Ultrasonic guided wave damage identification method for multi-structure data feature migration

Abstract

The invention discloses an ultrasonic guided wave damage identification method for multi-structure data feature migration, which comprises the following steps: collecting a single-structure ultrasonic guided wave scattering signal as a source domain data set, and collecting a multi-structure ultrasonic guided wave scattering signal as a target domain data set; obtaining feature information of the structural damage by using a convolutional neural network; constructing a quantitative evaluation mechanism of guided wave signal differences of different structures by adopting a maximum mean value difference method; the feature matching loss and the label loss of the source domain training are used for optimizing the deep network model; constructing a multi-layer scattering feature mapping relation for source domain and target domain signals; realizing scattering feature adaptation in a plurality of hidden layers; and finally, verifying the effectiveness of the damage characteristics by combining the multi-sensor damage index and utilizing the structure image and the damage positioning. Modeling is carried out by utilizing real detection data of various structures, so that migration detection and evaluation of damage conditions of different structures are realized, and the universality and generalization ability of an ultrasonic guided wave big data detection model are enhanced.

Description

Technical Field

[0001] The present invention belongs to the technical field of ultrasonic guided wave structural health monitoring, and in particular relates to an ultrasonic guided wave damage identification method for multi-structure data feature migration. Background Art

[0002] A rapid detection system suitable for multi-structure materials is one of the engineering goals of ultrasonic guided wave structural health monitoring. There are the following main problems in damage identification of multi-structure materials: (1) For traditional mechanism analysis methods, the dispersion characteristics, wave velocity and propagation characteristics of different materials are different. The monitoring data collected from existing materials are not suitable for damage analysis of multi-materials, and the generality and generalization of the model are poor. (2) The big data training set of deep network modeling is difficult to meet the requirements of multi-structure materials. The damage patterns of these new materials have certain correlations, but are not completely consistent, which makes it difficult for the same monitoring model to adapt to data from different monitoring structures. (3) Changes in material properties directly affect the ultrasonic propagation speed and attenuation characteristics, which in turn affect the amplitude and phase of the detection signal, making it difficult to distinguish between damage characteristics and environmental characteristics. Therefore, establishing a damage feature transfer strategy and utilizing damage data of different structures in actual detection is a new idea to solve this problem. In this way, a feature-based transfer learning method is used to reduce the feature gap between the source domain and the target domain. Combined with the multi-sensor damage index, the structural image and damage location are obtained.

[0003] The CFRP material structure damage area positioning method based on transfer learning (CN118380079A) uses composite plate simulation data as the source domain and the actual data of the composite plate as the target domain, and uses the transfer learning conditional operator to fine-tune the deep learning model to successfully achieve damage positioning of the composite plate in the target domain; however, this method requires that the distribution of the target domain data and the measured data is completely consistent, so this method cannot perform structural health monitoring on other panels of the same type, that is, it cannot achieve transfer adaptation without the participation of the target domain data in training; and this method uses the original signal modeling, and this method uses the scattered signal modeling to extract the inherent scattering characteristics of the signal for damage identification. Summary of the invention

[0004] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art and provide an ultrasonic guided wave damage identification method for multi-structure data feature migration. The present invention collects single-structure ultrasonic guided wave scattering signals as a source domain data set, collects multi-structure ultrasonic guided wave scattering signals as a target domain data set, and inputs its hidden layer multi-structure information into a detection model, thereby realizing the monitoring of multi-structure cross-workpiece damage using ultrasonic guided waves.

[0005] The present invention is achieved by at least one of the following technical solutions.

[0006] The ultrasonic guided wave damage identification method based on multi-structure data feature migration includes the following steps:

[0007] S1, collect single-structure ultrasonic guided wave scattering signals as source domain data sets, and collect multi-structure ultrasonic guided wave scattering signals as target domain data sets;

[0008] S2, using convolutional neural network to obtain scattering characteristic information of structural damage;

[0009] S3. Use the maximum mean difference method to construct a quantitative evaluation mechanism for the difference in guided wave scattering signals of different structures; use the feature matching loss and the label loss of source domain training to optimize the convolutional neural network; construct a multi-layer scattering feature mapping relationship for source domain and target domain signals; and achieve scattering feature adaptation in multiple hidden layers;

[0010] S4. Combined with the multi-sensor damage index, the effectiveness of the damage characteristics was verified using structural images and damage localization.

[0011] Furthermore, in step S1, the source domain dataset is in are the sth sample in the source domain and the corresponding damage condition, s∈n s , n s is the number of labeled samples; the target domain dataset is where n t is the number of unlabeled samples, For nth t target samples; a bandpass filter is used to denoise the signal, and the airborne signal in the monitoring signal is removed to adapt to the influence of sensors at different positions in the array sensor. The i-th element X in each waveguide sample X i The normalization factor N f Normalize:

[0012]

[0013] Furthermore, in step S1, the source domain data is a single-structure ultrasonic guided wave scattering signal, and the target domain data is a multi-structure ultrasonic guided wave scattering signal; the single structure uses isotropic material structural parts, and a single sensor experiment is used to construct a structural health monitoring source domain data set; the multi-structure uses anisotropic material or composite material structural parts, and a single sensor experiment is used to construct a structural health monitoring target domain data set.

[0014] Furthermore, in step S2, a convolutional neural network is used to obtain characteristic information of structural damage and construct a neural network with N L A multi-layer convolutional neural network with convolutional layers to obtain the original signal X i Hidden mapping features in different domains; for the convolutional network, the kth element h of the convolution result obtained by the i-th convolution kernel in the first layer i,k for:

[0015]

[0016]

[0017]

[0018] in m1 is the convolution kernel size, l is the convolution layer number, is the zero-filled vector of the input samples, K i,j is the jth element of the ith convolution kernel, n1 is the number of convolution kernels; the convolution result Use Relu function activation.

[0019] Furthermore, the maximum mean difference method is used in step S3 to construct a quantitative evaluation mechanism for the difference of guided wave signals of different structures. The specific process is: the maximum mean difference is used to calculate the source domain X S With the target domain X T The scattering characteristic distance between samples in is:

[0020]

[0021] Where f(·) represents the feature mapping function, each sample is mapped to the Hilbert space through this function, n s is the number of labeled samples, n t is the number of unlabeled samples; X i , X j represents the i-th and j-th elements in each waveguide sample X.

[0022] Furthermore, the feature matching loss in step S3 and the label loss of source domain training are used to optimize the deep network model, and the source domain feature h S and the target domain feature h T The hidden feature distance between Calculated by:

[0023]

[0024] The overall loss function of the scattered feature adaptive deep transfer learning method is calculated as follows:

[0025]

[0026] Where L total is the regression loss, Convolutional layer N l The weight coefficient of the hidden feature distance loss, c is the amplification factor of the regression loss, N S is the source domain, N T is the target domain, y preis the network output, y s is a known tag.

[0027] Furthermore, in step S3, a multi-layer feature mapping relationship is constructed for the source domain and the target domain signals to achieve feature adaptation in multiple hidden layers. The parameter features of the source domain and the target domain are extracted through the convolutional neural network feature, the model loss is reduced to optimize the feature distance, the feature distance between the source domain and the target domain is reduced, and the hidden layer scattering feature adaptation is achieved.

[0028] Furthermore, in step S4, the structural image and damage location are obtained by combining the multi-sensor damage index, and the specific steps are: by discrete point-by-point calculation, the anisotropic probabilistic imaging method of the annular array with excitation end correction is improved, and the pixel value of the discrete point depends on the damage index and the transient imaging weight P(x, y):

[0029]

[0030] Among them, the damage probability P at the discrete coordinate point (x, y) r (x,y) through all transmission paths N P The superposition calculation is used to estimate DI r is the damage index obtained by the convolutional neural network, W D (x, y) is the excitation end imaging correction coefficient, W r (x,y) is the linear attenuation imaging weight. The difference between the damaged signal and the normal signal is used to obtain the flight time, and then the imaging path where the damage is located is determined.

[0031] Furthermore, in step S4, the damage state is evaluated by using the scattering characteristics in combination with the multi-sensor damage index, thereby obtaining a structural image and damage location, and an array sensor is used to arrange a damage monitoring platform. The structural damage is detected based on the damage assessment results of the scattering signal using a convolutional neural network and probabilistic imaging technology.

[0032] Furthermore, in step S4, the multi-sensor damage index is combined and the damage state is evaluated using the scattering characteristics to obtain a structural image and damage location. The monitored samples are structural parts made of different materials, specifically anisotropic materials or composite materials, which are different from the samples of the same material collected in the target domain.

[0033] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0034] The present invention adopts feature migration technology to realize ultrasonic guided wave structural health monitoring, extracts hidden layer multi-structure features through convolutional neural network and maximum mean difference algorithm, uses feature optimization algorithm to reduce the feature gap between source domain and target domain, optimizes the model, and finally combines multi-sensor damage index to obtain structural image and damage location, thus realizing ultrasonic guided wave structural health monitoring of multi-structure data. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0036] Figure 1 It is a flow chart of the ultrasonic guided wave damage identification method of multi-structure data feature migration according to an embodiment of the present invention;

[0037] Figure 2 This is a flow chart of the convolutional neural network acquiring characteristic information of structural damage according to an embodiment of the present invention;

[0038] Figure 3 The present invention is a flowchart for implementing multi-hidden layer feature adaptation in an embodiment of the present invention. DETAILED DESCRIPTION

[0039] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present application.

[0040] Reference to "embodiments" in this application means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in this application may be combined with other embodiments.

[0041] like Figure 1 As shown, a method for ultrasonic guided wave damage identification based on multi-structure data feature migration in this embodiment includes the following steps:

[0042] S1, collect single-structure ultrasonic guided wave scattering signals as source domain data sets, and collect multi-structure ultrasonic guided wave scattering signals as target domain data sets;

[0043] S2. Use convolutional neural network to obtain the scattering characteristic information of structural damage.

[0044] S3. The maximum mean difference method is used to construct a quantitative evaluation mechanism for the difference of guided wave signals of different structures. The feature matching loss and the label loss of source domain training are used to optimize the deep network model. A multi-layer scattering feature mapping relationship is constructed for the source domain and target domain signals. The scattering feature adaptation is realized in multiple hidden layers.

[0045] S4, combined with multi-sensor damage index, the effectiveness of damage characteristics was verified using structural images and damage localization;

[0046] In waveguide structures made of different materials, the differences in ultrasonic guided wave modal responses are reflected in the number, type, amplitude and position of ultrasonic guided wave modes. The proposed feature transfer method uses the regenerated Hilbert space to obtain the feature distance and feature distribution distance of different materials, and constructs a similarity measure of multi-layer mapping features to improve the consistency of feature representation. The method can be divided into four stages: signal preprocessing, convolution feature extraction, scattering feature adaptation and multi-sensor probabilistic imaging.

[0047] In this embodiment, the signal collected from an aluminum plate is used as the source domain and the signal collected from a multi-sensor composite plate is used as the target domain. In step S1, the single structure is an aluminum plate, which simulates artificial crack damage. The thickness of the aluminum plate sample is 7 mm, and the damage size made on the back is 5 mm long and 1 mm wide. Different damages are set by gradually increasing the damage depth. For every increase of 1 mm, the crack damage depth increases from 3 mm to 6 mm. The PZT exciter and receiver are respectively installed on the damaged side. A total of 150 samples were collected, of which 30 data samples were in normal state and 4 were in damaged state. Source domain data set where x Si and Si are the i-th sample in the source domain and the corresponding damage condition, n s is the number of labeled samples.

[0048] As an embodiment, the multi-structure is a multi-material composite plate. This embodiment collects samples of 150 T300 composite plate structures in normal state and 4 damaged states. The specimen size is 300*300mm and the thickness is 3mm. A total of 15 layers of carbon fiber layers in the layering direction are used to construct the specimen, and the thickness of each layer of carbon fiber is 0.2mm. Due to the layering of this composite structure, the dispersion characteristics of the ultrasonic guided wave and the speed in each direction are different. The dispersion characteristics and anisotropy of the specimen were further analyzed. At a frequency of 200kHz, the S0 modal velocity propagating along the 0° direction is 7289m / s, while the S0 modal velocity propagating along the 90° direction is 6917m / s. Set the same artificial crack damage and PZT installation. Set different damage degrees with crack depths of 1.5mm to 3mm by increasing the crack damage depth (0.5mm each time). Target domain data set where n t is the number of unlabeled samples, x Tj is the jth target sample. The signal needs to be normalized to balance the signals from multiple sensors. A 20-800kHz bandpass filter is used to denoise the signal. The airborne signal in the monitoring signal is removed to adapt to the influence of sensors at different positions in the array. The i-th element X in each waveguide sample X i The normalization factor N f Normalization is performed. As an example, N f Set to 10.

[0049]

[0050] The test target domain detection experiment was carried out in the same type of composite plate. The specimen was a plate structure with a side length of 600 mm and a thickness of 3 mm. 16 PZT sensors were installed on the composite plate to form a ring array, in which PZT 1 was used as the driving PZT. Artificial crack damage (four degrees, including 5 mm, 10 mm, 15 mm and 20 mm) was set on the specimen. Under each condition, each of the 15 receiving PZT sensors collected 30 samples. A total of 2250 guided wave signals were collected, including 450 normal samples and 1800 damaged samples under four damage conditions.

[0051] As Figure 2 As shown, in this embodiment, the convolutional neural network described in step S2 is used to obtain the characteristic information of the structural damage, and a multi-layer feature mapping relationship is constructed for the source domain and the target domain signals. L A multi-layer convolutional neural network with convolutional layers to obtain the original signal X i Hidden mapping features in different domains.

[0052] As an embodiment, the convolutional neural network model adopts 5 convolutional layers and 3 fully connected layers, in which feature distance matching is performed in the fully connected layer. The data samples are intercepted and the signal length is set to 1024. The feature distance loss weight of each layer is set to 0.01. In terms of optimizer selection, the Adam optimizer with a learning rate of 0.001 is used. Finally, the total loss of back propagation is obtained by the label distance and the feature distance. During the training process, the single PZT monitoring data of the aluminum plate and the composite plate are used as the source domain and the target domain for network training respectively. The PCA algorithm is used to extract the main components of the hidden features and reduce the feature dimension for visualization. For the convolutional network, the kth element h of the convolution result obtained by the i-th convolution kernel of the first layer i,k for:

[0053]

[0054] i∈[1,n1],k∈[1,l]

[0055]

[0056] in m1 is the convolution kernel size, l is the convolution layer number, is the zero-filled vector of the input samples, K i,j is the jth element of the ith convolution kernel, and n1 is the number of convolution kernels. As an example, the convolution result Use Relu function activation. In damage assessment, the model uses regression algorithm instead of classification algorithm, so small fluctuations can be seen under different damage levels.

[0057] In this embodiment, feature adaptation is implemented in multiple hidden layers in step S3. The specific process is: first, the source domain and target domain data are input into the convolutional neural network respectively, and then the maximum mean difference loss is calculated for each layer of the network data in the two domains, and the maximum mean difference loss is added to the model training loss. Finally, the model is used to optimize the model regression loss and the maximum mean difference loss to achieve transfer learning. The maximum mean difference is used to calculate the source domain X S With the target domain X T The characteristic distance between samples in:

[0058]

[0059] Where f(·) represents the feature mapping function, and each sample is mapped to the Hilbert space through this function. In the reproduction kernel Hilbert space, MMD can be obtained by transforming the source domain data matrix [x i S x i* S ] and the target domain data matrix [xj T x j* T ] into the Gaussian kernel k(x i ,x j )accomplish:

[0060]

[0061] The source domain feature h S and the target domain feature h T The hidden feature distance between is calculated as follows:

[0062]

[0063] In order to enhance the similarity of features learned from different fields and ensure the consistency of model recognition effect, the overall loss function of the scattered feature adaptive deep transfer learning method can be calculated in the following way:

[0064]

[0065] Where L total is the regression loss, Convolutional layer N l The weight coefficient of the hidden feature distance loss in . c is the amplification factor of the regression loss, N S is the source domain, y pre is the network output, y s is a known tag.

[0066] Therefore, the similarity of features learned from different fields is enhanced to ensure the consistency of model recognition effect.

[0067] like Figure 3 As shown, in this embodiment, the combination of multi-sensor damage index described in step S4 is used to obtain structural images and damage location. In this verification, another similar composite plate with multiple sensors is used as a new monitoring sample, and the damage index is obtained after model training on the source domain samples of the aluminum plate and the target domain samples. In the pre-training process, the aluminum plate and the composite plate are used as the source domain and the target domain, and the deep learning network for waveguide feature extraction is calibrated to obtain an automatic damage identification model. The specific steps are: through discrete point-by-point calculation, the anisotropic probabilistic imaging method of the annular array with excitation end correction is improved, and the pixel value of the discrete point depends on the damage index and the transient imaging weight:

[0068]

[0069] Among them, the damage probability P at the discrete coordinate point (x, y) r (x,y) can pass through all transmission paths N P The superposition calculation is used to estimate, WD (x, y) is the excitation end imaging correction coefficient, W r (x,y) is the linear attenuation imaging weight. DI r The damage index obtained for the model. The higher the DI value, the higher the probability that the structural damage is located in the corresponding imaging path. For each receiving PZT sensor, the damage index can accurately reflect four different damage levels. The difference between the damaged signal and the normal signal is used to obtain the flight time, and then the imaging path where the damage is located is determined. By discretizing the monitoring structure, the conversion time from the discrete coordinate point to the driving PZT and the receiving PZT can be obtained. Imaging time coefficient C T Defined as:

[0070] C T =(t s '-t s ) / t0=[(t1′+t2′)-(t1+t2)] / t0

[0071] Where t0 is the flight time of the direct signal. t1 and t2 are the time it takes for the wave signal to reach the damage location from the launch and the time it takes for the wave signal to reach the receiving PZT from the damage location. T The linear attenuation imaging weight γ is calculated by using the value and scaling factor. When the discrete coordinate point is far away from the actual damage point, the damage probability decreases, and C T The value becomes larger:

[0072]

[0073] The amplitude threshold is used to obtain the time of flight (ToF), and the damage path is determined by ToF. The damage index is superimposed on the damage path to form a structural image of the composite plate. In traditional probabilistic imaging, the imaging path is an ellipse because the speed of sound is the same in all directions. The composite plate used in this work is 0 / 90 layered, and the speed difference in each direction is large, so the imaging path forms a rectangular path instead of an elliptical path. After setting a 96% threshold to filter the image, the true damage position can be accurately located in the structural image. After calculating the damage probability value of each discrete point through each monitoring path, the structural image of the monitored structure is finally obtained. The position with the largest pixel value in the image should be the damage position.

[0074] It should be noted that, for the sake of convenience, the aforementioned method embodiments are all expressed as a series of action combinations, but those skilled in the art should know that the present invention is not limited to the described order of actions, because according to the present invention, certain steps can be performed in other orders or simultaneously.

[0075] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0076] The above embodiments are preferred implementation modes of the present invention, but the implementation modes of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications that do not deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.

Claims

1. An ultrasonic guided wave damage identification method based on multi-structure data feature migration, characterized in that: The steps include: S1, collect single-structure ultrasonic guided wave scattering signals as source domain data sets, and collect multi-structure ultrasonic guided wave scattering signals as target domain data sets; S2, using convolutional neural network to obtain scattering characteristic information of structural damage; S3. The maximum mean difference method is used to construct a quantitative evaluation mechanism for the difference of guided wave scattering signals of different structures. The feature matching loss and the label loss of source domain training are used to optimize the convolutional neural network. A multi-layer scattering feature mapping relationship is constructed for the source domain and target domain signals. Implement scattering feature adaptation in multiple hidden layers; S4. Combined with the multi-sensor damage index, the effectiveness of damage characteristics was verified using structural images and damage localization.

2. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: In step S1, the source domain dataset is in are the sth sample in the source domain and the corresponding damage, s∈n s , n s is the number of labeled samples; the target domain dataset is where n t is the number of unlabeled samples, For nth t target samples; a bandpass filter is used to denoise the signal, and the airborne signal in the monitoring signal is removed to adapt to the influence of sensors at different positions in the array sensor. The i-th element X in each waveguide sample X i The normalization factor N f Normalize:

3. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: In step S1, the source domain data is a single-structure ultrasonic guided wave scattering signal, and the target domain data is a multi-structure ultrasonic guided wave scattering signal; the single structure uses isotropic material structural parts, and a single sensor experiment is used to construct a structural health monitoring source domain data set; the multi-structure uses anisotropic material or composite material structural parts, and a single sensor experiment is used to construct a structural health monitoring target domain data set.

4. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: In step S2, a convolutional neural network is used to obtain characteristic information of structural damage and construct a neural network with N L A multi-layer convolutional neural network with convolutional layers to obtain the original signal X i Hidden mapping features in different domains; for the convolutional network, the kth element h of the convolution result obtained by the i-th convolution kernel in the first layer i,k for: i∈[1,n1],k∈[1,l] in m1 is the convolution kernel size, l is the convolution layer number, is the zero-filled vector of the input samples, K i,j is the jth element of the ith convolution kernel, n1 is the number of convolution kernels; the convolution result Use Relu function activation.

5. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: The maximum mean difference method is used in step S3 to construct a quantitative evaluation mechanism for the difference of guided wave signals of different structures. The specific process is: the maximum mean difference is used to calculate the source domain X S With the target domain X T The scattering characteristic distance between samples in is: Where f(·) represents the feature mapping function, each sample is mapped to the Hilbert space through this function, n s is the number of labeled samples, n t is the number of unlabeled samples; X i , X j represents the i-th and j-th elements in each waveguide sample X.

6. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: The feature matching loss and the label loss of source domain training in step S3 are used to optimize the deep network model. The source domain feature h S and the target domain feature h T The hidden feature distance between Calculated by: The overall loss function of the scattered feature adaptive deep transfer learning method is calculated as follows: Where L total is the regression loss, Convolutional layer N l The weight coefficient of the hidden feature distance loss, c is the amplification factor of the regression loss, N S is the source domain, N T is the target domain, y pre is the network output, y s is a known tag.

7. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: In step S3, a multi-layer feature mapping relationship is constructed for the source domain and target domain signals to achieve feature adaptation in multiple hidden layers; the parameter features of the source domain and the target domain are extracted through the convolutional neural network features, the model loss is reduced to optimize the feature distance, the feature distance between the source domain and the target domain is shortened, and the hidden layer scattering feature adaptation is achieved.

8. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1 is characterized in that: In step S4, the structural image and damage location are obtained by combining the multi-sensor damage index. The specific steps are: by discrete point-by-point calculation, the anisotropic probabilistic imaging method of the annular array with excitation end correction is improved, and the pixel value of the discrete point depends on the damage index and the transient imaging weight P(x, y): Among them, the damage probability P at the discrete coordinate point (x, y) r (x,y) through all transmission paths N P The superposition calculation is used to estimate DI r is the damage index obtained by the convolutional neural network, W D (x, y) is the excitation end imaging correction coefficient, W r (x, y) is the linear attenuation imaging weight, which uses the difference between the damaged signal and the normal signal to obtain the flight time, and then determines the imaging path where the damage is located.

9. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1, characterized in that: In step S4, the damage state is evaluated by using the scattering characteristics in combination with the multi-sensor damage index, and then the structural image and damage location are obtained. The damage monitoring platform is arranged using array sensors, and the structural damage is detected based on the damage assessment results of the scattering signal using the convolutional neural network and the probabilistic imaging technology.

10. The ultrasonic guided wave damage identification method based on multi-structure data feature migration according to claim 1, characterized in that: In step S4, the multi-sensor damage index is combined and the damage state is evaluated using the scattering characteristics to obtain a structural image and damage location. The monitored samples are structural parts made of different materials, specifically anisotropic materials or composite materials, which are different from the samples of the same material collected in the target domain.

Images