Structure detection device and application set thereof

Abstract

The invention discloses a structure detection device and an application kit thereof. The structure detection device is used for receiving at least one to-be-detected sound wave signal of a to-be-detected structure and judging the structure state of the to-be-detected structure and comprises a communication interface, a filtering and amplifying circuit and a logic circuit. The communication interface is used for receiving at least one sound wave signal to be detected. The filtering and amplifying circuit is used for filtering the at least one sound wave signal to be detected and outputting at least one filtering and amplifying signal. The logic circuit includes a diagnostic model constructed by a convolutional neural network, and determines a structural state of the at least one structure to be measured according to the at least one filtered and amplified signal.

Description

Technical Field

[0001] This disclosure relates to a structure detection device and its application set, and particularly to a device and its application set that uses acoustic vibration diagnosis technology to detect the state of a structure to be measured. Background Art

[0002] In recent years, industrial structure or pipeline-related accidents have occurred frequently. When an industrial structure or pipeline deteriorates or leaks due to anomalies, it usually leads to major disasters, such as casualties and property losses. The main reasons for the anomalies in industrial structures or pipelines are human factors, followed by material deterioration of structures / pipelines / equipment. To avoid disasters, monitoring and analyzing the deterioration or leakage of industrial structures or pipelines has become one of the important topics in this field.

[0003] Although various monitoring systems and technologies for industrial structures or pipelines have been developed in the industry, they still have many drawbacks. For example, they lack a safety diagnosis module for appropriate logical judgment and analysis, and require professional personnel for evaluation; they lack a monitoring technology for remotely sensing deterioration and are only applicable to local positions where sensors are located for inspection and monitoring; and they cannot issue early warning signals when deterioration occurs, etc. They still rely on manual walking inspections to listen to the changes in pipeline sound waves.

[0004] Therefore, there is a need to provide an advanced structure detection device and its application set. Summary of the Invention

[0005] An embodiment of this specification discloses a structure detection device for receiving at least one acoustic wave signal to be measured of a structure to be measured and determining the structural state of the structure to be measured. This structure detection device includes: a communication interface, a filter amplification circuit, and a logic circuit. The communication interface is used to receive this at least one acoustic wave signal to be measured. The filter amplification circuit is used to filter this at least one acoustic wave signal to be measured and output at least one filtered and amplified signal. The logic circuit includes a diagnostic model constructed by Convolutional Neural Networks (CNN), and determines the structural state of this at least one structure to be measured according to this at least one filtered and amplified signal.

[0006] Another embodiment of this specification discloses a structural detection set for receiving at least one acoustic wave signal to be measured of a structure to be measured and determining the structural state of the structure to be measured. This structural detection set includes: an acoustic wave sensing device and a structural detection device. This structural detection device includes a communication interface, a filter amplification circuit, and a logic circuit. The acoustic wave sensing device is used to extract and output at least one acoustic wave signal to be measured of at least one structure to be measured. The communication interface is electrically connected to the acoustic wave sensing device and receives this acoustic wave signal to be measured. The filter amplification circuit is used to filter this at least one acoustic wave signal to be measured and output at least one filtered and amplified signal. The logic circuit includes a diagnostic model constructed by a convolutional neural network, and determines the structural state of the structure to be measured according to this at least one filtered and amplified signal.

[0007] According to the above embodiment, this specification provides a device and its application set for detecting the state of a structure to be measured by using acoustic vibration diagnosis technology. Integrate at least one communication interface, a filter amplification circuit, and a logic circuit into a structural detection device. Receive at least one acoustic wave signal to be measured output by a distal acoustic wave sensing unit or a proximal acoustic wave sensing unit for at least one area to be measured in a contact or non-contact manner. Filter the acoustic wave signal to be measured by the filter amplification circuit and output a filtered and amplified signal, and then use the diagnostic model constructed by a convolutional neural network in the logic circuit to immediately determine the structural state of the structure to be measured according to the filtered and amplified signal output by the filter amplification circuit. Realize remote sensing diagnosis and monitoring of the state of the structure to be measured (for example, thinning and leakage of pipelines) to ensure the safe operation of the structure to be measured (pipelines). Brief Description of the Drawings

[0008] To have a better understanding of the above and other aspects of this specification, the following specific embodiments are given and described in detail in conjunction with the accompanying drawings as follows:

[0009] Figure 1 is a configuration block diagram of a structural detection set shown according to an embodiment of this specification;

[0010] Figure 2 is a time-frequency spectrum data diagram obtained by filtering the acoustic wave signal to be measured extracted by an acoustic wave sensing unit by using a single-chip audio control circuit shown according to an embodiment of this specification;

[0011] Figure 3 is shown according to an embodiment of this specification Figure 1 of the structural detection set to detect the structural state of the structure to be measured;

[0012] Figure 4 is a block diagram of a deep autoencoder shown according to an embodiment of this specification;

[0013] Figure 5 It shows the frequency domain band of the acoustic wave signal to be measured in a leakage state according to an embodiment of the present invention;

[0014] Figure 6 It shows, according to an embodiment of the present invention, a comparison diagram of the amplitude-position (length) relationship curves of multiple sets of sections to be measured stored in a database, which have the same structural degradation (leakage) characteristics but different characteristic positions;

[0015] Figure 7 It shows, according to an embodiment of the present invention, a comparison diagram of multiple amplitude-position (length) relationship curves corresponding to a specific characteristic frequency stored in a database;

[0016] Figure 8 It is a configuration block diagram of a structure detection set shown according to another embodiment of this specification

[0017] Figure 9 It shows, according to an embodiment of this specification, a layout schematic diagram of a structure detection device arranged on a printed circuit board; and

[0018] Figure 10 It shows, according to an embodiment of this specification, an exploded view of a proximal / distal acoustic wave sensing unit.

[0019]

Symbol Explanation

[0020] 10: Structure detection set

[0021] 11: Acoustic wave sensing device

[0022] 11a: Proximal acoustic wave sensing unit

[0023] 11t: Audio input / output interface

[0024] 12: Vibration and sound diagnosis module

[0025] 12m: Diagnosis model

[0026] 12b: Feature label

[0027] 12t: Training model

[0028] 13: Communication module

[0029] 14: Structure to be measured

[0030] 14A: Filtered and amplified signal

[0031] 14d: Structural degradation feature

[0032] 14f: Filtered signal

[0033] 14s: Section to be measured

[0034] 14k: Acoustic signal to be measured

[0035] 14t: Training acoustic signal

[0036] 14v: Verification acoustic signal

[0037] 14w: Vibration acoustic wave

[0038] 15: Database

[0039] 16: Single-chip audio control circuit

[0040] 16a: Audio transcoder

[0041] 16b: Filter amplifier

[0042] 17: Graphical user interface

[0043] 80: Structure detection set

[0044] 81: Acoustic sensing device

[0045] 81a - 81e: Remote acoustic sensing unit

[0046] 81t: Audio input / output interface

[0047] 87: Data preprocessing module

[0048] 84: Structure to be measured

[0049] 84s: Section to be measured

[0050] 84A: Filtered and amplified signal

[0051] 84k: Acoustic signal to be measured

[0052] 84f: Filtered signal

[0053] 84w: Vibration acoustic wave

[0054] 100: Structure detection device

[0055] 100a: Communication interface

[0056] 100b: Logic circuit

[0057] 100c: Image processing circuit

[0058] 100f: Filter amplifier circuit

[0059] 100g: Soil parameter sensor

[0060] 100m: Memory unit

[0061] 111: Outer cover

[0062] 112: High-sensitivity ceramic patch

[0063] 113: Electroacoustic induction material

[0064] 114: Sealing rubber ring

[0065] 131: Handheld device

[0066] 400: Deep autoencoder

[0067] 401: Encoder

[0068] 402: Decoder

[0069] 501a: Characteristic frequency

[0070] 501b: Characteristic frequency

[0071] 601: Amplitude-position (length) relationship curve

[0072] 800: Structure detection device

[0073] 800a: Communication interface

[0074] 800b: Logic circuit

[0075] 800c: Image processing circuit

[0076] 800f: Filtering and amplifying circuit

[0077] 800m: Memory unit

[0078] 900: Printed circuit board

[0079] 901: Edge computing intelligent networking server

[0080] Q1-Q16: Curve

[0081] Z: Feature

[0082] WL: Signal line

[0083] S31: Construct a training model based on a deep neural network

[0084] S32: Input at least two training acoustic signals into the training model for training

[0085] S33: Construct a diagnostic model based on a convolutional neural network according to the training results

[0086] S34: Input the acoustic signal to be measured into the diagnostic model to judge the structural deterioration state of the section to be measured Detailed implementation manner

[0087] This specification provides a device and its application set for detecting the state of a structure to be measured using acoustic vibration diagnosis technology, which can achieve remote sensing diagnosis and monitoring and interpretation of the structural state of the structure to be measured to ensure the safe operation of the structure to be measured (pipeline). In order to make the above-mentioned embodiments and other purposes, features, and advantages of this specification more obvious and understandable, multiple preferred embodiments are specifically given below and described in detail in conjunction with the drawings.

[0088] However, it must be noted that these specific implementation cases and methods are not used to limit the present invention. The present invention can still be implemented using other features, elements, methods, and parameters. The presentation of the preferred embodiments is only used to illustrate the technical features of the present invention and not to limit the claims of the present invention. Those skilled in the art will be able to make equivalent modifications and changes within the spirit of the present invention based on the following description of the specification. In different embodiments and drawings, the same elements will be represented by the same element symbols.

[0089] Please refer to Figure 1 , Figure 1 which is a configuration block diagram of a structure detection set 10 shown according to an embodiment of this specification. The structure detection set 10 includes at least one acoustic wave sensing device 11 and a structure detection device 100. The structure detection device 100 includes a communication interface 100a, a filter amplification circuit 100f, a memory unit 100m, and a logic circuit 100b.

[0090] The acoustic wave sensing device 11 is used to extract a measured acoustic wave signal 14k from a measured section 14s of a structure to be measured 14. In some embodiments of this specification, the structure to be measured 14 can be (but is not limited to) a pipeline structure, such as an oil pipeline, a water pipeline, or other pipelines for transporting liquids or gases; it can also be a solid structure, such as a floor structure, a road filling structure, a steel frame structure, or other structures that can generate vibration wave signals through sound resonance. In one embodiment, the acoustic wave sensing device 11 extracts a measured acoustic wave signal 14k from the measured section 14s of the structure to be measured 14 in a non-contact / at a distance manner. In one embodiment, the acoustic wave sensing device 11 extracts a measured acoustic wave signal 14k from the measured section 14s of the structure to be measured 14 in a direct contact or indirect contact manner.

[0091] For example, in an embodiment of the present specification, the acoustic wave sensing device 11 may include a proximal acoustic wave sensing unit 11a and an audio input / output interface 11t. Among them, the proximal acoustic wave sensing unit 11a may be, for example (but not limited to), a portable high-sensitivity piezoelectric probe, which can be carried by a leak detection personnel to different positions of the structure to be measured 14 (pipe structure) for detection, so as to extract the acoustic wave signal 14k to be measured of the structure to be measured 14 (pipe structure). The audio input / output interface 11t of the acoustic wave sensing device 11 can be used for input and output of the acoustic wave signal 14k to be measured. For example, in this embodiment, the proximal acoustic wave sensing unit 11a of the acoustic wave sensing device 11 may not be in direct contact with the structure to be measured 14 (pipe structure), but is separated from the structure to be measured 14 (pipe structure) by a certain distance. The acoustic wave signal 14k to be measured may be, for example (but not limited to), a time domain vibration waveform (time waveform).

[0092] The memory unit 100m of the structure detection device 100 may be, for example (but not limited to), a memory chip. For example, in this embodiment, the memory unit 100m includes a memory array composed of a plurality of non-volatile memory cells (non-volatile memory cell) and a peripheral circuit (peripheral circuits) (not shown) including a decoder, a data buffer, and a sense amplifier, and can construct a database 15 for storing the acoustic wave signal 14k to be measured and various data for performing subsequent steps.

[0093] The communication interface 100a of the structure detection device 100 is electrically connected to the audio input / output interface 11t of the acoustic wave sensing device 11 and receives the acoustic wave signal 14k to be measured extracted by the acoustic wave sensing device 11. In this embodiment, the communication interface 100a includes a signal line WL, which connects a handheld leak detection probe of the proximal acoustic wave sensing unit 11a and the filter amplification circuit 100f of the structure detection device 100. The audio input / output interface 11t of the acoustic wave sensing unit device 11 may include a bus and an input / output circuit connected to the signal line WL.

[0094] The filtering and amplifying circuit 100f of the structure detection device 100 can be a single-chip audio control circuit 16, which is used to perform a filtering step, including performing a square-wave fast Fourier transform (FFT) on the acoustic wave signal 14k to be measured and then performing a Mel-frequency cepstrum (MFC) analysis, and providing it to the logic circuit 100b for subsequent structure detection (hereinafter referred to as: acoustic vibration diagnosis). The single-chip audio control circuit 16 includes an audio transcoder 16a and a filtering amplifier 16b. Among them, the audio transcoder 16a is used to perform a transformation step, and the filtering amplifier 16b is used to perform a spectrum analysis step, so as to obtain a filtered and amplified signal 14A from each acoustic wave signal 14k extracted from the structure to be measured 14 (pipe structure).

[0095] Specifically, the transformation step performed by the audio transcoder 16a includes performing a time-frequency domain transformation, transforming the time-domain vibration waveform of the acoustic wave signal 14k to be measured into a frequency-domain waveform and extracting a part of the frequency-domain waveform as a filtering signal 14f. The spectrum analysis step performed by the filtering amplifier 16b includes extracting the filtering signal 14f and amplifying it, and outputting a frequency-domain band as the filtered and amplified signal 14A for subsequent acoustic vibration diagnosis.

[0096] In some embodiments of this specification, the frequency of the frequency-domain band used for acoustic vibration diagnosis is substantially between 10 Hz and 1800 Hz; preferably substantially between 30 Hz and 1600 Hz. The audio transcoder 16a performs a time-frequency domain transformation, transforming the acoustic wave signal 14k with an original time-domain vibration waveform into a frequency-domain waveform; and then extracting a part of the frequency-domain waveform, so that the acoustic wave signal 14k after being filtered by the filtering amplifier 16b has a frequency-domain band between 200 Hz and 700 Hz.

[0097] For example, in this embodiment, the filtering step includes using the audio transcoder 16a to perform a Mel-frequency cepstrum analysis on the acoustic wave signal 14k extracted by the acoustic wave sensing device 11 through a discrete square-wave fast Fourier transform. Among them, for example, the number of filters (single-chip audio control circuit 16) is 30, the Mel-frequency cepstrum parameter is 20 dimensions, the frequency range is 0 Hz to 44100 Hz, the Fourier transform is 2048 points, and the frame size of the audio file is 5 seconds (s). In addition, in order to avoid too drastic changes between frames, a 20-millisecond (ms) overlap can be taken between two frames. Obtain Figure 2 The shown time-frequency spectrum data graph 200, with the three axes being amplitude, frequency, and time respectively.

[0098] Figure 2 The acquisition method and process are as follows. The filtered signal 14f processed by the audio transcoder 16a can be handed over to the filter amplifier 16b to be further segmented into 5 frequency domain bands according to the time axis (seconds), and then each frequency domain band is divided into 2000 equal parts. The frequency and amplitude of each equal part are transformed into two-dimensional vectors (output filtered amplified signal 14A), forming matrix data with a size of 5 (time) × 2000 (frequency and amplitude). For the signal data obtained from each on-site inspection measurement, the above steps are continuously repeated. The to-be-detected acoustic wave signal 14k (including the signal representing the waveform of the vibration acoustic wave, hereinafter referred to as the vibration acoustic wave 14w) obtained from each measurement (extraction) is transformed into a piece of matrix data. Eventually, approximately 430,000 pieces of 5 × 2000 matrix data can be obtained, sorted according to the time point sequence, and stored in the database 15 as training data (training acoustic wave signal 14t and verification acoustic wave signal 14v) for training to establish a vibro-acoustic diagnosis module 12.

[0099] The logic circuit 100b of the structure detection device 100 includes a vibro-acoustic diagnosis module 12, which can judge the structural state of the to-be-detected structure 14 based on the filtered amplified signal 14A after filtering the to-be-detected acoustic wave signal 14k and the vibration acoustic wave 14w (currently stored in the database 15) as training data (for example, training acoustic wave signal 14t and verification acoustic wave signal 14v). In some embodiments of this specification, the logic circuit 100b may include an artificial intelligence (AI) chip, a Graphics Processing Unit (GPU), a Microcontroller Unit (MCU), etc., to construct the vibro-acoustic diagnosis module 12, and then judge the structural state of the to-be-detected section according to the to-be-detected acoustic wave signal 14k (frequency domain band) after filtering.

[0100] In addition, each vibration acoustic wave 14w must be processed by normalization before it can be trained using deep learning algorithms. For example, in this embodiment, the normalization method is min-max normalization. At a certain time point n, the readings obtained from 13 samplings form a vector (or a one-dimensional array); the maximum and minimum values of the measurement readings at each time are respectively composed of two vectors and , and the normalization is as follows in formula (1):

[0101] ……(1)

[0102] Using the difference method, subtract the normalized measurement reading at the previous time point from the measurement reading at the current time point as shown in Equation (2):

[0103] …(2)

[0104] where is the differential signal.

[0105] And calculate the sum of the differential signal while setting a threshold. As shown in Equation (3):

[0106] …(3)

[0107] If the sum of the differential signals is greater than the threshold, it can be determined that the input vibration acoustic wave 14w is a transient signal containing a waveform with drastic changes; otherwise, it can be determined that the input vibration acoustic wave 14w is a steady-state signal with a stable and gentle waveform change. (The vibration acoustic wave 14w after normalization processing includes the training acoustic wave signal 14t and the verification acoustic wave signal 14v after normalization processing, which may be transient signals or steady-state signals).

[0108] The acoustic vibration diagnosis module 12 is used to execute the method for acoustic vibration diagnosis of the structure to be measured. Please refer to Figure 3 Figure 3 which is a flowchart of the steps for detecting the structural state of the structure to be measured by the structure detection set 10 shown in Figure 1 . This acoustic vibration diagnosis method includes the following steps: First, construct a training model 12t based on a deep neural network (DNN) using unsupervised learning (as in step S31). Then, input at least two training acoustic wave signals 14t (stored in the database 15) into the training model 12t for training (as in step S32) (the vibration acoustic wave 14w after normalization processing includes the training acoustic wave signal 14t and the verification acoustic wave signal 14v after normalization processing). After that, construct a diagnosis model 12m based on a convolutional neural network (CNN) according to the training results (as in step S33). Then, input the real-time (current) detected acoustic wave signal 14k into the diagnosis model 12m to determine the structural degradation state of the section to be measured 14s (as in step S34).

[0109] ​Specifically, the training model 12t in the acoustic vibration diagnosis module 12 may include a deep autoencoder 400 based on a deep product neural network. Please refer to Figure 4 , Figure 4 which is a block diagram of a deep autoencoder 400 illustrated according to an embodiment of this specification. Among them, the architecture of the deep autoencoder 400 can be subdivided into two parts: an encoder 401 and a decoder 402, which respectively perform the work of compressing and decompressing the training acoustic wave signal 14t. In this embodiment, the deep autoencoder 400 is based on multi-layer fully connected layers. The number of neurons is the largest starting from the input layer of the encoder 401, and the number of neurons in each layer of the encoder 401 gradually decreases. The original training acoustic wave signal 14t data extracts features or reduces the data dimension through linear transformation or highly nonlinear transformation.

[0110] The decoder 402 then uses the code output by the encoder 401 for decompression to restore the input data. In other words, the input data and the output data of the deep autoencoder 400 will be the same. Since the fully connected layer can only accept a one-dimensional array as input, it is necessary to first flatten the 5×2000 input matrix of each vibration acoustic wave 14w to be input into the deep autoencoder 400, and change it into a one-dimensional array with a length of 10,000 and input it into the encoder 401. For example, the number of neurons in the encoder 401 decreases layer by layer from 10,000 to 5,000 and 2,500, and there are three consecutive fully connected layers in the encoder 401. There are three consecutive fully connected layers in the decoder 402, and the number of neurons in each layer is 2,500, 5,000, and 10,000 respectively, and finally 10,000 output values are output.

[0111] The training of the acoustic vibration diagnosis module 12 includes the following steps: First, 80% of the vibration sound waves 14w stored in the database 15 (for example, the data classified as steady state in the vibration sound waves 14w, obtained by equations (2) and (3)) are selected and input into the deep automatic encoder 400 of the training model 12t, and feature value extraction is performed by the encoder 401; a representative feature (feature) Z is extracted from the original training sound wave signal 14t, and a plurality of feature labels (labeling) 12b are pre-selected. After adjustment, it can be verified that the steady-state training data still has a good restoration effect after the compression and decompression process of the deep automatic encoder 400. In this embodiment, after the feature extraction of the deep automatic encoder 400, the training sound wave signal 14t can be roughly divided into feature labels 12b of four states: leakage frequency, metal frequency, environmental frequency, and noise frequency.

[0112] Subsequently, according to the feature labels 12b of the training model 12t, a diagnostic model 12m including a convolutional autoencoder is constructed based on a convolutional neural network. The remaining 20% ​​of the vibration sound waves 14w (for example, other vibration sound waves 14w including transients, obtained by equations (2) and (3)) are input into the convolutional autoencoder of the diagnostic model 12m as verification data (verification sound wave signal 14v) to test whether the diagnostic model 12m can successfully detect the occurrence of transients. Among them, the criterion for determining the transient state is the error between the signal restored by the convolutional autoencoder of the diagnostic model 12m and the original signal. When the error exceeds a preset threshold (the threshold is the signal-to-noise ratio value: 500), the input training data will be determined as transient. In this embodiment, the algorithm used by the convolutional autoencoder includes k-means clustering.

[0113] The output result of the diagnostic model 12m is compared with the verification data, and the weight of the diagnostic model 12m and the number of feature labels 12b are adjusted to complete the training of the acoustic vibration diagnostic module 12. After the training is completed, the sum of the feature values ​​of the feature labels 12b of the diagnostic model 12m is equal to 1. In this embodiment, the four feature labels are leakage frequency, metal frequency, environment frequency, and noise frequency.

[0114] After completing the training of the acoustic vibration diagnosis module 12, the real-time (currently) extracted sound wave signal 14k to be tested is input into the diagnosis model 12m of the acoustic vibration diagnosis module 12, and then according to the characteristic value output by each characteristic label 12b of the diagnosis model 12m, it can be determined what kind of pipeline structure the test section 14s from which the test sound wave signal 14k to be tested is currently extracted from the test structure 14 (pipeline structure), and the current structural state of the test section 14s.

[0115] In some embodiments of this specification, when the diagnostic model 12m determines whether there is a leak in the to-be-detected section 14s of the to-be-detected acoustic wave signal 14k extracted in real time (currently) from the to-be-detected structure 14 (pipe structure), the acoustic vibration diagnosis module 12 can also cross-compare the frequency domain band of the to-be-detected acoustic wave signal 14k with the historical data of the acoustic wave signals of multiple pipe structures of the same type stored in the database 15 and having different leak characteristics, so as to identify the relative position of the structure degradation feature 14d in the to-be-detected section 14s of the to-be-detected structure 14 (pipe structure), and estimate the degree of degradation of the structure degradation feature 14d.

[0116] In some embodiments of this specification, when the to-be-detected section 14s is determined to be in a leak state, multiple historical data of acoustic wave signals with different structure degradation (leak) characteristics 14d of the same pipe structure in the database 15 can be further compared to generate a frequency tracking diagram, a time-frequency diagram, a classification diagnosis result, and calibrate the position of the structure degradation (leak) feature 14d in the to-be-detected section 14s.

[0117] Specifically, when the to-be-detected section 14s is determined to be in a leak state, the frequency domain band of this to-be-detected acoustic wave signal 14k after time-frequency domain transformation will have at least one characteristic frequency (peak). For example, please refer to Figure 5 , Figure 5 which shows the frequency domain band of a to-be-detected acoustic wave signal 14k in a leak state according to an embodiment of the present invention. In this embodiment, according to the label value output by the characteristic label 12b of the diagnostic model 12m, the to-be-detected section 14s is a metal pipe determined to be in a leak state, and the frequency domain band of the to-be-detected acoustic wave signal 14k generates a characteristic frequency 501a and 501b at frequencies 290 Hz and 580 Hz respectively.

[0118] Then, based on the characteristic frequencies 501a and 501b and the characteristic amplitude values of the characteristic frequencies, multiple amplitude-position (length) relationship curves of the to-be-detected sections 14s in the database 15 with the same structure degradation (leak) characteristics 14d but different characteristic positions are compared to identify the position of the structure degradation (leak) feature 14d in the to-be-detected section 14s. For example, please refer to Figure 6 , Figure 6 which shows a comparison diagram of the amplitude (dB)-position (length) relationship curves of multiple to-be-detected sections 14s stored in the database 15 with the same structure degradation (leak) characteristics 14d but different characteristic positions according to an embodiment of the present invention.

[0119] In this embodiment, according to the characteristic frequencies 501a and 501b, an amplitude (mdB)-position (length, meter) relationship curve 601 can be obtained in the database 15. Among them, the amplitude-position (length) relationship curve 601 represents an amplitude-position (length) relationship curve with a frequency of 580Hz. Subsequently, according to the relationship curve 601 transformed by the characteristic amplitude value 340dB of the characteristic frequency 501b, the peak position of the relationship curve 601 is close to the peak position of the pipe length 4 / 8L, and the structural degradation (leakage) feature 14d can be calibrated to be roughly the relative position of the pipe length 4 / 8L in the section to be tested 14s.

[0120] In addition, the degradation degree of the structural degradation (leakage) feature 14d can be estimated by comparing the multiple amplitude-degradation degree relationship curves 601 corresponding to the specific characteristic frequencies in the database 15 according to the characteristic amplitude values ​​of the characteristic frequencies 501a and 501b. Figure 7 , Figure 7 1 is a graph showing a comparison of multiple amplitude-position (length) relationship curves corresponding to specific characteristic frequencies stored in the database 15 according to an embodiment of the present invention. Curves Q1-Q16 represent the amplitude-deterioration degree relationships of different defect sizes. Figure 7 At the intersection of the horizontal and vertical dashed lines, the structural degradation (leakage) feature 14d of the current section 14s to be tested can be estimated based on the characteristic frequency 302 (580 Hz), and the degree of degradation is approximately 70%.

[0121] In some embodiments of the present specification, the structure detection device 100 further includes a graphical user interface (GUI) 17 constructed by an image processing circuit 100c (e.g., including an embedded image processing card, an input / output circuit, an image processing chip, etc. (not shown)), which directly displays the diagnosis results obtained by the acoustic vibration diagnosis module 12, the acoustic wave signal extracted by the acoustic wave sensing device 11 (e.g., a tracking graph and a time-frequency graph), the detection position of the structure to be detected 14 (pipeline structure) and the map marking of the leakage point, etc., in a graphical manner on the display screen of the structure detection device 100.

[0122] In addition, the communication interface 100a of the structure detection device 100 also includes a communication module 13. Through the communication module 13 (e.g., a wireless network communication module), the graphical user interface 17 of the structure detection device 100 can be displayed on the display interface of the handheld device 131 (e.g., a smart phone or a notebook computer, etc.) of on-site leak detection personnel or other experts at a distance, so as to construct a human-machine integration interface. This enables on-site leak detection personnel or other experts at a distance to immediately grasp the acoustic vibration diagnosis results of the structure to be measured and share the detection information and historical records stored in the monitoring and management cloud platform.

[0123] In some embodiments, on-site leak detection personnel and experts can, according to their respective authorities, provide corrective opinions or instructions to the acoustic vibration diagnosis module 12 through the handheld device 131 of the communication module 13 to modify and update the diagnosis model 12m of the acoustic vibration diagnosis module 12.

[0124] Please refer to Figure 8 , Figure 8 is a configuration block diagram of the structure detection set 80 illustrated according to an embodiment of this specification. Figure 8 illustrated by the example of the structure detection set 80 and Figure 1 is similar to the structure detection set 10 illustrated by the example. The difference is that the structure detection set 80 includes an acoustic wave sensing device 81 composed of a plurality of remote acoustic wave sensing units 81a - 81e and a plurality of audio input / output interfaces 81t, and can be electrically connected to the communication interface 800a of the structure detection device 800 through a communication module 13 (e.g., a wireless network communication module).

[0125] In this embodiment, the remote acoustic wave sensing units 81a-81e can be a plurality of remote leak detection probes respectively arranged in different sections 84s to be tested of the structure 84 to be tested (e.g., a pipeline structure), and are used to extract at least one acoustic wave signal 84k to be tested. The structure detection device 800 also includes a global positioning system (GPS), and the global positioning system (not shown) can locate the different sections 84s to be tested where the remote acoustic wave sensing units 81a-81e are located. The audio input / output interface 81t can input and output at least one acoustic wave signal 84k to be tested (including a signal representing a waveform of a vibration acoustic wave, referred to as a vibration acoustic wave 84w) extracted by the remote acoustic wave sensing units 81a-81e, and transmit it to the structure detection device 800 through the communication module 13 and the communication interface 800a for subsequent structural detection, and store it in the database 15. Specifically, the remote acoustic wave sensing units 81a - 81e include a remote leak detection probe respectively, and the communication module 13 is electrically connected to the remote leak detection probe and the communication interface 800a , so that at least one acoustic wave signal 84k to be detected can be transmitted from the acoustic wave sensing device 81 to the structure detection device 800 .

[0126] The logic circuit 800b of the structure detection device 800 further includes a data preprocessing module 87, which uses artificial intelligence (AI) technology to filter out invalid data for structure detection from the sound wave signal 84k to be detected, so as to reduce the computational load and improve the performance (e.g., diagnostic accuracy) of the acoustic vibration diagnosis module 12. For example, in this embodiment, the data preprocessing module 87 applies machine learning, such as support vector machine (SVM) learning, to remove or replace invalid data for detecting structural degradation from the sound wave signal 84k to be detected. For example, from the three-dimensional vector (e.g., Figure 2 Invalid data for detecting structural degradation is removed or replaced in the 5 (time) × 2,000 (amplitude) time-frequency spectrum matrix shown in FIG.

[0127] Specifically, when the acoustic vibration diagnosis module 12 receives the sound wave signal 84k to be tested, the single-chip audio control circuit 16 performs a time domain to frequency domain transformation (e.g., discrete square wave fast Fourier transform (FFT) and / or Mel frequency cepstrum (MFC) analysis) to transform the original time waveform of the sound wave signal 84k to be tested into a frequency domain. Figure 2The shown three-axis (amplitude, frequency, and time) spectrogram 200 is a 2D / 3D feature plan view. That is, the filter amplification circuit 800f of the structure detection device 800 transforms and separates the acoustic wave signal 84k to be measured by the single-chip audio control circuit 16 and then outputs the filtered signal 84f. Then, a part of the frequency-domain waveform in the filtered signal 84f is extracted to output a frequency-domain band as the filtered amplification signal 84A. After the data preprocessing module 87 removes the environmental noise, the logic circuit 800b continues with the subsequent structure detection.

[0128] And the various data of the above steps are stored in the memory unit 800m, and the diagnostic results diagnosed by the acoustic vibration diagnosis module 12, the acoustic wave signals extracted by the acoustic wave sensing device 81, the detection positions of the structure to be measured 84 (pipe structure), and the map markings of the leakage points, etc., are directly displayed in a graphical manner on the display screen of the structure detection device 800 through the image processing circuit 800c.

[0129] In some embodiments of this specification, the communication interfaces 100a / 800a, logic circuits 100b / 800b, image processing circuits 100c / 800c, memory units 100m / 800m, and filter amplification circuits 100f / 800f in the structure detection devices 100 / 800 are constructed on a printed circuit board (PCB) 900 through a surface mount technology (SMT). For example, please refer to Figure 9 , Figure 9 is a schematic layout diagram showing the integration of the structure detection device 800 on the printed circuit board 900 according to an embodiment of this specification.

[0130] As Figure 9As shown, in some embodiments of the present specification, the audio input / output interface 11t / 81t in the structure detection device 100 / 800 can be disposed on the printed circuit board 900 and electrically connected to the edge computing artificial intelligence of things (AIoT) server 901 in the proximal acoustic sensing unit 11a of the acoustic sensing device 11 through the signal line WL; or through a wireless network and electrically connected to the edge computing artificial intelligence of things server 901 in the distal acoustic sensing units 81a - 81e of the acoustic sensing device 81. By means of the edge computing artificial intelligence of things server 901, the acoustic wave signals 14k / 84k to be measured collected by the proximal acoustic sensing unit 11a or the distal acoustic sensing units 81a - 81e can be preliminarily discriminated in data. In addition to saving the bandwidth for transmitting back to the monitoring and management cloud platform, it can also accelerate the operation speed of the acoustic vibration diagnosis module 12 data analysis and improve the application of the structure detection set 10 / 80 in the Internet of Things (IOT).

[0131] In some embodiments of the present specification, the structure detection device 100 / 800 further includes a soil parameter sensor 100g disposed on the printed circuit board 900 and electrically connected to the communication interface 100a / 800a for sensing the soil moisture and gas content around the structure to be measured 14 / 84. And the sensed data is transmitted to the diagnosis model 12m through electrical connection to the communication interface 100a / 800a. Furthermore, the structural state of the structure to be measured 14 / 84 can be judged according to the filtered and amplified signal 14A / 84A and the soil moisture and gas content. In this embodiment, the soil parameter sensor 100g may include an induction layer (not shown) with nano microchannels constructed using a liquid electric induction material 113 (such as graphene oxide).

[0132] Please refer to Figure 10 , in some embodiments of the present specification, the proximal acoustic sensing unit 11a of the acoustic sensing device 11 or the distal acoustic sensing units 81a - 81e of the structure detection set 80 include a housing 111, a high - sensitivity ceramic patch 112, a liquid electric induction material 113, and a sealing rubber ring 114. Among them, the aforementioned printed circuit board 900, the high - sensitivity ceramic patch 112 for sensing acoustic vibration, and the liquid electric induction material 113 for sensing soil moisture and gas content are disposed in the housing 111, and the sealing rubber ring 114 is disposed at one end of the housing 111. Therefore, the printed circuit board 900, the high - sensitivity ceramic patch 112, and the liquid electric induction material 113 are encapsulated in the housing 111, and the resonance cavity design of the housing 111 can strengthen the signal. In this embodiment, the housing 111 can be a full - cover protective cover. Therefore, the proximal acoustic sensing unit 11a or the distal acoustic sensing units 81a - 81e can be a multi - high - sensitivity intelligent detection probe with functions such as wind protection, anti - noise high - sensitivity, and instant data upload.

[0133] According to the above embodiments, the present specification provides a device and its application set for detecting the state of a structure to be measured by using acoustic vibration diagnosis technology. At least one communication interface, a filter amplification circuit, and a logic circuit are integrated into a structure detection device. By means of contact or non-contact, at least one measured acoustic wave signal output by at least one area to be measured is received from a remote acoustic wave sensing unit or a proximal acoustic wave sensing unit. The measured acoustic wave signal is filtered by the filter amplification circuit and a filtered amplified signal is output. Then, a diagnostic model constructed by a convolutional neural network in the logic circuit instantaneously determines the structural state of the structure to be measured according to the filtered amplified signal output by the filter amplification circuit. Remote sensing diagnosis and monitoring of the state of the structure to be measured (for example, thinning and leakage of a pipeline) are realized to ensure the safe operation of the structure to be measured (pipeline).

[0134] Although the present invention has been disclosed above with preferred embodiments, it is not intended to limit the present invention. Those skilled in the art can make some modifications and refinements without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the appended claims.

Claims

1. A structure detection device, for receiving at least one acoustic wave signal to be detected of at least one structure to be detected, and determining the structural state of the at least one structure to be detected, the structure detection device comprising: A communication interface, used for receiving the at least one sound wave signal to be measured; A filter amplifier circuit, used for filtering the at least one sound wave signal to be measured and outputting at least one filtered amplified signal; as well as The logic circuit includes a diagnostic model constructed by a convolutional neural network (CNN) to determine the structural state of the at least one structure to be tested according to the at least one filtered and amplified signal.

2. The structure detection device as claimed in claim 1, wherein the filter amplifier circuit is used to perform a square wave fast Fourier transform (FFT) on the at least one sound wave signal to be detected and then perform a Mel-frequency cepstrum analysis.

3. The structure detection device as claimed in claim 2, wherein the filter amplifier circuit comprises: An audio transcoder, configured to perform a time-frequency domain transformation on the at least one sound wave signal to be measured, transform the time domain waveform of the at least one sound wave signal to be measured into a frequency domain waveform, and extract a portion of the frequency domain waveform as at least one filtered signal; and The filter amplifier is used to extract and amplify the at least one filtered signal, and output at least one frequency domain band as the at least one filtered amplified signal. 4 . The structure detection device as claimed in claim 3 , wherein the filter amplifier circuit is a single-chip audio control circuit. 5 . The structure detection device as claimed in claim 3 , wherein the filter amplifier is a digital filter amplifier.

6. The structural detection device as claimed in claim 3, wherein the logic circuit comprises a microcontroller unit (MCU), and the structural state of the structure to be detected is determined by the diagnostic model according to the at least one filtered and amplified signal.

7. The structural detection device as described in claim 6 further includes a memory unit for storing a plurality of structural characteristic values; wherein the step of determining the structural state of the structure to be detected includes determining the type of the structure to be detected based on the plurality of structural characteristic values.

8. The structural detection device as described in claim 3 further includes an audio input / output interface, wherein the communication interface, the filter amplifier circuit and the logic circuit are constructed on a printed circuit board (PCB), and the audio input / output interface is arranged on the printed circuit board for electrical and / or telecommunication connection to an edge computing intelligent networking server located in the acoustic wave sensing device.

9. The structural detection device as described in claim 8 further includes a soil parameter sensor, which is arranged on the printed circuit board and is electrically connected to the communication interface, and is used to sense the soil moisture and gas content around the structure to be tested; wherein the diagnostic model determines the structural state of the at least one structure to be tested based on the at least one filtered amplified signal and the soil moisture and gas content.

10. The structure detection device as claimed in claim 9, wherein the soil parameter sensor comprises a liquid-electric sensing material.

11. A structural detection kit, used for extracting at least one acoustic wave signal to be detected from at least one structure to be detected, and determining the structural state of the at least one structure to be detected, the structural detection kit comprising: An acoustic wave sensing device, used for extracting and outputting the at least one acoustic wave signal to be detected; as well as Structural testing device, comprising: A communication interface is electrically connected to the acoustic wave sensing device and receives the at least one acoustic wave signal to be detected; A filter amplifier circuit, used for filtering the at least one sound wave signal to be measured and outputting at least one filtered amplified signal; as well as The logic circuit includes a diagnostic model constructed by a convolutional neural network, and determines the structural state of the structure to be tested according to the at least one filtered and amplified signal.

12. The structural inspection kit as claimed in claim 11, wherein the acoustic wave sensing device comprises: The far-end acoustic wave sensing unit and / or the near-end acoustic wave sensing unit are used to extract the at least one acoustic wave signal to be detected; and The audio input / output interface is used for inputting and outputting the at least one sound wave signal to be measured, and transmitting the sound wave signal to the filter amplifier circuit through the communication interface. 13 . The structural inspection kit as claimed in claim 11 , wherein the filter amplifier circuit is used for performing square wave fast Fourier transform on the at least one acoustic wave signal to be detected and then performing Mel-frequency cepstrum analysis.

14. The structural detection kit as claimed in claim 11, wherein the filter amplifier circuit comprises: An audio transcoder, used for performing a time-frequency domain transformation on the at least one sound wave signal to be measured, transforming the time-domain vibration waveform of the at least one sound wave signal to be measured into a frequency-domain waveform, and extracting a part of the frequency-domain waveform as at least one filtering signal; and The filter amplifier is used to extract and amplify the at least one filtered signal, and output at least one frequency domain band as the at least one filtered amplified signal. 15 . The structure detection kit as claimed in claim 14 , wherein the filter amplifier circuit is a single-chip audio control circuit. 16 . The structural inspection kit as claimed in claim 14 , wherein the logic circuit comprises a microcontroller, and determines the structural state of the structure to be inspected according to the at least one filtered and amplified signal by the diagnostic model.

17. The structure detection kit as claimed in claim 14, wherein the filter amplifier is a digital filter amplifier.

18. The structural detection kit as claimed in claim 11, further comprising a memory unit for storing a plurality of structural characteristic values; wherein the step of determining the structural state of the structure to be detected comprises determining the type of the structure to be detected according to the plurality of structural characteristic values.

19. The structural inspection kit as claimed in claim 12, further comprising a communication module, wherein the remote acoustic wave sensing unit comprises a remote leak detection probe, and the communication module is electrically connected to the remote leak detection probe and the communication interface.

20. The structural inspection kit as claimed in claim 12, wherein the proximal acoustic wave sensing device further comprises a handheld leak detection probe, and the communication interface comprises a signal line connecting the handheld leak detection probe and the filter amplifier circuit.

21. The structural inspection kit as described in claim 12, wherein the communication interface, the filter amplifier circuit, the logic circuit and the audio input / output interface are constructed on a printed circuit board, and the audio input / output interface is electrically and / or telecommunicationally connected to an edge computing intelligent networking server located at the remote acoustic wave sensing unit and / or the proximal acoustic wave sensing unit.

22. The structural inspection kit as claimed in claim 12, wherein the distal acoustic wave sensing unit and / or the proximal acoustic wave sensing unit comprises a high-sensitivity ceramic patch.

23. The structural detection kit as described in claim 11 further includes a soil parameter sensor, which is electrically connected to the communication interface and is used to sense soil moisture and gas content around the structure to be tested; wherein the diagnostic model determines the structural state of the at least one structure to be tested based on the at least one filtered amplified signal and the soil moisture and gas content.

24. The structural inspection kit of claim 23, wherein the soil parameter sensor comprises a hydro-electric sensing material.

Images