Aluminum sheath corrosion detection method, system and medium based on variational autoencoder network

Abstract

This invention relates to the field of nondestructive testing technology, and particularly to a method, system, and medium for detecting corrosion of aluminum sheaths based on variational autoencoder networks. The method provided by this invention acquires axial and circumferential guide wave signals, uses discrete wavelet transform to obtain the time-frequency diagrams of these signals, constructs a variational autoencoder network model for detecting corrosion of cable aluminum sheaths, and uses this model to locate defects axially and circumferentially, achieving high-precision detection of cable aluminum sheath corrosion.

Description

Technical Field

[0001] This invention relates to the field of nondestructive testing technology, and in particular to a method, system and medium for detecting corrosion of aluminum sheaths based on variational autoencoder networks. Background Art

[0002] In power systems, high-voltage cables are a crucial component and the most in-demand part. They consist of four parts: conductor, insulation layer, shielding layer, and protective layer. The protective layer is divided into an outer sheath and an inner sheath. The outer sheath is generally made of PVC insulation material, while the inner sheath is usually a metal layer, commonly a corrugated aluminum sheath. Due to installation and long-term exposure to complex operating conditions, the cable's waterproofing ability decreases significantly. Electrochemical corrosion occurs during operation, causing corrosion defects on the surface of the corrugated aluminum sheath. In severe cases, this can lead to short circuits and major accidents. Since the aluminum sheath is encased in the insulation layer, traditional detection methods cannot detect corrosion damage to the aluminum sheath. Therefore, the introduction of ultrasonic guided wave detection technology to detect damage to the cable's aluminum sheath is a current research focus.

[0003] With the rapid development of artificial intelligence, methods such as classifiers and neural networks have been widely applied. However, due to the multimodal and dispersion characteristics of guided waves, as well as the interference of noise on guided wave signals, traditional data-driven algorithms cannot capture the complex relationship between ultrasonic guided wave signals and key features. To address this challenge, deep learning algorithms based on autoencoders, with their feature learning and global optimization capabilities, have been successfully and effectively applied to various fields of computer vision, including damage recognition. However, existing autoencoders only need to learn one encoding to reproduce the input, but the distribution of data in the latent space is uneven, meaning that some regions in the latent space cannot represent any observed data. Furthermore, due to the high noise level in the acquired ultrasonic guided wave signals, existing methods struggle to extract effective feature information from them. Summary of the Invention

[0004] In view of this, the first objective of the present invention is to provide a method for detecting aluminum sheath corrosion based on variational autoencoder networks. By acquiring axial and circumferential guided wave signals, using discrete wavelet transform to obtain the time-frequency diagrams of the axial and circumferential guided wave signals, constructing a variational autoencoder network model for detecting aluminum sheath corrosion of cables, and using the variational autoencoder network model for detecting aluminum sheath corrosion of cables to locate defects in the axial and circumferential directions, thereby achieving high-precision detection of aluminum sheath corrosion of cables.

[0005] Based on the same inventive concept, the second objective of this invention is to provide an aluminum sheath corrosion detection system based on variational autoencoder networks.

[0006] Based on the same inventive concept, a third objective of this invention is to provide a storage medium.

[0007] The first objective of this invention can be achieved through the following technical solution:

[0008] A method for detecting corrosion of aluminum sheaths based on variational autoencoders includes the following steps:

[0009] A simulated sample of cable aluminum sheath corrosion was constructed, and the original axial and circumferential guide wave signals of the simulated sample of cable aluminum sheath corrosion were obtained.

[0010] The original axial and circumferential guided wave signals are converted into time-frequency diagrams to obtain axial and circumferential time-frequency diagrams.

[0011] A variational autoencoder network model for detecting corrosion of aluminum sheaths in cables was constructed, including a variational autoencoder network model for detecting corrosion of aluminum sheaths in axial cables and a variational autoencoder network model for detecting corrosion of aluminum sheaths in circumferential cables.

[0012] The axial time-frequency plot is input into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training.

[0013] The circumferential time-frequency graph is input into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training;

[0014] The original axial guided wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the aluminum sheath of the cable are located axially using the trained variational autoencoder network model for axial corrosion detection of aluminum sheath of cable.

[0015] The original circumferential guide wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the cable aluminum sheath are circumferentially located using a trained circumferential cable aluminum sheath corrosion detection variational autoencoder network model.

[0016] Furthermore, the axial and circumferential guide wave signals are converted into time-frequency diagrams using discrete wavelet transform, including the following steps:

[0017] The original axial signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original axial signal is calculated by mother wavelet calculation and signal conversion.

[0018] The original circumferential signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original circumferential signal is calculated by mother wavelet calculation and signal transformation.

[0019] Furthermore, the variational autoencoder network model for detecting aluminum sheath corrosion in cables includes an encoder and a decoder. The variational autoencoder network is trained to minimize the reconstruction error between the encoded / decoded data and the initial data, wherein:

[0020] The encoder is used to compress vectors in the input feature space into the hidden feature space to obtain hidden variables, and the distribution of the hidden variables tends to a normal distribution.

[0021] The decoder is used to restore the representation of the hidden feature space to the input feature space.

[0022] Furthermore, multiple convolutional and pooling layers are added to the encoder part of the variational autoencoder network model for detecting aluminum sheath corrosion of cables, and the Dropout method is used to construct the encoder.

[0023] Furthermore, the performance of the reconstructed signal from the variational autoencoder network model is evaluated using the ELBO loss function LOSS, expressed as:

[0024]

[0025] Where x is the time-frequency graph input of the variational autoencoder network model, To reconstruct the signal, μ is the mean of the time-frequency plot, and σ 2 Let be the variance of the time-frequency plot, KL be the KL divergence, and N be the normal distribution.

[0026] The second objective of this invention can be achieved through the following technical solution:

[0027] A corrosion detection system for aluminum sheaths based on variational autoencoders includes:

[0028] The sample construction module is used to construct simulation samples of cable aluminum sheath corrosion.

[0029] The data acquisition module is used to acquire the raw signals of the axial and circumferential guide waves;

[0030] The time-frequency conversion module is used to convert the original axial and circumferential guided wave signals into time-frequency diagrams, resulting in axial and circumferential time-frequency diagrams.

[0031] The model building module is used to build variational autoencoder network models for detecting corrosion of aluminum sheaths in cables, including variational autoencoder network models for detecting corrosion of axial aluminum sheaths in cables and variational autoencoder network models for detecting corrosion of circumferential aluminum sheaths in cables.

[0032] The model training module is used to input the axial time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training; and to input the circumferential time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training.

[0033] The model recognition module is used to locate corrosion defects in the aluminum sheath of cables axially using a trained variational autoencoder network model for corrosion detection of axial cable aluminum sheaths; and to locate corrosion defects in the aluminum sheath of cables circumferentially using a trained variational autoencoder network model for corrosion detection of circumferential cable aluminum sheaths.

[0034] Furthermore, the data acquisition module includes: a computer, a signal generator, a signal amplifier, a signal acquisition card, and a solid coupling medium, wherein the solid coupling medium is matched with the acoustic impedance of the object under test.

[0035] Furthermore, the solid coupling medium is epoxy resin.

[0036] The third objective of this invention can be achieved through the following technical solution:

[0037] A storage medium storing a program that, when executed by a computer, implements the above-described method for detecting corrosion of aluminum sheaths based on variational autoencoder networks.

[0038] The present invention has the following advantages over the prior art:

[0039] (1) This invention acquires axial and circumferential guide wave signals, uses discrete wavelet transform to obtain time-frequency diagrams of axial and circumferential guide wave signals, constructs a variational autoencoder network model for detecting corrosion of cable aluminum sheath, and uses the variational autoencoder network model for detecting corrosion of cable aluminum sheath to locate defects in the axial and circumferential directions, thereby extracting effective feature information from the signals, improving the detection and positioning accuracy of cable aluminum sheath detection, and facilitating timely maintenance of cable aluminum sheath.

[0040] (2) The variational autoencoder network model provided by the present invention adds multiple convolutional and pooling layers to the encoder part and uses the Dropout method to construct the encoder, which can improve the generalization ability of the model and thus ensure that the latent space has good properties that can carry out the data generation process.

[0041] (3) The distribution of the hidden variables of the self-encoding network of the present invention tends to the standard normal distribution, which makes the hidden variables have stable statistical properties, thereby improving the convenience and application efficiency of the decoder.

[0042] (4) The present invention enables the excited ultrasonic signal to enter the test object smoothly through solid coupling, and at the same time, it further reduces the influence of noise in the aluminum sheath corrosion detection method based on variational autoencoder network provided by the present invention. Attached Figure Description

[0043] Figure 1 This is a flowchart of the cable aluminum sheath corrosion detection method based on variational autoencoder network according to Embodiment 1 of the present invention;

[0044] Figure 2 This is a flowchart of the discrete wavelet transform in Embodiment 1 of the present invention;

[0045] Figure 3 This is a structural diagram of the variational autoencoder network model of Embodiment 1 of the present invention;

[0046] Figure 4 This is a schematic diagram of the internal operation of the variational autoencoder network in Embodiment 1 of the present invention. Detailed Implementation

[0047] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are some embodiments of the present invention, but not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0048] Example 1:

[0049] like Figure 1 As shown, this embodiment provides a method for detecting corrosion of cable aluminum sheaths based on variational autoencoder networks, including the following steps:

[0050] S10. Construct a simulation sample of cable aluminum sheath corrosion and obtain the original axial and circumferential guide wave signals of the simulation sample of cable aluminum sheath corrosion.

[0051] In this embodiment, the corrosion defect samples of the cable aluminum sheath include uniform corrosion, pitting corrosion, and filiform corrosion, which are obtained by simulating the corrosion defect samples manually using heavy objects or by collecting actual corrosion defect samples.

[0052] S20. Convert the original axial and circumferential guide wave signals into a time-frequency diagram;

[0053] In this embodiment, the axial and circumferential guide wave signals are converted into time-frequency diagrams through discrete wavelet transform, including the following steps:

[0054] S21. The original axial signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original axial signal is calculated by mother wavelet calculation and signal conversion.

[0055] S22. The original circumferential signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original circumferential signal is calculated by mother wavelet calculation and signal conversion.

[0056] In steps S210 and S220 of this embodiment, the time-frequency diagram of the signal is obtained under finite computation by using the discrete scale parameter a and the translation parameter b. a and b can be discretized using the following formula:

[0057]

[0058] In the further discretization process, assuming a0 = 2 and b0 = 1, the mother wavelet can be expressed as:

[0059]

[0060] At this point, the discrete wavelet transform can be expressed as:

[0061]

[0062] Discretization refers to the discretization of the scale parameter a and the translation parameter b, while the time variable t in a space can still be defined as continuous.

[0063] S30. Construct a variational autoencoder network model for detecting corrosion of aluminum sheath of cables, including a variational autoencoder network model for detecting corrosion of axial aluminum sheath of cables and a variational autoencoder network model for detecting corrosion of circumferential aluminum sheath of cables.

[0064] In this embodiment, the variational autoencoder network model for detecting aluminum sheath corrosion in cables includes an encoder and a decoder. The variational autoencoder network is trained to minimize the reconstruction error between the encoded and decoded data and the initial data, wherein:

[0065] The encoder is used to compress vectors in the input feature space into the hidden feature space to obtain hidden variables. The distribution of the hidden variables tends to a normal distribution. The encoder part incorporates multiple convolutional layers and pooling layers, and uses the Dropout method to construct the encoder.

[0066] The decoder is used to restore the representation of the hidden feature space to the input feature space.

[0067] In this embodiment, the internal operation of the variational autoencoder network model is as follows:

[0068] In this embodiment, the performance of the reconstructed signal from the variational autoencoder network model is evaluated using the ELBO loss function, which is calculated as follows:

[0069] The encoder part of the entire neural network is designed to represent a conditional probability q(z|x), where x is the time-frequency graph data input to the model, and z is a latent variable. The decoder then attempts to recover the original data using the latent variable z. Specifically, a neural network represents a conditional probability p(x|z).

[0070] First, after introducing the latent variable z, p(x) can be represented by z:

[0071] p(x)=∫ z p(x|z)p(z)dz

[0072] Autoencoder networks also have encoding and decoding processes, and for the same input, they should have the most similar output. In the above formula, p(x) is represented by z, so for any input data, we should try to ensure that there are hidden variables that transform the output data into output data that is as equal as possible to the input data.

[0073] Transform log p(x) by the following steps:

[0074] log p(x)=∫ z q(z|x)log p(x)dz

[0075] Where q(z|x) can be any distribution.

[0076] To involve the encoding process, we introduce q(z|x), resulting in the following equation:

[0077]

[0078] in Let KL divergence be the KL divergence. Let L be the objective function of the variational autoencoder. b .

[0079] After deformation, L can be obtained. b The expression is shown below:

[0080] L b =-KL(q(z|x)‖p(z))+∫ z q(z|x)log(p(x|z))dz

[0081] To maximize p(x), we need to find p(x|z) and q(z|x) such that L b As large as possible;

[0082] Minimize the KL value by adjusting q(z|x) and make ∫ z Maximize q(z|x)log(p(x|z))dz. After parameter tuning via the neural network, KL(q(z|x)‖p(z|x))=0, and we consider L to be... b That is, ELBO, as shown in the following formula:

[0083]

[0084] Where E refers to the mean, that is, an integral is transformed into the mean of log(p(x|z)) over the range of q(z|x); θ represents the weights of the encoder, and θ represents the weights of the neural network.

[0085] Finally, the expression for the ELBO loss function LOSS is:

[0086]

[0087] Where x is the time-frequency graph input of the variational autoencoder network model, To reconstruct the signal, μ is the mean of the time-frequency plot, and σ 2 Let be the variance of the time-frequency plot, KL be the KL divergence, and N be the normal distribution.

[0088] S40. Input the axial time-frequency diagram into the variational autoencoder network model for axial cable aluminum sheath corrosion detection for training.

[0089] S50. Input the circumferential time-frequency diagram into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training.

[0090] S60. Obtain the original axial guide wave signal of the aluminum sheath of the cable under test, and use the trained axial cable aluminum sheath corrosion detection variational autoencoder network model to locate the corrosion defects of the cable aluminum sheath axially.

[0091] In this embodiment, step S60 includes the following steps:

[0092] S61. Obtain the original axial guided wave signal of the aluminum sheath of the cable under test;

[0093] S62. Convert the axial guide wave signal into a time-frequency diagram using discrete wavelet transform;

[0094] S63. Input the axial time-frequency map into the variational autoencoder network model for axial cable aluminum sheath corrosion detection trained in step S40 for feature recognition, thereby realizing the axial positioning of corrosion defects in the cable aluminum sheath.

[0095] S70. Obtain the original circumferential guide wave signal of the aluminum sheath of the cable under test, and use the trained circumferential cable aluminum sheath corrosion detection variational autoencoder network model to circumferentially locate the corrosion defects of the cable aluminum sheath.

[0096] In this embodiment, step S70 includes the following steps:

[0097] S71. Obtain the original circumferential guide wave signal of the aluminum sheath of the cable under test;

[0098] S72. Convert the circumferential guide wave signal into a time-frequency diagram using discrete wavelet transform;

[0099] S73. Input the circumferential time-frequency diagram into the circumferential cable aluminum sheath corrosion detection variational autoencoder network model trained in step S50 for feature recognition, so as to realize the circumferential location of corrosion defects in the cable aluminum sheath.

[0100] In summary, this embodiment acquires axial and circumferential guide wave signals, uses discrete wavelet transform to obtain the time-frequency diagrams of these signals, constructs a variational autoencoder network model for detecting aluminum sheath corrosion in cables, and utilizes this model to locate defects axially and circumferentially. This allows for the extraction of effective feature information from the signals, improving the detection and location accuracy of aluminum sheath corrosion detection and facilitating timely maintenance of the cable sheath. The variational autoencoder network model provided in this embodiment incorporates multiple convolutional and pooling layers in the encoder section and employs the Dropout method for encoder construction, enhancing the model's generalization ability and ensuring the latent space possesses desirable properties for data generation. Furthermore, the distribution of the latent variables in the autoencoder network of this embodiment tends towards a standard normal distribution, resulting in stable statistical characteristics and improving the convenience and application efficiency of the decoder.

[0101] Example 2:

[0102] This embodiment provides an aluminum sheath corrosion detection system based on a variational autoencoder network, including:

[0103] The sample construction module is used to construct simulation samples of cable aluminum sheath corrosion.

[0104] The data acquisition module is used to acquire the raw signals of the axial and circumferential guide waves;

[0105] The time-frequency conversion module is used to convert the original axial and circumferential guided wave signals into time-frequency diagrams, resulting in axial and circumferential time-frequency diagrams.

[0106] The model building module is used to build variational autoencoder network models for detecting corrosion of aluminum sheaths in cables, including variational autoencoder network models for detecting corrosion of axial aluminum sheaths in cables and variational autoencoder network models for detecting corrosion of circumferential aluminum sheaths in cables.

[0107] The model training module is used to input the axial time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training; and to input the circumferential time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training.

[0108] The model recognition module is used to locate corrosion defects in the aluminum sheath of cables axially using a trained variational autoencoder network model for corrosion detection of axial cable aluminum sheaths; and to locate corrosion defects in the aluminum sheath of cables circumferentially using a trained variational autoencoder network model for corrosion detection of circumferential cable aluminum sheaths.

[0109] In this embodiment, the data acquisition module includes: a computer, a signal generator, a signal amplifier, a signal acquisition card, and a solid coupling medium, wherein the solid coupling medium is matched with the acoustic impedance of the object under test.

[0110] In this embodiment, the solid coupling medium is epoxy resin.

[0111] In other words, in the modules described above in this embodiment, the sample construction module and the data acquisition module cooperate to implement step S10 of embodiment 1; the time-frequency graph conversion module is used to implement step S20 of embodiment 1, and the model construction module is used to implement step S30 of embodiment 1; the model training module is used to implement steps S40 and S50 of embodiment 1; and the model recognition module and the data acquisition module cooperate to implement steps S60 and S70 of embodiment 1. Since steps S10-S70 have been described in detail in embodiment 1, for the sake of brevity, the detailed implementation process of the above modules in this embodiment is referred to in embodiment 1 and will not be repeated here.

[0112] This embodiment uses solid coupling to enable the excited ultrasonic signal to smoothly enter the object under test, while further reducing the influence of noise in the aluminum sheath corrosion detection method based on variational autoencoder network provided by this invention.

[0113] Example 3:

[0114] This embodiment provides a storage medium storing a program. When the program is executed by a computer, it implements the aluminum sheath corrosion detection method based on variational autoencoder networks according to this invention, including the following steps:

[0115] A simulated sample of cable aluminum sheath corrosion was constructed, and the original axial and circumferential guide wave signals of the simulated sample of cable aluminum sheath corrosion were obtained.

[0116] The original axial and circumferential guided wave signals are converted into time-frequency diagrams to obtain axial and circumferential time-frequency diagrams.

[0117] A variational autoencoder network model for detecting corrosion of aluminum sheaths in cables was constructed, including a variational autoencoder network model for detecting corrosion of aluminum sheaths in axial cables and a variational autoencoder network model for detecting corrosion of aluminum sheaths in circumferential cables.

[0118] The axial time-frequency plot is input into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training.

[0119] The circumferential time-frequency graph is input into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training;

[0120] The original axial guided wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the aluminum sheath of the cable are located axially using the trained variational autoencoder network model for axial corrosion detection of aluminum sheath of cable.

[0121] The original circumferential guide wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the cable aluminum sheath are circumferentially located using a trained circumferential cable aluminum sheath corrosion detection variational autoencoder network model.

[0122] It should be noted that the computer-readable storage medium in this embodiment can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof.

[0123] In this embodiment, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in connection with an instruction execution system, apparatus, or device. In this embodiment, the computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying a computer-readable program. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable storage medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The computer program contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0124] The computer-readable storage medium described above can be used to write computer programs for executing this embodiment in one or more programming languages ​​or combinations thereof. These programming languages ​​include object-oriented programming languages—such as Java, Python, and C++—and conventional procedural programming languages—such as C or similar programming languages. The program can be executed entirely on the user's computer, partially on the user's computer, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0125] Obviously, the above-described embodiments are only some embodiments of the present invention, and not all embodiments. The present invention is not limited to the details of the above embodiments. Any appropriate changes or modifications made by those skilled in the art are considered to be within the scope of the patent of the present invention.

Claims

1. A method for detecting corrosion of aluminum sheaths based on variational autoencoders, characterized in that, The following steps are involved: A simulated sample of cable aluminum sheath corrosion was constructed, and the original axial and circumferential guide wave signals of the simulated sample of cable aluminum sheath corrosion were obtained. The original axial and circumferential guided wave signals are converted into time-frequency diagrams to obtain axial and circumferential time-frequency diagrams. A variational autoencoder network model for detecting corrosion of aluminum sheaths in cables was constructed, including a variational autoencoder network model for detecting corrosion of aluminum sheaths in axial cables and a variational autoencoder network model for detecting corrosion of aluminum sheaths in circumferential cables. The axial time-frequency plot is input into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training. The circumferential time-frequency graph is input into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training; The original axial guided wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the aluminum sheath of the cable are located axially using the trained variational autoencoder network model for axial corrosion detection of aluminum sheath of cable. The original circumferential guide wave signal of the aluminum sheath of the cable under test is obtained, and the corrosion defects of the aluminum sheath of the cable are located circumferentially using the trained circumferential cable aluminum sheath corrosion detection variational autoencoder network model. The variational autoencoder network model for detecting aluminum sheath corrosion in cables includes an encoder and a decoder. The variational autoencoder network is trained to minimize the reconstruction error between the encoded and decoded data and the initial data. Specifically, the encoder compresses vectors in the input feature space into the hidden feature space to obtain hidden variables, and the distribution of the hidden variables tends to a normal distribution. The decoder restores the representation in the hidden feature space to the input feature space. The encoder part of the variational autoencoder network model for detecting aluminum sheath corrosion in cables incorporates multiple convolutional layers and pooling layers, and the Dropout method is used to construct the encoder.

2. The aluminum sheath corrosion detection method based on variational autoencoder network according to claim 1, characterized in that, The axial and circumferential guide wave signals are converted into time-frequency diagrams by means of discrete wavelet transform.

3. The aluminum sheath corrosion detection method based on variational autoencoder network according to claim 2, characterized in that, The axial and circumferential guide wave signals are converted into time-frequency diagrams using discrete wavelet transform, including the following steps: The original axial signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original axial signal is calculated by mother wavelet calculation and signal conversion. The original circumferential signal is decomposed into a series of finite attenuated signals by passing it through a high-pass filter and a low-pass filter, and then through a downsampling filter. The time-frequency diagram of the original circumferential signal is calculated by mother wavelet calculation and signal transformation.

4. The aluminum sheath corrosion detection method based on variational autoencoder network according to claim 1, characterized in that, The performance of the reconstructed signal of the variational autoencoder network model is evaluated using the ELBO loss function. The evaluation is performed using the following expression: Where x is the time-frequency graph input of the variational autoencoder network model, To reconstruct the signal, μ The mean of the time-frequency graph. σ 2 Let be the variance of the time-frequency plot, KL be the KL divergence, and N be the normal distribution.

5. A corrosion detection system for aluminum sheaths based on variational autoencoders, characterized in that, include: The sample construction module is used to construct simulation samples of cable aluminum sheath corrosion. The data acquisition module is used to acquire the raw signals of the axial and circumferential guide waves; The time-frequency conversion module is used to convert the original axial and circumferential guided wave signals into time-frequency diagrams, resulting in axial and circumferential time-frequency diagrams. The model building module is used to build variational autoencoder network models for detecting corrosion of aluminum sheaths in cables, including variational autoencoder network models for detecting corrosion of axial aluminum sheaths in cables and variational autoencoder network models for detecting corrosion of circumferential aluminum sheaths in cables. The model training module is used to input the axial time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of axial cable for training; and to input the circumferential time-frequency map into the variational autoencoder network model for corrosion detection of aluminum sheath of circumferential cable for training. The model recognition module is used to locate corrosion defects in the aluminum sheath of cables axially using a trained variational autoencoder network model for corrosion detection of axial cable aluminum sheaths; and to locate corrosion defects in the aluminum sheath of cables circumferentially using a trained variational autoencoder network model for corrosion detection of circumferential cable aluminum sheaths.

6. The aluminum sheath corrosion detection system based on variational autoencoder network according to claim 5, characterized in that, The data acquisition module includes: a computer, a signal generator, a signal amplifier, a signal acquisition card, and a solid coupling medium, wherein the solid coupling medium is matched with the acoustic impedance of the object under test.

7. The aluminum sheath corrosion detection system based on variational autoencoder network according to claim 6, characterized in that, The solid coupling medium is epoxy resin.

8. A storage medium storing a program, characterized in that, When the program is executed by a computer, it implements the aluminum sheath corrosion detection method based on variational autoencoder network according to any one of claims 1-4.

Images