Ultrasonic guided wave path response distribution adaptive defect quantification identification method and system
By combining deep learning and differentiable physics constraints, an adaptive quantitative defect identification method based on ultrasonic guided wave path response distribution is developed. This method solves the problem of insufficient adaptability in defect identification in existing technologies and achieves high-precision defect localization and accurate quantification of geometric parameters.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CHANGCHUN INST OF OPTICS FINE MECHANICS & PHYSICS CHINESE ACAD OF SCI
- Filing Date
- 2026-06-11
- Publication Date
- 2026-07-14
AI Technical Summary
Existing ultrasonic guided wave defect identification methods rely on empirical assumptions and manually set parameters, which are not adaptable enough, make it difficult to accurately obtain defect geometric size information, and have limited quantitative identification capabilities.
An adaptive quantitative defect identification method based on ultrasonic guided wave path response distribution is adopted. This method combines deep learning and differentiable physical constraints. The parameters of the path response distribution model are predicted by a parameter prediction network, and the geometric parameters of the defect are calculated by combining a differentiable physical layer.
It improves the accuracy of defect location and the quantitative characterization of geometric parameters, and enhances the physical interpretability of the model.
Smart Images
Figure CN122385779A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of mechanical engineering, optical engineering, nondestructive testing, and automated testing, and in particular to an adaptive method and system for quantitative identification of defects in ultrasonic guided wave path response distribution. Background Technology
[0002] Large plate structures are widely used in aerospace, shipbuilding, and rail transportation. During service, they are prone to defects such as cracks, corrosion thinning, and debonding, leading to a decrease in structural load-bearing capacity and causing safety accidents. Ultrasonic guided wave testing technology has advantages such as long propagation distance, large detection range, and suitability for rapid screening, and is widely used in the defect detection of large plate structures. Existing ultrasonic guided wave methods are usually based on multi-path detection signals, combined with differences in energy changes or damage characterization along different paths, and rely on a pre-set path space response model to reconstruct and determine the possible locations of defects.
[0003] In recent years, research has introduced deep learning methods into ultrasonic guided wave defect identification. By learning from guided wave signals or damage characterization features, defect classification, localization, or parameter prediction can be achieved. Existing ultrasonic guided wave defect detection methods are typically based on multi-path detection signals, combining energy changes or damage characterization differences along different paths, and relying on a pre-set path spatial response model to reconstruct the defect location. Their main drawbacks are: first, they rely on empirical assumptions and manually set parameters, resulting in insufficient parameter adaptability to different defect sizes, shapes, and operating conditions; second, their ability to characterize the defect spatial response is limited, easily leading to defect location errors and boundary ambiguity; and third, they focus on defect location judgment, making it difficult to accurately obtain defect geometric dimension information, resulting in insufficient quantitative identification capabilities. Summary of the Invention
[0004] This invention aims to address the technical problems of existing ultrasonic guided wave defect identification methods, which rely solely on data-driven approaches and lack physical constraints combined with the guided wave path response mechanism. These methods also suffer from insufficient interpretability of defect detection results and inadequate quantitative characterization of defect geometric parameters. The invention provides an adaptive ultrasonic guided wave path response distribution-based quantitative defect identification method and system suitable for detecting defects of various sizes and shapes in large-scale plates. This method balances positioning accuracy, quantitative characterization of geometric parameters, and physical interpretability.
[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:
[0006] An adaptive quantitative identification system for ultrasonic guided wave path response distribution defects includes: a control and testing system, specifically comprising:
[0007] A robotic arm, the front end of which is connected to an ultrasonic probe; the ultrasonic probe is used to test the sample.
[0008] The host computer is connected to the robotic arm controller and the signal generator / receiver. The signal generator / receiver is also connected to the preamplifier and the postamplifier. The preamplifier and the postamplifier are connected to the ultrasonic probe. The robotic arm controller is used to control the robotic arm's movements under the control of the host computer. The signal generator / receiver is used to generate ultrasonic signals under the control of the host computer and transmit them to the preamplifier. After being amplified by the preamplifier, the ultrasonic signals are emitted by the ultrasonic probe. The signal generator / receiver is also used to receive ultrasonic signals collected by the ultrasonic probe and then amplified by the postamplifier under the control of the host computer.
[0009] In the above technical solution, the ultrasonic probe includes: an ultrasonic receiving sensor and an ultrasonic transmitting sensor; the ultrasonic transmitting sensor is connected to a preamplifier, and the ultrasonic receiving sensor is connected to a postamplifier.
[0010] Under the control of the host computer, the signal generator / receiver generates an ultrasonic signal and transmits it to the preamplifier. After being amplified by the preamplifier, it is emitted by the ultrasonic transmitting sensor.
[0011] Under the control of the host computer, the signal generator / receiver receives the ultrasonic signal collected by the ultrasonic receiving sensor and then amplifies it through a post-amplifier.
[0012] An adaptive quantitative identification method for defects in ultrasonic guided wave path response distribution is provided, applicable to the aforementioned adaptive quantitative identification system for ultrasonic guided wave path response distribution. The method includes the following steps:
[0013] Step S1: Set the detection parameters and collect the health status signal and the guided wave signal of the tested area;
[0014] Step S2: Signal preprocessing and calculation of damage intensity curves for each path;
[0015] Step S3: Input the damage intensity curve into the parameter prediction network to predict the path response distribution model parameters for each path;
[0016] Step S4: Input the predicted path response distribution model parameters into the differentiable physical layer to calculate the geometric parameters of the defect;
[0017] Step S5: Output the defect identification results.
[0018] In the above technical solution, step S1 specifically includes:
[0019] The host computer controls the robotic arm controller, sets the scanning direction and scanning step size according to the requirements of the detection task, and the robotic arm drives the ultrasonic probe to complete the scanning of the test sample;
[0020] The host computer controls the signal generator / receiver, and uses a preamplifier and a postamplifier to complete the acquisition of health status signals and guided wave signals of the measured area.
[0021] In the above technical solution, step S2 specifically includes:
[0022] After acquiring the detection signal, the damage intensity value in each scanning direction is obtained based on the difference between the health status signal and the guided wave signal of the tested area, and the damage intensity curve is formed by arranging them in order of scanning distance.
[0023] In the above technical solution, step S3 specifically includes:
[0024] The damage intensity curves in each scanning direction are input into the parameter prediction network; the parameter prediction network extracts features from the multi-directional damage intensity curves and predicts the path response distribution model parameters corresponding to each effective path.
[0025] In the above technical solution, step S4 specifically includes:
[0026] The predicted path response distribution model parameters are input into the differentiable physical layer, which performs spatial mapping and multi-path fusion calculation on the response of each detection path based on the predicted path response distribution model parameters.
[0027] In the above technical solution, step S4 specifically includes:
[0028] No. The coordinates of the distribution centers in each scanning direction are shown below:
[0029]
[0030] in, Indicates the first The first scanning direction The original location of the path. Indicates the path number. Indicates the total number of paths. Indicates the scan direction number. Indicates the first The center offset parameter of the path, This represents the center position of the path response distribution after offset correction; based on this center position vector and the distribution width parameter vector, a path response distribution matrix is constructed as follows:
[0031]
[0032] in, Indicates the first The first scanning direction The position coordinates of a discrete point in space Indicates the number of discrete points in space. This represents the total number of discrete points in space. Indicates the first Distribution width parameter in each scanning direction, Indicates the cutoff radius. Indicates the first The path in the first The response distribution values at discrete points in space; This represents the step function, used to truncate the Gaussian response distribution. The value is 1 at time. The value is 0 at that time.
[0033] Based on this, the path fusion result for a single scanning direction is obtained, as shown below:
[0034]
[0035] in, Indicates the first The first scanning direction Path damage intensity value; Indicates the first The scanning direction in the first... Fusion response values at discrete points in space;
[0036] Weighted fusion of multiple scanning directions yields the final defect characterization result, as shown below:
[0037]
[0038] in, Indicates the first Weights are fused along each scanning direction. Indicates the first The final defect characterization value at each spatial discrete point This indicates the total number of scanning directions.
[0039] In the above technical solution, in step S4, the geometric parameters of the defect include: defect length, defect width, and defect area.
[0040] In the above technical solution, step S5 specifically includes:
[0041] The final defect characterization results are segmented by threshold, and high-response regions are extracted and used as candidate defect regions.
[0042] The defect location is determined based on the position of the defect candidate region in spatial coordinates, the defect length is calculated based on the projection range of the defect candidate region in one direction, the defect width is calculated based on the projection range in another direction perpendicular to the aforementioned direction, and the defect area is calculated based on the spatial area covered by the defect candidate region.
[0043] The present invention has the following beneficial effects:
[0044] The ultrasonic guided wave path response distribution adaptive defect quantitative identification method and system of the present invention adaptively predicts the path response distribution parameters of each effective path through a parameter prediction network, avoiding the dependence of traditional methods on fixed empirical parameters, and can better adapt to the response differences corresponding to defects of different sizes and shapes, thereby improving the defect location accuracy.
[0045] The ultrasonic guided wave path response distribution adaptive defect quantitative identification method and system of the present invention constructs the path response distribution, multi-path fusion and defect parameter extraction process into a differentiable physical layer, which can directly output the defect location and length and width size parameters, thereby enhancing the quantitative characterization ability of defect geometric information.
[0046] The adaptive defect quantitative identification method and system for ultrasonic guided wave path response distribution of the present invention combines deep learning parameter prediction with physical constraint calculation process, thereby improving the physical interpretability of the model. Attached Figure Description
[0047] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.
[0048] Figure 1 This is a schematic diagram of a system to which the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention is applicable.
[0049] Figure 2 This is a schematic diagram of the X-direction scanning direction in the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention.
[0050] Figure 3 This is a schematic diagram of the scanning direction in the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention, specifically in the Y-direction.
[0051] Figure 4 This is a schematic diagram illustrating an example of the health status signal and the guided wave signal of the measured area in the adaptive defect quantitative identification method of ultrasonic guided wave path response distribution of the present invention.
[0052] Figure 5 This is a schematic diagram illustrating an example of the health status signal and the envelope of the guided wave signal in the measured area in the adaptive defect quantitative identification method of ultrasonic guided wave path response distribution of the present invention.
[0053] Figure 6 This is an example diagram of the damage intensity curve involved in the present invention.
[0054] Figure 7 This is a schematic diagram of a one-dimensional convolutional neural network with physical layers in the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention.
[0055] Figure 8 This is an example diagram of the X and Y direction scanning recognition results in the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention.
[0056] Figure 9 This is a flowchart illustrating the steps of the adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to the present invention.
[0057] The reference numerals in the figure are:
[0058] 1-Control and testing system; 2-Test sample;
[0059] 11-Host computer; 12-Robot arm controller; 13-Signal generator / receiver; 14-Preamplifier;
[0060] 15-Post-amplifier; 16-Robotic arm; 17-Ultrasonic probe; 18-Scanning direction;
[0061] 171 - Ultrasonic receiving sensor; 172 - Ultrasonic transmitting sensor;
[0062] 21 - Defect; 22 - Ultrasonic beam propagation path. Detailed Implementation
[0063] The inventive concept of this invention is as follows:
[0064] The purpose of this invention is to provide an adaptive quantitative identification method and system for ultrasonic guided wave path response distribution based on deep learning and differentiable physical constraints, so as to improve the positioning accuracy, geometric parameter quantitative characterization ability and model interpretability of defects of different sizes.
[0065] The present invention will now be described in detail with reference to the accompanying drawings.
[0066] An adaptive defect quantitative identification system for ultrasonic guided wave path response distribution, applicable to the ultrasonic guided wave path response distribution adaptive defect quantitative identification method of the present invention, such as... Figure 1 As shown ( Figure 1 In this system, Tx represents the ultrasonic transmitter and Rx represents the ultrasonic receiver, including: a control and testing system 1, which specifically includes:
[0067] A robotic arm 16 is provided, the front end of which is connected to an ultrasonic probe 17. The ultrasonic probe 17 is used to detect the sample 2. The ultrasonic probe 17 mainly consists of two parts: an ultrasonic receiving sensor 171 and an ultrasonic transmitting sensor 172.
[0068] The host computer 11 is connected to the robotic arm controller 12 and the signal generator / receiver 13 respectively. The robotic arm controller 12 is used to control the movement of the robotic arm 16 under the control of the host computer 11. The signal generator / receiver 13 is also connected to the preamplifier 14 and the postamplifier 15 respectively. The preamplifier 14 is also connected to the ultrasonic transmitting sensor 172, and the postamplifier 15 is also connected to the ultrasonic receiving sensor 171.
[0069] The signal generator / receiver 13 is used to generate ultrasonic signals under the control of the host computer 11 and transmit them to the preamplifier 14. After being amplified by the preamplifier 14, the signals are emitted by the ultrasonic transmitting sensor 172. The signal generator / receiver 13 is also used to receive ultrasonic signals collected by the ultrasonic receiving sensor 171 and then amplified by the postamplifier 15 under the control of the host computer 11.
[0070] In this embodiment, the test sample 2 is an aluminum plate. During the testing process, the ultrasonic probe 17 detects the defect 21 on the test sample 2 according to the scanning direction 18; the ultrasonic wave propagates between the ultrasonic transmitting sensor 172 and the ultrasonic receiving sensor 171 along the ultrasonic beam propagation path 22. Schematic diagrams of the scanning directions 18 in the X and Y directions are shown below. Figure 2 and Figure 3 As shown.
[0071] The deep learning-based adaptive defect quantitative identification method for ultrasonic guided wave path response distribution of the present invention is applicable to the aforementioned ultrasonic guided wave path response distribution adaptive defect quantitative identification system, such as... Figure 9 As shown, it includes the following steps:
[0072] Step S1: Set the detection parameters and collect the health status signal and the guided wave signal of the tested area;
[0073] The host computer 11 and the robotic arm controller 12 set the scanning direction 18 and scanning step size according to the detection task requirements, and control the robotic arm 16 to drive the ultrasonic probe 17 to complete the scan. Simultaneously, the host computer 11 controls the signal generator / receiver 13, using the preamplifier 14 and postamplifier 15 to complete the acquisition of health status signals and guided wave signals from the tested area, such as... Figure 1 As shown. The scanning direction 18 can be multiple directions; in this embodiment, the mutually orthogonal X and Y directions are selected, such as... Figure 2 and Figure 3As shown, from the top-down view of sample 2, the X direction is horizontal to the left, and the Y direction is vertically upward. The scanning step size in each scanning direction 18 can be set to 1 mm. Each step corresponds to a detection path, i.e., the ultrasonic beam propagation path 22, obtaining the detection signals of all detection paths in each direction, as shown... Figure 4 As shown, the solid black line represents an example of a healthy guided wave signal (i.e., when there is no defect), and the dashed red line represents an example of a signal received when the detection path passes through a defect (i.e., when there is a defect).
[0074] Step S2: Signal preprocessing and calculation of damage intensity curves for each path;
[0075] The acquired multipath guided wave signals are preprocessed. The purpose of this preprocessing is to suppress noise and extract energy representations reflecting differences in path response. Here, bandpass filtering for noise reduction and calculating the integral energy from the signal envelope are used as examples. An example of the signal envelope is shown below. Figure 5 As shown, the solid black line represents an example of the guided wave signal envelope in a healthy state (i.e., without defects), and the dashed red line represents an example of the signal envelope when the detection path passes through a defect. The integrated energy is calculated by intercepting the envelope signal between the start gate and the stop gate and integrating the energy (i.e., when there is a defect).
[0076] After acquiring the detection signal, based on the difference between the guided wave signal energy in the healthy state and the guided wave signal energy along each path in the tested area, the damage intensity values in each scanning direction 18 are obtained, and arranged in order of scanning distance to form a damage intensity curve, such as... Figure 6 As shown, when the detection path passes through a defect, the difference between the energy of the guided wave signal in the healthy state and the energy of the guided wave signal in each path of the tested area increases, and the damage intensity curve rises, which is the defect-affected area in the figure. The damage intensity curve is used to characterize the impact of the defect on the response of different detection paths and serves as the input to the subsequent parameter prediction network.
[0077] Step S3: Input the damage intensity curve into the parameter prediction network to predict the path response distribution model parameters for each path;
[0078] The damage intensity curves on each scanning direction 18 are input into the parameter prediction network; the parameter prediction network extracts features from the multi-directional damage intensity curves and predicts the path response distribution parameters corresponding to each effective path. Here, a one-dimensional convolutional neural network is used as an example. Figure 7As shown, the network consists of input, a parameter prediction network, output parameters, differentiable physical layers, and output. The damage intensity curve [B,2,120] represents the data input to the network, where "B" is the number of samples, "2" represents two scanning directions, and "120" represents the number of detection paths in each direction. The effective path mask [B,120] is used to mark the effective paths participating in parameter prediction. Convolutional modules 1, 2, and 3 are used to extract path features. Convolutional kernels 7, 5, and 3 represent the lengths of the one-dimensional convolutional kernels for each module. Channels (2→16), (16→32), and (32→32) represent the changes in the number of feature channels, used to progressively map the bidirectional inputs to high-dimensional path features. Batch normalization is used to stabilize network training, the activation function is used to enhance nonlinear expressive power, and random deactivation (0.1) is used to reduce the risk of overfitting. [B,16,120] and [B,32,120] represent the feature sizes for the corresponding stages, where "16" and "32" are the number of feature channels, respectively. Path features are the features extracted by the convolutional module for each detection path; effective path features are the path features filtered according to the effective path mask. Fully connected layer 1 (32→16) and fully connected layer 2 (16→2) are used to regress the path response distribution parameters. Taking a Gaussian distribution as an example, the output parameter, i.e., the prediction parameter, is the offset parameter. and standard deviation ,in, For the first There are 18 scanning directions, including the X and Y directions.
[0079] Step S4: Input the predicted parameters into the differentiable physical layer to calculate the geometric parameters of defect 21, such as length, width, and area.
[0080] The predicted path response distribution model parameters are input into the differentiable physical layer. Based on these parameters, the differentiable physical layer performs spatial mapping and multi-path fusion calculations on the responses of each detection path, including: spatial weight distribution of each path → multi-path superposition calculation → extraction of defect 21 geometric parameters, as detailed below:
[0081] No. The coordinates of the distribution center of each scanning direction 18 can be represented by formula (1).
[0082] (1)
[0083] in, Indicates the first The 18th scanning direction The original location of the path. Indicates the path number. Indicates the total number of paths. Indicates the scan direction number. Indicates the first The center offset parameter of the path, The offset correction indicates the center position of the path response distribution; based on the center position vector and the distribution width parameter vector, the path response distribution matrix is established as shown in formula (2).
[0084] (2)
[0085] in, Indicates the first The 18th scanning direction The position coordinates of a discrete point in space Indicates the number of discrete points in space. This represents the total number of discrete points in space. Indicates the first Distribution width parameters in each scanning direction 18, Indicates the cutoff radius. Indicates the first The path in the first The response distribution values at discrete points in space; This represents the step function, used to truncate the Gaussian response distribution. The value is 1 at time. The value is 0 at that time.
[0086] Based on this, the path fusion result of a single scanning direction 18 can be obtained, as shown in formula (3).
[0087] (3)
[0088] in, Indicates the first The first scan direction 18 on the Path damage intensity value; Indicates the first The scanning direction 18 in the first scan direction The fusion response value at a discrete point in space.
[0089] Weighted fusion of multiple scanning directions 18 is performed to obtain the final characterization result of defect 21, as shown in formula (4).
[0090] (4)
[0091] in, Indicates the first 18 fusion weights in each scanning direction, Indicates the first The final defect 21 characterization value at each spatial discrete point, This indicates the total number of scanning directions.
[0092] Step S5: Output the defect identification results;
[0093] After obtaining the final characterization result of defect 21, threshold segmentation is first performed on the final characterization result of defect 21 to extract high-response regions, which are then used as candidate regions for defect 21. Next, the location of defect 21 is determined based on the position of the candidate region in spatial coordinates. The length of defect 21 is calculated based on the projection range of the candidate region in the X direction, the width of defect 21 is calculated based on the projection range in the Y direction, and the area of defect 21 is calculated based on the spatial area covered by the candidate region. This achieves quantitative identification of the location and size information of defect 21.
[0094] Examples of X and Y direction scanning recognition results Figure 8 As shown in the figure, the horizontal axis X and the vertical axis Y represent the spatial coordinates of the detection area, and the color represents the spatial response intensity after multipath superposition. The high response bands in the horizontal and vertical directions correspond to the path response results in the two scanning directions, and the high response area formed by their overlap is identified as the candidate region for defect 21. The white dashed box represents the identified defect 21 region, and the length of defect 21 in the X direction, the width in the Y direction, and the area it covers are calculated accordingly.
[0095] Regarding parameter prediction networks, the above embodiments of the present invention take a one-dimensional convolutional neural network as an example. In other specific embodiments, recurrent neural networks, multi-branch feature fusion networks, or other deep learning models that can output path response distribution model parameters can also be used.
[0096] Regarding the path response distribution model, the above embodiments of the present invention take a Gaussian distribution with offset as an example. In other specific embodiments, a truncated Gaussian distribution, an elliptical distribution, or other parameterizable spatial distribution models can also be used, and the corresponding parameters can also be extended to characterization parameters such as amplitude, skewness, and truncation radius.
[0097] Regarding the differentiable physical layer, the above embodiments of the present invention integrate the path response distribution construction, multi-path fusion, and defect parameter extraction processes into the physical constraint calculation. In other specific embodiments, unidirectional fusion, multi-directional fusion, and defect 21 parameter extraction can employ other equivalent calculation methods, as long as the location and size parameters of defect 21 can be output, they are all alternative implementations of the present invention.
[0098] Furthermore, the above embodiments of the present invention are illustrated using air-coupled ultrasonic guided wave detection as an example. In other specific embodiments, any other detection method capable of achieving single-transmitter-receiver guided wave detection can be used, such as patch transducer detection, contact detection using a coupling agent, or other guided wave excitation and reception methods. Input features are not limited to damage intensity curves; energy difference curves, damage index curves, or other features capable of characterizing multi-path response differences can also be used. Scanning methods are not limited to the two orthogonal X and Y directions; they can be extended to multiple directions and other comprehensive scanning paths. The detection object is not limited to aluminum plates; it can also be used for quantitative identification of defects in other plate and shell structures.
[0099] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.
Claims
1. An adaptive quantitative identification system for ultrasonic guided wave path response distribution defects, characterized in that, include: The control and testing system (1) specifically includes: A robotic arm (16) is connected at its front end to an ultrasonic probe (17); the ultrasonic probe (17) is used to test the sample (2); The host computer (11) is connected to the robotic arm controller (12) and the signal generator / receiver (13) respectively; the signal generator / receiver (13) is also connected to the preamplifier (14) and the postamplifier (15) respectively; the preamplifier (14) and the postamplifier (15) are connected to the ultrasonic probe (17) respectively; the robotic arm controller (12) is used to control the robotic arm (16) under the control of the host computer (11); the signal generator / receiver (13) is used to generate ultrasonic signals under the control of the host computer (11) and transmit them to the preamplifier (14), and after being amplified by the preamplifier (14), they are emitted by the ultrasonic probe (17); the signal generator / receiver (13) is also used to receive the ultrasonic signals collected by the ultrasonic probe (17) and then amplified by the postamplifier (15) under the control of the host computer (11).
2. The adaptive defect quantitative identification system for ultrasonic guided wave path response distribution according to claim 1, characterized in that, The ultrasonic probe (17) includes an ultrasonic receiving sensor (171) and an ultrasonic transmitting sensor (172); the ultrasonic transmitting sensor (172) is connected to the preamplifier (14), and the ultrasonic receiving sensor (171) is connected to the postamplifier (15). Under the control of the host computer (11), the signal generator / receiver (13) generates an ultrasonic signal and transmits it to the preamplifier (14). After being amplified by the preamplifier (14), the ultrasonic signal is emitted by the ultrasonic transmitting sensor (172). Under the control of the host computer (11), the signal generator / receiver (13) receives the ultrasonic signal collected by the ultrasonic receiving sensor (171) and then amplified by the post-amplifier (15).
3. A method for adaptive quantitative identification of defects in ultrasonic guided wave path response distribution, applicable to the ultrasonic guided wave path response distribution adaptive quantitative identification system as described in claim 1 or 2, characterized in that, The method includes the following steps: Step S1: Set the detection parameters and collect the health status signal and the guided wave signal of the tested area; Step S2: Signal preprocessing and calculation of damage intensity curves for each path; Step S3: Input the damage intensity curve into the parameter prediction network to predict the path response distribution model parameters for each path; Step S4: Input the predicted path response distribution model parameters into the differentiable physical layer to calculate the geometric parameters of the defect (21); Step S5, output the defect (21) identification result.
4. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 3, characterized in that, Step S1 is as follows: The host computer (11) controls the robotic arm controller (12) to set the scanning direction (18) and scanning step size according to the requirements of the detection task. The robotic arm (16) drives the ultrasonic probe (17) to complete the scanning of the detection sample (2). The host computer (11) controls the signal generator / receiver (13), and uses the preamplifier (14) and postamplifier (15) to complete the acquisition of health status signals and guided wave signals of the measured area.
5. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 3, characterized in that, Step S2 is as follows: After acquiring the detection signal, the damage intensity value in each scanning direction (18) is obtained based on the difference between the health status signal and the guided wave signal of the tested area, and the damage intensity curve is formed by arranging them in order of scanning distance.
6. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 5, characterized in that, Step S3 is as follows: The damage intensity curves in each scanning direction (18) are input into the parameter prediction network; the parameter prediction network extracts features from the multi-directional damage intensity curves and predicts the path response distribution model parameters corresponding to each effective path.
7. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 6, characterized in that, Step S4 is as follows: The predicted path response distribution model parameters are input into the differentiable physical layer, which performs spatial mapping and multi-path fusion calculation on the response of each detection path based on the predicted path response distribution model parameters.
8. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 7, characterized in that, Step S4 is as follows: No. The coordinates of the distribution centers in each scanning direction (18) are shown below: in, Indicates the first The first scan direction (18) The original location of the path, Indicates the path number. Indicates the total number of paths. Indicates the scanning direction (18) number, Indicates the first The center offset parameter of the path, This represents the center position of the path response distribution after offset correction; based on this center position vector and the distribution width parameter vector, a path response distribution matrix is constructed as follows: in, Indicates the first The first scan direction (18) The position coordinates of a discrete point in space Indicates the number of discrete points in space. This represents the total number of discrete points in space. Indicates the first The distribution width parameter in each scanning direction (18), Indicates the cutoff radius. Indicates the first The path in the first The response distribution values at discrete points in space; This represents the step function, used to truncate the Gaussian response distribution. The value is 1 at time. The value is 0 at that time; Based on this, the path fusion result for a single scanning direction (18) is obtained, as shown below: in, Indicates the first The first scan direction (18) on the Path damage intensity value; Indicates the first The scanning direction (18) in the first scanning direction (18) Fusion response values at discrete points in space; Weighted fusion of multiple scanning directions (18) yields the final defect characterization result (21), as shown below: in, Indicates the first The fusion weights for each scanning direction (18) are calculated. Indicates the first The final defect (21) characterization value at each spatial discrete point, This indicates the total number of scanning directions.
9. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 3, characterized in that, In step S4, the geometric parameters of the defect (21) include: the length of the defect (21), the width of the defect (21), and the area of the defect (21).
10. The adaptive defect quantitative identification method for ultrasonic guided wave path response distribution according to claim 9, characterized in that, Step S5 is as follows: Threshold segmentation is performed on the final defect (21) characterization results to extract high response regions, and these regions are used as candidate regions for defects (21); The position of the defect (21) is determined based on the position of the candidate region of the defect (21) in the spatial coordinates. The length of the defect (21) is calculated based on the projection range of the candidate region of the defect (21) in a certain direction. The width of the defect (21) is calculated based on the projection range in another direction perpendicular to the aforementioned direction. The area of the defect (21) is calculated based on the spatial area covered by the candidate region of the defect (21).