A single-channel panel damage localization method based on Lamb wave coded metamaterial interface

Abstract

The present invention discloses a single-channel wall panel damage location method based on Lamb wave coded metamaterial interface, comprising: constructing a priori database for wall panel damage detection; the priori database stores the layout form of the damage monitoring area on the wall panel, the layout parameters of the coding unit, the Lamb wave parameters of the excitation signal applied during damage detection, the echo signal of the wall panel in a healthy state, and a neural network model for damage azimuth identification; for the wall panel to be damaged, the coding unit and the exciter are arranged according to the information in the priori database, and the echo signal is collected after the excitation signal is applied; the polar angle area corresponding to the damage is obtained based on the neural network model to realize the azimuth identification of the damage; the flight time of the Lamb wave is obtained based on signal processing, so as to determine the distance from the damaged part to the center of the damage monitoring area, and realize the determination of the damage distance; finally, the wall panel damage location under single-channel data acquisition is realized.

Description

Technical Field

[0001] The present invention relates to the field of wall panel damage identification and positioning, and in particular to a single-channel wall panel damage positioning method based on Lamb wave coding metamaterial interface. Background Art

[0002] Structural health detection is one of the key technologies to ensure the safe service of engineering structures. Among many detection methods, Lamb waves have been widely used in damage detection and positioning of siding structures due to their advantages such as long propagation distance and sensitivity to damage. Lamb waves are ultrasonic guided waves that propagate in plate-like structures. Their unique dispersion characteristics and multimodal properties enable them to carry rich structural status information. By analyzing the differences in the propagation characteristics of Lamb waves in healthy sidings and damaged sidings, early warning and precise positioning of siding damage can be achieved.

[0003] Traditional Lamb wave panel damage location methods, such as time-of-flight method, scattered wave imaging method and array imaging method, usually require the deployment of multiple sensors to determine the location of the damage on the panel by analyzing the differences in multi-channel acquisition signals. Although this sensor array-based positioning strategy can achieve high positioning accuracy, it also has some shortcomings. First, the deployment of multiple sensors on large-scale panel structures will significantly increase the complexity and cost of the detection system, and the number and layout optimization of sensors become a major challenge. Reducing the number of sensors often comes at the expense of positioning accuracy. Secondly, the acquisition and processing of multi-channel signals puts higher requirements on data transmission and computing power, increasing the difficulty of data analysis. Efficient processing and storage of massive detection data is also a thorny issue. In addition, due to the multi-modal conversion and dispersion effects that occur during the propagation of Lamb waves in the panel, the comparative analysis and damage feature extraction between signals from different channels also become more complicated. Modal conversion and dispersion will distort the waveform, resulting in deviations in the positioning results. Therefore, the traditional multi-channel Lamb wave damage location method still faces limitations such as high cost and complexity in actual panel structure health detection. In order to promote the further application of Lamb wave detection technology in the engineering field, it is urgent to develop a new method for efficient, economical and low-complexity Lamb wave wall panel damage detection and location. Summary of the invention

[0004] The purpose of the present invention is to provide a single-channel wall panel damage location method based on Lamb wave coded metamaterial interface, so as to solve the problems of high cost and great complexity existing in the existing damage location methods.

[0005] In order to achieve the above tasks, the present invention adopts the following technical solutions:

[0006] A single-channel panel damage location method based on Lamb wave coded metamaterial interface, comprising:

[0007] Step 1, constructing a priori database for siding damage detection; the priori database stores the layout of damage monitoring areas on sidings, layout parameters of coding units, Lamb wave parameters of excitation signals applied during damage detection, echo signals of sidings in a healthy state, and a neural network model for damage location identification;

[0008] Step 2: for the wall panel to be damaged, the damage monitoring area and the coding unit are arranged based on the arrangement form of the damage monitoring area and the arrangement parameters of the coding unit in the prior database;

[0009] Step 3, arrange an exciter at the center of the damage monitoring area, apply a Lamb wave excitation signal through the exciter according to the Lamb wave parameters recorded in the priori database, and collect echo signals; calculate the difference between the echo signal and the echo signal of the wall panel in a healthy state as the differential signal to be identified, and input it into the neural network model, and output the corresponding polar angle area through the neural network model to realize the azimuth identification of the damage under single-channel data acquisition; for the differential signal to be identified, the flight time of the Lamb wave and the Lamb wave parameters recorded in the priori database are obtained based on signal processing, so as to determine the distance from the damaged part to the center of the damage monitoring area, and realize the determination of the damage distance under single-channel data acquisition.

[0010] Furthermore, the arrangement of the damage monitoring area includes position, size and shape, wherein the shape includes circle and polygon;

[0011] A polar coordinate system is constructed in the damage monitoring area. The pole of the polar coordinate system is arranged at the center of the damage monitoring area. The polar axis of the polar coordinate system is a ray extending from the pole along the wall panel surface in any direction, and the counterclockwise direction is defined as the positive direction. The position of the damaged part is described in the form of polar coordinates.

[0012] Furthermore, the coding units are distributed around the damage monitoring area, arranged in one or more circles; the coding units are columnar structures, and the material is the same as the wall panel or a magnetic material, and the coding units are fixed to the wall panel surface by pasting or adsorption.

[0013] Furthermore, the arrangement parameters of the encoding unit are determined by finite element simulation, including:

[0014] Finite element simulation software is used to perform finite element simulation modeling on the wall panel and the coding unit. When modeling, a low-reflection boundary is set around the established wall panel model. Lamb wave excitation signal It is applied to the center of the damage monitoring area on the panel model and applied in the direction perpendicular to the panel plane according to formula (1) for a time length of Incentives:

[0015] (1);

[0016] in, is the time variable, is the center frequency of the Lamb wave excitation signal; wherein the center frequency adopts the Lamb wave parameters of the excitation signal applied during damage detection stored in the priori database;

[0017] n acquisition nodes are arranged on a circle with the pole of the polar coordinate system as the center and R as the radius in the damage monitoring area, for receiving response signals;

[0018] Calculate the information entropy shown in formula (2):

[0019] (2);

[0020] in, Represents random events The probability of Representative The collection nodes respond to the signal at time j The normalized result of Represents the information entropy of the displacement field at the acquisition node at time j;

[0021] Define the average information entropy of the displacement field , by selecting the time length N( ) at all times The cumulative average is expressed as:

[0022] (3);

[0023] After arranging the acquisition nodes, the size, layout position and layout quantity of the coding unit are taken as variables. The wall panel model constructed under each set of variables is simulated by applying Lamb wave excitation signals, and the average information entropy of the displacement field corresponding to each set of variables is calculated. ,A set of variables with the largest average information entropy of the displacement field is selected as the final arrangement parameters of the coding unit and is stored in the priori database.

[0024] Furthermore, the method for determining the Lamb wave parameters of the excitation signal applied during damage detection is:

[0025] Raleigh-Lamb dispersion equations constructed for wall panels;

[0026] Estimate the phase velocities of symmetric and antisymmetric modes under different frequency-thickness products, and draw dispersion curves based on frequency-thickness products and phase velocities;

[0027] After the dispersion curve is obtained, the dispersion curve is analyzed to determine the center frequency, mode and velocity of the Lamb wave applied for panel damage detection.

[0028] Furthermore, the dispersion curve is analyzed to determine the center frequency, mode and velocity of the Lamb wave applied for panel damage detection, including:

[0029] (a) Determine the Lamb wave frequency range corresponding to the portion with a single modal component in the dispersion curve, and divide the Lamb wave frequency range into M frequency points; then, extract the Lamb wave modes and Lamb wave velocities corresponding to these frequency points in the dispersion curve in turn to form M groups of Lamb wave parameters; each group of Lamb wave parameters includes the Lamb wave center frequency, mode and Lamb wave velocity;

[0030] (b) Determine the final Lamb wave parameters through the panel damage test;

[0031] (b-1) When the wall panel is not provided with a coding unit, an exciter is arranged at the center of the wall panel damage monitoring area; when the wall panel is in a healthy state without damage, Lamb wave excitation signals under M groups of Lamb wave parameters are respectively applied by the exciter, and the echo signals collected by the exciter are respectively recorded;

[0032] (b-2) Prefabricate a damage in the damage monitoring area, and record the actual distance between the damage and the center of the damage monitoring area on the wall panel; then apply Lamb wave excitation signals under M groups of Lamb wave parameters through the exciter, and record the echo signals collected by the exciter;

[0033] (b-3) performing differential calculation on the echo signals collected when the wall panel is undamaged and damaged under each set of Lamb wave parameters to obtain the differential signal corresponding to the set of Lamb wave parameters; performing continuous wavelet transform on the differential signal, and then extracting the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the time domain distribution diagram, so as to determine the arrival time of the Lamb wave; and obtaining the flight time of the Lamb wave based on the arrival time and the emission time of the Lamb wave;

[0034] (b-4) Calculating the distance from the damaged part to the center of the damage monitoring area based on the flight time of the Lamb wave and the velocity information of the Lamb wave;

[0035] (b-5) Compare the distances from the damage site to the center of the damage monitoring area calculated by the M groups of Lamb wave parameters, and select the Lamb wave parameter corresponding to the distance closest to the actual distance recorded as the final Lamb wave parameter and record it in the prior database.

[0036] Furthermore, the process of determining the echo signal in the healthy state of the wall panel is as follows:

[0037] A healthy and undamaged siding is selected, and a damage monitoring area and coding units are arranged on the siding according to the information in the prior database. An exciter is arranged at the center of the damage monitoring area, and an excitation signal is applied by the exciter according to the Lamb wave parameters recorded in the prior database to collect echo signals. The echo signals are collected and averaged by repeating this step multiple times, and the finally obtained average echo signal is recorded in the prior database as the echo signal of the siding in a healthy state.

[0038] Furthermore, the neural network model for damage location identification is specifically:

[0039] The neural network model includes a feature extractor and a classifier, wherein the feature extractor includes four convolutional layers connected in sequence, and a maximum pooling layer is set after each convolutional layer; the feature extractor is used to extract features of the differential sample signal, so as to obtain the "directional fingerprint" features contained in the differential signal; the classifier includes two fully connected layers connected in sequence, which are used to map the "directional fingerprint" features extracted by the feature extractor to different polar angle areas.

[0040] Furthermore, the construction process of the differential signal data set for training the neural network model is as follows:

[0041] The damage monitoring area is divided into a plurality of polar angle areas at preset intervals with the pole as the center; damage is prefabricated in the damage monitoring area of ​​the wall panel, and the polar angle area of ​​the prefabricated damage is recorded; after the damage is prefabricated, the Lamb wave excitation signal is applied again by the exciter according to the determined Lamb wave parameters and the echo signals are collected multiple times, and multiple groups of echo signals obtained by the wall panel in the damaged state are recorded;

[0042] The difference between multiple groups of echo signals obtained when the wall panel is in a damaged state and the echo signals of the wall panel in a healthy state recorded in a priori database is used as the differential sample signal of the wall panel; this step is repeated multiple times, and the polar angle area where the prefabricated damage on the wall panel is located is different in each operation, thereby obtaining different differential sample signals, and constructing a differential signal data set based on these differential sample signals; wherein each sample in the differential signal data set is the differential sample signal, and the label is the corresponding polar angle area.

[0043] Furthermore, the flight time of the Lamb wave is obtained based on signal processing, specifically:

[0044] The differential signal to be identified is subjected to continuous wavelet transform, and then the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the Lamb wave parameters recorded in the priori database is extracted in the time domain distribution diagram, so as to determine the arrival time of the Lamb wave; based on the arrival time and the emission time of the Lamb wave, the flight time of the Lamb wave is obtained, and based on the flight time and the Lamb wave velocity in the Lamb parameters, the distance from the damaged part to the center of the damage monitoring area is obtained, so as to realize the determination of the distance of the damage.

[0045] Compared with the prior art, the present invention has the following technical features:

[0046] The present invention proposes a method framework for detecting and locating wall panel damage using a single sensor, which preliminarily solves the limitations of the traditional Lamb wave positioning method and provides a new idea for realizing efficient and economical detection of wall panel structural damage. Compared with the prior art, the present invention can greatly improve the recognition accuracy of different damage locations under single-channel signal acquisition by rationally designing the coded metamaterial interface. At the same time, the layout of the single-channel signal collector and the simplification of the signal processing method greatly reduce the complexity of the detection system and the difficulty of data processing, improve the feasibility of practical application, and is expected to promote the further development of structural health detection technology. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 It is a schematic diagram of the process of the present invention;

[0048] Figure 2 Aluminum alloy siding with coded metamaterial interface designed for embodiments of the present invention;

[0049] Figure 3 Schematic diagram for constructing a polar coordinate system within the damage monitoring area;

[0050] Figure 4 is the out-of-plane displacement field of the metamaterial interface with the number of coding units being 0, 10, 20, 30, 40 and 50 respectively in the embodiment of the present invention;

[0051] Figure 5 is the time domain response of the normalized out-of-plane displacement at the specified node under different numbers of coding units;

[0052] Figure 6 The maximum, minimum and average values ​​of the information entropy of the coding metamaterial interface calculated according to five different arrangements with different numbers of coding units;

[0053] Figure 7 Lamb wave dispersion curve of an aluminum alloy wall plate with a thickness of 1.5 mm provided in an embodiment of the present invention;

[0054] Figure 8 Schematic diagram of a test device in an embodiment of the present invention;

[0055] Fig. 9 The normalized results of Lamb wave differential signals at different damage orientations obtained by measurement in an embodiment of the present invention;

[0056] Fig.10 The wavelet envelope results of the differential signal obtained under two damage experimental cases provided in the embodiments of the present invention;

[0057] Fig.11 The experimental results of orientation identification of two damage cases provided by the embodiments of the present invention. DETAILED DESCRIPTION

[0058] See attached Figure 1 The present invention provides a single-channel plate damage localization method (SDL) based on Lamb wave coded metamaterial interface. The method cleverly integrates Lamb wave coded metamaterial interface, deep learning algorithm and wave velocity equation; by designing a coded metamaterial interface composed of coding units on the wall plate, utilizing its modulation effect on the propagation direction of Lamb waves, combining with deep learning algorithm to identify the damage azimuth of single-channel acquisition signals, and using Lamb wave velocity equation to estimate the damage distance, finally achieving the purpose of simultaneously obtaining the azimuth information and distance information of wall plate damage using only a single sensing channel. This method effectively simplifies the wall plate damage detection system and reduces its detection cost. The specific implementation process of the present invention is described in detail below in conjunction with the accompanying drawings.

[0059] Step 1: construct a priori database for siding damage detection; the priori database stores the layout of the damage monitoring area on the siding, the layout parameters of the coding unit, the Lamb wave parameters of the excitation signal applied during damage detection, the echo signal of the siding in a healthy state, and the neural network model for damage orientation identification. Using the data in the priori database, damage location can be performed on sidings of the same type under single-channel data acquisition.

[0060] Step 1.1, layout of damage monitoring area and establishment of polar coordinates.

[0061] In this solution, the wall panels are made of the same homogeneous material, and the wave speeds in different directions are roughly the same. In practical applications, they can be metal plates, etc. The position, size and shape of the damage monitoring area on the wall panel are specified by the user, such as circular, square, rectangular, polygonal, etc., which can be set according to the shape of the wall panel; the subsequent work of this solution is to determine the position of the damaged part in the polar coordinate system and the distance from the pole in the damage monitoring area of ​​the wall panel to be damaged.

[0062] In order to facilitate identification and positioning, this scheme constructs a polar coordinate system in the damage monitoring area, specifically:

[0063] The pole of the polar coordinate system is arranged at the center of the damage monitoring area, and the polar axis of the polar coordinate system is a ray extending from the pole along the wall panel surface in any direction, and the counterclockwise direction is defined as the positive direction.

[0064] Polar coordinates are used to describe the location of the injury, expressed as ; where the polar diameter Indicates the distance from the damage site to the pole, the polar angle It represents the angle between the line between the damage site and the pole and the polar axis.

[0065] Step 1.2, encoding unit arrangement parameters.

[0066] In this solution, the coding units are distributed around the damage monitoring area, arranged in one or more circles; the coding units are columnar structures, for example, cylindrical, and the material can be the same as the wall panel or a magnetic material. The coding units are fixed to the wall panel surface by pasting or adsorption.

[0067] After arranging the coding unit array around the damage monitoring area, a coded metamaterial interface is formed; the coding unit array has a unique wave scattering boundary condition and can directionally encode the Lamb wave propagating in the wall panel; through this directional coding, the response signals collected from different damage orientations can show obvious differences in propagation characteristics, just like each orientation has its own unique fingerprint information; therefore, this scheme figuratively calls this difference the "directional fingerprint" feature.

[0068] The layout parameters of the encoding unit are determined by finite element simulation, as follows:

[0069] Finite element simulation modeling of the wall panel and the coding unit was performed using the finite element simulation software COMSOL Multiphysics. During modeling, a low-reflection boundary was set around the established wall panel model to eliminate the interference of boundary reflection on the simulation results. The model grid size Should meet , is the minimum wavelength of the excitation signal; the time sampling step Should meet , is the maximum frequency of the excitation signal; Lamb wave excitation signal It is applied to the center (pole) of the damage monitoring area on the panel model and applied in the direction perpendicular to the panel plane according to formula (1) for a time length of Incentives:

[0070] (1);

[0071] in, is the time variable, is the center frequency of the Lamb wave excitation signal; wherein the center frequency adopts the Lamb wave frequency of the excitation signal applied during damage detection.

[0072] In addition, n acquisition nodes are arranged on a circle with the pole of the polar coordinate system as the center and R as the radius in the damage monitoring area to receive response signals. During the simulation, the time domain responses of these acquisition nodes over a period of time are extracted to evaluate the quality of the "directional fingerprint" features generated under the current coding unit arrangement parameters. Among them, R cannot be too small or too large, and is generally selected to be about half of the maximum radius of the damage monitoring area.

[0073] In this scheme, the above advantages and disadvantages are judged by the information entropy shown in formula (2):

[0074] (2);

[0075] in, Represents random events The probability of Representative The collection nodes respond to the signal at time j The normalized result of It represents the information entropy of the displacement field at the acquisition node at time j, which reflects the difference in the characteristics of the Lamb wave "directional fingerprint" at each acquisition node.

[0076] Define the average information entropy of the displacement field , by selecting the time length N( ) at all times The accumulated average is used to quantitatively evaluate the quality of the "directional fingerprint" characteristics of the Lamb wave given by the coding metamaterial interface over a period of time. The expression is:

[0077] (3);

[0078] The larger it is, the more effective the encoding unit layout parameters are, the more significant the "directional fingerprint" feature is, and the more conducive it is to the subsequent single-channel damage location identification.

[0079] After arranging the acquisition nodes, the size, arrangement position and arrangement quantity of the coding unit are taken as variables. The wall panel model containing the coding metamaterial interface constructed under each set of variables is simulated by applying the Lamb wave excitation signal, and the average information entropy of the displacement field corresponding to each set of variables is calculated. ,A set of variables with the largest average information entropy of the displacement field is selected as the final arrangement parameters of the coding unit and is stored in the priori database.

[0080] Step 1.3, Lamb wave parameters of the excitation signal applied during damage detection.

[0081] (1) The dispersion curve is drawn based on the Rayleigh-Lamb dispersion equation. Through experimental analysis, the Lamb wave parameters of the excitation signal applied for panel damage detection are determined, including the Lamb wave mode, center frequency and Lamb wave velocity.

[0082] (1-1) Lamb wave is an ultrasonic guided wave propagating in a plate-like structure. Its unique dispersion characteristics make the phase velocity and group velocity vary with frequency. Within a given frequency range, Lamb waves may propagate in multiple modes. In order to simplify the signal processing and analysis steps, choosing the right frequency and mode is crucial to achieve accurate damage location. As shown in formula (4), the Raleigh-Lamb dispersion equation constructed for the wall panel is as follows:

[0083] (4);

[0084] Where d represents half the wall thickness, and p and q are given by:

[0085] (5);

[0086] in, is the angular frequency, , f is the frequency; k is the wave number, phase velocity ; is the longitudinal wave velocity, ; is the shear wave velocity, ; and They are Lamé constants; is the density of the siding material.

[0087] Rewrite formula (4) as follows:

[0088] (6);

[0089] By solving equation (6), the phase velocities of the symmetric mode and antisymmetric mode under different frequency-thickness products are estimated, and the dispersion curve is drawn based on the frequency-thickness product and phase velocity.

[0090] (1-2) After obtaining the dispersion curve, the dispersion curve is analyzed to determine the center frequency, mode and Lamb wave velocity v of the Lamb wave applied for panel damage detection.

[0091] (a) Determine the Lamb wave frequency range (part of the horizontal axis of the dispersion curve) corresponding to the portion with a relatively simple modal component in the dispersion curve, and divide the Lamb wave frequency range into M frequency points; then, extract the Lamb wave modes and Lamb wave velocities corresponding to these frequency points in the dispersion curve in turn to form M groups of Lamb wave parameters; each group of Lamb wave parameters includes the Lamb wave center frequency, mode and Lamb wave velocity.

[0092] (b) The final Lamb wave parameters are determined through the wall panel damage test.

[0093] (b-1) When the wall panel is not provided with a coding unit, an exciter (such as a piezoelectric ceramic) is arranged at the center of the damage monitoring area on the wall panel; when the wall panel is in a healthy state without damage, Lamb wave excitation signals under M groups of Lamb wave parameters are applied by the exciter, and the echo signals S collected by the exciter are recorded respectively. 01 ~S 0M , where S 01 , S 0M Respectively represent the first and last group of echo signals.

[0094] (b-2) A damage is prefabricated in the damage monitoring area, and the actual distance between the damage and the center of the damage monitoring area on the wall panel is recorded; then Lamb wave excitation signals under M groups of Lamb wave parameters are applied by the exciter, and the echo signals S collected by the exciter are recorded respectively. 11 ~S 1M ; where S 11 , S 1M Respectively represent the first and last group of echo signals.

[0095] (b-3) The echo signals collected when the wall panel is intact and damaged under each set of Lamb wave parameters are differentially calculated (e.g. ), and obtain the differential signal corresponding to the group of Lamb wave parameters; perform continuous wavelet transform on the differential signal, and then extract the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the time domain distribution diagram, so as to determine the arrival time of the Lamb wave; based on the arrival time and emission time of the Lamb wave, obtain the flight time T of the Lamb wave.

[0096] (b-4) Based on the flight time T of the Lamb wave and the velocity v of the Lamb wave, calculate the distance from the damaged part to the center of the damage monitoring area. , the calculation formula is as follows:

[0097] (7);

[0098] (b-5) Comparison of the distance from the damage site to the center of the damage monitoring area calculated by the Lamb wave parameters of group M , select the closest actual distance to the record The corresponding Lamb wave parameters are recorded in the priori database as the final Lamb wave parameters.

[0099] Step 1.4, determination of the echo signal of the panel in a healthy state.

[0100] A healthy and undamaged siding is selected, and a damage monitoring area and coding units are arranged on the siding according to the information in the prior database. An exciter is arranged at the center of the damage monitoring area, and an excitation signal is applied by the exciter according to the Lamb wave parameters recorded in the prior database to collect echo signals. The echo signals are collected and averaged by repeating this step multiple times, and the finally obtained average echo signal is recorded in the prior database as the echo signal of the siding in a healthy state.

[0101] Step 1.5, construction of differential signal data set.

[0102] The damage monitoring area is divided into multiple polar angle areas at preset intervals (for example, 10°) with the pole as the center; damage is prefabricated in the damage monitoring area of ​​the wall panel (by adsorbing a magnet on the wall panel), and the polar angle area of ​​the prefabricated damage is recorded; after the damage is prefabricated, the Lamb wave excitation signal is applied again through the exciter according to the determined Lamb wave parameters and the echo signals are collected multiple times, and the multiple groups of echo signals obtained when the wall panel is in the damaged state are recorded.

[0103] The difference between multiple groups of echo signals obtained when the wall panel is in a damaged state and the echo signals of the wall panel in a healthy state recorded in a priori database is used as the differential sample signal of the wall panel; this step is repeated multiple times, and the polar angle area where the prefabricated damage on the wall panel is located is different in each operation, thereby obtaining different differential sample signals, and constructing a differential signal data set based on these differential sample signals; wherein each sample in the differential signal data set is the differential sample signal, and the label is the corresponding polar angle area.

[0104] Step 1.6, damage location identification neural network model.

[0105] The neural network model in this scheme includes a feature extractor and a classifier, wherein the feature extractor includes four convolutional layers connected in sequence, and a maximum pooling layer is set after each convolutional layer; the feature extractor is used to extract features of the differential sample signal, so as to obtain the "directional fingerprint" features contained in the differential signal; the convolutional layer is used to extract local features, and the maximum pooling layer is used to reduce the feature dimension and reduce the computational complexity; the classifier includes two fully connected layers connected in sequence, which are used to map the "directional fingerprint" features extracted by the feature extractor to different polar angle areas.

[0106] The input and output of the damage orientation identification neural network model should be adapted to the length of the differential sample signal and the number of polar angle regions. In one embodiment of the present invention, the length of the differential sample signal is 48500, and the number of polar angle regions is 36 (with an interval of 10°, 0° and 360° overlap), and the constructed neural network model is as follows:

[0107]

[0108] The PyTorch deep learning framework is used to program the neural network model during training, and the Adam optimizer and cross entropy loss function are selected. The learning rate is set to 0.000001, the batch size is 32, and the number of iterations is 50. The neural network model is trained using the differential signal dataset, and the trained neural network model is saved in the prior database.

[0109] Step 2: for the wall panel to be damaged, the damage monitoring area and the coding unit are arranged based on the arrangement form of the damage monitoring area and the arrangement parameters of the coding unit in the priori database.

[0110] Step 3: Arrange an exciter at the center of the damage monitoring area, apply a Lamb wave excitation signal through the exciter according to the Lamb wave parameters recorded in the prior database, and collect the echo signal; calculate the difference between the echo signal and the echo signal of the wall panel in a healthy state as the differential signal to be identified, and input it into the neural network model, and output the corresponding polar angle area through the neural network model , to realize the position identification of damage under single-channel data acquisition; for the differential signal to be identified, the flight time T of the Lamb wave and the velocity v of the Lamb wave recorded in the prior database are obtained based on signal processing, so as to determine the distance from the damaged part to the center of the damage monitoring area , to realize the damage distance determination under single-channel data acquisition.

[0111] The flight time T of the Lamb wave obtained based on signal processing is specifically:

[0112] The differential signal to be identified is subjected to continuous wavelet transform. Then, in the time domain distribution diagram, the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the Lamb parameters recorded in the priori database is extracted to determine the arrival time of the Lamb wave. Based on the arrival time and emission time of the Lamb wave, the flight time T of the Lamb wave is obtained. Based on the flight time T and the Lamb wave velocity v recorded in the priori database, the distance from the damaged part to the center of the damage monitoring area is obtained according to formula (7), thereby realizing the determination of the damage distance.

[0113] Example:

[0114] In one embodiment of the present invention, when determining the arrangement parameters of the encoding unit by finite element simulation, Figure 4 The simulation results of the out-of-plane displacement field of the coding metamaterial interface with the number of coding units of 0, 10, 20, 30, 40 and 50 at a frequency of 30 kHz are given. mm, height mm; Figure 5 The extracted and normalized time domain responses of node out-of-plane displacements on a circle with a radius of 100 mm and the center of the wall panel are shown; the simulation results of these time domain responses intuitively show the modulation effect of the designed metamaterial interface on the propagation direction of the Lamb wave. When the number of coding units is small (such as 0 or 10), the propagation of Lamb waves is relatively regular, and the out-of-plane displacement field shows a relatively simple distribution feature. However, when the number of coding units increases to 50, the scattering effect of the Lamb waves modulated by the coding units becomes stronger and stronger, and the scattered waves in different directions interfere with each other to form a complex wave field distribution. This result lays the foundation for providing a recognizable "directional fingerprint" feature for Lamb waves containing damage information.

[0115] In addition, according to formula (3), the average information entropy of the displacement field under different numbers of coding units and five different coding unit distribution modes is calculated: , and counted its maximum, minimum and average values, such as Figure 6 As shown. The following conclusions can be drawn:

[0116] (1) The number of coding units in the coding metamaterial interface significantly affects the "directional fingerprint" feature of the Lamb wave. As the number of coding units increases, the information entropy of the "directional fingerprint" feature shows an upward trend, which indicates that the coding effect of the coding metamaterial interface is constantly increasing; (2) The arrangement of the coding units will also affect the effect of the "directional fingerprint" feature to a certain extent. However, when the number of coding units reaches a certain threshold (such as 50 in the embodiment), the effect of the arrangement on the "directional fingerprint" feature becomes relatively small. This means that when designing an efficient coding metamaterial interface, priority should be given to increasing the number of coding units rather than paying too much attention to their arrangement.

[0117] Based on the arrangement parameters of the coding units determined by the above simulation, in one embodiment, the arrangement is as follows:

[0118] The dimensions of the wall panels (length × width × thickness) are as follows: mm 3 , made of 6061 aluminum alloy siding, its elastic modulus GPa, shear modulus GPa, Poisson's ratio ,density kg / m³. In this example, 50 cylindrical coding units are used, with a radius of mm, height mm; the specific layout position is at the outer boundary of the circular damage monitoring area of ​​the wall panel, distributed in the circle with the center of the wall panel as the center and the radius is mm, On the two circumferences of mm, 25 coding units are randomly arranged on the outer and inner circumferences to form a coding metamaterial interface; Figure 2 The polar coordinate system established on the aluminum alloy wall plate containing the coded metamaterial interface is shown in Figure 3 The specific test setup and equipment are shown in Figure 8 shown.

[0119] See attached Figure 7 , which is the Lamb wave dispersion curve of the aluminum alloy wall panel with a thickness of 1.5 mm at different frequencies; This embodiment selects 5 groups of Lamb wave parameters for testing, and finally determines that 30kHz ( kHz) is the center frequency excitation Modal Lamb wave is used to detect damage to the wall panel; at a frequency of 30kHz, the aluminum alloy wall panel The wave velocity of the modal Lamb wave is approximately m / s.

[0120] See attached Figure 8 When locating damage on a wall panel, the coding unit on the aluminum alloy wall panel is composed of a pair of magnets with a radius of 5mm and a height of 8mm adsorbed on the front and back sides of the wall panel, totaling 50 pairs; the single piezoelectric ceramic located at the center of the wall panel simultaneously undertakes the two tasks of exciting the Lamb wave excitation signal and receiving the echo signal. The magnet simulating the wall panel damage is placed as follows Figure 8 On the dot array circle with a radius of 100 mm and a central angle interval of 10°, echo signals are collected according to step 1.5, and a differential signal data set is constructed; in this case, echo signals are collected 50 times at each damage orientation.

[0121] exist Figure 8 Two damage cases are set in the damage monitoring area to verify the effectiveness of the proposed method. The damage distances and orientations of the two damage cases are as follows:

[0122] Case 1: The damage distance is 147 mm, and the damage direction is 50° counterclockwise, that is, .

[0123] Case 2: The damage distance is 74 mm, and the damage direction is 150° counterclockwise, that is, .

[0124] Fig. 9 The results of normalized Lamb wave differential signals obtained from experimental measurements at 36 different damage orientations are shown. Fig. 9 It can be clearly observed that the Lamb wave differential signals under different damage orientations show significant differences. This result proves that the coded metamaterial interface can effectively give Lamb waves unique "directional fingerprint" characteristics under different damage orientations, providing a basis for realizing single-channel damage orientation identification.

[0125] Fig.10 The wavelet envelope of the difference signal obtained under two damage cases is shown, where the two cases are marked The arrival time of the modal Lamb wave is then subtracted from the excitation start time of the Lamb wave in the corresponding case (83 and 91 ), we get two cases The flight time T of the modal Lamb wave, the specific calculation results are shown in Table 2; after obtaining the flight time T, the damage distances of the two damage cases are calculated using the wave velocity v of the Lamb wave ,The calculation results are summarized in Table 3. It can be seen from Table 3 that the estimated damage distance obtained by experimental calculation is very close to the actual damage distance.

[0126]

[0127] This embodiment of the present invention realizes the identification of damage location based on the neural network model constructed based on the parameters in Table 1. The model achieved a cross entropy loss of 0.0035 on the test set, showing excellent damage location identification performance. Fig.11 The position identification results of two damage cases are shown and summarized in Table 4. For damage case 1 and damage case 2, the damage position identification results given by the position identification network are They are 50° and 150° respectively, which are completely consistent with the actual damage orientation.

[0128]

[0129] The above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application, and should all be included in the protection scope of the present application.

Claims

1. A single-channel panel damage location method based on Lamb wave coded metamaterial interface, characterized in that: include: Step 1, construct a priori database for wall panel damage detection; The a priori database stores the layout of the damage monitoring area on the wall panel, the layout parameters of the coding unit, the Lamb wave parameters of the excitation signal applied during damage detection, the echo signal of the wall panel in a healthy state, and the neural network model for damage location identification; Step 2: for the wall panel to be damaged, the damage monitoring area and the coding unit are arranged based on the arrangement form of the damage monitoring area and the arrangement parameters of the coding unit in the prior database; Step 3, arrange an exciter at the center of the damage monitoring area, apply a Lamb wave excitation signal through the exciter according to the Lamb wave parameters recorded in the priori database, and collect echo signals; calculate the difference between the echo signal and the echo signal of the wall panel in a healthy state as the differential signal to be identified, and input it into the neural network model, and output the corresponding polar angle area through the neural network model to realize the azimuth identification of the damage under single-channel data acquisition; for the differential signal to be identified, obtain the flight time of the Lamb wave and the Lamb wave parameters recorded in the priori database based on signal processing, so as to determine the distance from the damage site to the center of the damage monitoring area, and realize the determination of the damage distance under single-channel data acquisition; The construction process of the differential signal data set used to train the neural network model is: The damage monitoring area is divided into a plurality of polar angle areas at preset intervals with the pole as the center; damage is prefabricated in the damage monitoring area of ​​the wall panel, and the polar angle area of ​​the prefabricated damage is recorded; after the damage is prefabricated, the Lamb wave excitation signal is applied again by the exciter according to the determined Lamb wave parameters and the echo signals are collected multiple times, and multiple groups of echo signals obtained by the wall panel in the damaged state are recorded; The difference between multiple groups of echo signals obtained when the wall panel is in a damaged state and the echo signals of the wall panel in a healthy state recorded in a priori database is used as the differential sample signal of the wall panel; this step is repeated multiple times, and the polar angle area where the prefabricated damage on the wall panel is located is different in each operation, thereby obtaining different differential sample signals, and constructing a differential signal data set based on these differential sample signals; wherein each sample in the differential signal data set is the differential sample signal, and the label is the corresponding polar angle area.

2. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The arrangement of the damage monitoring area includes position, size and shape, wherein the shape includes circle and polygon; A polar coordinate system is constructed in the damage monitoring area. The pole of the polar coordinate system is arranged at the center of the damage monitoring area. The polar axis of the polar coordinate system is a ray extending from the pole along the wall panel surface in any direction, and the counterclockwise direction is defined as the positive direction. The position of the damaged part is described in the form of polar coordinates.

3. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The coding units are distributed around the damage monitoring area and arranged in one or more circles; the coding units are columnar structures, and the material is the same as that of the wall panels. The coding units are fixed to the surface of the wall panels by pasting or adsorption.

4. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The layout parameters of the encoding unit are determined by finite element simulation, including: Finite element simulation software is used to perform finite element simulation modeling on the wall panel and the coding unit. When modeling, a low-reflection boundary is set around the established wall panel model. Lamb wave excitation signal It is applied to the center of the damage monitoring area on the panel model and applied in the direction perpendicular to the panel plane according to formula (1) for a time length of Incentives: (1)； in, is the time variable, is the center frequency of the Lamb wave excitation signal; wherein the center frequency adopts the Lamb wave parameters of the excitation signal applied during damage detection stored in the priori database; A circle with the pole of the polar coordinate system as the center in the damage monitoring area. R Arrange on a circle with a radius of n A collection node, used for receiving a response signal; Calculate the information entropy shown in formula (2): (2)； in, Represents random events The probability of Representative The collection nodes are j Always respond to signals The normalized result of Representatives in j The information entropy of the displacement field at the acquisition node at the time; Define the average information entropy of the displacement field , by selecting the time length N ( ) at all times The cumulative average is expressed as: (3)； After arranging the acquisition nodes, the size, layout position and layout quantity of the coding unit are taken as variables. The wall panel model constructed under each set of variables is simulated by applying Lamb wave excitation signals, and the average information entropy of the displacement field corresponding to each set of variables is calculated. ,A set of variables with the largest average information entropy of the displacement field is selected as the final arrangement parameters of the coding unit and is stored in the priori database.

5. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The method for determining the Lamb wave parameters of the excitation signal applied during damage detection is: Raleigh-Lamb dispersion equations constructed for wall panels; Estimate the phase velocities of symmetric and antisymmetric modes under different frequency-thickness products, and draw dispersion curves based on frequency-thickness products and phase velocities; After the dispersion curve is obtained, the dispersion curve is analyzed to determine the center frequency, mode and velocity of the Lamb wave applied for panel damage detection.

6. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 5 is characterized in that: The dispersion curve is analyzed to determine the center frequency, mode and velocity of the Lamb wave applied for panel damage detection, including: (a) Determine the Lamb wave frequency range corresponding to the single modal component part of the dispersion curve, and divide the Lamb wave frequency range to obtain M frequency points; then, the Lamb wave modes and Lamb wave velocities corresponding to these frequency points are extracted in turn from the dispersion curve to form M A group of Lamb wave parameters; each group of Lamb wave parameters includes the Lamb wave center frequency, mode and Lamb wave velocity; (b) Determine the final Lamb wave parameters through the panel damage test; (b-1) When the wall panel is not equipped with a coding unit, an exciter is arranged at the center of the wall panel damage monitoring area; when the wall panel is in a healthy state without damage, the exciter is used to apply M The Lamb wave excitation signal under the Lamb wave parameters is set, and the echo signals collected by the exciter are recorded respectively; (b-2) Prefabricate a damage in the damage monitoring area and record the actual distance between the damage and the center of the damage monitoring area on the wall plate; then apply M The Lamb wave excitation signal under the Lamb wave parameters is set, and the echo signals collected by the exciter are recorded respectively; (b-3) performing differential calculation on the echo signals collected when the wall panel is undamaged and damaged under each set of Lamb wave parameters to obtain the differential signal corresponding to the set of Lamb wave parameters; performing continuous wavelet transform on the differential signal, and then extracting the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the time domain distribution diagram, so as to determine the arrival time of the Lamb wave; and obtaining the flight time of the Lamb wave based on the arrival time and the emission time of the Lamb wave; (b-4) Calculating the distance from the damaged part to the center of the damage monitoring area based on the flight time of the Lamb wave and the velocity information of the Lamb wave; (b-5) Comparison M The distance from the damage site to the center of the damage monitoring area calculated by the Lamb wave parameters is selected, and the Lamb wave parameter corresponding to the distance closest to the actual distance recorded is selected as the final Lamb wave parameter and recorded in the prior database.

7. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The process of determining the echo signal in the healthy state of the wall panel is as follows: A healthy and undamaged siding is selected, and a damage monitoring area and coding units are arranged on the siding according to the information in the prior database. An exciter is arranged at the center of the damage monitoring area, and an excitation signal is applied by the exciter according to the Lamb wave parameters recorded in the prior database to collect echo signals. The echo signals are collected and averaged by repeating this step multiple times, and the finally obtained average echo signal is recorded in the prior database as the echo signal of the siding in a healthy state.

8. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The neural network model for damage location identification is specifically: The neural network model includes a feature extractor and a classifier, wherein the feature extractor includes four convolutional layers connected in sequence, and a maximum pooling layer is set after each convolutional layer; the feature extractor is used to extract features of the differential sample signal, so as to obtain the "directional fingerprint" features contained in the differential signal; the classifier includes two fully connected layers connected in sequence, which are used to map the "directional fingerprint" features extracted by the feature extractor to different polar angle areas.

9. The single-channel wall panel damage location method based on Lamb wave coded metamaterial interface according to claim 1 is characterized in that: The flight time of the Lamb wave obtained based on signal processing is specifically: The differential signal to be identified is subjected to continuous wavelet transform, and then the time information under the peak value of the Lamb wave envelope spectrum corresponding to the Lamb wave mode in the Lamb wave parameters recorded in the priori database is extracted in the time domain distribution diagram, so as to determine the arrival time of the Lamb wave; based on the arrival time and the emission time of the Lamb wave, the flight time of the Lamb wave is obtained, and based on the flight time and the Lamb wave velocity in the Lamb parameters, the distance from the damaged part to the center of the damage monitoring area is obtained, so as to realize the determination of the distance of the damage.

Images