Piezoelectricity-based tunnel lining active and passive monitoring method and system

Abstract

The invention provides a tunnel lining active and passive monitoring method and system based on piezoelectricity, and relates to the technical field of tunnel lining monitoring, the method comprises the steps that a plurality of monitoring faces are arranged in a tunnel lining, each monitoring face is provided with a plurality of piezoelectric sensors, and the monitoring faces are used for dividing effective sections of the tunnel lining; performing passive monitoring based on the piezoelectric sensor, and combining temperature correction and time-frequency diagram conversion to obtain an abnormal section; the abnormal section is actively monitored, piezoelectric driving and model construction are combined, the damage result of the tunnel lining is obtained, and the damage result comprises the damage position, the damage type and the damage degree. According to the invention, the problem that the existing periodic test is insensitive to the acquisition of the damage occurring at the period interval is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tunnel lining monitoring, in particular to a piezoelectric-based tunnel lining active and passive monitoring method and system. BACKGROUND

[0002] During the operation period of the tunnel, the tunnel lining structure is long-term affected by various external environments and operating loads, such as geological changes, vehicle vibration, hydrological conditions, and ground subsidence, which may cause progressive damage to the structure. In order to ensure the safe operation of the tunnel and prolong its service life, damage monitoring is crucial. Damage monitoring can obtain the health status of the tunnel structure in real time, and through intelligent sensor technology, data acquisition and analysis system, it can accurately identify the occurrence and development of problems such as cracks, deformation, and leakage. This not only helps to discover potential risks in a timely manner and develop scientific and reasonable maintenance measures, but also reduces maintenance costs and avoids major accidents caused by sudden structural failure.

[0003] At present, in the field of monitoring of tunnels in operation, intelligent monitoring systems based on sensor networks have gradually become mainstream, such as distributed optical fiber sensor monitoring of overall deformation of long-distance tunnels, or detection of local damage by intelligent piezoelectric sensors. The above methods are all achieved by periodic testing to achieve the purpose of monitoring, however, periodic testing is not sensitive to damage that occurs during the interval, and damage that occurs during this period needs to be discovered until the next test. In addition, for large structures such as tunnels, maintaining the active excitation state of the sensor not only consumes a large amount of power, but also has a great impact on the service life of the sensor. Therefore, there is an urgent need for a tunnel damage monitoring method that combines active and passive monitoring and can achieve immediate feedback of damage. SUMMARY

[0004] The purpose of the present application is to provide a piezoelectric-based tunnel lining active and passive monitoring method and system to improve the above problems. In order to achieve the above purpose, the technical solutions adopted by the present application are as follows:

[0005] In a first aspect, the present application provides a piezoelectric-based tunnel lining active and passive monitoring method, comprising:

[0006] A plurality of monitoring surfaces are arranged in the tunnel lining, a plurality of piezoelectric sensors are arranged on each monitoring surface, and the monitoring surface is used to divide the effective section of the tunnel lining;

[0007] Passive monitoring is performed based on the piezoelectric sensors, combined with temperature correction and time-frequency diagram conversion to obtain an abnormal section;

[0008] The active monitoring unit is used for actively monitoring the abnormal section, and damage results of the tunnel lining are obtained by combining piezoelectric driving and model construction, the damage results including damage positions, damage types and damage degrees.

[0009] In a second aspect, the application further provides a piezoelectric-based tunnel lining active and passive monitoring system, comprising:

[0010] The setting unit is used for setting a plurality of monitoring surfaces in the tunnel lining, a plurality of piezoelectric sensors are arranged on each monitoring surface, and the monitoring surfaces are used for dividing effective sections of the tunnel lining.

[0011] The passive monitoring unit is used for performing passive monitoring based on the piezoelectric sensors, and an abnormal section is obtained by combining temperature correction and time-frequency diagram conversion.

[0012] The active monitoring unit is used for actively monitoring the abnormal section, and damage results of the tunnel lining are obtained by combining piezoelectric driving and model construction, the damage results including damage positions, damage types and damage degrees.

[0013] The application has the following beneficial effects:

[0014] (1) The application realizes efficient and accurate identification of tunnel lining damage by combining active and passive monitoring. The passive monitoring uses a loaded vehicle to excite stress waves, combines temperature correction and time-frequency diagram conversion, quickly locates an abnormal section, does not need additional active excitation, and reduces the energy consumption and complexity of long-term monitoring. The active monitoring focuses on detection of the abnormal section, combines piezoelectric driving, image recognition and a multi-feature fusion model, accurately outputs damage positions, types and degrees, forms an efficient process of wide-area screening and fine diagnosis, and greatly improves the pertinence and accuracy of monitoring.

[0015] (2) The reliability and adaptability of monitoring are ensured by multi-dimensional signal processing. The application eliminates the influence of environmental interference and material characteristics on signals by temperature correction and noise reduction processing, solves the time difference and waveform distortion problems of array piezoelectric sensors by beam forming and phase correction, and realizes effective extraction of high and low frequency features by wavelet decomposition and discrete feature extraction. These technologies work together to ensure the stability of signals under complex working conditions and improve the intelligent level of damage identification through a deep learning model, providing reliable technical support for the whole life cycle health management of tunnel lining.

[0016] Other features and advantages of the application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application as described in the written description and claims. The purposes and other advantages of the application can be realized and attained by the structure particularly pointed out in the written description, claims, and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings needed to be used in the embodiments will be briefly introduced as follows. It should be understood that the following drawings only show some of the embodiments of the present application, and therefore should not be considered as limiting the scope. For those skilled in the art, other related drawings can also be obtained without creative labor on the basis of these drawings.

[0018] Figure 1 The flow chart of the piezoelectric-based tunnel lining active and passive monitoring method described in the embodiments of the present application;

[0019] Figure 2 The structural schematic diagram of the piezoelectric sensor in the embodiments of the present application;

[0020] Figure 3 The setting schematic diagram of the monitoring surface in the embodiments of the present application;

[0021] Figure 4 The setting schematic diagram of the first scheme in the embodiments of the present application;

[0022] Figure 5 The setting schematic diagram of the second scheme in the embodiments of the present application.

[0023] Markings in the drawings: 1, temperature sensor; 2, BNC connector; 3, PVC circuit board; 4, piezoelectric sheet. DETAILED DESCRIPTION

[0024] In order to make the purpose, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application, but only represents selected embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0025] It should be noted that: similar labels and letters represent similar items in the following drawings, and therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms "first", "second", etc. are only used to distinguish the description, and cannot be understood as indicating or implying relative importance.

[0026] Embodiment 1:

[0027] This embodiment provides a piezoelectric-based active and passive monitoring method for tunnel lining.

[0028] See Figure 1 The figure shows that the method includes steps S1, S2 and S3.

[0029] Step S1: Set multiple monitoring surfaces in the tunnel lining, and set multiple piezoelectric sensors on each monitoring surface. The monitoring surfaces are used to divide the effective sections of the tunnel lining.

[0030] In this embodiment, the piezoelectric sensor is as follows: Figure 2 As shown, it includes: a temperature sensor 1, a BNC connector 2, a PVC circuit board 3, and a piezoelectric element 4, all connected by soldering. The monitoring surface is set as follows. Figure 3 As shown, a set of piezoelectric sensors is installed at 3-meter intervals along the monitoring surface.

[0031] Step S2: Based on the piezoelectric sensor, perform passive monitoring, and combine temperature correction and time-frequency graph conversion to obtain abnormal sections;

[0032] Step S3: Actively monitor the abnormal section, and combine piezoelectric drive with model construction to obtain the damage results of the tunnel lining. The damage results include the damage location, damage type and damage degree.

[0033] In step S2, obtaining the abnormal segment includes:

[0034] Step S21: Set a loading scheme, wherein a loading vehicle of the same model and load capacity passes through the tunnel at preset time intervals;

[0035] In this embodiment, a heavy-duty electric vehicle of the same model and load capacity is used as a loading vehicle to pass through the tunnel every 8 hours.

[0036] Step S22: Based on the loading scheme, obtain the first stress wave signal of the piezoelectric sensor in the effective section, wherein the effective section is the section within a preset length from the front and rear of the vehicle when the loaded vehicle passes by;

[0037] In this embodiment, when a vehicle passes by, the sections within 6 meters of the front and rear of the vehicle are selected as the effective sections for passive monitoring.

[0038] Step S23: Perform temperature correction and noise reduction processing on the first stress wave signal to obtain the second stress wave signal;

[0039] Step S24: Convert the second stress wave signal into a three-channel passive time-frequency diagram using continuous wavelet transform;

[0040] In this embodiment, the stress wave signal after noise reduction processing is converted into a time-frequency graph by using continuous wavelet transform (CWT), and saved as a three-channel image, forming a two-dimensional CWT image (passive time-frequency graph).

[0041] Step S25: performing abnormality identification on the passive time-frequency graph through a damage passive identification model to obtain an abnormal section, wherein the damage passive identification model adopts a CNN network.

[0042] In this embodiment, after the overall maintenance of the tunnel is completed, the passive monitoring is performed multiple times through the loading vehicle, and the passive nondestructive data set is established to train the CNN network, and the damage passive identification model is obtained. When performing the passive monitoring, the passive time-frequency graph of different effective sections is input into the damage passive identification model established based on the CNN network, so that whether the corresponding section has an abnormal state can be identified through the passive monitoring method, and once the abnormal state exists, the active monitoring of the corresponding effective section is started to perform confirmation and positioning.

[0043] The step S3 comprises:

[0044] Step S31: performing image acquisition on the abnormal section to obtain a tunnel surface video;

[0045] Step S32: inputting the tunnel surface video into a preset YOLO model to perform damage identification, and obtaining a surface damage type and a surface damage position of the tunnel lining;

[0046] In this embodiment, a trolley carrying multiple starlight cameras slowly passes through the corresponding abnormal section and records the tunnel surface, and the recorded video is input into the trained YOLO model to perform surface damage identification and positioning. The training of the YOLO model based on image recognition is performed by recording a large amount of video after the tunnel maintenance is completed and establishing a data set.

[0047] Step S33: performing signal acquisition and feature extraction based on the monitoring surface, combining temperature correction and noise reduction processing, and obtaining an active time-frequency graph, a low-frequency feature and a high-frequency feature;

[0048] Step S34: performing damage identification on the active time-frequency graph, the low-frequency feature and the high-frequency feature through a damage active identification model to obtain an internal damage position and a corresponding internal damage degree of the tunnel lining;

[0049] Step S35: taking the surface damage type, the surface damage position, the internal damage position and the internal damage degree as a damage result of the tunnel lining.

[0050] In step S33, the obtaining of the active time-frequency graph, the low-frequency feature and the high-frequency feature comprises:

[0051] Step S331: The monitoring surface of the abnormal section is set according to a first scheme, and the first scheme is to set the monitoring surface adjacent to the driving surface and the receiving surface, and the piezoelectric sensors of the driving surface are used to emit stress waves, and the piezoelectric sensors of the receiving surface are used to receive stress waves.

[0052] In this embodiment, the active monitoring of internal damage requires at least one piezoelectric sensor as a driver to emit stress waves, and multiple sensors can be used to receive stress waves, and each stress wave can monitor the health status of the corresponding transceiver section.

[0053] The first scheme is to set the monitoring surface adjacent to the driving surface and the receiving surface, as shown in Figure 4 The purpose of the driving surface is to further confirm the damage location, and the range is reduced to 2 sections containing 3 monitoring surfaces, and each receiving surface can monitor the damage of the adjacent front and rear two sections.

[0054] Step S332: Signal emission and collection are performed through the first scheme, and temperature correction and noise reduction processing are performed on the received signals to obtain a third stress wave signal;

[0055] Step S333: The third stress wave signal is converted into a first time-frequency graph through continuous wavelet transform, and the first time-frequency graph is input into a preset LSTM-CNN model for identification to obtain a preliminary damage section, and the preliminary damage section includes two effective sections and three monitoring surfaces;

[0056] In this embodiment, the wave signal collected by each receiver is converted into a time-frequency graph through CWT and input into the trained LSTM-CNN model with time characteristics for identification, so as to determine whether there is damage in the 2 effective sections before and after each receiving surface.

[0057] Step S334: Signal emission and collection are performed based on the preliminary damage section, and active time-frequency graphs, low-frequency features and high-frequency features are obtained through discrete decomposition.

[0058] In step S334, the active time-frequency graphs, low-frequency features and high-frequency features obtained through discrete decomposition include:

[0059] Step A1: The monitoring surface of the preliminary damage section is set according to a second scheme, and the second scheme is that each monitoring surface is used as a driving surface, and the adjacent piezoelectric sensors on the adjacent surface are sequentially and alternately used as drivers to emit stress waves, and other piezoelectric sensors are used to receive stress waves;

[0060] In this embodiment, the second scheme is shown in Figure 5 Each monitoring surface in the two effective sections is a driving surface, and the drivers on the adjacent surfaces are different, and the adjacent piezoelectric sensors on the adjacent surfaces are sequentially used as drivers.

[0061] This is because the tunnel belongs to a large structure, and there may be noise that cannot be eliminated by noise reduction processing due to factors such as geological activity. If the first scheme is not preliminarily determined and directly positioned and evaluated by the second scheme, the influence of this part cannot be eliminated, which may cause damage position deviation, inaccurate damage condition, and false damage that does not exist.

[0062] Step A2: Based on the second scheme, signal emission and collection are performed, and the received signal is temperature corrected and noise reduced to obtain a fourth stress wave signal;

[0063] Step A3: The fourth stress wave signal is converted into an active time-frequency diagram through continuous wavelet transform;

[0064] Step A4: The fourth stress wave signal is filtered by a low-pass filter and a high-pass filter to obtain approximation coefficients and detail coefficients;

[0065] In this embodiment, since each piezoelectric sensor receives stress waves emitted by multiple drivers, a certain deviation may be caused by using a continuous wave signal processing method, and therefore, a discrete decomposition method is needed to evaluate the wave signal to correct the result of the identification model.

[0066] The collected fourth stress wave signal is convolved and down-sampled by a low-pass filter (scale filter) and a high-pass filter (wavelet filter), specifically:

[0067]

[0068] j max =log2 N

[0069] In the formula, A j [·] represents the approximation coefficient of the jth layer, h[·] represents the low-pass filter, D j [·] represents the detail coefficient of the jth layer, g[·] represents the high-pass filter, N represents the signal length, j max represents the maximum decomposition layer number, A j-1 [·] represents the approximation coefficient of the j-1th layer, k represents the signal index of the jth layer, and n represents the signal index of the j-1th layer, wherein A0[·] represents the fourth stress wave signal.

[0070] Step A5: Calculate a single-layer index based on the approximation coefficient and the detail coefficient, wherein the single-layer index includes a single-layer low-frequency overall feature and a single-layer high-frequency local feature;

[0071] In this embodiment, the calculation formula of the single-layer index is:

[0072]

[0073] wherein I j denotes the single-layer low-frequency global feature of the jth layer, E j denotes the single-layer high-frequency local feature of the jth layer, A j denotes the approximation coefficient of the jth layer, K denotes the total number of approximation coefficients of the jth layer, D j denotes the detail coefficient of the jth layer, k denotes the signal index of the jth layer, and |·| denotes taking an absolute value.

[0074] Step A6: calculating a global index based on the single-layer index, the global index including a low-frequency feature and a high-frequency feature.

[0075] In this embodiment, the low-frequency feature and the high-frequency feature of the global are calculated in combination with the single-layer index of each layer:

[0076]

[0077] wherein I denotes the low-frequency feature, E denotes the high-frequency feature, I j denotes the single-layer low-frequency global feature of the jth layer, I j0 denotes the reference low-frequency global feature, E j denotes the single-layer high-frequency local feature of the jth layer, E j0 denotes the reference high-frequency local feature, j max denotes the maximum number of decomposition layers, and |·| denotes taking an absolute value.

[0078] In this embodiment, the active time-frequency map carrying the continuity feature and the two discrete features (the low-frequency feature and the high-frequency feature) are introduced into the trained damage active identification model for identification.

[0079] In step S34, the construction step of the damage active identification model is:

[0080] Step B1: constructing a first feature extraction layer, the first feature extraction layer adopting a DenseNet-GRU network, and the input of the first feature extraction layer being the low-frequency feature;

[0081] Step B2: constructing a second feature extraction layer, the second feature extraction layer adopting a Transformer model, and the input of the second feature extraction layer being the high-frequency feature;

[0082] Step B3: constructing a third feature extraction layer, the third feature extraction layer including a CNN-SEBlock network, a global average pooling layer and an LSTM network connected in sequence, and the input of the third feature extraction layer being the active time-frequency map;

[0083] Step B4: a cross-attention layer, a full connection layer and a Softmax function are sequentially constructed, the cross-attention layer includes a cross-attention mechanism, the input of the cross-attention layer includes the output features of the first feature extraction layer, the second feature extraction layer and the third feature extraction layer, and the output of the Softmax function is the output of the damage active identification model.

[0084] In this embodiment, the low-frequency features are input into the DenseNet-GRU network for processing, the high-frequency features are input into the Transformer model for processing, and the active time-frequency graph is input into the CNN-SEBlock network for processing and then into the LSTM network for processing after global average pooling processing. All of them can retain the continuity of the time factor, and after processing, the cross-attention mechanism is used to establish the attention relationship among them, and then the full connection layer is used to identify and evaluate the damage through the softmax function.

[0085] The training steps of the damage active identification model include:

[0086] In the tunnel, a certain thickness of concrete layer is poured around the lining, and a certain damage is applied by knocking, bending and compression. The stress wave signals at different degrees of damage are monitored by the active monitoring method, and the corresponding time-frequency graph and feature index are generated. The piezoelectric sensor corresponding position is sampled and the traditional mechanical test method is used to measure its compressive strength. The time-frequency graph, feature index and actual traditional mechanical test strength obtained by monitoring are used to train the damage active identification model.

[0087] Therefore, the surface and internal identification and positioning method combining passive monitoring and active monitoring can realize comprehensive real-time monitoring of the tunnel lining.

[0088] It should be emphasized that in all steps of steps S1-S3, the acquired stress wave signals need to be temperature corrected and denoised. The temperature correction is a temperature adaptability improvement. Since the tunnel is a large structure, its temperature is distributed in a gradient along the tunnel axis, and local heat sources exist, causing local high temperature in the tunnel. The change of temperature will affect the positive and negative piezoelectric effect of the piezoelectric sensor, so that the same vibration signal shows different characteristics. Therefore, a temperature sensor is installed on the piezoelectric sensor to obtain the temperature parameter at the location to calculate the influence on the piezoelectric coefficient and the dielectric constant, so as to adjust the received wave signal. After receiving the stress wave signal, it needs to be converted to the standard state at 25°C before denoising.

[0089] The stress wave denoising preprocessing is because the tunnel is a large structure, and local vibration signals may be generated in some areas, so that the stress wave signals collected by the active or passive collection are changed by the noise, reducing the accuracy of monitoring, and therefore the wave signals collected by the active and passive monitoring methods need to be denoised first. The collected stress wave signals are first denoised by using the shrinkage model, and then the signals are corrected and reconstructed based on the array distribution.

[0090] Specifically, the temperature correction and denoising processing includes:

[0091] Step C1: obtaining first information, the first information including a temperature change, a temperature coefficient of piezoelectric constant, a temperature coefficient of dielectric constant and a current temperature signal amplitude, the temperature change being a difference between a current temperature and a standard temperature, and the current temperature signal amplitude being a signal amplitude of the piezoelectric sensor at the current temperature;

[0092] In this embodiment, the standard temperature is taken as 25℃.

[0093] Step C2: temperature correction is performed on the current temperature signal amplitude based on the first information to obtain a first signal at the standard temperature;

[0094] In this embodiment, the formula for temperature correction is:

[0095]

[0096] In the formula, V out (T0) represents the signal amplitude converted into the standard temperature T0, V out (T) represents the signal amplitude collected at the current temperature T, a d represents the temperature coefficient of piezoelectric constant, a ∈ represents the temperature coefficient of dielectric constant, and ΔT represents the temperature change.

[0097] Step C3: eliminating the influence of the concrete material on the first signal based on the stress wave attenuation coefficient of the concrete material to obtain a second signal;

[0098] In this embodiment, the formula for eliminating the influence of the concrete material on the first signal is:

[0099]

[0100] In the formula, A(d) represents the second signal about d, A0 represents the amplitude before attenuation, d represents the minimum distance between the driver and the receiver, and n' represents the stress wave attenuation coefficient of the concrete material.

[0101] Step C4: Fourier transform the second signal to obtain a frequency domain signal, and phase correct each frequency component in the frequency domain signal to obtain a corrected frequency domain signal;

[0102] In this embodiment, the waveform broadening and distortion phenomenon caused by the variation of the phase velocity and group velocity of the stress wave in the concrete with frequency also needs to be corrected. Therefore, Fourier transform is performed on the time domain signal, and phase correction is performed.

[0103] The formula for phase correction is:

[0104]

[0105] v p (ω) = ω / k(ω)

[0106]

[0107] In the formula, X(ω) represents the frequency domain signal with a frequency of ω, X c (ω) represents the phase-corrected frequency domain signal with a frequency of ω, exp[·] represents the exponential function, j' represents the imaginary number, d represents the minimum distance between the driver and the receiver, v p (ω) represents the phase velocity of the stress wave, v g (ω) represents the group velocity of the stress wave, v g,ref represents the reference group velocity, and k(ω) represents the wave number with respect to the frequency ω.

[0108] Step C5: inverse Fourier transform the corrected frequency domain signal to obtain a third signal;

[0109] Step C6: perform beamforming on the third signal to correct the time difference of the piezoelectric sensors at different positions, and obtain a signal after temperature correction and noise reduction processing.

[0110] In this embodiment, for the array distribution, beamforming is used to improve the signal-to-noise ratio and correct the time difference of different piezoelectric sensors. The formula for beamforming is:

[0111]

[0112] In the formula, y(t) represents the noise reduction signal with respect to time t, M represents the number of piezoelectric sensors, w i represents the array weight of the i-th piezoelectric sensor, w i = d / d i , d represents the minimum distance between the driver and the receiver, d i represents the distance from the i-th piezoelectric sensor to the driver, x i (·) represents the received stress wave emitted by the i-th piezoelectric sensor, and c represents the propagation speed of the stress wave in the concrete.

[0113] Embodiment 2:

[0114] The embodiment provides a piezoelectric-based tunnel lining active and passive monitoring system, and the system comprises:

[0115] A setting unit is configured to set a plurality of monitoring surfaces in a tunnel lining, a plurality of piezoelectric sensors are arranged on each monitoring surface, and the monitoring surfaces are used to divide effective sections of the tunnel lining.

[0116] A passive monitoring unit is configured to perform passive monitoring based on the piezoelectric sensors, combine temperature correction and time-frequency map conversion, and obtain an abnormal section.

[0117] An active monitoring unit is configured to perform active monitoring on the abnormal section, combine piezoelectric driving and model construction, and obtain a damage result of the tunnel lining, wherein the damage result comprises a damage position, a damage type and a damage degree.

[0118] The passive monitoring unit comprises:

[0119] A first setting subunit is configured to set a loading scheme, wherein the loading scheme is to pass a same model and load of a loading vehicle in the tunnel every preset time interval.

[0120] A first acquisition subunit is configured to obtain a first stress wave signal of the piezoelectric sensors in the effective section based on the loading scheme, wherein the effective section is a section within a preset length from a front end and a tail end of the loading vehicle when the loading vehicle passes.

[0121] A first processing subunit is configured to perform temperature correction and noise reduction processing on the first stress wave signal, and obtain a second stress wave signal.

[0122] A first conversion subunit is configured to convert the second stress wave signal into a three-channel passive time-frequency map through continuous wavelet transform.

[0123] A first identification subunit is configured to perform abnormal identification on the passive time-frequency map through a damage passive identification model, and obtain an abnormal section, wherein the damage passive identification model adopts a CNN network.

[0124] The active monitoring unit comprises:

[0125] A second acquisition subunit is configured to perform image acquisition on the abnormal section, and obtain a tunnel surface video.

[0126] A second identification subunit is configured to input the tunnel surface video into a preset YOLO model to perform damage identification, and obtain a surface damage type and a surface damage position of the tunnel lining.

[0127] The third acquisition subunit is configured to acquire and extract signals based on the monitoring surface, and obtain an active time-frequency diagram, low-frequency features and high-frequency features by combining temperature correction and noise reduction processing.

[0128] The fourth identification subunit is configured to identify damages by using the active damage identification model to process the active time-frequency diagram, the low-frequency features and the high-frequency features, so as to obtain the internal damage position and the corresponding internal damage degree of the tunnel lining.

[0129] The second processing subunit is configured to take the surface damage type, the surface damage position, the internal damage position and the internal damage degree as the damage result of the tunnel lining.

[0130] The third acquisition subunit comprises:

[0131] The second setting subunit is configured to set the monitoring surface of the abnormal section according to a first scheme, and the first scheme is to set the monitoring surface adjacent to the driving surface and the receiving surface, and the piezoelectric sensor of the driving surface is configured to emit stress waves, and the piezoelectric sensor of the receiving surface is configured to receive stress waves.

[0132] The fourth acquisition subunit is configured to emit and collect signals by using the first scheme, and perform temperature correction and noise reduction processing on the collected signals, so as to obtain a third stress wave signal.

[0133] The second conversion subunit is configured to convert the third stress wave signal into a first time-frequency diagram by using continuous wavelet transform, and input the first time-frequency diagram into a preset LSTM-CNN model to identify, so as to obtain a preliminary damage section, and the preliminary damage section comprises two effective sections and three monitoring surfaces.

[0134] The fifth acquisition subunit is configured to emit and collect signals based on the preliminary damage section, and obtain an active time-frequency diagram, low-frequency features and high-frequency features by using discrete decomposition.

[0135] It should be noted that, as for the system in the above-mentioned embodiments, the specific manner in which each module performs operations has been described in detail in the embodiments related to the method, and will not be described in detail here.

[0136] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

[0137] The above merely illustrates the specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of the changes or replacements within the technical range disclosed by the present application, which should be covered in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A piezoelectric-based method for active and passive monitoring of tunnel lining, characterized by, The application relates to a tunnel lining damage detection method and device. The application comprises the following steps: A plurality of monitoring surfaces are arranged in a tunnel lining, and a plurality of piezoelectric sensors are arranged on each monitoring surface, the monitoring surfaces are used to divide effective sections of the tunnel lining; Passive monitoring is carried out based on the piezoelectric sensors, temperature correction and time-frequency diagram conversion are combined, and an abnormal section is obtained; 2. The piezoelectric based tunnel lining active-passive monitoring method according to claim 1, characterized in that Active monitoring is carried out on the abnormal section, piezoelectric driving and model construction are combined, and a damage result of the tunnel lining is obtained, wherein the damage result comprises a damage position, a damage type and a damage degree. The abnormal section is obtained by the following steps: A loading scheme is arranged, the loading scheme is that a same type and load of a loading vehicle is arranged in the tunnel every preset time interval; First stress wave signals of the piezoelectric sensors in the effective section are obtained based on the loading scheme, the effective section is a section within a preset length from a vehicle head and a vehicle tail when the loading vehicle passes through; Temperature correction and noise reduction processing are carried out on the first stress wave signals, and second stress wave signals are obtained; The second stress wave signals are converted into a three-channel passive time-frequency diagram through continuous wavelet transform; 3. The piezoelectric based tunnel lining active-passive monitoring method according to claim 1, characterized in that The passive time-frequency diagram is subjected to abnormal identification through a damage passive identification model, and an abnormal section is obtained, wherein the damage passive identification model adopts a CNN network. The active monitoring is carried out on the abnormal section, piezoelectric driving and model construction are combined, and the damage result of the tunnel lining is obtained, and the method comprises the following steps: A tunnel surface video is obtained through image acquisition on the abnormal section; The tunnel surface video is input into a preset YOLO model for damage identification, and a surface damage type and a surface damage position of the tunnel lining are obtained; Active time-frequency diagrams, low-frequency features and high-frequency features are obtained through signal acquisition and feature extraction based on the monitoring surfaces, combined with temperature correction and noise reduction processing; The surface damage type, the surface damage position, the internal damage position and the internal damage degree are taken as the damage result of the tunnel lining.

4. The piezoelectric based tunnel lining active-passive monitoring method according to claim 3, characterized in that The active time-frequency diagrams, the low-frequency features and the high-frequency features are obtained by the following steps: The monitoring surfaces of the abnormal section are arranged according to a first scheme, the first scheme is that the monitoring surfaces are arranged adjacent to driving surfaces and receiving surfaces, piezoelectric sensors of the driving surfaces are used to emit stress waves, and piezoelectric sensors of the receiving surfaces are used to receive stress waves; Signal emission and acquisition are carried out through the first scheme, and temperature correction and noise reduction processing are carried out on the received signals, and third stress wave signals are obtained; The third stress wave signals are converted into first time-frequency diagrams through continuous wavelet transform, and the first time-frequency diagrams are input into a preset LSTM-CNN model for identification, and a preliminary damage section is obtained, wherein the preliminary damage section comprises two effective sections and three monitoring surfaces; Active time-frequency diagrams, low-frequency features and high-frequency features are obtained through signal emission and acquisition based on the preliminary damage section, combined with discrete decomposition.

5. The piezoelectric based tunnel lining active-passive monitoring method according to claim 4, characterized in that The active time-frequency diagrams, the low-frequency features and the high-frequency features are obtained through discrete decomposition. The monitoring surface of the preliminary damage section is set according to a second scheme, the second scheme is that each monitoring surface is used as a driving surface, and adjacent piezoelectric sensors on the adjacent surface are sequentially and alternately used as a driver to emit a stress wave, and other piezoelectric sensors receive the stress wave; Signal emission and collection are performed based on the second scheme, and the received signals are temperature corrected and denoised to obtain a fourth stress wave signal; The fourth stress wave signal is converted into an active time-frequency diagram through continuous wavelet transform; The fourth stress wave signal is filtered through a low-pass filter and a high-pass filter to obtain approximation coefficients and detail coefficients; Single-layer indexes are calculated based on the approximation coefficients and the detail coefficients, the single-layer indexes include single-layer low-frequency overall characteristics and single-layer high-frequency local characteristics; Overall indexes are calculated based on the single-layer indexes, the overall indexes include low-frequency characteristics and high-frequency characteristics.

6. The piezoelectric based tunnel lining active-passive monitoring method according to claim 3, characterized in that The construction steps of the damage active identification model are: A first feature extraction layer is constructed, the first feature extraction layer adopts a DenseNet-GRU network, and the input of the first feature extraction layer is the low-frequency characteristics; A second feature extraction layer is constructed, the second feature extraction layer adopts a Transformer model, and the input of the second feature extraction layer is the high-frequency characteristics; A third feature extraction layer is constructed, the third feature extraction layer includes a CNN-SEBlock network, a global average pooling layer and an LSTM network connected in sequence, and the input of the third feature extraction layer is the active time-frequency diagram; A cross-attention layer, a fully connected layer and a Softmax function connected in sequence are constructed, the cross-attention layer includes a cross-attention mechanism, the input of the cross-attention layer includes the output features of the first feature extraction layer, the second feature extraction layer and the third feature extraction layer, and the output of the Softmax function is the output of the damage active identification model.

7. A piezoelectric based active and passive monitoring system for tunnel lining characterized in that, It comprises: A setting unit is configured to set a plurality of monitoring surfaces in a tunnel lining, each of the monitoring surfaces is provided with a plurality of piezoelectric sensors, and the monitoring surfaces are used to divide effective sections of the tunnel lining; A passive monitoring unit is configured to perform passive monitoring based on the piezoelectric sensors, and to obtain an abnormal section by combining temperature correction and time-frequency diagram conversion; An active monitoring unit is configured to perform active monitoring on the abnormal section, and to obtain a damage result of the tunnel lining by combining piezoelectric driving and model construction, the damage result includes a damage position, a damage type and a damage degree.

8. The piezoelectric based tunnel lining active-passive monitoring system according to claim 7, wherein, The passive monitoring unit comprises: A first setting subunit is configured to set a loading scheme, the loading scheme is to drive a same model and load vehicle in the tunnel at a preset time interval; A first acquisition subunit is configured to obtain a first stress wave signal of the piezoelectric sensors in the effective section based on the loading scheme, the effective section is a section within a preset length from the front and rear of the vehicle when the vehicle passes through; A first processing subunit is configured to perform temperature correction and denoising processing on the first stress wave signal to obtain a second stress wave signal; A first conversion subunit is configured to convert the second stress wave signal into a three-channel passive time-frequency diagram through continuous wavelet transform; The first identification subunit is configured to perform abnormal identification on the passive time-frequency diagram by using a passive damage identification model to obtain an abnormal section, and the passive damage identification model adopts a CNN network.

9. The piezoelectric based tunnel lining active-passive monitoring system of claim 7, wherein, The active monitoring unit comprises: The second acquisition subunit is configured to perform image acquisition on the abnormal section to obtain a tunnel surface video. The second identification subunit is configured to input the tunnel surface video into a preset YOLO model to perform damage identification, and obtain a surface damage type and a surface damage position of the tunnel lining. The third acquisition subunit is configured to perform signal acquisition and feature extraction based on the monitoring surface, and obtain an active time-frequency diagram, a low-frequency feature and a high-frequency feature by combining temperature correction and noise reduction processing. The fourth identification subunit is configured to perform damage identification on the active time-frequency diagram, the low-frequency feature and the high-frequency feature by using a damage active identification model to obtain an internal damage position and a corresponding internal damage degree of the tunnel lining. The second processing subunit is configured to take the surface damage type, the surface damage position, the internal damage position and the internal damage degree as damage results of the tunnel lining.

10. The piezoelectric based tunnel lining active-passive monitoring system of claim 9, wherein, The third acquisition subunit comprises: The second setting subunit is configured to set the monitoring surface of the abnormal section according to a first scheme, and the first scheme is to set the monitoring surface adjacent to a driving surface and a receiving surface, and the piezoelectric sensor of the driving surface is used to emit stress waves, and the piezoelectric sensor of the receiving surface is used to receive stress waves. The fourth acquisition subunit is configured to perform signal emission and acquisition by using the first scheme, and perform temperature correction and noise reduction processing on the received signal to obtain a third stress wave signal. The second conversion subunit is configured to convert the third stress wave signal into a first time-frequency diagram by using continuous wavelet transform, and input the first time-frequency diagram into a preset LSTM-CNN model to perform identification, and obtain a preliminary damage section, and the preliminary damage section comprises two effective sections and three monitoring surfaces. The fifth acquisition subunit is configured to perform signal emission and acquisition based on the preliminary damage section, and obtain the active time-frequency diagram, the low-frequency feature and the high-frequency feature by combining discrete decomposition.

Images