Wave form signal processing system, structure evaluation system and wave form signal processing method

Abstract

To provide a waveform signal processing system, a structure evaluation system, and a waveform signal processing method that are capable of obtaining a desired signal even in a noisy environment.SOLUTION: A waveform signal processing system according to an embodiment of the present invention includes a neural network, a learning unit, and an extraction unit. The neural network generates at least first time series data related to noise and second time series data related to a signal other than noise. The learning unit updates parameters of the neural network based on a loss function including a primal constraint term whose value decreases as the similarity between synthesized time series data obtained by adding the first time series data and the second time series data generated by the neural network and observed time series waveform including noise increases. The extraction unit extracts at least one of the first time series data or the second time series data generated based on the updated parameters, as a target signal.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] An embodiment of the present invention relates to a waveform signal processing system, a structure evaluation system, and a waveform signal processing method. [Background technology]

[0002] In recent years, problems associated with the deterioration of industrial equipment and structures have become evident. Since the damage caused by accidents in industrial equipment and structures is immeasurable, technologies for monitoring the condition of industrial equipment and structures have been developed. For example, a technology is known that detects damage to structures by using the acoustic emission (AE) method, which uses a highly sensitive sensor to detect elastic waves generated by the occurrence or progression of internal cracks. Acoustic emission is an elastic wave generated by the progression of fatigue cracks in materials. In the AE method, elastic waves are detected as voltage signals (elastic waves) by an AE sensor using a piezoelectric element. Elastic waves are detected as a sign before the breakage of materials occurs, so the frequency of occurrence and signal strength of elastic waves are useful as indicators of the soundness of materials. For this reason, research into technologies that use the AE method to detect signs of deterioration of structures is actively being conducted.

[0003] AE detection technology is widely used, mainly in Europe and the United States, in corrosion diagnosis of oil tanks and in the manufacturing process of industrial equipment, and AE detection technology is also being standardized for some targets. As shown in Patent Document 1, it is also widely used to detect damage to gears and other mechanical parts that are often used in industrial equipment, and it is known that there is a correlation between abnormalities in industrial equipment and AE parameters extracted from elastic wave signal waveforms. Representative AE parameters include energy, RMS (Root Mean Square) value, and crest value (crest factor, the value obtained by dividing the absolute value of the amplitude by the RMS value of the amplitude).

[0004] Elastic waves are generally weak signals, and it is necessary to increase the signal level and detect them using a highly sensitive sensor and a highly amplified amplifier. Therefore, the AE method has an aspect of being vulnerable to noise. In conventional technology, depending on the environment, there is a possibility that noise may be superimposed to the extent that elastic wave detection is misjudged, making it difficult to obtain correct evaluation results. A representative noise filtering method is a method that improves the S / N (signal-noise) ratio by filtering in the frequency domain (low-pass filter or band-pass filter). However, it is difficult to obtain an effect when the frequency bands of noise and elastic waves overlap. Furthermore, filtering the noise may change the shape of the signal. In this case, the evaluation accuracy may decrease. As described above, in the past, it was sometimes impossible to obtain the desired signal in a noisy environment. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2001-153784 A [Patent Document 2] Patent Publication No. 2022-53728 Summary of the Invention [Problem to be solved by the invention]

[0006] An object of the present invention is to provide a waveform signal processing system, a structure evaluation system, and a waveform signal processing method that are capable of obtaining a desired signal even in a noisy environment. [Means for solving the problem]

[0007] A waveform signal processing system according to an embodiment includes a neural network, a learning unit, and an extraction unit. The neural network generates at least first time series data related to noise and second time series data related to a signal other than noise based on input random noise. The learning unit updates parameters of the neural network based on a loss function including a main constraint term whose value decreases as a result of a similarity between synthesized time series data obtained by adding the first time series data and the second time series data generated by the neural network and an observed time series waveform including noise. The extraction unit extracts at least one of the first time series data or the second time series data generated by the neural network as a target signal based on the parameters updated by the learning unit. [Brief description of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram showing an example of the configuration of a waveform signal processing device according to a first embodiment. [Diagram 2] FIG. 13 is a diagram showing an example of waveform data of a typical elastic wave. [Diagram 3] 3 is a diagram for explaining processing performed by the waveform signal processing device according to the first embodiment. [Figure 4] FIG. 2 is a diagram showing the flow of processing performed by the waveform signal processing device according to the first embodiment. [Diagram 5] FIG. 4 is a diagram showing the results of noise removal using the technique according to the first embodiment. [Figure 6] FIG. 4 is a diagram showing the results of noise removal using the technique according to the first embodiment. [Figure 7] A diagram showing each waveform and the time-frequency waveform obtained by short-time Fourier transform. [Figure 8] FIG. 13 is a graph showing the relationship between the added noise level and the peak signal-to-noise ratio. [Figure 9] 13 is a graph showing the relationship between the added noise level and the feature amount of an elastic wave. [Figure 10] 13 is a graph showing the relationship between the added noise level and the feature amount of an elastic wave. [Figure 11]FIG. 1 is a diagram showing an example of the configuration of a structure evaluation system according to an embodiment. [Figure 12] FIG. 2 is a diagram showing an example of the configuration of a signal processing unit in the embodiment. [Figure 13] FIG. 4 is a sequence diagram showing the flow of a deterioration state evaluation process performed by the structure evaluation system according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0009] Hereinafter, a waveform signal processing system, a structure evaluation system, and a waveform signal processing method according to embodiments will be described with reference to the drawings.

[0010] (First embodiment) Fig. 1 is a diagram showing an example of the configuration of a waveform signal processing device 1 in a first embodiment. The waveform signal processing device 1 is a device that receives observation data including noise as an input and generates a target signal based on the input observation data. Here, the observation data including noise is, for example, a digital signal obtained based on elastic waves detected by one or more sensors installed in industrial equipment, structures, etc. that are targets for anomaly detection. The target signal is a signal of either elastic waves or noise. The industrial equipment is, for example, a compressor, a prime mover, an electric motor, a pump, etc.

[0011] The sensor detects elastic waves (AE waves) generated from industrial equipment, structures, etc. The sensor converts the detected elastic waves into an electric signal such as a voltage signal and outputs the signal. For example, a piezoelectric element having sensitivity in the range of 10 kHz to 1 MHz is used for the sensor. There are various types of sensors, such as a resonance type having a resonance peak within a frequency range and a wideband type with suppressed resonance, but any type of sensor may be used. The method by which the sensor detects elastic waves includes a voltage output type, a resistance change type, and a capacitance type, but any detection method may be used. An acceleration sensor may be used instead of the sensor. In this case, the acceleration sensor detects elastic waves generated from industrial equipment, structures, etc. Then, the acceleration sensor converts the detected elastic waves into an electric signal by performing the same process as the sensor.

[0012] Between the waveform signal processing device 1 and one or more sensors, for example, an amplifier, an analog filter, and an AD (Analog Digital) converter are provided. The amplifier amplifies the electrical signal output from the sensor. The amplifier amplifies the electrical signal to a level that allows processing in, for example, an AD converter. The amplifier outputs the amplified electrical signal to the analog filter. The analog filter removes noise components outside a predetermined band. For example, the analog filter is a band pass filter (BPF). It is desirable to use a filter with a pass band width sufficiently wide so as not to distort the shape of the elastic wave (AE signal) as the band pass filter used here. The electrical signal from which noise has been removed by the analog filter is input to the AD converter. The AD converter quantizes the electrical signal from which noise has been removed and converts it into a digital signal. The AD converter outputs the digital signal to the waveform signal processing device 1.

[0013] The waveform signal processing device 1 receives a digital signal output from an AD converter as an input. The waveform signal processing device 1 uses a neural network that receives random noise as an input and outputs at least two signals, and the input digital signal to extract, as a target signal, either an elastic wave from which noise has been removed, or the noise, from the input digital signal.

[0014] The waveform signal processing device 1 includes a noise extraction unit 2, a neural network 3, a learning unit 4, and a target signal extraction unit 5. The waveform signal processing device 1 is configured using a processor such as a CPU (Central Processing Unit) and a memory. The waveform signal processing device 1 functions as a device including the noise extraction unit 2, the neural network 3, the learning unit 4, and the target signal extraction unit 5 by executing a program.

[0015] Some or all of the functional units of the noise extraction unit 2, the neural network 3, the learning unit 4, and the target signal extraction unit 5 may be realized by hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA (Field Programmable Gate Array), or may be realized by a combination of software and hardware. The program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a non-transitory storage medium such as a portable medium such as a flexible disk, a magneto-optical disk, a ROM (Read Only Memory), or a CD-ROM, or a storage device such as a hard disk built into a computer system. The program may be transmitted via a telecommunication line.

[0016] Some of the functions of the noise extraction unit 2, the neural network 3, the learning unit 4 and the target signal extraction unit 5 do not need to be pre-installed in the waveform signal processing device 1, and may be realized by installing additional application programs in the waveform signal processing device 1.

[0017] The noise extraction unit 2 identifies a noise region from the input digital signal. The noise region is a region where noise is dominant. Here, a method for identifying a noise region from a digital signal will be described. The timing of occurrence of an elastic wave is usually indefinite. Therefore, the elastic wave is stored going back a predetermined number of pre-triggers, with the timing when the amplitude of the signal exceeds a predetermined voltage threshold as the reference (arrival time = 0). The recording of the elastic wave ends when the amplitude or envelope of the elastic wave is deemed to be below a predetermined threshold, or when a predetermined number of samples has passed. In other words, the data stored as the elastic wave is added with data from the pre-trigger period immediately before the arrival time. As a result, the elastic wave is recorded as a series of burst waveforms.

[0018] 2 is a diagram showing an example of waveform data of a typical elastic wave. As shown in FIG. 2, in the waveform data of the elastic wave, the timing when the amplitude of the signal exceeds a predetermined voltage threshold +Vth is the arrival time of the elastic wave, and the region preceding the arrival time by a predetermined number of pre-triggers is the pre-trigger region N. p Pre-trigger area N p is the area that contains the data waveform before the arrival time. Pre-trigger area N p In this case, the data waveform just before the arrival time may contain not only noise but also elastic waves, and the further back in time from the arrival time, the more likely it is that only noise will be included. With this in mind, the pre-trigger region N p As shown in Figure 2, the noise region N s and buffer area N b The noise region N s is the region dominated by noise. b is a region where noise and elastic waves are likely to coexist. Note that region R1 in Figure 2 is a region where elastic waves are dominant.

[0019] The above-mentioned measurement method is only one example, and other measurement algorithms may be used as long as a signal containing an elastic wave and the period before and after it is recorded as a hit. In other words, if the arrival time is set to 0, there will be an area in the data before the arrival time 0 that does not contain elastic waves. Strictly speaking, there is a possibility that a minute elastic wave is included even just before the threshold is reached. The area that may contain this minute elastic wave is the buffer area N shown in Figure 2. b From time 0 to buffer area N b Previous data in noise region N s By determining the above, the noise region can be specified with high accuracy.

[0020] The noise extraction section 2 has a pre-trigger region N p and the buffer area N b The noise extraction unit 2 uses the input digital signal to extract the pre-trigger region N p Period and buffer area N bBased on the information of the period considered, the noise region N s A mask signal is generated that sets the area to 1 and the other areas to 0.

[0021] The neural network 3 is a network that uses noise as an input and generates at least two one-dimensional data sequence signals. The neural network 3 can be realized by, for example, configuring a fully connected layer, a transposed convolution layer that transposes the input to increase the resolution, and a combination of an upsample layer and a convolution layer. The activation function can be a sigmoid function, a ReLU function, a sin function, a tanh function, or the like. As long as the input and output conditions are satisfied, the configuration of the layers that configure the neural network 3 is not limited. In the following description, an example will be described in which the neural network 3 generates an elastic wave (first time series data) from which noise has been removed and a noise signal (second time series data) as two one-dimensional data sequence signals.

[0022] The learning unit 4 updates parameters (e.g., weights and biases) constituting the neural network 3 based on the digital signal input to the waveform signal processing device 1, the mask signal generated by the noise extraction unit 2, and the two one-dimensional data string signals generated by the neural network 3. The learning unit 4 is composed of a loss function calculation unit 6 and a parameter update unit 7.

[0023] The loss function calculation unit 6 calculates a loss based on the digital signal input to the waveform signal processing device 1, the mask signal generated by the noise extraction unit 2, and the two one-dimensional data string signals generated by the neural network 3. The parameter update unit 7 updates the parameters constituting the neural network 3 so that the loss calculated by the loss function calculation unit 6 approaches zero. The parameter update unit 7 performs parameter updates at least 1000 times, for example.

[0024] The target signal extraction unit 5 extracts at least one of the two one-dimensional data string signals generated by the neural network 3 as a target signal. It is assumed that the target signal to be at least either the elastic wave from which noise has been removed or the noise signal is set in advance by the target signal extraction unit 5. Therefore, the target signal extraction unit 5 extracts the elastic wave from which noise has been removed as the target signal, extracts the noise signal as the target signal, or extracts the elastic wave from which noise has been removed and the noise signal as the target signal, depending on the setting.

[0025] (Explanation of neural network processing) Next, a procedure for generating a target signal by the neural network 3 based on the mask signal generated by the noise extraction unit 2 and the input digital signal will be described. The number of recorded samples of a certain elastic wave waveform is assumed to be N. The observed time series data y(t) is considered to be a noise-free elastic wave x(t) to which noise n(t) has been added. In other words, the time series data y(t) can be expressed as in equation (1).

[0026]

number

[0027] The sampled signal is i ,x i ,n i (i = 0, 1, ..., N-1). Using neural network 3, x i ,n i The corresponding estimate ^x i , ^n i Let us consider generating each of these independently. Note that ^ is placed above x and n.

[0028] 3 is a diagram for explaining the processing performed by the waveform signal processing device 1 in the first embodiment. As shown in FIG. 3, the neural network 3 receives noise as an input and generates at least two one-dimensional data strings. Here, the two one-dimensional data strings generated by the neural network 3 are respectively referred to as the one-dimensional data string of noise ^n i and one-dimensional data sequence of elastic waves ^x i Let the noise one-dimensional data sequence ^n i and one-dimensional data sequence of elastic waves ^x i The constraint shown in the following formula (2) is set for the noise one-dimensional data sequence ^n i is one form of the first time series data, and is a one-dimensional data sequence of elastic waves ^x i is one aspect of the second time series data.

[0029]

number

[0030] The average value of the noise estimation signal (e.g., the input digital signal) within a time window of any length (starting sample position a, length w) is -n a,w Let the variance be σ a,w 2 Then, it is expressed as the following formula (3) and formula (4). Note that - is added above n.

[0031]

number

[0032]

number

[0033] Assuming that the noise n(t) is weakly stationary and that the mean and variance are invariant to time shifts, the mean and variance over a time window of any length are the mean over the entire noise minus n 0,N , variance σ 0,NThis condition can be expressed as the following formula (5), where k is the stride width for moving through a time window of any width w, and i is the number of steps.

[0034]

number

[0035] Here, I=floor((Nw) / k). The floor is a function that rounds down to the nearest whole number. As described above, since noise is dominant in the pre-trigger period, a mask signal M is set to 1 during the pre-trigger period and 0 during other periods. i Using (i=0, 1, ..., N-1), the mask signal M i represents the i-th element in the entire mask signal sequence (mask signal M).

[0036]

number

[0037] The MSE (Mean Square Error) in equation (6) represents the mean square error. On the other hand, since elastic waves are a physical vibration phenomenon, they are expected to be non-stationary but to change continuously on the time axis. This is expressed by restricting the average absolute value of the second derivative to be equal to or less than the reference value δ as shown in the following equation (7).

[0038]

number

[0039] In addition, assuming that δ is small enough for practical purposes, the loss function can be defined as the following equation (8). When incorporating it into the loss function described later, the coefficients can be appropriately adjusted to lower the priority relative to other constraints.

[0040]

number

[0041] These conditions are integrated to obtain the loss function L loss is defined as the following equation (9), where C1, C2, and C3 are coefficients of 1 or more.

[0042]

number

[0043] As shown in equation (9), the loss function L loss is the primary constraint L r and the noise constraint L n and AE constraint L AE and AC (Alternating Current) coupling constraint L ac The primary constraint L r is a one-dimensional data sequence of multiple signals (noise) generated by neural network 3. i and one-dimensional data sequence of elastic waves ^x i ) coincides with the time series data y(t) of the digital signal input to the waveform signal processing device 1. r is a term whose value decreases as the composite signal and the time series data y(t) of the digital signal become more similar.

[0044] Noise constraint L n (Noise constraint term) is a constraint that expresses the characteristics of noise (stationarity + pre-trigger constraint). Noise constraint L n is a one-dimensional noise data sequence ^n i is expressed so that the value becomes smaller the more similar it is to the noise region extracted by the noise extraction unit 2, and the one-dimensional data sequence of noise ^n i The AE constraint L includes a term in which the average or variance of the first range is small enough to match the average or variance of a second range different from the first range. Here, the first range and the second range are the stride width k. AE(Signal constraint term) is a constraint (continuity constraint) that expresses the characteristics of the elastic wave waveform. AE constraint condition L AE is a one-dimensional data sequence of elastic waves ^x i The term is expressed so as to be proportional to the sum of the magnitudes of the first-order differentials or second-order differentials of . The use of second-order differentials suppresses the amount of change in slope compared to the use of first-order differentials, so it is possible to express a more gentle signal. In this embodiment, the one-dimensional data sequence of elastic waves ^x i The term is expressed to be proportional to the sum of the magnitudes of the second derivatives of AC coupling constraint L ac is the condition where the average value is 0, assuming AC coupling. ac is the loss function L loss The AE constraint L in Eq. (9) does not have to be included in AE As such, equation (7) may be used.

[0045] The loss function calculation unit 6 calculates the loss based on the formula (9) using the input digital signal, the multiple signals generated by the neural network 3, and the mask signal M output from the noise extraction unit 2. Note that the loss function calculation unit 6 calculates the loss based on the one-dimensional data sequence ^n i At least one of the first range and the second range is set to a different range each time the parameters are updated, thereby adjusting the noise constraint condition L n The parameter update unit 7 updates the parameters of the neural network 3 so as to minimize the loss calculated by the loss function calculation unit 6.

[0046] Fig. 4 is a diagram showing a flow of processing performed by the waveform signal processing device 1 in the first embodiment. In the following explanation, the processing shown in Fig. 4 will be explained as a target signal extraction processing. The processing in Fig. 4 is executed when a digital signal is input to the waveform signal processing device 1.

[0047] The noise extraction unit 2 generates a mask signal M using the digital signal input to the waveform signal processing device 1 (step S11). Specifically, the noise extraction unit 2 extracts a noise region Ns This is to identify the noise region N s The noise extraction unit 2 identifies data corresponding to the identified noise region N s The data corresponding to the noise region N s The data in the other regions is set to 0 to generate a mask signal M. The noise extraction unit 2 outputs the generated mask signal M to the loss function calculation unit 6.

[0048] Random noise is input to the neural network 3. Initial parameters are set in the neural network 3 at the start of processing. The neural network 3 receives the random noise as input and generates a plurality of signals based on the initial parameters (step S12). As a result, the neural network 3 generates a one-dimensional data sequence of noise ^n i and the one-dimensional data sequence of elastic waves ^x i The neural network 3 generates the one-dimensional data sequence of the generated noise, i and the one-dimensional data sequence of elastic waves ^x i is output to the loss function calculation unit 6.

[0049] The loss function calculation unit 6 calculates the one-dimensional data sequence ^n of the digital signal, the mask signal M output from the noise extraction unit 2, and the noise output from the neural network 3. i and the one-dimensional data sequence of elastic waves ^x i The loss function calculation unit 6 receives the input digital signal, the mask signal M output from the noise extraction unit 2, and the one-dimensional data sequence ^n of noise output from the neural network 3. i and the one-dimensional data sequence of elastic waves ^x i The loss function calculation unit 6 calculates the loss based on the formula (9) using and (step S13). The loss function calculation unit 6 outputs the calculated loss to the parameter update unit 7.

[0050] The parameter update unit 7 updates the parameters of the neural network 3 based on the loss output from the loss function calculation unit 6 (step S14). Any method may be used by the parameter update unit 7 to update the parameters. For example, the parameter update unit 7 may update the parameters of the neural network 3 using a gradient method. Thereafter, the parameter update unit 7 updates the parameters of the neural network 3 until an update end condition is satisfied. The update end condition is a condition for ending the update of the parameters of the neural network 3, and may be, for example, that a predetermined number of updates (e.g., 1000 times) have been performed, or that the error has become less than a threshold value.

[0051] The parameter update unit 7 optimizes the parameters of the neural network 3 by updating the parameters of the neural network 3. The neural network 3 receives random noise as input and generates a plurality of signals based on the optimized parameters (step S15). The target signal extraction unit 5 extracts one or a plurality of signals as target signals from the plurality of signals generated in the processing of step S15 (step S16).

[0052] (Noise removal result) Figures 5 and 6 are diagrams showing the results of noise removal using the method of the first embodiment. (A) in Figure 5 shows the waveform of the true elastic wave (which would normally be unknown). (B) in Figure 5 shows the observed waveform obtained by adding white noise to the waveform of the true elastic wave (the waveform shown in (A) in Figure 5). (C) in Figure 5 shows the waveform of the elastic wave generated by the neural network 3 using the method of the first embodiment, and (D) in Figure 5 shows the waveform of the noise generated by the neural network 3 using the method of the first embodiment. As can be seen from (C) in Figure 5, this method is able to generate a waveform close to the waveform of the true elastic wave.

[0053] (A) to (H) in FIG. 6 show the waveforms of elastic waves generated according to the number of times the parameters of the neural network 3 are updated. The number of times the parameters are updated is "0", "10", "50", "100", "500", "1000", "5000", and "10000" in the order of (A), (B), (C), (D), (E), (F), (G), and (H) in FIG. 6. In (A) in FIG. 6, the number of times the parameters of the neural network 3 are updated is "0", so the waveform of the elastic wave generated by the neural network 3 based on the initial parameters is shown. As can be seen from (A) to (H) in FIG. 6, it can be seen that the noise is removed by increasing the number of times the parameters are updated. Furthermore, it can be seen that a waveform close to (A) in FIG. 5 is obtained when the number of times the parameters are updated is "1000" or more, as shown in (F) in FIG. 6. Therefore, by updating the parameters of the neural network 3 1000 times or more in the parameter update unit 7, it is possible to accurately obtain the waveform of the elastic wave from which the noise has been removed.

[0054] FIG. 7 is a diagram showing each waveform and a time-frequency waveform obtained by short time Fourier transform (STFT). FIG. 7(A) shows the true elastic wave waveform, FIG. 7(B) shows the elastic wave waveform with noise added, FIG. 7(C) shows the elastic wave waveform with noise removed by the method in the first embodiment, and FIG. 7(D) shows the elastic wave waveform with noise removed by the conventional method (band pass filter). As shown in FIG. 7(D), the conventional method performs frequency domain filtering, so it can be seen that the spectrum of the elastic wave waveform itself has changed. In contrast, it can be seen that the method in the first embodiment shown in FIG. 7(C) can remove noise while maintaining the frequency of the elastic wave waveform.

[0055] FIG. 8 is a diagram showing the relationship between the added noise level and the peak signal-to-noise ratio. The horizontal axis of FIG. 8 represents the added noise level (Noise amplitude), and the vertical axis represents the peak signal-to-noise ratio (PSNR) for the true elastic wave waveform. PSNR is a value obtained by multiplying the reciprocal of the mean square error with the true elastic wave waveform by the square of the peak amplitude and taking the logarithm, and is an evaluation index of signal quality. The smaller the error, the larger the PSNR value. R used in calculating PSNR is a normalization coefficient, and the maximum value that a signal can take is used. Generally, 20 dB or more is considered to be acceptable quality.

[0056] Here, the PSNR was evaluated for three types of waveforms: a waveform containing noise, a waveform in which noise was removed by the method in the first embodiment, and a waveform in which noise was removed by a Wiener filter that reduces additive noise. As shown in FIG. 8, the PSNR of a waveform containing noise decreases as the noise level increases. In contrast, it was confirmed that the SN ratio can be improved by up to 6 dB or more when noise is removed by the method in the first embodiment. It can be seen that in areas with high noise levels, a higher effect can be obtained than with the Wiener filter.

[0057] 9 and 10 are diagrams showing the relationship between the added noise level and the feature amount of the elastic wave. The horizontal axis of FIG. 9 represents the added noise level (Noise amplitude), and the vertical axis represents the AE energy. The horizontal axis of FIG. 10 represents the added noise level (Noise amplitude), and the vertical axis represents the peak value. The AE energy and the peak value are known to have a high correlation with bearing damage, and are AE parameters often used in bearing abnormality detection. It can be seen that the values ​​of all AE parameters deviate from the true value as the noise level increases. It can be seen that the AE parameters calculated after removing noise by the method of the first embodiment show values ​​close to the true value even when the noise level increases. From this, it is considered that the method of the first embodiment can improve the S / N ratio even in a noisy environment, and improve the evaluation accuracy.

[0058] According to the waveform signal processing device 1 configured as above, there are provided a neural network 3 that generates, based on input random noise, at least time series data of noise and time series data of elastic waves, a noise extraction unit 2 that extracts a noise region from an observed time series waveform (digital signal) containing noise, and a main constraint L that reduces the value of the time series data of a composite signal obtained by adding the time series data of noise generated by the neural network 3 and the time series data of elastic waves, as well as the observed time series waveform. r Loss function L including loss and a target signal extraction unit 5 that extracts at least one of the time series data of noise or the time series data of elastic waves generated by the neural network 3 as a target signal based on the parameters updated by the learning unit 4.

[0059] This satisfies the main constraint L for making the time series data of the composite signal, which is the sum of multiple signals generated by the neural network 3, closer to the digital signal input in real time. r Loss function L including loss The parameters of the neural network 3 are trained based on the above. Furthermore, a loss function L including constraints expressing noise characteristics and constraints expressing elastic wave characteristics is loss The parameters of the neural network 3 are learned based on the above. This allows the neural network 3 to generate time-series data of noise and time-series data of elastic waves with high accuracy. Therefore, it becomes possible to obtain a desired signal even in a noisy environment.

[0060] Furthermore, there is no need to perform learning in advance in the waveform signal processing device 1. This makes it possible to reduce the effort required to prepare learning data necessary for learning and the time required for learning.

[0061] Second Embodiment In the first embodiment, the loss function L lossHowever, the primary constraint L r and the noise constraint L n and AE constraint L AE and AC coupling constraint L ac As mentioned above, the AE constraint L AE has a lower priority than other constraint conditions. Therefore, in the second embodiment, the AE constraint condition L AE is set to "0" and the loss function L loss A configuration for obtaining the above will be described.

[0062] The configuration of the waveform signal processing device 1 in the second embodiment is similar to that in the first embodiment. The waveform signal processing device 1 in the second embodiment differs from the first embodiment in the processing in the loss function calculation unit 6. The following description will focus on the differences from the first embodiment.

[0063] The loss function calculation unit 6 calculates the loss based on the above formula (9) using the input digital signal, the multiple signals generated by the neural network 3, and the mask signal M output from the noise extraction unit 2. At this time, the loss function calculation unit 6 calculates the loss based on the above formula (9) using the AE constraint condition L AE is set to 0. This effectively reduces the loss function L loss However, the primary constraint L r and the noise constraint L n and AC coupling constraint L ac In addition, as in the first embodiment, the AC coupling constraint L ac is the loss function L loss It does not have to be included in the

[0064] According to the waveform signal processing device 1 in the second embodiment configured as above, when calculating the loss, the AE constraint L AE We use a loss function with 0. This allows us to achieve the AE constraint L AE Therefore, it is possible to reduce the calculation load compared to the first embodiment.

[0065] (Third embodiment) In the third embodiment, the AE constraint L different from that in the first embodiment is used. AE A configuration using the above will be described.

[0066] The loss function calculation unit 6 calculates the loss based on the above formula (9) using the input digital signal, the multiple signals generated by the neural network 3, and the mask signal M output from the noise extraction unit 2. At this time, the loss function calculation unit 6 calculates the loss based on the above formula (9) using the AE constraint condition L AE is set as the condition shown in the following equation (10). Note that, as in the first embodiment, the AC coupling constraint L ac is the loss function L loss It does not have to be included in the

[0067]

number

[0068] The AE constraint L shown in Eq. (10) AE is a condition that imposes a constraint that the target signal follows the observed signal (input digital signal) when the signal amplitude exceeds a predetermined voltage threshold. As shown in FIG. 2, the region where the signal amplitude exceeds the predetermined voltage threshold (region R1 in FIG. 2) is a region where elastic waves are dominant. Therefore, it is considered that there is a relatively small possibility that noise is included in the region where the signal amplitude exceeds the predetermined voltage threshold, and elastic waves are considered to change continuously in the region excluding region R1 after the arrival time in FIG. 2. Based on this idea, the loss function calculation unit 6 calculates the AE constraint condition L AE By using the above formula (10), it is possible to define a constraint condition that more accurately expresses the characteristics of the elastic wave. Formula (7) or formula (8) can be used as a condition for the elastic wave to change continuously.

[0069] According to the waveform signal processing device 1 of the third embodiment configured as above, when calculating the loss, the AE constraint L AEis expressed by classifying the cases where the amplitude of the signal exceeds a predetermined voltage threshold and where it does not. This makes it possible to calculate the loss using constraint conditions that more accurately express the characteristics of the elastic wave. This makes it possible to generate the target signal more stably.

[0070] (Modification 1 of the waveform signal processing device 1 in the first to third embodiments) The waveform signal processing device 1 in each of the above-described embodiments may be configured as a waveform signal processing system using a plurality of information processing devices. When configured in this manner, some of the functional units of the waveform signal processing device 1 are implemented in each of the plurality of information processing devices. For example, the noise extraction unit 2 may be implemented in a first information processing device, and the functional units other than the noise extraction unit 2 may be implemented in the first information processing device, or the target signal extraction unit 5 may be implemented in the first information processing device, and the functional units other than the target signal extraction unit 5 may be implemented in a second information processing device.

[0071] As an example, when the target signal extraction unit 5 is implemented in the first information processing device, the second information processing device transmits a plurality of signals generated by the neural network 3 whose parameters have been updated to the first information processing device. The first information processing device extracts, according to settings, an elastic wave from which noise has been removed as a target signal, a noise signal as a target signal, or an elastic wave from which noise has been removed and the noise signal as target signals, from the plurality of signals transmitted from the second information processing device.

[0072] (Modification 2 of the waveform signal processing device 1 in the first to third embodiments) The waveform signal processing device 1 in the first to third embodiments may be implemented in combination. Specifically, any of the configurations shown in the first to third embodiments may be used in the calculation of loss in the loss function calculation unit 6. The difference between the first to third embodiments is that the AE constraint condition L AEIn this way, when the waveform signal processing devices 1 according to the first to third embodiments are implemented in combination, the AE constraint L AE The method may be selected by the user or may be set in advance.

[0073] (System using waveform signal processing device) Next, a system to which the above-mentioned waveform signal processing device 1 is applied will be described. An example of a system to which the waveform signal processing device 1 is applied is a system for evaluating the soundness of industrial equipment or structures. In the following, a structure evaluation system 100 for evaluating the soundness of a structure will be described as an example, but the structure evaluation system 100 can also be similarly applied to evaluating the soundness of industrial equipment, although the target is different.

[0074] FIG. 11 is a diagram showing an example of the configuration of a structure evaluation system 100 in an embodiment. The structure evaluation system 100 is used to evaluate the soundness of a structure. In the following description, evaluation means determining the degree of soundness of a structure, that is, the deterioration state of the structure, based on a certain criterion. The deterioration state of a structure may be indicated in two stages, sound and deteriorated, or may be indicated by three or more evaluations according to the degree of deterioration (for example, in order of the least deteriorated, soundness level "1", soundness level "2", soundness level "3", ...). Soundness is a state in which there is no damage, or even if damage occurs, it is considered that no immediate action is required. Deterioration is a state in which damage occurs and immediate action is required.

[0075] In the following description, the structure is a bridge, but the structure does not have to be limited to a bridge. The structure may be any structure in which elastic waves are generated due to the occurrence or growth of cracks, or due to external impacts (e.g., rain, artificial rain, etc.). Note that bridges are not limited to structures erected over rivers, valleys, etc., but also include various structures (e.g., highway viaducts) that are erected above the ground.

[0076] Damage that affects the evaluation of the deterioration state of a structure includes damage inside the structure that impedes the propagation of elastic waves, such as cracks, cavities, and sedimentation. Here, cracks include vertical cracks, horizontal cracks, and diagonal cracks. A vertical crack is a crack that occurs in a direction perpendicular to the road surface. A horizontal crack is a crack that occurs in a direction horizontal to the road surface. A diagonal crack is a crack that occurs in a direction other than horizontal and vertical to the road surface. Sedimentation is a deterioration in which concrete turns into sediment, mainly at the boundary between the asphalt and the concrete deck.

[0077] The structure evaluation system 100 includes a plurality of sensors 20-1 to 20-U (U is an integer equal to or greater than 2), a plurality of amplifiers 21-1 to 21-U, a plurality of filters 22-1 to 22-U, a plurality of AD converters 23-1 to 23-U, a waveform signal processing device 1, a signal processing unit 30, and a structure evaluation device 40. The sensor 20-U (1≦u≦U) and the amplifier 21-u, the amplifier 21-u and the filter 22-u, and the filter 22-u and the AD converter 23-u are connected by wires to enable communication. Furthermore, each of the plurality of AD converters 23-1 to 23-U and the waveform signal processing device 1, and the waveform signal processing device 1 and the signal processing unit 30 are connected by wires to enable communication. The signal processing unit 30 and the structure evaluation device 40 are connected by wire or wirelessly to enable communication.

[0078] In the following description, sensors 20-1 to 20-U will be referred to as sensor 20 when there is no particular distinction, amplifiers 21-1 to 21-U will be referred to as sensor 20 and amplifier 21 when there is no particular distinction, filters 22-1 to 22-U will be referred to as filter 22 when there is no particular distinction, and AD converters 23-1 to 23-U will be referred to as AD converter 23 when there is no particular distinction.

[0079] The sensor 20 has a piezoelectric element and detects elastic waves generated from inside the structure. The sensor 20 is installed at a position where it can detect elastic waves on the surface of the structure. Specifically, the sensor 20 is installed on a surface different from the surface on which the impact is applied to the structure, at the same or different intervals in the vehicle travel axis direction and the vehicle travel axis orthogonal direction. The surface different from the surface on which the impact is applied is, for example, any of the road surface, side surface, and bottom surface. The vehicle travel axis direction refers to the direction in which the vehicle travels on the road surface. The vehicle travel axis orthogonal direction refers to the direction perpendicular to the vehicle travel axis direction. The sensor 20 converts the detected elastic waves into an electric signal. In the following description, a case in which the sensor 20 is installed on the bottom surface of the structure will be described as an example.

[0080] The impact given to the structure may be, for example, the collision of numerous objects such as rain with the structure, or a vehicle passing over the structure. When a vehicle passes over the structure, a load is applied to the road surface due to contact between the running part of the vehicle and the road surface. A large number of elastic waves are generated inside the structure due to deflection caused by the load. Each sensor 20 installed on the bottom surface of the structure can detect the elastic waves generated inside the structure.

[0081] For example, a piezoelectric element having sensitivity in the range of 10 kHz to 1 MHz is used for the sensor 20. There are various types of sensors 20, such as a resonance type having a resonance peak within a frequency range and a wideband type with suppressed resonance, but any type of sensor 20 may be used. The method of detecting elastic waves by the sensor 20 includes a voltage output type, a resistance change type, and a capacitance type, but any detection method may be used.

[0082] An acceleration sensor may be used instead of the sensor 20. In this case, the acceleration sensor detects elastic waves generated inside the structure. The acceleration sensor then converts the detected elastic waves into an electrical signal by performing the same process as the sensor 20.

[0083] The amplifier 21 amplifies the electrical signal output from the sensor 20. The amplifier 21 amplifies the electrical signal to a level that allows processing in, for example, an AD converter 23. The amplifier 21 outputs the amplified electrical signal to the filter 22.

[0084] The filter 22 removes noise components outside a predetermined band. The filter 22 is, for example, a band-pass filter. The electrical signal from which noise has been removed by the filter 22 is input to an AD converter 23.

[0085] The AD converter 23 quantizes the electrical signal from which the noise has been removed and converts it into a digital signal. The AD converter 23 outputs the digital signal to the waveform signal processing device 1.

[0086] The waveform signal processing device 1 receives as input the digital signal output from the AD converter 23. The waveform signal processing device 1 generates a target signal from the input digital signal. The waveform signal processing device 1 outputs the generated target signal to the signal processing unit 30. As described above, the target signal is at least one of an elastic wave from which noise has been removed and noise. In the following, an example will be described in which the target signal is an elastic wave from which noise has been removed.

[0087] The signal processing unit 30 receives the target signal output from the waveform signal processing device 1. The signal processing unit 30 performs signal processing on the input target signal. The signal processing performed by the signal processing unit 30 includes, for example, noise removal and extraction of elastic wave features. The signal processing unit 30 generates transmission data including the target signal after signal processing. The signal processing unit 30 outputs the generated transmission data to the structure evaluation device 40.

[0088] The signal processing unit 30 is configured using a digital circuit. The digital circuit is realized, for example, by an FPGA or a microcomputer. The digital circuit may be realized by a dedicated LSI (Large-Scale Integration). The signal processing unit 30 may be equipped with a non-volatile memory such as a flash memory or a removable memory.

[0089] The structure evaluation device 40 analyzes the transmission data collected from the signal processing unit 30 and estimates the deterioration state and the deterioration location of the structure, thereby enabling a diagnosis of the soundness of the structure and efficient maintenance and management.

[0090] 12 is a diagram showing an example of the configuration of the signal processing unit 30 in the embodiment. The signal processing unit 30 includes a waveform shaping filter 301, a gate generating circuit 302, an arrival time determining unit 303, a feature extracting unit 304, a transmission data generating unit 305, a memory 306, and an output unit 307.

[0091] The waveform shaping filter 301 removes noise components outside a predetermined band from the input target signal. The waveform shaping filter 301 is, for example, a digital band pass filter (BPF). The waveform shaping filter 301 outputs the target signal after the noise components have been removed (hereinafter referred to as the "noise-removed signal") to the gate generation circuit 302 and the feature extraction unit 304.

[0092] The gate generation circuit 302 receives the noise removal signal output from the waveform shaping filter 301. The gate generation circuit 302 generates a gate signal based on the input noise removal signal. The gate signal is a signal that indicates whether the waveform of the noise removal signal is sustained or not.

[0093] The gate generation circuit 302 is realized by, for example, an envelope detector and a comparator. The envelope detector detects the envelope of the noise-removed signal. The envelope is extracted, for example, by squaring the noise-removed signal and performing a predetermined process (for example, a process using a low-pass filter or a Hilbert transform) on the squared output value. The comparator determines whether the envelope of the noise-removed signal is equal to or greater than a predetermined threshold.

[0094] When the envelope of the noise-removed signal is equal to or greater than a predetermined threshold, the gate generation circuit 302 outputs a first gate signal indicating that the waveform of the noise-removed signal is sustained to the arrival time determination unit 303 and the feature extraction unit 304. On the other hand, when the envelope of the noise-removed signal is less than a predetermined threshold, the gate generation circuit 302 outputs a second gate signal indicating that the waveform of the noise-removed signal is not sustained to the arrival time determination unit 303 and the feature extraction unit 304. Note that, although the gate generation circuit 302 has been configured to determine whether the waveform of the noise-removed signal is sustained based on the envelope, the gate generation circuit 302 may process the noise-removed signal itself or a signal to which an absolute value is applied. The threshold used in this gate generation is referred to as a measurement threshold.

[0095] The arrival time determination unit 303 receives as input a clock output from a clock source such as a crystal oscillator (not shown) and a gate signal output from the gate generation circuit 302. The arrival time determination unit 303 determines the elastic wave arrival time using the clock input while the first gate signal is being input. The arrival time determination unit 303 outputs the determined elastic wave arrival time to the transmission data generation unit 305 as time information. The arrival time determination unit 303 does not perform processing while the second gate signal is being input. The arrival time determination unit 303 generates cumulative time information from the time of power-on based on the signal from the clock source. Specifically, the arrival time determination unit 303 may be a counter that counts edges of the clock, and the value of the register of the counter may be the time information. The register of the counter is determined to have a predetermined bit length.

[0096] The feature extraction unit 304 receives the noise-removed signal output from the waveform shaping filter 301 and the gate signal output from the gate generation circuit 302. The feature extraction unit 304 extracts the feature of the noise-removed signal by using the noise-removed signal input while the first gate signal is being input. The feature extraction unit 304 does not perform processing while the second gate signal is being input. The feature is information indicating the feature of the noise-removed signal. In other words, the feature of the noise-removed signal is the feature of the elastic wave detected by the sensor 20.

[0097] The feature amount is, for example, the amplitude [mV] of the waveform, the rise time [usec] of the waveform, the duration [usec] of the gate signal, the number of zero cross counts [times], the energy [arb.] of the waveform, the frequency [Hz], and the RMS value. The feature amount extraction unit 304 outputs parameters related to the extracted feature amount to the transmission data generation unit 305. When outputting the parameters related to the feature amount, the feature amount extraction unit 304 associates a sensor ID with the parameters related to the feature amount. The sensor ID represents identification information for identifying the sensor 20 installed in the area (hereinafter referred to as the "evaluation area") to be evaluated for the soundness of the structure.

[0098] The amplitude of the waveform is, for example, the maximum amplitude value in the noise elimination signal. The rise time of the waveform is, for example, the time T1 from when the gate signal starts to rise until the noise elimination signal reaches its maximum value. The duration of the gate signal is, for example, the time from when the gate signal starts to rise until the amplitude becomes smaller than a preset value. The zero cross count is, for example, the number of times the noise elimination signal crosses a reference line that passes through a zero value.

[0099] The energy of the waveform is, for example, a value obtained by integrating the squared amplitude of the noise-removing signal at each time point over time. The definition of the energy is not limited to the above example, and may be, for example, an approximation using the envelope of the waveform. The frequency is the frequency of the noise-removing signal. The RMS value is, for example, a value obtained by squaring the amplitude of the noise-removing signal at each time point and taking the square root.

[0100] The sensor ID, time information, and parameters related to the feature amount are input to the transmission data generation unit 305. The transmission data generation unit 305 generates transmission data including the input sensor ID, time information, and parameters related to the feature amount.

[0101] The memory 306 stores one or more pieces of transmission data generated by the transmission data generating unit 305. The memory 306 is, for example, a dual-port RAM (Random Access Memory).

[0102] The output unit 307 sequentially outputs one or more pieces of transmission data stored in the memory 306 to the structure evaluation device 40. For example, when the signal processing unit 30 and the structure evaluation device 40 are connected by a wire, the output unit 307 outputs one or more pieces of transmission data stored in the memory 306 to the structure evaluation device 40 via a wired cable. When the signal processing unit 30 and the structure evaluation device 40 are connected wirelessly, the output unit 307 outputs one or more pieces of transmission data stored in the memory 306 to the structure evaluation device 40 wirelessly.

[0103] 11, the structure evaluation device 40 includes a communication unit 41, a control unit 42, a storage unit 43, and a display unit 44. The communication unit 41 receives one or more pieces of transmission data output from the signal processing unit 30.

[0104] The control unit 42 controls the entire structure evaluation device 40. The control unit 42 is configured using a processor such as a CPU (Central Processing Unit) and a memory. The control unit 42 functions as an acquisition unit 421, an event extraction unit 422, a position determination unit 423, a distribution generation unit 424, and an evaluation unit 425 by executing a program.

[0105] Some or all of the functional units of the acquisition unit 421, the event extraction unit 422, the positioning unit 423, the distribution generation unit 424, and the evaluation unit 425 may be realized by hardware such as an ASIC (Application Specific Integrated Circuit), a PLD (Programmable Logic Device), or an FPGA, or may be realized by cooperation between software and hardware. The program may be recorded on a computer-readable recording medium. The computer-readable recording medium is, for example, a non-transitory storage medium such as a portable medium such as a flexible disk, a magneto-optical disk, a ROM (Read Only Memory), or a CD-ROM, or a storage device such as a hard disk built into a computer system. The program may be transmitted via an electric communication line.

[0106] Some of the functions of the acquisition unit 421, the event extraction unit 422, the position determination unit 423, the distribution generation unit 424, and the evaluation unit 425 do not need to be pre-installed in the structure evaluation device 40, and may be realized by installing additional application programs in the structure evaluation device 40.

[0107] The acquisition unit 421 acquires various information. For example, the acquisition unit 421 acquires transmission data received by the communication unit 41. The acquisition unit 421 acquires transmission data for an evaluation target period. The acquisition unit 421 stores the acquired transmission data in the storage unit 43.

[0108] The event extraction unit 422 extracts transmission data for one event from the transmission data for the evaluation period stored in the storage unit 43. An event refers to an elastic wave generating event that occurs in a structure. In this embodiment, the elastic wave generating event is the passage of a vehicle over a road surface. Note that the elastic wave generating event is not limited to the passage of a vehicle over a road surface, but may also be the application of an impact to a structure. The impact to a structure may be caused by the collision of countless minute objects, or may be caused by an artificial action such as spraying a drug or water, or a large number of strikes using a device, etc. The countless minute objects are objects that are generated by meteorological phenomena such as raindrops, hail, and sleet. It is desirable that the impact to a structure is applied uniformly to the evaluation target area.

[0109] When one event occurs, elastic waves are detected by multiple sensors 20 at approximately the same time. That is, the storage unit 43 stores transmission data related to elastic waves detected at approximately the same time. Therefore, the event extraction unit 422 sets a predetermined time window and extracts all transmission data whose arrival time is within the range of the time window as transmission data for one event. The event extraction unit 422 outputs the extracted transmission data for one event to the position location unit 423.

[0110] The time window range Tw may be determined using the elastic wave propagation velocity v in the target structure and the maximum sensor interval dmax so as to be in the range of Tw≧dmax / v. In order to avoid erroneous detection, it is desirable to set Tw to a value as small as possible, so that Tw=dmax / v can be practically satisfied. The elastic wave propagation velocity v may be obtained in advance.

[0111] The position locating unit 423 locates the position of the elastic wave source based on the sensor position information and the sensor ID and time information included in each of the multiple pieces of transmission data extracted by the event extracting unit 422.

[0112] The sensor position information includes information on the installation position of the sensor 20 in association with the sensor ID. The sensor position information includes information on the installation position of the sensor 20, such as latitude and longitude, or horizontal and vertical distances from a reference position of the structure. The positioning unit 423 holds the sensor position information in advance. The sensor position information may be stored in the positioning unit 423 at any timing before the positioning unit 423 performs positioning of the elastic wave source.

[0113] The sensor position information may be stored in the storage unit 43. In this case, the position locating unit 423 acquires the sensor position information from the storage unit 43 at the timing of performing position locating. A Kalman filter, a least squares method, or the like may be used to locate the position of the elastic wave source. The position locating unit 423 outputs the position information of the elastic wave source obtained during the evaluation period to the distribution generating unit 424.

[0114] The distribution generating unit 424 receives as input the position information of the multiple elastic wave sources output from the position determining unit 423. The distribution generating unit 424 generates an elastic wave source distribution using the input position information of the multiple elastic wave sources. The elastic wave source distribution represents a distribution indicating the positions of the elastic wave sources. More specifically, the elastic wave source distribution is a distribution in which the horizontal axis represents the distance in the traffic direction and the vertical axis represents the distance in the width direction, and indicates points indicating the positions of the elastic wave sources on virtual data representing the structure to be evaluated.

[0115] The distribution generating unit 424 generates an elastic wave source density distribution using the elastic wave source distribution. The elastic wave source density distribution represents a distribution in which a density value calculated according to the number of elastic wave sources included in each region is indicated for each region determined in advance in the elastic wave source distribution. Specifically, the distribution generating unit 424 first divides the elastic wave source distribution into a plurality of regions by dividing the elastic wave source distribution into predetermined sections. Next, the distribution generating unit 424 calculates the density of each region by dividing the number of elastic wave sources located in the region by the area of ​​the region. Then, the distribution generating unit 424 generates an elastic wave source density distribution by allocating the calculated density value of each region to each region. In this way, the distribution generating unit 424 generates an elastic wave source density distribution by calculating the density for the evaluation target region.

[0116] The evaluation unit 425 evaluates the deterioration state of the structure by using the elastic wave source density distribution generated by the distribution generation unit 424. For example, the evaluation unit 425 evaluates a region in the elastic wave source density distribution where the density of the elastic wave source is equal to or greater than a threshold as a healthy region, and evaluates a region where the density of the elastic wave source is less than the threshold as a damaged region.

[0117] The storage unit 43 stores the transmission data for the evaluation period acquired by the acquisition unit 421. The storage unit 43 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device.

[0118] The display unit 44 displays the evaluation results under the control of the evaluation unit 425. For example, the display unit 44 may display the corrected elastic wave source density distribution as the evaluation result, or may display the area considered to be a damaged area in a different display mode from other areas. The display unit 44 is an image display device such as a liquid crystal display or an organic EL (Electro Luminescence) display. The display unit 44 may be an interface for connecting the image display device to the structure evaluation device 40. In this case, the display unit 44 generates a video signal for displaying the evaluation result and outputs the video signal to the image display device connected to itself.

[0119] Fig. 13 is a sequence diagram showing the flow of the evaluation process of the deterioration state by the structure evaluation system 100 in the embodiment. In Fig. 13, the sensor 20, the amplifier 21, the filter 22, and the AD converter 23 are collectively referred to as the sensor etc.

[0120] Each of the multiple sensors 20 detects elastic waves generated within the structure (step S101). Each of the multiple sensors 20 converts the detected elastic waves into an electrical signal and outputs it to each amplifier 21. Each of the multiple amplifiers 21 amplifies the electrical signal output from the connected sensor 20 (step S102). Each of the multiple amplifiers 21 outputs the amplified electrical signal to each filter 22.

[0121] The amplified electrical signals output from each of the multiple amplifiers 21 are filtered by filters 22-u connected to each amplifier 21-u (step S103). This removes noise from the amplified electrical signals. The electrical signals from which the noise has been removed are input to each AD converter 23. Each of the multiple AD converters 23 converts the electrical signal filtered by the connected filter 22 into a digital signal (step S104). Each of the multiple AD converters 23 outputs the digital signal to the waveform signal processing device 1 (step S105).

[0122] The waveform signal processing device 1 receives as input the digital signals output from each AD converter 23. Since the waveform signal processing device 1 cannot process multiple digital signals simultaneously, it sequentially takes in the digital signals output from each AD converter 23 one by one and processes the taken-in digital signals. The waveform signal processing device 1 performs a target signal extraction process on the input digital signals (step S106). Details of the target signal extraction process have been explained in FIG. 4 and will not be described here. As a result, the waveform signal processing device 1 obtains an elastic wave from which noise has been removed as a target signal. The waveform signal processing device 1 outputs the target signal to the signal processing unit 30 (step S106).

[0123] The signal processing unit 30 inputs a digital signal of an elastic wave, which is a target signal output from the waveform signal processing device 1. The signal processing unit 30 inputs, for example, the noise-removed digital signals output from the waveform signal processing device 1 in order. The arrival time determination unit 303 of the signal processing unit 30 determines the arrival time of each elastic wave (step S108). Specifically, the arrival time determination unit 303 determines the elastic wave arrival time using a clock input while the first gate signal is input. The arrival time determination unit 303 outputs the determined elastic wave arrival time to the transmission data generation unit 305 as time information. The arrival time determination unit 303 performs this process on all input noise-removed digital signals.

[0124] The feature extraction unit 304 of the signal processing unit 30 extracts features of the noise-removed signal, which is a digital signal input while the first gate signal is being input (step S109). The feature extraction unit 304 outputs parameters related to the extracted features to the transmission data generation unit 305. The transmission data generation unit 305 generates transmission data including a sensor ID, time information, and parameters related to the features (step S110). The output unit 307 sequentially outputs the transmission data to the structure evaluation device 40 (step S111).

[0125] The communication unit 41 of the structure evaluation device 40 receives the transmission data output from the signal processing unit 30. The acquisition unit 421 acquires the transmission data received by the communication unit 41. The acquisition unit 421 records the acquired transmission data in the storage unit 43 (step S112). The event extraction unit 422 extracts transmission data for one event from the transmission data stored in the storage unit 43. The event extraction unit 422 outputs the extracted transmission data for one event to the position determination unit 423.

[0126] The positioning unit 423 locates the position of the elastic wave source based on the sensor ID and time information included in the transmission data output from the event extraction unit 422 and pre-stored sensor position information (step S113). Specifically, the positioning unit 423 first calculates the difference in arrival time of the elastic wave to each of the multiple sensors 20. Next, the positioning unit 423 locates the position of the elastic wave source using the sensor position information and information on the difference in arrival time.

[0127] The position locating unit 423 executes the process of step S113 every time transmission data of one event is output from the event extracting unit 422 during the measurement period. In this way, the position locating unit 423 locates the positions of the multiple elastic wave sources. The position locating unit 423 outputs the position information of the multiple elastic wave sources to the distribution generating unit 424. The distribution generating unit 424 generates an elastic wave source distribution using the position information of the multiple elastic wave sources output from the position locating unit 423. Specifically, the distribution generating unit 424 generates the elastic wave source distribution by plotting the positions of the elastic wave sources indicated by the obtained position information of the multiple elastic wave sources on virtual data.

[0128] The distribution generating unit 424 generates an elastic wave source density distribution using the generated elastic wave source distribution (step S114). Specifically, the distribution generating unit 424 first divides the elastic wave source distribution into a plurality of areas by dividing the elastic wave source distribution into predetermined sections. Next, the distribution generating unit 424 calculates the density of the elastic wave sources for each area. Then, the distribution generating unit 424 generates an elastic wave source density distribution by assigning, to each area, the calculated elastic wave source density value for each area. The distribution generating unit 424 outputs the generated elastic wave source density distribution to the evaluation unit 425.

[0129] The evaluation unit 425 evaluates the deterioration state of the structure using the elastic wave source density distribution output from the distribution generation unit 424 (step S115). The evaluation unit 425 outputs the evaluation result to the display unit 44. The display unit 44 displays the evaluation result output from the evaluation unit 425 (step S116). For example, the display unit 44 may display the elastic wave source density distribution as the evaluation result, or may display an area considered to be a damaged area in a different display mode from other areas.

[0130] In the structure evaluation system 100 configured as described above, the waveform signal processing device 1 generates an elastic wave from which noise has been removed using a neural network as a target signal. This makes it possible to improve the S / N ratio. In this manner, in the structure evaluation system 100, the waveform signal processing device 1 appropriately removes noise while minimizing changes in the shape of the signal waveform. Then, the structure evaluation device 40 evaluates damage to the structure using feature quantities obtained by performing signal processing on the elastic wave from which noise has been removed by the waveform signal processing device 1. This makes it possible to improve the accuracy of evaluation of the deterioration state of the structure.

[0131] (Variation 1 in the structure evaluation system) In the above embodiment, a configuration has been shown in which a plurality of AD converters 23-1 to 23-U are connected to one waveform signal processing device 1. The structure evaluation system 100 may include U waveform signal processing devices 1, and each of the AD converters 23-1 to 23-U may be connected to a different waveform signal processing device 1. When M waveform signal processing devices 1 are included, the structure evaluation system 100 may include U signal processing devices 30. In this case, each of the U waveform signal processing devices 1 may be connected to a different signal processing device 30.

[0132] (Variation 2 in the structure evaluation system) Some or all of the functional units included in the structure evaluation device 40 may be included in another device. For example, the display unit 44 included in the structure evaluation device 40 may be included in the other device. When configured in this manner, the structure evaluation device 40 transmits the evaluation results to the other device equipped with the display unit 44. The other device equipped with the display unit 44 displays the received evaluation results.

[0133] (Variation 3 in the structure evaluation system) When the structure evaluation device 40 evaluates the deterioration state of industrial equipment, the structure evaluation system 100 may include one or more sensors 20, and the control unit 42 of the structure evaluation device 40 may include an acquisition unit 421 and an evaluation unit 425. When the structure evaluation system 100 includes one sensor 20, one amplifier 21, one filter 22, and one AD converter 23 are also included. The one or more sensors 20 are installed around the arc spot of the industrial equipment. The one or more sensors 20 detect elastic waves generated by the industrial equipment and output the detected elastic waves to the amplifier 21 as an electrical signal. Thereafter, the processes performed by the amplifier 21, the filter 22, the AD converter 23, the waveform signal processing device 1, and the signal processing unit 30 are the same as those in the above-mentioned embodiment.

[0134] The acquisition unit 421 acquires the transmission data output from the signal processing unit 30. The evaluation unit 425 evaluates an abnormality of the industrial equipment based on the transmission data acquired by the acquisition unit 421. Specifically, the evaluation unit 425 evaluates an abnormality of the industrial equipment based on a combination of multiple feature amounts obtained from elastic waves. As an example, the evaluation unit 425 evaluates an abnormality of the industrial equipment when a rate at which the correlations of the multiple feature amounts of the multiple elastic waves deviate is equal to or greater than a threshold value.

[0135] According to at least one of the embodiments described above, there is provided a neural network 3 that generates at least first time series data related to noise and second time series data related to a signal other than noise based on input random noise, a learning unit 4 that updates parameters of the neural network 3 based on synthetic time series data obtained by adding the first time series data and the second time series data generated by the neural network 3, and a loss function including a main constraint term whose value becomes smaller as the observed time series waveform including noise becomes more similar, and a target signal extraction unit 5 that extracts at least one of the first time series data or the second time series data generated by the neural network 3 as a target signal based on the parameters updated by the learning unit 4, thereby making it possible to obtain a desired signal even in a noisy environment.

[0136] Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, and modifications can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope of the invention and its equivalents described in the claims, as well as in the scope and spirit of the invention. [Explanation of symbols]

[0137] 1...waveform signal processing device, 2...noise extraction unit, 3...neural network, 4...learning unit, 5...target signal extraction unit, 6...loss function calculation unit, 7...parameter update unit, 20, 20-1 to 20-U...sensor, 30...signal processing unit, 40...structure evaluation device, 41...communication unit, 42...control unit, 43...storage unit, 44...display unit, 100...structure evaluation system, 301...waveform shaping filter, 302...gate generation circuit, 303...arrival time determination unit, 304...feature extraction unit, 305...transmission data generation unit, 306...memory, 307...output unit, 421...acquisition unit, 422...event extraction unit, 423...position location unit, 424...distribution generation unit, 425...evaluation unit

Claims

1. a neural network that generates, based on input random noise, at least first time series data related to the noise and second time series data related to a signal other than the noise; a learning unit that updates parameters of the neural network based on a loss function including a primal constraint term whose value decreases as a similarity between a synthetic time series data obtained by adding the first time series data and the second time series data generated by the neural network increases and an observed time series waveform including noise increases; an extracting unit that extracts, as a target signal, at least one of the first time series data and the second time series data generated by the neural network based on the parameters updated by the learning unit; A waveform signal processing system comprising:

2. The learning unit updates parameters of the neural network based on the loss function further including a noise constraint term which is a constraint condition expressing a noise characteristic.

2. The waveform signal processing system of claim 1.

3. A noise extraction unit that extracts a noise region from the observed time-series waveform, the learning unit uses, as the noise constraint term, a term whose value becomes smaller as the first time-series data becomes more similar to the noise region extracted by the noise extraction unit; 3. The waveform signal processing system of claim 2.

4. the noise extraction unit extracts, as time series data in a noise region, at least a portion of time series data in a pre-trigger period recorded before the amplitude of the observed time series waveform exceeds a predetermined threshold; 4. The waveform signal processing system of claim 3.

5. the noise constraint term further includes a term that is small enough that a mean or variance of a first range of the first time series data matches a mean or variance of a second range different from the first range; 5. A waveform signal processing system according to any one of claims 2 to 4.

6. the learning unit sets at least one of the first range and the second range to a different range each time a parameter is updated.

6. The waveform signal processing system of claim 5.

7. the learning unit updates parameters of the neural network based on the loss function further including a signal constraint term which is a constraint condition expressing a feature of the signal other than the noise; 5. A waveform signal processing system according to any one of claims 1 to 4.

8. The learning unit uses, as the signal constraint term, a term proportional to a sum of magnitudes of first-order differentials or second-order differentials of the second time series data.

8. The waveform signal processing system of claim 7.

9. The learning unit updates parameters at least 1000 times.

5. A waveform signal processing system according to any one of claims 1 to 4.

10. The observed time series waveform is a signal waveform obtained as a result of observing elastic waves generated due to internal damage of a structure or industrial equipment with a sensor.

5. A waveform signal processing system according to any one of claims 1 to 4.

11. One or more sensors that detect elastic waves generated within a structure or industrial equipment; 2. The waveform signal processing system according to claim 1, further comprising: a waveform processing unit for extracting second time series data related to a signal other than noise as a target signal based on the elastic waves detected by the one or more sensors; an evaluation unit that evaluates a deterioration state of the structure or the industrial equipment based on the target signal extracted by the waveform signal processing system; A structure evaluation system comprising:

12. generating at least first time series data related to the noise and second time series data related to a signal other than the noise using a neural network based on the input random noise; updating parameters of the neural network based on a loss function including a primal constraint term whose value decreases as a similarity between synthesized time series data obtained by adding the first time series data and the second time series data generated by the neural network and an observed time series waveform including noise increases; extracting, as a target signal, at least one of the first time series data and the second time series data generated by the neural network based on the updated parameters; Waveform signal processing method.

Images