A method for laser ultrasonic imaging detection of metal composite plate interface based on PINN

Abstract

This invention discloses a laser ultrasonic imaging detection method and apparatus for metal composite plate interfaces based on PINN, belonging to the field of laser ultrasonic nondestructive testing technology. The method includes: extracting the first arrival travel time data of ultrasonic P-waves from the original time-domain ultrasonic signal data based on variational mode decomposition, and constructing a training dataset by combining the spatial coordinate information of the transmitting and receiving points; calculating based on a multi-component loss function using preset velocity structure data, travel time factor prediction values, velocity structure prediction values, travel time prediction values, travel time background values, sampling point spatial coordinate information, and the training dataset; performing inversion prediction using an optimized laser ultrasonic imaging network based on global spatial coordinate information; and extracting the interface contour based on the global velocity structure prediction values ​​using a velocity step feature detection method to obtain laser ultrasonic imaging. This invention is a laser ultrasonic imaging method that combines physical mechanism constraints and data-driven capabilities, adapts to sparse measurement points, and does not require prior velocity structure information.

Description

Technical Field

[0001] This invention relates to the field of laser ultrasonic nondestructive testing technology, and in particular to a laser ultrasonic imaging testing method and device for metal composite plate interfaces based on PINN. Background Technology

[0002] Laser ultrasound technology, as a core method of non-contact non-destructive testing, is widely used for the characterization of the internal structure and defect detection of various materials due to its advantages of not requiring coupling agents, adaptability to complex structures, and high detection accuracy. The closest existing technologies to this invention are traditional laser ultrasound imaging detection methods and pure data-driven neural network imaging methods.

[0003] Traditional laser ultrasonic imaging typically relies on numerical solutions to wave equations, such as the finite-difference time-domain method or the spectral element method, or ray tracing methods. These methods require a dense array of laser excitation and detection points on the surface of the component under test to acquire sufficient spatiotemporal wavefield information. Furthermore, these methods heavily depend on prior knowledge of the sound velocity distribution structure within the material, necessitating the establishment of an accurate velocity model as a physical constraint for imaging inversion. However, in industrial field testing scenarios, the material properties, internal structure, and stress state of the component under test are often unknown or difficult to accurately calibrate. This lack of prior knowledge of the velocity structure leads to severely malformed imaging inversion problems, significantly reduced imaging accuracy, and even the appearance of structural distortions and artifacts.

[0004] To reduce computational complexity, some methods use procedural equations to approximate the high-frequency propagation behavior of ultrasound, obtaining the travel time field by solving the procedural equations and thus achieving imaging. However, procedural equations are nonlinear partial differential equations, and their numerical solutions face problems such as large computational load, slow convergence speed, and difficulty in multi-path travel time calculation in complex geometric boundaries and non-uniform media, making it difficult to meet the real-time requirements of detection efficiency in industrial settings.

[0005] In recent years, deep learning-based end-to-end imaging methods have been introduced into the field of laser ultrasonic testing. These methods train neural networks with a large number of labeled samples to establish a mapping relationship between observation data and imaging results. However, such purely data-driven models lack physical mechanism constraints, and their generalization performance is highly dependent on the distribution coverage and labeling quality of the training data. Under conditions of sparse measurement points and limited samples in industrial settings, the models are prone to overfitting to noise, generating false imaging features, leading to the risk of missed and false detections, and making it difficult to meet the requirements of high-reliability non-destructive testing.

[0006] In the existing technology, there is a lack of a laser ultrasound imaging method that combines physical mechanism constraints and data-driven capabilities, is adaptable to sparse measurement points, and does not require velocity structure priors. Summary of the Invention

[0007] To address the technical problems of existing technologies, such as reliance on dense excitation measurement points, high dependence on prior knowledge of material velocity and structure, low imaging accuracy in scenarios with unknown velocity and structure, low efficiency of traditional numerical solutions to functional equations, and poor robustness and susceptibility to generating false imaging features in purely data-driven neural network models under sparse data, this invention provides a laser ultrasonic imaging detection method and device for metal composite plate interfaces based on PINN. The technical solution is as follows: On the one hand, a laser ultrasonic imaging inspection method for metal composite plate interfaces based on PINN is provided. This method is implemented by a laser ultrasonic imaging inspection device for metal composite plate interfaces, and includes: Based on laser ultrasonic testing equipment, information is collected from the metal composite plate to be tested to obtain the original time-domain ultrasonic signal data and the spatial coordinate information of the transmitting and receiving points. Based on the variational mode decomposition method, the first arrival travel time data of ultrasonic longitudinal wave is extracted from the original time-domain ultrasonic signal data, and a training dataset is constructed by combining the spatial coordinate information of the transmitter and receiver points. Based on the training dataset, forward inference is performed using a laser ultrasound imaging network to obtain predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction, and spatial coordinates of sampling points. Based on the multi-component loss function, the training loss is calculated according to the preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information and training dataset. Based on the training loss, the parameters of the laser ultrasound imaging network are optimized to obtain an optimized laser ultrasound imaging network. The system acquires the global spatial coordinates of the metal composite plate to be inspected; based on the global spatial coordinates, it uses an optimized laser ultrasonic imaging network to perform inversion prediction and obtain the global velocity structure prediction value; based on the velocity step feature detection method, it extracts the interface contour based on the global velocity structure prediction value and obtains the laser ultrasonic image.

[0008] On the other hand, a PINN-based laser ultrasonic imaging detection device for metal composite plate interfaces is provided. This device is applied to the PINN-based laser ultrasonic imaging detection method for metal composite plate interfaces. The device includes: The signal acquisition module is used to acquire information from the metal composite plate to be inspected based on the laser ultrasonic testing equipment, and obtain the original time-domain ultrasonic signal data and the spatial coordinate information of the transmitting and receiving points. The signal processing module is used to extract the first arrival travel time data of ultrasonic P-waves based on the variational mode decomposition method and the original time-domain ultrasonic signal data, and to construct a training dataset by combining the spatial coordinate information of the transmitter and receiver points. The network training module is used to perform forward inference using the laser ultrasound imaging network based on the training dataset to obtain the predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction, and spatial coordinate information of sampling points. The loss calculation module is used to calculate the training loss based on a multi-component loss function, according to preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information, and training dataset. The network optimization module is used to optimize the parameters of the laser ultrasound imaging network based on the training loss, so as to obtain an optimized laser ultrasound imaging network. The imaging inversion module is used to obtain the global spatial coordinate information of the metal composite plate to be inspected; based on the global spatial coordinate information, an optimized laser ultrasonic imaging network is used for inversion prediction to obtain the global velocity structure prediction value; based on the velocity step feature detection method, the interface contour is extracted according to the global velocity structure prediction value to obtain laser ultrasonic imaging.

[0009] On the other hand, a laser ultrasonic imaging detection device for metal composite plate interfaces is provided, the metal composite plate interface laser ultrasonic imaging detection device comprising: a processor; a memory, the memory storing computer-readable instructions, which, when executed by the processor, implement any of the methods in the above-described PINN-based metal composite plate interface laser ultrasonic imaging detection method.

[0010] On the other hand, a computer-readable storage medium is provided, wherein at least one instruction is stored therein, the at least one instruction being loaded and executed by a processor to implement any of the methods in the above-described PINN-based laser ultrasonic imaging detection method for metal composite plate interfaces.

[0011] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following: This invention proposes a laser ultrasonic imaging detection method for metal composite plate interfaces based on PINN. By combining physical information neural networks with laser ultrasound, it can complete velocity structure inversion and interface imaging with only sparse excitation points and a small amount of known velocity structure information. It does not require dense arrangement of excitation points, which greatly reduces the amount of data collection, lowers the detection cost, and improves the detection efficiency. The initial velocity structure at the start of the inversion is generated by random parameter initialization, without the need for prior knowledge of the approximate distribution of the velocity structure. It can still achieve high-precision imaging in scenarios where the velocity structure distribution is unknown, thus solving the problem of limited imaging accuracy of existing methods under unknown velocity structures. Based on a multi-component loss function, combined with data-driven and physical constraints, and by introducing a gradient regularization term to constrain the piecewise constant characteristics of the velocity model, this method can accurately reconstruct various complex interface structures, such as smooth curve interfaces and wavy parametric curve interfaces. This makes the method applicable to composite plates with various metal combinations, suitable for applications in aerospace, automotive, and energy fields. Furthermore, the imaging system can achieve integrated operation, demonstrating strong practicality and ease of engineering implementation. This invention is a laser ultrasonic imaging method that combines physical mechanism constraints with data-driven capabilities, adapts to sparse measurement points, and requires no prior knowledge of velocity structures. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a laser ultrasonic imaging detection method for the interface of a metal composite plate based on PINN, provided by an embodiment of the present invention. Figure 2 This is a block diagram of a PINN-based laser ultrasonic imaging detection device for metal composite plate interfaces provided in an embodiment of the present invention. Figure 3 This is a schematic diagram of the structure of a laser ultrasonic imaging detection device for the interface of a metal composite plate provided in an embodiment of the present invention. Detailed Implementation

[0014] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0015] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0016] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0017] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0018] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0019] This invention provides a PINN-based laser ultrasonic imaging method for detecting the interface of metal composite plates. This method can be implemented using a laser ultrasonic imaging device for the interface of metal composite plates, which can be a terminal or a server. Figure 1 The flowchart shown is for a laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN. The processing flow of this method may include the following steps: S1. Based on the laser ultrasonic testing equipment, information is collected from the metal composite plate to be tested to obtain the original time-domain ultrasonic signal data and the spatial coordinate information of the transmitting and receiving points. In one feasible implementation, the present invention arranges laser excitation points on the upper surface of the metal composite plate to be tested, and laser detection points on the lower surface. Ultrasonic signals are acquired using a multi-point excitation and single-point detection method: the detection point position is fixed, and the excitation points are sequentially moved to different positions for excitation, recording the ultrasonic signal corresponding to each excitation point. Utilizing the reversibility of the sound wave propagation path, this acquisition method is equivalent to single-point excitation and multi-point detection, ensuring the integrity of the acquired data.

[0020] The metal composite plate to be tested is a copper-aluminum bimetallic composite plate, 18 mm in length and 2.8 mm in thickness, with copper as the upper layer and aluminum as the lower layer. Excitation points are arranged on the upper surface of the copper side, and detection points are arranged on the lower surface of the aluminum side. The excitation points are evenly spaced along the length, with a spacing of 2 mm between adjacent excitation points, for a total of 10 excitation points. The detection points are aligned with each excitation point, with a horizontal offset ranging from -2 mm to +2 mm and an offset step of 0.4 mm. The original time-domain ultrasonic signal corresponding to each excitation point is acquired using a laser ultrasonic testing system to obtain the raw signal data required for imaging.

[0021] The dimensions and spacing described above are for illustrative purposes only and can be adjusted according to testing requirements and material specifications in actual applications. Multi-point excitation and single-point detection methods can also be replaced with single-point excitation and multi-point detection methods; both methods are equivalent in terms of data integrity.

[0022] S2. Based on the variational mode decomposition method, the ultrasonic longitudinal wave first arrival travel time data is extracted from the original time-domain ultrasonic signal data, and a training dataset is constructed by combining the spatial coordinate information of the transmitter and receiver points. Optionally, based on the variational mode decomposition method, the first arrival travel time data of the ultrasonic P-wave is extracted from the original time-domain ultrasonic signal data, and a training dataset is constructed by combining the spatial coordinate information of the transmitter and receiver points, including: Based on the variational mode decomposition method, multi-band adaptive decomposition is performed on the original time-domain ultrasonic signal data to obtain multiple band-limited intrinsic mode component signal data. Time-frequency joint analysis was performed on multiple band-limited intrinsic mode component signal data, and the components containing the complete effective features of ultrasonic longitudinal wave were selected for superposition and reconstruction to obtain denoised ultrasonic signal data. The ultrasonic P-wave first arrival travel time data are extracted from the denoised ultrasonic signal data; the dataset is constructed based on the ultrasonic P-wave first arrival travel time data and the spatial coordinate information of the transmitting and receiving points to obtain the training dataset.

[0023] In one feasible implementation, based on a constrained variational model, the original ultrasound signal is decomposed into k band-limited intrinsic mode functions (IMFs), each IMF component is distributed around its center frequency, and the sum of all IMF components is consistent with the original signal; the variational optimization problem is solved iteratively to make the sum of the estimated bandwidths of all IMF components converge to the minimum value, and the optimal number of decomposed modes k is adaptively determined; A joint time-domain and frequency-domain analysis was performed on each IMF component, and the component containing the complete effective features of the ultrasonic P-wave was selected to reconstruct the denoised signal. The first arrival travel time data of the P-wave was extracted from the denoised signal, and a training dataset was constructed by combining the spatial coordinate information of the excitation point and the detection point.

[0024] S3. Based on the training dataset, use the laser ultrasound imaging network to perform forward inference to obtain the predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction and sampling point spatial coordinate information. Among them, the laser ultrasound imaging network is a physical information neural network with a dual neural network architecture; Laser ultrasound imaging networks include a travel time factor prediction network and a velocity structure prediction network; The travel time factor prediction network contains 10 first hidden layers; each first hidden layer contains 64 neurons. The structural prediction network contains 10 second hidden layers; each second hidden layer contains 32 neurons.

[0025] In one feasible implementation, the present invention is based on the PyTorch framework to build a dual neural network architecture of Physics-Informed Neural Network (PINN).

[0026] Two neural networks independently calculate the travel time factor τ(x) and velocity structure v(x) at each spatial point of the composite plate. The travel time factor prediction network contains four variables (x, y, x). s ,y s The first network has 10 hidden layers with 64 neurons each; the second network has 10 hidden layers with 32 neurons each. The third network has 10 hidden layers with 64 neurons each. The fourth network has 10 hidden layers with 32 neurons each. Where x... s y s This refers to the coordinates of the excitation point, where x s The horizontal coordinates of the laser ultrasonic excitation emission point are y. s The vertical depth coordinates of the laser ultrasonic excitation emission point; x and y refer to the coordinates of the imaging point within the specimen; where x is the horizontal coordinate of any imaging point within the specimen under test, and y is the vertical depth coordinate of any imaging point within the specimen under test.

[0027] Both the travel time factor prediction network and the velocity structure prediction network use Layer-Adaptive ELU as the activation function in their hidden layers. Each hidden layer is assigned an independent learnable parameter 'a', which is then activated via elu(a). h) Calculate the activation value, where h is the layer input and elu is the standard exponential linear unit function; the output layers of both the travel time factor prediction network and the velocity structure prediction network adopt the Sigmoid activation function and are multiplied by a preset output scaling factor to constrain the predicted values ​​of the travel time factor and ultrasonic propagation velocity to a preset reasonable range.

[0028] Optionally, based on the training dataset, a laser ultrasound imaging network is used for forward inference to obtain predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction, and spatial coordinates of sampling points, including: Randomly select sampling points in the computational domain based on the training dataset to obtain the spatial coordinate information of the sampling points; Based on the preset wave velocity, the propagation time from the excitation point to the receiving point is calculated according to the spatial coordinate information of the transmitting and receiving points in the training dataset, and the travel time background value is obtained. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed according to the spatial coordinate information of the transmitting and receiving points in the training dataset to obtain the predicted value of the travel time factor of the ultrasonic longitudinal wave. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed to obtain the predicted value of the velocity structure. Based on the predicted value of the travel time factor, the travel time prediction value is obtained by weighting the travel time background value.

[0029] In one feasible implementation, spatial coordinate points are randomly generated according to a uniform distribution or an adaptive density distribution, and the spatial coordinate information of the sampling points and the spatial coordinate information of the transmitting and receiving points together constitute the network input coordinate set. Based on the preset wave velocity, the propagation time from the excitation point to the receiving point is calculated according to the spatial coordinate information of the transmitting and receiving points in the training dataset; based on the spatial coordinate information of the sampling point, a nonlinear mapping is performed according to the spatial coordinate information of the transmitting and receiving points in the training dataset, and the predicted value of the ultrasonic longitudinal wave travel time factor at the sampling point is output. The predicted value of the travel time factor represents the degree of deviation of the actual travel time from the background travel time. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed to output the velocity structure prediction value at the sampling points; based on the travel time factor prediction value, a weighted calculation is performed according to the travel time background value, and the travel time factor prediction value and the travel time background value are multiplied point by point to obtain the travel time prediction value.

[0030] For those located at (x s ,y s The laser excitation point is determined by pre-obtaining the preset wave velocity v of the medium at that point. s Then the background travel time of any point (x,y) within the specimen T 0 (x,y) is calculated as follows (1): (1); S4. Based on the multi-component loss function, the training loss is calculated according to the preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information and training dataset. Among them, the multi-component loss function uses the decomposed form of the equation as the physical constraint of the ultrasonic longitudinal wave; the multi-component loss is used to make the training of the laser ultrasonic imaging network both data-driven and supervised by physical laws.

[0031] In one feasible implementation, the ultrasonic longitudinal wave propagates in an isotropic metal composite plate, satisfying the original equation (2) as follows: (2); in, T (x,y) represents the travel time of the ultrasonic longitudinal wave at the corresponding spatial point, and v(x,y) represents the ultrasonic propagation velocity at the corresponding spatial point. For gradient operators; to avoid the singularity problem of the laser point source position, the travel time value is... T (x,y) is decomposed into background travel time T 0 The product of (x,y) and the travel time factor τ(x,y), where (x,y) are the same as x and y in the previous text, referring to the coordinates of the imaging point within the specimen. That is, the ultrasonic longitudinal wave physical constraint is a decomposed functional equation as shown in equation (3): (3); The physical constraints of ultrasonic longitudinal waves are used to constrain the physical rationality of the output of the travel time factor prediction network in the physical information neural network, and to ensure that the velocity structure inversion results conform to the propagation law of ultrasonic longitudinal waves in isotropic media.

[0032] Optionally, based on a multi-component loss function, the training loss is calculated using preset velocity structure data, travel time factor prediction values, velocity structure prediction values, travel time prediction values, travel time background values, sampling point spatial coordinate information, and the training dataset. This training loss includes: The mean squared error is calculated based on the travel time predictions and the travel time observations in the training dataset to obtain the data error loss. Based on the physical constraints of ultrasonic longitudinal waves, the difference of the equation is calculated according to the predicted values ​​of travel time factor, velocity structure, travel time background and sampling point spatial coordinates to obtain the physical error loss. The boundary condition deviation penalty is calculated based on the travel time factor prediction value and the spatial coordinate information of the excitation point in the training dataset to obtain the boundary condition loss; The mean square error at spatial coordinates is calculated based on the predicted velocity structure value and the preset velocity structure data to obtain the velocity structure constraint loss. Based on the predicted velocity structure values ​​and the spatial coordinate information of the sampling points, the gradient penalty of the velocity field is calculated to obtain the gradient regularization loss. The training loss is calculated by weighting the data error loss, physical error loss, boundary condition loss, velocity structure constraint loss, and gradient regularization loss.

[0033] In one feasible implementation, the multi-component loss function is composed of the data error loss function. L data Physical error loss function L PDE Boundary condition loss function L boundary Velocity-structure constraint loss function L know and gradient regularization loss function L grad The weighted sum of the five parts is expressed as follows (4): (4); Among them, α, β, γ, δ and w are adjustable weight parameters used to balance the contributions of each loss term.

[0034] The data error loss function uses the mean squared error to calculate the error between the network prediction time and the experimental observation time, ensuring the consistency between the model prediction and the experimental data. The loss function is as follows (5): (5); in, N s The total number of excitation points; N r This represents the total number of receiving points. These are the predicted travel times. These are travel time observations; x n,i , y n,i For the first n The excitation point, the first i The horizontal and vertical depth coordinates of the specimen imaging point corresponding to each receiving point; subscript n Corresponding excitation point number, subscript i Corresponding receiving point number.

[0035] The physical error loss function is used to constrain the travel time factor to satisfy the decomposed form of the equation, ensuring the physical rationality of the prediction results. The loss function is as follows (6): (6); in, N t The number of sampling points in the computational domain; Eik( ) This is a function for calculating the residuals of the functional equation, used to output the physical constraint residual values ​​of the functional equation at the coordinate points.

[0036] The boundary condition loss function forces the travel factor τ(x) of the laser excitation point position to be 1, ensuring that the output at the excitation point conforms to physical consistency. This loss function is as follows (7): (7); in, x n , y n The first n The horizontal coordinates and vertical depth coordinates of each laser ultrasonic excitation point, with the subscript n representing the excitation point number; The velocity structure constraint loss function uses the one-dimensional velocity profile at the center of the excitation point and the receiver point as a constraint to improve the velocity inversion accuracy. The loss function is as follows (8): (8); in, N L The number of training points with a known velocity structure; x l For the first l The horizontal coordinates of a known velocity sampling point y l For the firstl Vertical depth coordinates of a known velocity sampling point; For speed networks in ( x l ,y l The output velocity structure prediction value at () is shown. for( x l ,y l The known true velocity structure at ().

[0037] The gradient regularization loss function uses the L1 norm to penalize the velocity gradient, adapting to the step velocity change at the composite plate interface. The loss function is as follows (9): (9); in, x v i and y v i These are the gradients of the velocity in the x and y directions, respectively.

[0038] S5. Based on the training loss, optimize the parameters of the laser ultrasound imaging network to obtain an optimized laser ultrasound imaging network. In one feasible implementation, the present invention uses the Adam optimization algorithm to iteratively update and optimize the parameters of the neural network. During the iteration process, the convergence of the total loss function is monitored in real time until the loss function value converges to a preset threshold and all error terms (data error terms, physical error terms, etc.) meet the preset accuracy requirements, at which point the iteration stops.

[0039] The learning rate was set to 0.001, the batch size to 1024, and the maximum number of iterations to 1000. This optimization strategy effectively prevents overfitting and improves the model's generalization performance. After iterative optimization, the ultrasonic propagation velocity at each coordinate point in the global space of the metal composite plate can be retrieved by using the output layer of the velocity structure prediction network in the neural network. By integrating the velocity data from all spatial points, a complete velocity structure distribution result is obtained.

[0040] S6. Obtain the global spatial coordinate information of the metal composite plate to be detected; based on the global spatial coordinate information, use an optimized laser ultrasonic imaging network to perform inversion prediction and obtain the global velocity structure prediction value; based on the velocity step feature detection method, extract the interface contour according to the global velocity structure prediction value to obtain laser ultrasonic imaging.

[0041] In one feasible implementation, the global spatial coordinate grid of the area to be detected is input into a trained velocity structure prediction network, which outputs the inverted global velocity distribution. Based on the velocity step features, the interface contour of the metal composite plate is extracted to complete laser ultrasonic imaging.

[0042] In one feasible implementation, to verify the imaging accuracy and reliability of the method of the present invention, the imaging results are quantitatively evaluated. Experimental results show that the maximum reconstruction error of the composite plate interface is 0.063 mm, the average relative error is 1.7%, the coefficient of determination R² between the inverted interface and the real interface is 0.992183, indicating excellent fitting. The overall structural similarity index (SSIM) between the inverted velocity structure and the real structure is 0.88, demonstrating high structural consistency and accurately reflecting the true internal structure of the composite plate.

[0043] Therefore, the laser ultrasonic imaging detection method for metal composite plate interfaces based on physical information neural networks proposed in this invention can complete the imaging of material structures without the need for preset velocity priors and using only sparse measurement points. Through laser ultrasonic signal acquisition, variational mode decomposition for noise reduction, and a dual-branch neural network with embedded process function equation constraints, velocity inversion and interface reconstruction are achieved. The detection process is non-contact and non-destructive, adaptable to complex corrugated interfaces, and provides an efficient and reliable technical solution for non-destructive testing and quality evaluation of metal composite plates.

[0044] This invention proposes a laser ultrasonic imaging detection method for metal composite plate interfaces based on PINN. By combining physical information neural networks with laser ultrasound, it can complete velocity structure inversion and interface imaging with only sparse excitation points and a small amount of known velocity structure information. It does not require dense arrangement of excitation points, which greatly reduces the amount of data collection, lowers the detection cost, and improves the detection efficiency. The initial velocity structure at the start of the inversion is generated by random parameter initialization, without the need for prior knowledge of the approximate distribution of the velocity structure. It can still achieve high-precision imaging in scenarios where the velocity structure distribution is unknown, thus solving the problem of limited imaging accuracy of existing methods under unknown velocity structures. Based on a multi-component loss function, combined with data-driven and physical constraints, and by introducing a gradient regularization term to constrain the piecewise constant characteristics of the velocity model, this method can accurately reconstruct various complex interface structures, such as smooth curve interfaces and wavy parametric curve interfaces. This makes the method applicable to composite plates with various metal combinations, suitable for applications in aerospace, automotive, and energy fields. Furthermore, the imaging system can achieve integrated operation, demonstrating strong practicality and ease of engineering implementation. This invention is a laser ultrasonic imaging method that combines physical mechanism constraints with data-driven capabilities, adapts to sparse measurement points, and requires no prior knowledge of velocity structures.

[0045] Figure 2This is a block diagram of a PINN-based laser ultrasonic imaging detection device for metal composite plate interfaces, provided in an embodiment of the present invention. This device is used in a PINN-based laser ultrasonic imaging detection method for metal composite plate interfaces. (Refer to...) Figure 2 The device includes a signal acquisition module 210, a signal processing module 220, a network training module 230, a loss calculation module 240, a network optimization module 250, and an imaging inversion module 260. Among them: Signal acquisition module 210 is used to acquire information of the metal composite plate to be inspected based on laser ultrasonic testing equipment, and obtain raw time-domain ultrasonic signal data and spatial coordinate information of the transmitting and receiving points; Signal processing module 220 is used to extract the first arrival travel time data of ultrasonic longitudinal wave based on the original time-domain ultrasonic signal data using the variational mode decomposition method, and to construct a training dataset by combining the spatial coordinate information of the transmitting and receiving points. The network training module 230 is used to perform forward inference using the laser ultrasound imaging network based on the training dataset to obtain the travel time factor prediction value, velocity structure prediction value, travel time background value, travel time prediction value and sampling point spatial coordinate information. The loss calculation module 240 is used to calculate the training loss based on a multi-component loss function, according to preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information and training dataset. The network optimization module 250 is used to optimize the parameters of the laser ultrasound imaging network based on the training loss to obtain an optimized laser ultrasound imaging network. The imaging inversion module 260 is used to acquire the global spatial coordinate information of the metal composite plate to be detected; based on the global spatial coordinate information, an optimized laser ultrasonic imaging network is used for inversion prediction to obtain the global velocity structure prediction value; based on the velocity step feature detection method, the interface contour is extracted according to the global velocity structure prediction value to obtain laser ultrasonic imaging.

[0046] Optionally, the signal processing module 220 is further configured to: Based on the variational mode decomposition method, multi-band adaptive decomposition is performed on the original time-domain ultrasonic signal data to obtain multiple band-limited intrinsic mode component signal data. Time-frequency joint analysis was performed on multiple band-limited intrinsic mode component signal data, and the components containing the complete effective features of ultrasonic longitudinal wave were selected for superposition and reconstruction to obtain denoised ultrasonic signal data. The ultrasonic P-wave first arrival travel time data are extracted from the denoised ultrasonic signal data; the dataset is constructed based on the ultrasonic P-wave first arrival travel time data and the spatial coordinate information of the transmitting and receiving points to obtain the training dataset.

[0047] Among them, the laser ultrasound imaging network is a physical information neural network with a dual neural network architecture; Laser ultrasound imaging networks include a travel time factor prediction network and a velocity structure prediction network; The travel time factor prediction network contains 10 first hidden layers; each first hidden layer contains 64 neurons. The structural prediction network contains 10 second hidden layers; each second hidden layer contains 32 neurons.

[0048] Optionally, the network training module 230 is further used for: Randomly select sampling points in the computational domain based on the training dataset to obtain the spatial coordinate information of the sampling points; Based on the preset wave velocity, the propagation time from the excitation point to the receiving point is calculated according to the spatial coordinate information of the transmitting and receiving points in the training dataset, and the travel time background value is obtained. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed according to the spatial coordinate information of the transmitting and receiving points in the training dataset to obtain the predicted value of the travel time factor of the ultrasonic longitudinal wave. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed to obtain the predicted value of the velocity structure. Based on the predicted value of the travel time factor, the travel time prediction value is obtained by weighting the travel time background value.

[0049] Among them, the multi-component loss function uses the decomposed form of the equation as the physical constraint of the ultrasonic longitudinal wave; the multi-component loss is used to make the training of the laser ultrasonic imaging network both data-driven and supervised by physical laws.

[0050] Optionally, the loss calculation module 240 is further used for: The mean squared error is calculated based on the travel time predictions and the travel time observations in the training dataset to obtain the data error loss. Based on the physical constraints of ultrasonic longitudinal waves, the difference of the equation is calculated according to the predicted values ​​of travel time factor, velocity structure, travel time background and sampling point spatial coordinates to obtain the physical error loss. The boundary condition deviation penalty is calculated based on the travel time factor prediction value and the spatial coordinate information of the excitation point in the training dataset to obtain the boundary condition loss; The mean square error at spatial coordinates is calculated based on the predicted velocity structure value and the preset velocity structure data to obtain the velocity structure constraint loss. Based on the predicted velocity structure values ​​and the spatial coordinate information of the sampling points, the gradient penalty of the velocity field is calculated to obtain the gradient regularization loss. The training loss is calculated by weighting the data error loss, physical error loss, boundary condition loss, velocity structure constraint loss, and gradient regularization loss.

[0051] This invention proposes a laser ultrasonic imaging detection method for metal composite plate interfaces based on PINN. By combining physical information neural networks with laser ultrasound, it can complete velocity structure inversion and interface imaging with only sparse excitation points and a small amount of known velocity structure information. It does not require dense arrangement of excitation points, which greatly reduces the amount of data collection, lowers the detection cost, and improves the detection efficiency. The initial velocity structure at the start of the inversion is generated by random parameter initialization, without the need for prior knowledge of the approximate distribution of the velocity structure. It can still achieve high-precision imaging in scenarios where the velocity structure distribution is unknown, thus solving the problem of limited imaging accuracy of existing methods under unknown velocity structures. Based on a multi-component loss function, combined with data-driven and physical constraints, and by introducing a gradient regularization term to constrain the piecewise constant characteristics of the velocity model, this method can accurately reconstruct various complex interface structures, such as smooth curve interfaces and wavy parametric curve interfaces. This makes the method applicable to composite plates with various metal combinations, suitable for applications in aerospace, automotive, and energy fields. Furthermore, the imaging system can achieve integrated operation, demonstrating strong practicality and ease of engineering implementation. This invention is a laser ultrasonic imaging method that combines physical mechanism constraints with data-driven capabilities, adapts to sparse measurement points, and requires no prior knowledge of velocity structures.

[0052] Figure 3 This is a schematic diagram of the structure of a laser ultrasonic imaging detection device for the interface of a metal composite plate provided in an embodiment of the present invention, as shown below. Figure 3 As shown, the laser ultrasonic imaging inspection equipment for metal composite plate interfaces may include the above-mentioned Figure 2 The illustrated device is a PINN-based laser ultrasonic imaging inspection apparatus for metal composite plate interfaces. Optionally, the metal composite plate interface laser ultrasonic imaging inspection apparatus 310 may include a first processor 2001.

[0053] Optionally, the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces may also include a memory 2002 and a transceiver 2003.

[0054] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0055] The following is combined Figure 3 A detailed description of each component of the 310 laser ultrasonic imaging inspection device for metal composite plate interfaces is provided below: The first processor 2001 is the control center of the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0056] Optionally, the first processor 2001 can perform various functions of the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0057] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 3 CPU0 and CPU1 are shown in the diagram.

[0058] In a specific implementation, as one example, the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces may also include multiple processors, for example... Figure 3 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0059] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0060] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the metal composite plate interface laser ultrasonic imaging detection device 310. Figure 3 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0061] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0062] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 3 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0063] Optionally, the transceiver 2003 can be integrated with the first processor 2001, or it can exist independently and be connected to the interface circuit of the metal composite plate interface laser ultrasonic imaging detection device 310. Figure 3 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0064] It should be noted that, Figure 3 The structure of the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces shown in the figure does not constitute a limitation on the router. Actual laser ultrasonic imaging inspection devices for metal composite plate interfaces may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0065] Furthermore, the technical effects of the laser ultrasonic imaging inspection device 310 for metal composite plate interfaces can be referred to the technical effects of the PINN-based laser ultrasonic imaging inspection method for metal composite plate interfaces described in the above method embodiments, and will not be repeated here.

[0066] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor, or it may be any conventional processor, etc.

[0067] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DRRAM).

[0068] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0069] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0070] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0071] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0072] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0073] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0074] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0075] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0076] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0077] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A laser ultrasonic imaging method for detecting the interface of a metal composite plate based on PINN, characterized in that, The method includes: Based on laser ultrasonic testing equipment, information is collected from the metal composite plate to be tested to obtain the original time-domain ultrasonic signal data and the spatial coordinate information of the transceiver point. Based on the variational mode decomposition method, the first arrival travel time data of ultrasonic P-wave is extracted from the original time-domain ultrasonic signal data, and a training dataset is constructed by combining the spatial coordinate information of the transmitter and receiver points. Based on the training dataset, forward inference is performed using a laser ultrasound imaging network to obtain predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction, and spatial coordinates of sampling points. Based on the multi-component loss function, the training loss is calculated according to the preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information and training dataset. Based on the training loss, the parameters of the laser ultrasound imaging network are optimized to obtain an optimized laser ultrasound imaging network. The system acquires the global spatial coordinates of the metal composite plate to be inspected; based on the global spatial coordinates, it uses an optimized laser ultrasonic imaging network to perform inversion prediction and obtain the global velocity structure prediction value; based on the velocity step feature detection method, it extracts the interface contour based on the global velocity structure prediction value and obtains the laser ultrasonic image.

2. The laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN according to claim 1, characterized in that, The variational mode decomposition-based method extracts the first arrival travel time of the ultrasonic P-wave from the original time-domain ultrasonic signal data, and constructs a training dataset by combining the spatial coordinate information of the transmitter and receiver points, including: Based on the variational mode decomposition method, multi-band adaptive decomposition is performed on the original time-domain ultrasonic signal data to obtain multiple band-limited intrinsic mode component signal data. Time-frequency joint analysis was performed on multiple band-limited intrinsic mode component signal data, and the components containing the complete effective features of ultrasonic longitudinal wave were selected for superposition and reconstruction to obtain denoised ultrasonic signal data. The ultrasonic P-wave first arrival travel time data are extracted from the denoised ultrasonic signal data; the dataset is constructed based on the ultrasonic P-wave first arrival travel time data and the spatial coordinate information of the transmitting and receiving points to obtain the training dataset.

3. The laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN according to claim 1, characterized in that, The laser ultrasound imaging network is a physical information neural network with a dual neural network architecture; The laser ultrasound imaging network includes a travel time factor prediction network and a velocity structure prediction network. The travel time factor prediction network contains 10 first hidden layers; each first hidden layer contains 64 neurons. The structure prediction network contains 10 second hidden layers; each second hidden layer contains 32 neurons.

4. The laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN according to claim 1, characterized in that, The process involves using a laser ultrasound imaging network to perform forward inference based on the training dataset to obtain predicted values ​​for travel time factors, velocity structure, travel time background, travel time predictions, and spatial coordinates of sampling points, including: Randomly select sampling points in the computational domain based on the training dataset to obtain the spatial coordinate information of the sampling points; Based on the preset wave velocity, the propagation time from the excitation point to the receiving point is calculated according to the spatial coordinate information of the transmitting and receiving points in the training dataset, and the travel time background value is obtained. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed according to the spatial coordinate information of the transmitting and receiving points in the training dataset to obtain the predicted value of the travel time factor of the ultrasonic longitudinal wave. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed to obtain the predicted value of the velocity structure. Based on the predicted value of the travel time factor, the travel time prediction value is obtained by weighting the travel time background value.

5. The laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN according to claim 1, characterized in that, The multi-component loss function uses a decomposed form of the equation as a physical constraint for ultrasonic longitudinal waves; the multi-component loss is used to make the training of the laser ultrasound imaging network both data-driven and supervised by physical laws.

6. The laser ultrasonic imaging detection method for the interface of metal composite plates based on PINN according to claim 1, characterized in that, The training loss, calculated based on the multi-component loss function, is derived from preset velocity structure data, travel time factor prediction values, velocity structure prediction values, travel time prediction values, travel time background values, sampling point spatial coordinate information, and the training dataset. The training loss includes: The mean squared error is calculated based on the travel time predictions and the travel time observations in the training dataset to obtain the data error loss. Based on the physical constraints of ultrasonic longitudinal waves, the difference of the equation is calculated according to the predicted values ​​of travel time factor, velocity structure, travel time background and sampling point spatial coordinates to obtain the physical error loss. The boundary condition deviation penalty is calculated based on the travel time factor prediction value and the spatial coordinate information of the excitation point in the training dataset to obtain the boundary condition loss; The mean square error at spatial coordinates is calculated based on the predicted velocity structure value and the preset velocity structure data to obtain the velocity structure constraint loss. Based on the predicted velocity structure values ​​and the spatial coordinate information of the sampling points, the gradient penalty of the velocity field is calculated to obtain the gradient regularization loss. The training loss is calculated by weighting the data error loss, physical error loss, boundary condition loss, velocity structure constraint loss, and gradient regularization loss.

7. A PINN-based laser ultrasonic imaging detection device for metal composite plate interfaces, wherein the PINN-based laser ultrasonic imaging detection device is used to implement the PINN-based laser ultrasonic imaging detection method for metal composite plate interfaces as described in any one of claims 1-6, characterized in that, The device includes: The signal acquisition module is used to acquire information from the metal composite plate to be inspected based on the laser ultrasonic testing equipment, and obtain the original time-domain ultrasonic signal data and the spatial coordinate information of the transmitting and receiving points. The signal processing module is used to extract the first arrival travel time data of ultrasonic P-waves based on the variational mode decomposition method and the original time-domain ultrasonic signal data, and to construct a training dataset by combining the spatial coordinate information of the transmitter and receiver points. The network training module is used to perform forward inference using the laser ultrasound imaging network based on the training dataset to obtain the predicted values ​​of travel time factor, velocity structure, travel time background, travel time prediction, and spatial coordinate information of sampling points. The loss calculation module is used to calculate the training loss based on a multi-component loss function, according to preset velocity structure data, travel time factor prediction value, velocity structure prediction value, travel time prediction value, travel time background value, sampling point spatial coordinate information, and training dataset. The network optimization module is used to optimize the parameters of the laser ultrasound imaging network based on the training loss, so as to obtain an optimized laser ultrasound imaging network. The imaging inversion module is used to obtain the global spatial coordinate information of the metal composite plate to be inspected; based on the global spatial coordinate information, an optimized laser ultrasonic imaging network is used for inversion prediction to obtain the global velocity structure prediction value; based on the velocity step feature detection method, the interface contour is extracted according to the global velocity structure prediction value to obtain laser ultrasonic imaging.

8. The PINN-based laser ultrasonic imaging detection device for metal composite plate interfaces according to claim 7, characterized in that, The network training module is further used for: Randomly select sampling points in the computational domain based on the training dataset to obtain the spatial coordinate information of the sampling points; Based on the preset wave velocity, the propagation time from the excitation point to the receiving point is calculated according to the spatial coordinate information of the transmitting and receiving points in the training dataset, and the travel time background value is obtained. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed according to the spatial coordinate information of the transmitting and receiving points in the training dataset to obtain the predicted value of the travel time factor of the ultrasonic longitudinal wave. Based on the spatial coordinate information of the sampling points, a nonlinear mapping is performed to obtain the predicted value of the velocity structure. Based on the predicted value of the travel time factor, the travel time prediction value is obtained by weighting the travel time background value.

9. A laser ultrasonic imaging inspection device for the interface of a metal composite plate, characterized in that, The laser ultrasonic imaging detection equipment for the interface of the metal composite plate includes: processor; A memory storing computer-readable instructions that, when executed by the processor, implement the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 6.

Images