STFT data (time-frequency analysis step). Next, the correction processing unit 52 performs a process of squaring the signal intensity of the STFT data to enhance and correct it for each frequency (correction processing step). Next, based on the corrected STFT data obtained through the correction processing step, spectrogram images G1 and G2 are generated (image generation step). As a result, spectrogram images G1 and G2, as shown in Figures 3(a) to 3(f), are displayed in color on the display screen of the display device 42 of the PC 41.

[0048] Figure 3(b) shows a typical example of spectrogram image G1 obtained when a 50 kHz inspection ultrasonic signal is input to a healthy concrete utility pole 1 (hereinafter also referred to as "healthy pole") that is free from deterioration. Figure 3(a) shows a typical example of spectrogram image G2 obtained when a 50 kHz inspection ultrasonic signal is input to a concrete utility pole 1 (hereinafter also referred to as "deteriorated pole") that has no visible changes in appearance, such as cracks on the surface, but has potential deterioration inside. In this context, the deteriorated pole refers to a concrete utility pole 1 to which a load 1.6 times the design load has been applied.

[0049] Furthermore, Figure 3(c) shows a typical example of spectrogram image G2 obtained when an 80 kHz inspection ultrasonic signal is input to a deteriorated pipe, and Figure 3(d) shows a typical example of spectrogram image G1 obtained when an 80 kHz inspection ultrasonic signal is input to a healthy pipe. Figure 3(e) shows a typical example of spectrogram image G2 obtained when a 100 kHz inspection ultrasonic signal is input to a deteriorated pipe, and Figure 3(f) shows a typical example of spectrogram image G1 obtained when a 100 kHz inspection ultrasonic signal is input to a healthy pipe.

[0050] Comparing Figure 3(a) and Figure 3(b), the spectrogram image G2 of the degraded tube shows only the line segment Lc, which indicates the signal strength in the natural frequency band (approximately 13 kHz). This line segment Lc is yellow, indicating a fairly strong signal. On the other hand, the spectrogram image G1 of the healthy tube shows a yellow line segment Ln around 43 kHz, indicating the signal strength in the input frequency band and a fairly strong signal. In addition, a line segment Lc appears, indicating the signal strength in the natural frequency band (approximately 13 kHz) and a fairly weak signal. Blue line segments Lc also appear around 20 kHz, 28 kHz, and 37 kHz.

[0051] Comparing Figure 3(c) and Figure 3(d), the spectrogram image G2 of the degraded tube shows the signal strength in the natural frequency band (approximately 13 kHz), and only the yellow line segment Lc, indicating a considerably strong signal, is visible. On the other hand, the spectrogram image G1 of the healthy tube shows the signal strength in the input frequency band around 80 kHz, and the yellow line segment Ln, indicating a considerably strong signal, is visible. Additionally, a yellow-green line segment Ln appears around 68 kHz, and a light blue line segment Ln appears around 71 kHz, and these are also understood to indicate the signal strength in the input frequency band. Furthermore, blue line segments appear around 48 kHz and 40 kHz, but these do not indicate the signal strength in the natural frequency band (approximately 13 kHz).

[0052] Comparing Figure 3(e) and Figure 3(f), the spectrogram image G2 of the degraded tube shows the signal strength in the natural frequency band (approximately 13 kHz), and only the yellow line segment Lc, indicating a considerably strong signal, is visible. On the other hand, in the healthy tube, the signal strength in the input frequency band is shown around 90 kHz, and only the yellow line segment Ln, indicating a considerably strong signal, is visible.

[0053] Then, in the determination step following the image generation step described above, the deterioration status of the concrete utility pole 1 is determined, for example, by an operator visually inspecting the spectrogram images G1 and G2 on the display screen. Specifically, the deterioration status is determined based on a line segment Ln that shows the signal strength in the input frequency band and a line segment Lc that shows the signal strength in the natural frequency band of the concrete utility pole 1 that may appear at frequencies lower than the input frequency band.

[0054] For example, in the determination step, the determination may be made using the first determination method described below. That is, in the first case, where only the line segment Ln indicating the signal strength of the input frequency band appears in the spectrogram images G1 and G2, and the line segment Lc indicating the signal strength of the natural frequency band does not appear, it is determined that the concrete utility pole 1 is not deteriorated or that the degree of deterioration is relatively small. For example, Figure 3(f) corresponds to the "first case," so an image similar to this would be considered a "healthy pole." In contrast, in the second case, where only the line segment Lc indicating the signal strength of the natural frequency band appears in the spectrogram images G1 and G2, or in the third case, where both the line segment Ln indicating the signal strength of the input frequency band and the line segment Lc indicating the signal strength of the natural frequency band appear, it is determined that the degree of deterioration of the concrete utility pole 1 is greater than in the first case. For example, Figure 3(e) corresponds to the "second case," so an image similar to this would be considered a "deteriorated pole."

[0055] Furthermore, in the determination step, the determination may be made using the second determination method described below. Specifically, in the first case, where both the line segment Ln indicating the signal strength of the input frequency band and the line segment Lc indicating the signal strength of the natural frequency band appear in the spectrogram images G1 and G2, and the signal strength of the input frequency band is relatively stronger, it is determined that there is no deterioration of the concrete utility pole 1 or that the degree of deterioration is small. For example, Figure 3(b) corresponds to the "first case," so images similar to this would be considered "healthy poles." In contrast, in the second case, where only the line segment Lc indicating the signal strength of the natural frequency band appears in the spectrogram images G1 and G2, or in the third case, where both the line segment Ln indicating the signal strength of the input frequency band and the line segment Lc indicating the signal strength of the natural frequency band appear, and the signal strength of the input frequency band is relatively weaker, it is determined that the degree of deterioration of the concrete utility pole 1 is greater than in the first case. For example, Figure 3(a) corresponds to "Case 2," so any image similar to this would be considered a "degraded tube."

[0056] Incidentally, as described above, the operator may perform the determination themselves using the first and second determination methods, but the system may also be configured to automatically perform the determination using the determination device 51 of the PC 41. For example, data for determining the degree of deterioration may be created, including algorithms related to the first and second determination methods, and this data for determining the degree of deterioration may be stored in the storage unit 55 in advance. Then, the determination unit 54 may determine the degree of deterioration based on this data for determining the degree of deterioration. In this case, the system may be configured to allow the operator to arbitrarily select either a self-determination mode in which the operator performs the determination themselves, or an automatic determination mode in which the determination device 51 performs the determination automatically.

[0057] Therefore, according to the above-described embodiment, the following effects can be obtained.

[0058] (1) The determination device 51 in the non-destructive testing system 11 of this embodiment is equipped with a correction processing unit 52, an image generation unit 53, a determination unit 54, etc. as described above. The non-destructive testing method using this system 11 includes a correction processing step performed by the correction processing unit 52, an image generation step performed by the image generation unit 53, and a determination step as described above. With this configuration, after the signal intensity of the STFT data is enhanced and corrected for each frequency in the correction processing step, spectrogram images G1 and G2 are generated in the image generation step. Therefore, spectrogram images G1 and G2 that are easy to determine in the next determination step can be obtained. Specifically, spectrogram images G1 and G2 can be obtained in which line segments Ln and Lc indicating signal intensity appear at positions corresponding to the input frequency band and positions corresponding to the natural frequency band of the concrete utility pole 1. Then, in the determination step, by making a determination based on these line segments Ln and Lc, it is possible to detect and accurately determine the state of deterioration that has not yet become apparent. Therefore, according to this embodiment, potential deterioration can be detected even before changes appear in the appearance of the concrete utility pole 1. In other words, because deterioration can be detected at an early stage, it becomes possible to take some kind of action on the concrete utility pole 1 before the deterioration progresses and damage becomes apparent.

[0059] (2) In the correction processing step of this embodiment, the correction processing unit 52 performs a process of squaring the signal strength of the STFT data for each frequency. As a result, the signal in the input frequency band is mainly enhanced, and if a signal in the natural frequency band of the concrete utility pole 1 appears at a frequency lower than the input frequency band, that signal is also enhanced. As a result, line segments Ln and Lc indicating signal strength appear clearly in the input frequency band and natural frequency band in the spectrogram images G1 and G2. Therefore, compared to a normal spectrogram image without such enhancement processing, it becomes easier to compare these line segments Ln and Lc by, for example, visually in the judgment step. Thus, it becomes easier to determine the deterioration status of the concrete utility pole 1.

[0060] [Second Embodiment]

[0061] Next, a non-destructive testing method for a concrete utility pole 1 and a non-destructive testing system 11A according to a second embodiment of the present invention will be described in detail with reference to Figures 4 to 6. Here, components common to the first embodiment will simply be given the same numbers, and their detailed descriptions will be omitted.

[0062] Figure 4 is a block diagram of the non-destructive testing system 11A of this embodiment. Figure 5 is a schematic diagram illustrating the autoencoder in the non-destructive testing system 11A of this embodiment. Figure 6 is a graph showing the rearrangement error and pixel difference converted into two-dimensional data in the non-destructive testing method of this embodiment.

[0063] As shown in Figure 4, the non-destructive testing system 11A of this embodiment is similar to the first embodiment in that it comprises an oscillating probe 21, a receiving probe 31, and a PC 41 including a determination device 51A. However, the determination device 51A of this non-destructive testing system 11A has a slightly different configuration from the determination device 51 of the non-destructive testing system 11 of the first embodiment. This determination device 51A comprises a correction processing unit 52, an image generation unit 53, a storage unit 55, and a determination unit 54. Of these, the correction processing unit 52 and the image generation unit 53 are the same as in the first embodiment, but the storage unit 55 and the determination unit 54 have different functions from those of the first embodiment.

[0064] The memory unit 55 of this embodiment stores various programs executed by the CPU in the PC 41 (for example, judgment programs using the first and second judgment methods described later), as well as a predetermined trained autoencoder (trained AE) 56. An autoencoder is a type of neural network algorithm that compresses input data, retains only the important features, and then restores it back to its original dimensions. The trained autoencoder 56 of this embodiment was trained by inputting a set of time-frequency images of healthy concrete utility poles 1 that have not deteriorated, but the details of this will be described later.

[0065] The determination unit 54 of this embodiment determines the deterioration status of the concrete utility pole 1 using a trained autoencoder 56, and has a first function unit and a second function unit. The first function unit calculates the Mahalanobis distance from the first output value and the second output value. The first output value refers to the output value obtained when a set of spectrogram images (time-frequency images) of a healthy concrete utility pole 1 without deterioration is input to the trained autoencoder 56. The second output value refers to the output value obtained when a spectrogram image (time-frequency image) of the concrete utility pole 1 to be inspected is input to the trained autoencoder 56. The second function unit determines the deterioration status of the concrete utility pole 1 to be inspected based on the calculated Mahalanobis distance.

[0066] Next, the procedure for non-destructive testing of the concrete utility pole 1 using this non-destructive testing system 11A will be described.

[0067] In this non-destructive testing method, it is necessary to train an autoencoder to generate a trained autoencoder 56 (trained AE). The upper part of Figure 5 illustrates the training process of the autoencoder 56A. The autoencoder 56A has an encoder and a decoder. A predetermined spectrogram image is input to the encoder as normal data. Specifically, a group of spectrogram images of a healthy concrete utility pole 1 without deterioration is input as normal data. This group of spectrogram images is generated through the steps of preparation, start, oscillation, reception, data transmission, time-frequency analysis, correction processing, and image generation, similar to the first embodiment (see Figures 3(b), (d), and (f)). The encoder compresses (reduces the dimensionality of) the input spectrogram image data, leaving only the important features. The compressed spectrogram image data is input to the decoder and restored to its original dimensions. Then, the autoencoder 56A adjusts the weighting of the compression by the encoder and the restoration by the decoder so that the difference between the restored data output from the decoder and the normal data is minimized. Through this learning process, a trained autoencoder 56 suitable for determining the deterioration status of the concrete utility pole 1 is generated. Note that there is no particular limit to the number of normal data (i.e., the number of spectrogram images) used in the above learning (unsupervised learning), but here, for example, it is set to about 100 to 300 images.

[0068] The lower part of Figure 5 illustrates the inference process using the trained autoencoder 56. In the inference process, the spectrogram image of the object to be inspected is input to the encoder. When a spectrogram image of a healthy concrete utility pole 1 without deterioration (normal data) is input, the decoder outputs reconstructed data A after compression by the encoder. The difference between the reconstructed data A generated in this way and the normal data becomes small. In contrast, when a spectrogram image of a deteriorated concrete utility pole 1 (abnormal data) is input, the decoder outputs reconstructed data B after compression by the encoder. The difference between the reconstructed data B generated in this way and the abnormal data becomes large.

[0069] In the non-destructive testing method of this embodiment, the first functional unit of the determination unit 54 calculates the Mahalanobis distance MD1 using the trained autoencoder 56 (calculation step). Specifically, a group of spectrogram images of healthy concrete utility poles 1 without deterioration are input to obtain the first output value. In this embodiment, the training data (normal data) used in the above training process is reused to obtain the first output value (reorganization error and pixel difference). The first functional unit then converts the reorganization error and pixel difference, which are the first output values, into two-dimensional data (see Figure 6). This is defined as the normal data group.

[0070] Meanwhile, the time-frequency image of the concrete utility pole 1, which is the object of inspection, is input to obtain the second output value (reorganization error and pixel difference). This is defined as the inspection data. The spectrogram image, which is the object of inspection, is also generated through the following steps: preparation, start, oscillation, reception, data transmission, time-frequency analysis, correction processing, and image generation (see Figures 3(a) to (f)). The first functional unit then converts the reorganization error and pixel difference, which are the second output values, into two-dimensional data (see Figure 6). The first functional unit then calculates the Mahalanobis distance MD1 from the center point of the normal data group to the point indicated by the inspection data. In the upper left of the graph in Figure 6, numerous points representing the normal data group are plotted. Also, in the graph in Figure 6, numerous points representing another data group are plotted at a position separated to the lower right from the normal data group D1. These represent the abnormal data group D2. Looking at this graph, the normal data group D1 falls within a range where the Mahalanobis distance MD1 is less than 3.5. On the other hand, the abnormal data group D2 has a Mahalanobis distance MD1 of 3.5 or greater. Therefore, the threshold TH1 for distinguishing between the normal data group D1 and the abnormal data group D2 is set to "3.5". This threshold TH1 may be set by the operator themselves, or it may be set automatically by the first function unit. The set threshold TH1 is stored in the storage unit 55.

[0071] Next, the second functional unit of the determination unit 54 determines the state of deterioration of the concrete utility pole 1, which is the subject of inspection, based on the calculated Mahalanobis distance MD1 (determination step). In the determination step, the determination may be made using, for example, the first determination method described below. That is, the second functional unit of the determination unit 54 compares the Mahalanobis distance MD1 calculated for the subject of inspection with a pre-set threshold TH1 (=3.5). If the calculated Mahalanobis distance MD1 is less than the threshold TH1, it is determined that there is no deterioration, and if it is greater than or equal to the threshold TH1, it is determined that there is deterioration.

[0072] In the judgment step, a second judgment method described below may also be used to make a judgment. Specifically, the second functional unit of the judgment unit 54 compares the Mahalanobis distance MD1 calculated for the object to be inspected with the threshold TH1 (=3.5). If the Mahalanobis distance MD1 is greater than or equal to the threshold TH1 (=3.5) and the difference between the two is small, it is determined that the degree of deterioration is small. For example, if the difference is 0 or more and less than 1 (3.5 ≤ MD1 < 4.5), it is determined that no damage is visible on the surface of the utility pole and the degree of deterioration is small. Also, if the calculated Mahalanobis distance MD1 is greater than or equal to the threshold TH1 (=3.5) and the difference between the two is large, it is determined that although no damage is visible on the surface of the utility pole, the degree of deterioration is large. For example, if the difference is 1 or more and less than 2 (4.5 ≤ MD1 < 5.5), it is determined that the degree of deterioration is large. Furthermore, if the difference is, for example, 2 or more (MD1 ≥ 5.5), it may be determined that the degree of deterioration is very large.

[0073] Therefore, according to the above-described embodiment, the following effects can be obtained.

[0074] (1) The determination device 51A in the non-destructive testing system 11A of this embodiment includes a correction processing unit 52, an image generation unit 53, a determination unit 54 having a first function unit and a second function unit, a storage unit 55 for storing a learned autoencoder 56, etc. The non-destructive testing method using this system 11A includes a correction processing step performed by the correction processing unit 52, an image generation step performed by the image generation unit 53, and a calculation and determination step performed by the determination unit 54, as described above. With this configuration, after the signal intensity of the STFT data is enhanced and corrected for each frequency in the correction processing step, a spectrogram image is generated in the image generation step. Therefore, a spectrogram image that facilitates calculation and determination in the subsequent calculation step and determination step can be obtained. Specifically, a spectrogram image can be obtained in which line segments Ln and Lc indicating signal intensity appear at positions corresponding to the input frequency band and positions corresponding to the natural frequency band of the concrete utility pole 1. Then, in the calculation step, a suitable learned autoencoder 56 can be obtained relatively easily by using the spectrogram image group including these line segments Ln and Lc. Furthermore, by inputting a set of spectrogram images containing these line segments Ln and Lc into the trained autoencoder 56 and outputting predetermined first and second output values, an appropriate Mahalanobis distance MD1 can be calculated relatively easily. In the judgment step, a judgment is made based on the calculated Mahalanobis distance MD1 (score), thereby enabling the detection and accurate determination of deterioration that has not yet become apparent. In addition, the degree of deterioration can be evaluated objectively and quantitatively without human intervention.

[0075] Furthermore, the determination device 51A in the above configuration is equipped with a trained autoencoder 56. The autoencoder can be trained using only the input of the normal data, and has the advantage of requiring only a relatively small amount of training data. In addition, compared to other similar neural networks, using an autoencoder reduces the computational load, which in turn makes it easier to miniaturize the determination device 51A. As a result, it becomes possible to adopt a configuration such as the modified non-destructive testing system 11B. Figure 7 is a block diagram showing a modified non-destructive testing system 11B of this embodiment. In the modified non-destructive testing system 11B, the determination device 51A equipped with the trained autoencoder 56 is provided on the receiving probe 31 instead of the PC 41. Therefore, in this modified version, the degree of degradation is determined within the receiving probe 31, and the determination result is transmitted to the PC 41 via the communication device 35.

[0076] The above embodiment may be modified as follows.

[0077] In the first and second embodiments described above, the time-frequency analysis data was STFT data, and the time-frequency images were spectrogram images G1 and G2 generated based on the corrected STFT data obtained through the enhancement step, but the embodiment is not limited to these. For example, in another embodiment, the time-frequency analysis data may be wavelet transform data, and the time-frequency images may be scaler images generated based on the corrected wavelet transform data obtained through the enhancement step.

[0078] In the calculation step of the second embodiment described above, the reorganization error and pixel difference were output as the first and second output values ​​from the trained autoencoder 56, and the Mahalanobis distance MB1 was calculated through the conversion of these two-dimensional data, but this is not limited to this. For example, in another embodiment, the Mahalanobis distance MB1 may be calculated by adding yet another parameter to the reorganization error and pixel difference and converting these into three-dimensional data.

[0079] • In the correction processing steps of the first and second embodiments described above, the signal intensity of the time-frequency analysis data was squared and then enhanced and corrected for each frequency. However, the method is not limited to this, and it is also possible to cube the signal intensity, for example, and then enhance and correct for each frequency.

[0080] In the second embodiment described above, a correction processing step was performed to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasound signal for inspection for each frequency, followed by an image generation step to generate a time-frequency image. However, the embodiment is not limited to this. For example, in other embodiments, the correction processing step may be omitted and the image generation step may be performed to generate the time-frequency image.

[0081] Next, in addition to the technical ideas described in the claims, the technical ideas that can be grasped by the embodiments described above are listed below.

[0082] (1) An oscillator-side probe installed on a concrete utility pole, which inputs an ultrasonic inspection signal of a predetermined input frequency to the concrete utility pole, A receiving probe installed on the aforementioned concrete utility pole, which acquires the received waveform of the aforementioned ultrasonic signal for inspection, The device includes a determination device that acquires time-frequency analysis data from the acquired received waveform and determines the deterioration of the concrete utility pole based on a time-frequency image generated from the time-frequency analysis data, The determination device is An image generation unit generates a time-frequency image based on the time-frequency analysis data obtained by calculating from the received waveform of the ultrasonic signal used for inspection, A storage unit for storing a trained autoencoder that has been trained by inputting a set of time-frequency images of healthy concrete utility poles that have not deteriorated, A determination unit having the function of calculating the Mahalanobis distance from a first output value obtained by inputting a group of time-frequency images of a healthy concrete utility pole without deterioration to the pre-trained autoencoder, and a second output value obtained by inputting a time-frequency image of the concrete utility pole to be inspected to the pre-trained autoencoder, and the function of determining the deterioration status of the concrete utility pole to be inspected based on the calculated Mahalanobis distance. A non-destructive testing system for concrete utility poles, characterized by comprising the following features. [Explanation of Symbols]

[0083] 1: Concrete utility pole 11, 11A, 11B: Non-destructive testing systems for concrete utility poles. 21: Oscillator probe 31: Receiver probe 51, 51A: Judgment device 52: Correction Processing Unit 53: Image generation unit 54: Judgment section 55: Storage section 56: Pre-trained autoencoder D1: Normal data group G1, G2: Spectrogram images as time-frequency images Ln: Line segment representing the signal strength in the input frequency band. Lc: Line segment indicating signal strength in the natural frequency band. MD1: Mahalanobis distance TH1: Threshold

Claims

1. A method for non-destructively inspecting a concrete utility pole, comprising: causing an oscillating probe installed on the concrete utility pole to oscillate and input an inspection ultrasonic signal of a predetermined input frequency to the concrete utility pole; acquiring the received waveform of the inspection ultrasonic signal with a receiving probe installed on the concrete utility pole; calculating time-frequency analysis data; and inspecting the concrete utility pole non-destructively based on a time-frequency image generated from the time-frequency analysis data, A correction processing step which performs a correction process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasonic signal for inspection for each frequency, An image generation step is performed to generate a time-frequency image based on the corrected time-frequency analysis data obtained through the correction processing step, A determination step in which the deterioration status of the concrete utility pole is determined based on a line segment showing the signal strength in the input frequency band and a line segment showing the signal strength in the natural frequency band of the concrete utility pole that may appear at a frequency lower than the input frequency band in the generated time-frequency image. A non-destructive testing method for concrete utility poles, characterized by including the following:

2. The non-destructive testing method for concrete utility poles according to claim 1, characterized in that the correction processing step involves raising the signal intensity of the time-frequency analysis data to a power for each frequency.

3. The time-frequency analysis data is STFT data. The aforementioned time-frequency image is a spectrogram image generated based on the corrected STFT data obtained through the enhancement processing step. The non-destructive testing method for concrete utility poles according to feature 2.

4. The data from the aforementioned time-frequency analysis is wavelet transform data. The aforementioned time-frequency image is a scaler image generated based on the corrected wavelet transform data obtained through the enhancement processing step. The non-destructive testing method for concrete utility poles according to feature 2.

5. In the aforementioned determination step, In the first case, where only line segments indicating the signal strength of the input frequency band appear in the time-frequency image and line segments indicating the signal strength of the natural frequency band do not appear, it is determined that there is no deterioration of the concrete utility pole or that the degree of deterioration is relatively small. In the second case, where only the line segment indicating the signal strength of the natural frequency band appears in the time-frequency image, or in the third case, where both the line segment indicating the signal strength of the input frequency band and the line segment indicating the signal strength of the natural frequency band appear, it is determined that the degree of deterioration of the concrete utility pole is greater than in the first case. A non-destructive testing method for concrete utility poles according to any one of claims 2 to 4.

6. In the aforementioned determination step, In the time-frequency image, if both a line segment indicating the signal strength of the input frequency band and a line segment indicating the signal strength of the natural frequency band appear, and the signal strength of the input frequency band is relatively stronger (first case), then it is determined that there is no deterioration of the concrete utility pole or that the degree of deterioration is small. In the second case, where only the line segment indicating the signal strength of the natural frequency band appears in the time-frequency image, or in the third case, where both the line segment indicating the signal strength of the input frequency band and the line segment indicating the signal strength of the natural frequency band appear, and the signal strength of the input frequency band is relatively weaker, it is determined that the degree of deterioration of the concrete utility pole is greater than in the first case. A non-destructive testing method for concrete utility poles according to any one of claims 2 to 4.

7. A method for non-destructively inspecting a concrete utility pole, comprising: causing an oscillating probe installed on the concrete utility pole to oscillate and input an inspection ultrasonic signal of a predetermined input frequency to the concrete utility pole; acquiring the received waveform of the inspection ultrasonic signal with a receiving probe installed on the concrete utility pole; calculating time-frequency analysis data; and inspecting the concrete utility pole non-destructively based on a time-frequency image generated from the time-frequency analysis data, A correction processing step which performs a correction process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasonic signal for inspection for each frequency, An image generation step is performed to generate a time-frequency image based on the corrected time-frequency analysis data obtained through the correction processing step, A calculation step to calculate the Mahalanobis distance from a trained autoencoder that has been trained by inputting a set of time-frequency images of healthy concrete utility poles that are not degraded, a first output value obtained by inputting a set of time-frequency images of healthy concrete utility poles that are not degraded, and a second output value obtained by inputting a time-frequency image of the concrete utility pole to be inspected to the trained autoencoder, Based on the calculated Mahalanobis distance, a determination step is taken to determine the state of deterioration of the concrete utility pole that is the subject of inspection. A non-destructive testing method for concrete utility poles, characterized by including the following:

8. The non-destructive testing method for concrete utility poles according to claim 7, characterized in that the correction processing step involves raising the signal intensity of the time-frequency analysis data to a power for each frequency.

9. The time-frequency analysis data is STFT data. The aforementioned time-frequency image is a spectrogram image generated based on the corrected STFT data obtained through the enhancement processing step. The non-destructive testing method for concrete utility poles according to feature 8.

10. The data from the aforementioned time-frequency analysis is wavelet transform data. The aforementioned time-frequency image is a scaler image generated based on the corrected wavelet transform data obtained through the enhancement processing step. The non-destructive testing method for concrete utility poles according to feature 8.

11. In the calculation step described above, When a set of time-frequency images of healthy concrete utility poles without degradation is input to the aforementioned trained autoencoder, the reorganization error and pixel difference are converted into two-dimensional data, which are then defined as the normal data set. For the trained autoencoder, the reorganization error and pixel difference obtained by inputting the time-frequency image of the concrete utility pole to be inspected are converted into two-dimensional data, which is used as the inspection data. The Mahalanobis distance is calculated from the center point of the normal data group to the point indicated by the test data. A non-destructive testing method for concrete utility poles according to any one of claims 7 to 10.

12. In the aforementioned determination step, The calculated Mahalanobis distance is compared with a pre-set threshold, If the calculated Mahalanobis distance is less than the threshold, it is determined that there is no deterioration; if it is greater than or equal to the threshold, it is determined that there is deterioration. The non-destructive testing method for concrete utility poles according to feature 11.

13. In the aforementioned determination step, If the calculated Mahalanobis distance is greater than or equal to the threshold and the difference from the threshold is small, it is determined that the degree of deterioration is small. If the calculated Mahalanobis distance is greater than or equal to the threshold and the difference from the threshold is large, it is determined that the degree of deterioration is high. The non-destructive testing method for concrete utility poles according to feature 11.

14. An oscillator-side probe installed on a concrete utility pole, which inputs an ultrasonic inspection signal of a predetermined input frequency to the concrete utility pole, A receiving probe installed on the aforementioned concrete utility pole, which acquires the received waveform of the aforementioned ultrasonic signal for inspection, The system includes a determination device that acquires time-frequency analysis data from the received waveform and determines the state of deterioration of the concrete utility pole based on a time-frequency image generated from the time-frequency analysis data, The determination device is A correction processing unit performs a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasonic signal used for the inspection, for each frequency. An image generation unit generates a time-frequency image based on the corrected time-frequency analysis data obtained by the processing of the correction processing unit, A determination unit determines the state of deterioration of the concrete utility pole based on a line segment showing the signal strength in the input frequency band and a line segment showing the signal strength in the natural frequency band of the concrete utility pole that may appear at a frequency lower than the input frequency band, in the generated time-frequency image. A non-destructive testing system for concrete utility poles, characterized by comprising the following features.

15. An oscillator-side probe installed on a concrete utility pole, which inputs an ultrasonic inspection signal of a predetermined input frequency to the concrete utility pole, A receiving probe installed on the aforementioned concrete utility pole, which acquires the received waveform of the aforementioned ultrasonic signal for inspection, The system includes a determination device that acquires time-frequency analysis data from the received waveform and determines the state of deterioration of the concrete utility pole based on a time-frequency image generated from the time-frequency analysis data, The determination device is A correction processing unit performs a process to enhance and correct the signal intensity of the time-frequency analysis data obtained from the received waveform of the ultrasonic signal used for the inspection, for each frequency. An image generation unit generates a time-frequency image based on the corrected time-frequency analysis data obtained by the processing of the correction processing unit, A storage unit for storing a trained autoencoder that has been trained by inputting a set of time-frequency images of healthy concrete utility poles that have not deteriorated, A determination unit having the function of calculating the Mahalanobis distance from a first output value obtained by inputting a group of time-frequency images of a healthy concrete utility pole without deterioration to the pre-trained autoencoder, and a second output value obtained by inputting a time-frequency image of the concrete utility pole to be inspected to the pre-trained autoencoder, and the function of determining the deterioration status of the concrete utility pole to be inspected based on the calculated Mahalanobis distance. A non-destructive testing system for concrete utility poles, characterized by comprising the following features.

Images