Guided wave full wave field inversion method based on physical information fusion generative adversarial neural network

Abstract

The invention provides a guided wave full wave field inversion method based on a physical information fusion generative adversarial neural network. The method comprises the steps of damage wave field feature extraction, network architecture design of the generative adversarial network, guided wave field and damage geometric image reconstruction and the like. According to the method, a guided wave field generator output signal of the generative adversarial network after training is matched with an actual measurement signal, and a guided wave field and a damage image are reconstructed through a random sequence corresponding to a matching result. The method can be flexibly applied to working conditions of various measuring point arrangement schemes, is high in reconstruction efficiency, and has certain robustness to noise.

Description

Technical Field

[0001] The present invention belongs to the field of machine learning and civil engineering technology, and in particular relates to a guided wave full wavefield inversion method based on physical information fusion generative adversarial neural network. The method is suitable for restoring the wavefield signal of ultrasonic guided waves and the geometric information of structural defects. Background Art

[0002] Nowadays, civil engineering construction has become increasingly saturated, and the inspection and maintenance of completed engineering structures have received more and more widespread attention. Ultrasonic non-destructive testing technology can effectively detect hidden defects in civil engineering structures, locate defects and estimate the severity of defects, so as to evaluate the safety status of completed engineering structures and take corresponding maintenance measures in time. Compared with other ultrasonic testing methods, ultrasonic guided wave non-destructive testing has the advantages of large detection range and high efficiency, which can greatly reduce manpower and time costs. Therefore, the development and use of new ultrasonic guided wave non-destructive testing technologies to efficiently detect practical problems in civil engineering structures is a research work that needs to be carried out urgently.

[0003] Guided waves have the natural characteristics of dispersion, multi-modality, and mode conversion at discontinuities during propagation, which usually results in a large number of overlapping wave packets in the guided wave signals collected from a small number of measuring points, making it very difficult to effectively interpret the signal, which seriously limits the practicality of ultrasonic guided wave nondestructive testing technology in actual engineering structure detection. If the wave field signal of the structure to be tested can be obtained in the entire domain, the direct wave and defect scattered wave of each mode can be intuitively observed from the wave field image, which greatly reduces the difficulty of interpreting the guided wave signal. Therefore, it is a very practical problem to study how to obtain the global wave field signal from the signals collected from a small number of measuring points. In addition, the overlapping wave packets in the guided wave signals collected from a small number of measuring points often contain a large amount of defect-related information. How to make full use of this information to obtain the location information and geometric information of the defects is also a problem worth studying.

[0004] As a deep generative model, the generative adversarial neural network realizes the training of the neural network through the process of adversarial game, can learn the inherent characteristics of the wave field in the waveguide signal data set, and establish a mapping relationship between the waveguide wave field, defect image and vector in the low-dimensional space, so as to realize the generation of diversified waveguide wave field signals and defect image data through randomly sampled low-dimensional vectors. In the present invention, a generative adversarial neural network with waveguide physics information fusion is trained using the waveguide signal data set, and the low-dimensional vector is inverted by minimizing the gap between the wavefield signal output by the generator and the measured waveguide signal, so as to realize the reconstruction of the waveguide wave field signal and the mining of defect location and geometric information. Summary of the invention

[0005] The purpose of the present invention is to solve the problems in the prior art to meet the needs of actual situations, and proposes a waveguide full-wavefield inversion method based on physical information fusion to generate adversarial neural networks.

[0006] The present invention is implemented by the following technical scheme. The present invention proposes a guided wave full wavefield inversion method based on physical information fusion generative adversarial neural network, and the method comprises the following steps:

[0007] Step 1: Establish an original data set that contains structural defects and has diversity, extract the waveguide wavefield features from the original data set to form a new data set for training the generative adversarial network;

[0008] Step 2: Design the generator G in the generative adversarial neural network θ,I and G θ,II and the discriminator D θ The network architecture is set; the initial values ​​of the network training hyperparameters are set, and the training of the generative adversarial neural network is carried out. By tuning the hyperparameters, a generator G that can accurately generate the guided wave field characteristics and defect images with diversity is obtained. θ,I and G θ,II ;

[0009] Step 3: Measure the guided wave signal of the target structure and record it as the actual measurement signal Y;

[0010] Step 4: The process of full wavefield inversion is to optimize the low-dimensional vector z and the defect position parameter x d Implementation: Generate the initial reconstructed full wavefield feature G through the generator θ,I (z), the full wavefield feature G will be reconstructed θ,I (z) is transformed into the reconstructed full wave field signal u[x d ,G θ,I (z)]; Based on the actual measurement signal Y, the measurement matrix Φ of the target structure and the reconstructed full wave field signal u[x d ,G θ,I (z)], based on the optimization of the low-dimensional vector z and the defect position x d The objective function Get the optimal low-dimensional vector and defect location The low-dimensional vector and defect location Substitute into generator G θ,I and G θ,II Reconstructing the full wavefield signal The corresponding reconstructed defect geometry image

[0011] Furthermore, the step 1 is specifically as follows:

[0012] Step 1.1: Use the physical numerical calculation method to calculate the guided wave measurement signal Y of the structure under different defect conditions FEM According to the physical equation of waveguide propagation, the waveguide signal v0[t] is excited at the structural excitation point s, the waveguide signal is scattered at the defect point r, and the measurement signal Y FEM Wave field characteristics in It can be extracted as follows:

[0013]

[0014] Among them, the propagation operator It can be calculated by the following formula:

[0015]

[0016] Where a and b are r, q or s, q or s, r respectively; operator and are Fourier transform and its inverse transform respectively; d ref is the reference distance, which can be taken as the average of the distances from the excitation point to each receiving point; c p (f) is the phase velocity of the guided wave at frequency f;

[0017] Step 1.2: Extract the wave field features from step 1.1 The corresponding defect image D d Composing a new dataset Where N is the number of samples.

[0018] Furthermore, in step 2, for the generation tasks of wave field features and defect images, convolutional neural networks that can effectively extract image features are selected to construct the generator G θ,I and G θ,II And the discriminator D θ The network architecture of θ,I and G θ,II Input is a low-dimensional vector z randomly sampled from a standard normal distribution i , the low-dimensional vector z i Through the generator G θ,I and G θ,II The network is projected to generate wave field features and defect images G θ,I (z i ) and G θ,II (z i ); Discriminator D θ Input samples from the real dataset or the output of the generator (G θ,I (z i ),G θ,II (z i )), through the discriminator Dθ Output a numerical value; the closer the numerical value corresponding to the overall output of the generator is to the numerical value corresponding to the overall sample of the real data set, the closer the data distribution output by the generator is to the sample distribution of the real data set.

[0019] Furthermore, in step 2, the adversarial training method is used to train the generative adversarial network, where the discriminator network aims to determine whether the input comes from the real data set, and its training error function is:

[0020]

[0021] in, With x i Represent the generator output and the real data set sample respectively, that is, (G θ,I (z i ),G θ,II (z i ))and is the interpolation value between the generator output and the real dataset sample, that is And ∈ i It obeys the uniform distribution U[0,1] from 0 to 1; m is the number of samples in each minibatch; λ is the gradient penalty sparsity.

[0022] Furthermore, in step 2, the generator network aims to generate as realistic wavefield features as possible to confuse the judgment of the defect image discriminator, and its training error function is:

[0023]

[0024] Among them, E d With E ud is the total energy of ultrasonic guided wave on defective / intact structure at time t1, which can be expressed as Calculate; ρ μ is the ratio of the guided wave energy to the signal energy in the μ mode, which can be obtained by finite element calculation; is the amplitude of the guided wave at the polar coordinate (r,θ) position in the defect / intact structure.

[0025] Furthermore, in step 2, the parameters of the discriminator network and the generator network are updated alternately, and the network's learning level of the characteristics of the real data set is continuously improved through a dynamic game process.

[0026] Furthermore, in step 2, the quality and diversity of the generated wave field characteristics and defect images are calculated to evaluate the generator G θ,I With G θ,IIThe learning degree of the real data set is used to optimize the network architecture and training hyperparameters of the generative adversarial network to obtain a generator G that effectively represents the wave field characteristics and defect geometric images. θ,I With G θ,II .

[0027] Furthermore, the step 4 is specifically as follows:

[0028] Step 4.1: Establish the waveguide energy conservation judgment condition and transform the wave field signal u[x d ,G θ,I (z)], the reconstructed full wavefield feature G is expressed as follows based on the physical model: θ,I (z) is converted into the full wave field signal u[x d ,G θ,I (z)]:

[0029]

[0030] Step 4.2: Based on the full wavefield signal obtained in step 4.1 and the actual measured signal, iteratively optimize x d and z to minimize the value of the following objective function:

[0031]

[0032] Remember that the optimal low-dimensional vector and defect position are and

[0033] Step 4.3: Based on the defect location obtained in step 4.2 With low-dimensional vector Substitute the following formula to reconstruct the full wavefield signal:

[0034]

[0035] Substitute the following formula to reconstruct the defect image:

[0036]

[0037] The present invention also proposes an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a guided wave full-wavefield inversion method based on physical information fusion to generate an adversarial neural network are implemented.

[0038] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of a guided wave full-wavefield inversion method based on physical information fusion to generate an adversarial neural network.

[0039] Beneficial effects of the present invention:

[0040] 1. The method described in the present invention uses a wavefield feature generator as a constraint to learn the mapping relationship between a low-dimensional vector and a (high-dimensional) guided wavefield, and realizes the reconstruction of the guided wavefield and the defect image by optimizing the low-dimensional vector, which significantly reduces the ill-posedness of the inversion problem of reconstructing the guided wavefield from a small number of measurement point signals, thereby reducing the difficulty of reconstructing the guided wavefield and the defect image.

[0041] 2. The method described in the present invention adds a waveguide energy conservation constraint to the generator training error function, and connects a wavefield physics model converted from wavefield characteristics to the generator, thereby integrating physical information into the generative adversarial neural network, making the output of the generator more consistent with the waveguide physics characteristics, thereby improving the accuracy of waveguide wavefield reconstruction.

[0042] 3. The method described in the present invention generates a waveguide wavefield through a generator and a physical model. The time required for this process is less than the time consumed by physical calculation models such as traditional finite element and finite difference methods, which can increase the speed of reconstruction of the waveguide wavefield and defect image.

[0043] 4. Compared with the method of using a neural network model to replace the entire full wavefield inversion, the method described in the present invention can be applied to different measurement point layout schemes. Compared with the method of outputting defect images based on the input signal of the neural network model, it can output the guided wave field and explain the source of each wave packet in the measurement signal, which is interpretable. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0045] Figure 1 It is a schematic diagram of the guided wave full wavefield inversion method based on generative adversarial neural network according to the present invention;

[0046] Figure 2 The architecture and training diagram of the physical information fusion generative adversarial neural network described in the present invention;

[0047] Figure 3 A schematic diagram of the arrangement of measuring points and defect locations in an embodiment of the present invention;

[0048] Figure 4Schematic diagram of the guided wave field, defect position and defect image obtained by the method of the present invention in an embodiment of the present invention. (a) is a schematic diagram of the guided wave field; (b) is a schematic diagram of the defect position; (c) is a schematic diagram of the defect image. DETAILED DESCRIPTION

[0049] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] The purpose of the present invention is to restore the guided wave field and obtain the defect position and geometry using a small number of measurement point signals, and to propose a guided wave full wave field inversion method based on a generative adversarial neural network with physical information fusion. First, a generative adversarial neural network with physical information fusion is designed, and it is trained using a guided wave field data set to learn the mapping relationship between low-dimensional vectors and (high-dimensional) guided wave fields. Then, by minimizing the difference between the generated guided wave field signal obtained by the generator and the actual measured signal, the low-dimensional vector and the defect position parameters are inverted. Finally, the guided wave full wave field signal and the defect image are reconstructed based on the inverted low-dimensional vector and defect position parameters.

[0051] Combination Figure 1 The present invention proposes a guided wave full wavefield inversion method based on physical information fusion generative adversarial neural network, which specifically includes the following steps:

[0052] Step 1: Establish an original data set that contains structural defects and has diversity, extract the waveguide wavefield features from the original data set to form a new data set for training the generative adversarial network;

[0053] Step 2: Design the generator G in the generative adversarial neural network θ,I and G θ,II and the discriminator D θ The network architecture is constructed by setting the initial values ​​of the network training hyperparameters, training the generative adversarial neural network, and obtaining a generator G that can accurately generate the guided wave field characteristics and defect images with diversity by tuning the hyperparameters. θ,I and G θ,II ;

[0054] Step 3: Measure the guided wave signal of the target structure and record it as the actual measurement signal Y;

[0055] Step 4: The process of full wavefield inversion is to optimize the low-dimensional vector z and the defect position parameter x dImplementation. Generate the initial reconstructed full wave field feature G through the generator θ,I (z), the full wavefield feature G will be reconstructed θ,I (z) is converted into the reconstructed full wavefield signal u|x d ,G θ,I According to the actual measurement signal Y, the measurement matrix Φ of the target structure and the reconstructed full wave field signal u[x d ,G θ,I (z)], based on the optimization of the low-dimensional vector z and the defect position x d The objective function Finally, the optimal low-dimensional vector is obtained and defect location The low-dimensional vector and defect location Substitute into generator G θ,I and G θ,II Reconstructing the full wavefield signal The corresponding reconstructed defect image

[0056] The step 1 is specifically as follows:

[0057] Step 1.1: Use the physical numerical calculation method to calculate the guided wave measurement signal Y of the structure under different defect conditions FEM According to the physical equation of waveguide propagation, the waveguide signal v0[t] is excited at the structural excitation point s, the waveguide signal is scattered at the defect point r, and the measurement signal Y FEM Wave field characteristics in It can be extracted as follows:

[0058]

[0059] Among them, the propagation operator It can be calculated by the following formula:

[0060]

[0061] Where a and b are r, q or s, q or s, r respectively; operator and are Fourier transform and its inverse transform respectively. ref is the reference distance, which can be taken as the average of the distances from the excitation point to each receiving point. p (f) is the phase velocity of the guided wave at frequency f.

[0062] Step 1.2: Extract the wave field features from step 1.1 The corresponding defect image D d Composing a new dataset Where N is the number of samples.

[0063] The step 2 is specifically as follows:

[0064] Step 2.1: For the generation tasks of wave field features and defect images, select convolutional neural networks that can effectively extract image features to construct the generator G. θ,I and G θ,II And the discriminator D θ The network architecture of θ,I and G θ,II Input is a low-dimensional vector z randomly sampled from a standard normal distribution i , the low-dimensional vector z i Through the generator G θ,I and G θ,II The network is projected to generate wave field features and defect images G θ,I (z i ) and G θ,II (z i ); Discriminator D θ Input samples from the real dataset or the output of the generator (G θ,I (z i ),G θ,II (z i )), through the discriminator D θ Output a value. The closer the value corresponding to the overall output of the generator is to the value corresponding to the overall sample of the real data set, the closer the data distribution of the generator output is to the sample distribution of the real data set;

[0065] Step 2.2: Use the adversarial training method to train the generative adversarial network. The discriminator network aims to determine whether the input comes from the real data set, and its training error function is:

[0066]

[0067] in, With x i Represent the generator output and the real data set sample respectively, that is, (G θ,I (z i ),G θ,II (z i ))and is the interpolation value between the generator output and the real dataset sample, that is And ∈ i It obeys the uniform distribution U[0,1] from 0 to 1; m is the number of samples in each min batch; λ is the gradient penalty sparsity, which defaults to 10.

[0068] The goal of the generator network is to generate as realistic wavefield features as possible to confuse the judgment of the defect image discriminator, and its training error function is:

[0069]

[0070] Among them, E d With E ud is the total energy of ultrasonic guided wave on defective / intact structure at time t1, which can be expressed as Calculate; ρ μ is the ratio of the guided wave energy to the signal energy in the μ mode, which can be obtained by finite element calculation; is the amplitude of the guided wave at the polar coordinate (r,θ) position in the defect / intact structure.

[0071] Alternately update the parameters of the discriminator network and the generator network, and continuously improve the network's learning of the characteristics of the real data set through a dynamic game process;

[0072] Step 2.3: Evaluate the generator G by calculating the quality and diversity indicators of the generated wave field characteristics and defect images. θ,I With G θ,II The learning degree of the real data set is used to optimize the network architecture and training hyperparameters of the generative adversarial network to obtain a generator G that effectively represents the wave field characteristics and defect images. θ,I With G θ,II .

[0073] The step 4 is specifically as follows:

[0074] Step 4.1: Establish the waveguide energy conservation judgment condition and transform the wave field signal u[x d ,G θ,I (z)], the reconstructed full wavefield feature G is expressed as follows based on the physical model: θ,I (z) is converted into the full wave field signal u[x d ,G θ,I (z)]:

[0075]

[0076] Step 4.2: Based on the full wavefield signal obtained in step 4.1 and the actual measured signal, iteratively optimize x d and z to minimize the value of the following objective function:

[0077]

[0078] Remember that the optimal low-dimensional vector and defect position are and

[0079] Step 4.3: Based on the defect location obtained in step 4.2 With low-dimensional vector Substitute the following formula to reconstruct the full wavefield signal:

[0080]

[0081] Substitute the following formula to reconstruct the defect image:

[0082]

[0083] Example

[0084] This embodiment is to apply the present invention to the defect detection of aluminum plate components. Figure 1 A flow chart of the method of the present invention is given. Figure 2 A schematic diagram of the training process of the neural network model is given. Figure 3 The layout of the measuring points and the schematic diagram of the actual defect position are given. The excitation signal is a Hanning window narrow-band five-peak sine signal with a center frequency of 100kHz. The plate material is aluminum with a thickness of 6mm.

[0085] The step 1 is specifically as follows: using the finite element method to calculate the guided wave measurement signal Y of the aluminum plate under different defect conditions FEM , and extract the corresponding wave field features The extracted wave field features The corresponding defect image D d Composing a new dataset

[0086] The specific step 2 is as follows: after obtaining the above data set, refer to the deep convolutional generative adversarial neural network to construct the generator G θ,I and G θ,II And the discriminator D θ . Generator G θ,I and G θ,II Both are composed of four deconvolution layers, batch normalization layers, and activation function layers. The low-dimensional vector z of the input layer is randomly sampled from a standard normal distribution, with a dimension of 100, and the resolution of the output image is 32 pixels × 32 pixels; the discriminator is composed of four convolution layers, batch normalization layers, and activation function layers. The resolution of the input image of the input layer is 32 pixels × 32 pixels, and the output is a value between 0 and 1. Using the Adam optimization algorithm, initialize various training parameters (batch size, number of training rounds, learning rate, etc.), use the adversarial training method to train the generative adversarial network, and tune the training parameters to obtain a generator G that effectively learns the wave field characteristics and defect images. θ,I With G θ,II , the network parameters θ are fixed and no longer change.

[0087] The step three is specifically as follows: in the numerical simulation example, the waveguide signal is excited at the measuring point 1, and the measurement signal is received at the other measuring points at the same time. Based on this example, the waveguide signal is excited at each measuring point in turn, and the measurement signal is received at the other measuring points, and finally 30 measurement signals are obtained to form the measurement signal matrix Y.

[0088] The step 4 is specifically as follows: according to the generator G obtained in step 2 θ,I The measured signal Y obtained in step 3 and based on the physical model u[x d ,G θ,I (z)], using the optimization algorithm to optimize the objective function Minimize and obtain the optimal low-dimensional vector and defect position: and Substitute:

[0089]

[0090] The reconstructed full wave field signal and the reconstructed defect image are obtained. The snapshots of the reconstructed full wave field signal, the defect position and the reconstructed defect image are shown in Figure 4 As shown in (a), (b) and (c).

[0091] The present invention proposes a guided wave full wavefield inversion method based on physical information fusion generative adversarial neural network, the method includes damage wavefield feature extraction, network architecture design of generative adversarial network and guided wavefield and damage geometric image reconstruction method. The method of the present invention uses the output signal of the guided wavefield generator of the generative adversarial network after training to match the actual measurement signal, and reconstructs the guided wavefield and damage image through the random sequence corresponding to the matching result. The method can be flexibly applied to the working conditions of various measurement point layout schemes, and has high reconstruction efficiency and certain robustness to noise.

[0092] The present invention also proposes an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a guided wave full-wavefield inversion method based on physical information fusion to generate an adversarial neural network are implemented.

[0093] The present invention also proposes a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the steps of a guided wave full-wavefield inversion method based on physical information fusion to generate an adversarial neural network.

[0094] The memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus RAM (DRRAM). It should be noted that the memory of the method described in the present invention is intended to include, but is not limited to, these and any other suitable types of memory.

[0095] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer instructions are loaded and executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions may be transmitted from a website site, computer, server or data center by wired (e.g., coaxial cable, optical fiber, digital subscriber line (digital subscriber line, DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) mode to another website site, computer, server or data center. The computer-readable storage medium may 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 available media integrated. The available medium may be a magnetic medium (eg, a floppy disk, a hard disk, a magnetic tape), an optical medium (eg, a high-density digital video disc (DVD)), or a semiconductor medium (eg, a solid state disc (SSD)).

[0096] In the implementation process, each step of the above method can be completed by an integrated logic circuit of hardware in a processor or an instruction in the form of software. The steps of the method disclosed in conjunction with the embodiment of the present application can be directly embodied as a hardware processor for execution, or a combination of hardware and software modules in a processor for execution. The software module can be located in a storage medium mature in the art such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in conjunction with its hardware. To avoid repetition, it is not described in detail here.

[0097] It should be noted that the processor in the embodiment of the present application can be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method embodiment can be completed by an integrated logic circuit of hardware in the processor or an instruction in the form of software. The above processor can be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components. The methods, steps and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the embodiment of the present application can be directly embodied as a hardware decoding processor to perform, or the hardware and software modules in the decoding processor can be combined and performed. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in a memory, and the processor reads the information in the memory and completes the steps of the above method in combination with its hardware.

[0098] The above is a detailed introduction to the waveguide full-wavefield inversion method based on physical information fusion to generate adversarial neural networks proposed in the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. A guided wave full wavefield inversion method based on physical information fusion generative adversarial neural network, characterized in that: The method comprises the following steps: Step 1: Establish an original data set that contains structural defects and has diversity, extract the waveguide wavefield features from the original data set to form a new data set for training the generative adversarial network; Step 2: Design the generator G in the generative adversarial neural network θ,I and G θ,II and the discriminator D θ The network architecture is set; the initial values ​​of the network training hyperparameters are set, and the training of the generative adversarial neural network is carried out. By tuning the hyperparameters, a generator G that can accurately generate the guided wave field characteristics and defect images with diversity is obtained. θ,I and G θ,II ; Step 3: Measure the guided wave signal of the target structure and record it as the actual measurement signal Y; Step 4: The process of full wavefield inversion is to optimize the low-dimensional vector z and the defect position parameter x d Implementation: Generate the initial reconstructed full wavefield feature G through the generator θ,I (z), the full wavefield feature G will be reconstructed θ,I (z) is transformed into the reconstructed full wave field signal u[x d ,G θ,I (z)]; Based on the actual measurement signal Y, the measurement matrix Φ of the target structure and the reconstructed full wave field signal u[x d ,G θ,I (z)], based on the optimization of the low-dimensional vector z and the defect position x d The objective function Get the optimal low-dimensional vector and defect location The low-dimensional vector and defect location Substitute into generator G θ,I and G θ,II Reconstructing the full wavefield signal The corresponding reconstructed defect geometry image 2. The method according to claim 1, characterized in that The step 1 is specifically as follows: Step 1.1: Use the physical numerical calculation method to calculate the guided wave measurement signal Y of the structure under different defect conditions FEM According to the physical equation of waveguide propagation, the waveguide signal v0[t] is excited at the structural excitation point s, the waveguide signal is scattered at the defect point r, and the measurement signal Y FEM Wave field characteristics in It can be extracted as follows: Among them, the propagation operator It can be calculated by the following formula: Where a and b are r, q or s, q or s, r respectively; operator and are Fourier transform and its inverse transform respectively; d ref is the reference distance, which can be taken as the average of the distances from the excitation point to each receiving point; c p (f) is the phase velocity of the guided wave at frequency f; Step 1.2: Extract the wave field features from step 1.1 The corresponding defect image D d Composing a new dataset Where N is the number of samples.

3. The method according to claim 2, characterized in that In step 2, for the generation tasks of wave field features and defect images, convolutional neural networks that can effectively extract image features are selected to construct the generator G θ,I and G θ,II And the discriminator D θ The network architecture of θ,I and G θ,II Input is a low-dimensional vector z randomly sampled from a standard normal distribution i , the low-dimensional vector z i Through the generator G θ,I and G θ,II The network is projected to generate wave field features and defect images G θ,I (z i ) and G θ,II (z i ); Discriminator D θ Input samples from the real dataset or the output of the generator (G θ,I (z i ),G θ,II (z i )), through the discriminator D θ Output a numerical value; the closer the numerical value corresponding to the overall output of the generator is to the numerical value corresponding to the overall sample of the real data set, the closer the data distribution output by the generator is to the sample distribution of the real data set.

4. The method according to claim 3, characterized in that In step 2, the adversarial training method is used to train the generative adversarial network, where the discriminator network aims to determine whether the input comes from the real data set, and its training error function is: in, With x i Represent the generator output and the real data set sample respectively, that is, (G θ,I (z i ),G θ,II (z i ))and is the interpolation value between the generator output and the real dataset sample, that is And ∈ i It obeys the uniform distribution U[0,1] from 0 to 1; m is the number of samples in each min batch; λ is the gradient penalty sparsity.

5. The method according to claim 4, characterized in that In step 2, the generator network aims to generate as realistic wavefield features as possible to confuse the judgment of the defect image discriminator, and its training error function is: Among them, E d With E ud is the total energy of ultrasonic guided wave on defective / intact structure at time t1, which can be expressed as Calculate; μ is the ratio of the guided wave energy to the signal energy in the μ mode, which can be obtained by finite element calculation; is the amplitude of the guided wave at the polar coordinate (r,θ) position in the defect / intact structure.

6. The method according to claim 5, characterized in that In step 2, the parameters of the discriminator network and the generator network are updated alternately, and the network's learning level of the real data set characteristics is continuously improved through a dynamic game process.

7. The method according to claim 6, characterized in that In step 2, the generator G is evaluated by calculating the quality and diversity indicators of the generated wave field characteristics and defect images. θ,I With G θ,II The learning degree of the real data set is used to optimize the network architecture and training hyperparameters of the generative adversarial network to obtain the generator G that effectively represents the wave field characteristics and defect geometric images. θ,I With G θ,II .

8. The method according to claim 7, characterized in that The step 4 is specifically as follows: Step 4.1: Establish the waveguide energy conservation judgment condition and transform the wave field signal u[x d ,G θ,I (z)], the reconstructed full wavefield feature G is expressed as follows based on the physical model: θ,I (z) is converted into the full wave field signal u[x d ,G θ,I (z)]: Step 4.2: Based on the full wavefield signal obtained in step 4.1 and the actual measured signal, iteratively optimize x d and z to minimize the value of the following objective function: Remember that the optimal low-dimensional vector and defect position are and Step 4.3: Based on the defect location obtained in step 4.2 With low-dimensional vector Substitute the following formula to reconstruct the full wavefield signal: Substitute the following formula to reconstruct the defect image:

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium for storing computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images