Data-driven high-voltage cable aluminum sheath defect identification method, system and medium

Abstract

The present invention relates to the field of nondestructive testing technology, and in particular to a data-driven high-voltage cable aluminum sheath defect recognition method, system and medium. The data-driven high-voltage cable aluminum sheath defect recognition method provided by the present invention reconstructs the original ultrasonic guided wave signal to achieve data dimensionality reduction by introducing a dimensionality reduction matrix, reduces redundant information in the original ultrasonic guided wave signal, and reduces the loss of relevant information about corrosion defects in the original ultrasonic guided wave signal. A neural network model based on cable aluminum sheath defect recognition is constructed to recognize cable aluminum sheath defects, thereby improving the calculation accuracy and efficiency of high-voltage cable aluminum sheath defect recognition.

Description

Technical Field

[0001] The present invention relates to the technical field of nondestructive testing, and in particular to a data-driven high-voltage cable aluminum sheath defect recognition method, system and medium. Background Art

[0002] In the power system, high-voltage cable is an important component and the part with the largest demand. It consists 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 PVC insulation material, and the inner sheath is generally a metal layer, commonly used is a corrugated aluminum sheath. Due to the installation process and long-term exposure to complex working conditions, the waterproof ability of the cable will be greatly reduced, and electrochemical corrosion will occur during the operation, causing corrosion defects on the surface of the corrugated aluminum sheath. In severe cases, the cable may short-circuit and cause major accidents. The aluminum sheath is wrapped in an insulating layer, and traditional detection methods cannot detect corrosion damage to the aluminum sheath. Therefore, the introduction of ultrasonic guided wave detection technology to detect cable aluminum sheath damage is a hot direction in existing technology.

[0003] With the rapid development of artificial intelligence, methods such as classifiers and neural networks have been widely used. However, due to the multimodal and dispersion characteristics of guided waves and the interference of noise on guided wave signals, guided wave data presents characteristics such as diversity, large scale, and high dimensionality. Although it contains rich rules and information, they are often covered by a large amount of redundant data and difficult to observe intuitively. This also leads to the inability of traditional data-driven algorithms to capture the complex relationship between ultrasonic guided wave signals and key features, and also brings many problems such as low recognition accuracy and excessive calculation. Summary of the invention

[0004] In view of this, the first object of the present invention is to provide a data-driven high-voltage cable aluminum sheath defect identification method, by introducing a dimensionality reduction matrix, the ultrasonic guided wave original signal is reconstructed to achieve data dimensionality reduction, reduce the redundant information in the ultrasonic guided wave original signal, and at the same time reduce the loss of relevant information of corrosion defects in the ultrasonic guided wave original signal, and construct a cable aluminum sheath defect identification neural network model to identify the cable aluminum sheath defects, thereby improving the computational accuracy and efficiency of high-voltage cable aluminum sheath defect identification.

[0005] Based on the same inventive concept, the second object of the present invention is to provide a data-driven high-voltage cable aluminum sheath defect identification system.

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

[0007] The first object of the present invention can be achieved by the following technical solutions:

[0008] A data-driven high-voltage cable aluminum sheath defect identification method comprises the following steps:

[0009] Construct a cable aluminum sheath corrosion simulation sample and obtain the ultrasonic guided wave original signal of the cable aluminum sheath corrosion simulation sample;

[0010] Preprocess the collected ultrasonic guided wave original signals to produce training samples and test samples;

[0011] Merge the training samples and the test samples, reduce the dimensions of the training samples and the test samples, and reconstruct them;

[0012] Construct a neural network model for cable aluminum sheath defect recognition, and use the reconstructed training samples and test samples to train the neural network model for cable aluminum sheath defect recognition;

[0013] The corrosion defects of cable aluminum sheath are identified using the neural network model for cable aluminum sheath defect recognition.

[0014] Furthermore, the cable aluminum sheath corrosion defect samples include uniform corrosion, pitting corrosion, and filamentary corrosion defect samples, which are obtained by artificially simulating using heavy blocks or collecting actual corrosion defect samples.

[0015] Furthermore, the collected ultrasonic guided wave original signal is preprocessed to produce training samples and test samples, including the following steps:

[0016] In order to make the amplitudes of the collected waveguide signals have the same numerical scale and avoid the signal reconstruction network from biasing the extraction of signals with larger amplitude characteristics, the amplitude scale of each signal is normalized. The normalized expression is as follows:

[0017]

[0018] Among them, B i represents the amplitude of the i-th signal, B max Indicates the maximum value of the signal, B min Indicates the minimum value of the signal;

[0019] Get N training samples, denoted as [X 1 ,X 2 ,...,X N ]; n test samples, denoted as [Y 1 ,Y 2 ,...,Y n ].

[0020] Further, the training samples and the test samples are merged, and the dimensions of the training samples and the test samples are reduced and reconstructed, including the following steps:

[0021] Calculate the distance matrix and fuse the time-frequency information of the guided wave signals of the training samples and the test samples;

[0022] Construct the objective function of the optimization problem and solve the dimension reduction matrix;

[0023] Reconstruct the training and test samples using the reduced dimension matrix.

[0024] Furthermore, the distance matrix is ​​calculated, and the time-frequency information of the guided wave signals of the training samples and the test samples is integrated. The distance matrix M is calculated by the following expression:

[0025]

[0026] Among them, M i,j represents the element in the i-th row and j-th column of the distance matrix M, represents the amplitude difference of time-frequency information between different samples, and m is the parameter weight parameter that controls the weight of the amplitude difference of time-frequency information.

[0027] Furthermore, the optimization problem is expressed as:

[0028]

[0029] sTP T =E

[0030] Among them, P represents the dimension reduction matrix, X represents the training sample, Y represents the test sample, A represents the representation coefficient of the discriminant feature obtained by the matrix P, ⊙ represents the Hadamard operator, M represents the distance matrix, J represents the auxiliary variable for constructing the distance matrix M, and λ 1 and λ 2 are two parameters for adjusting the weights of training samples and test samples, E is the unit matrix, F is the original dimension, τ is the parameter for adjusting the size of the auxiliary variable, I represents the signal time-frequency atlas, I = [X; Y] = [X 1 ,X 2 ,...,X N ; Y 1 ,Y 2 ,...,Y n ]∈R d ×(N+n) .

[0031] Furthermore, the cable aluminum sheath defect recognition neural network model is a GRU-based recurrent neural network model, or a LSTM-based recurrent neural network model.

[0032] Furthermore, the cable aluminum sheath defect recognition neural network model is a GRU-based recurrent neural network model.

[0033] The second object of the present invention can be achieved by the following technical solutions:

[0034] A data-driven high-voltage cable aluminum sheath defect recognition system, comprising:

[0035] Sample construction module, used to construct cable aluminum sheath corrosion simulation samples;

[0036] A data acquisition module, which acquires the original ultrasonic guided wave signal;

[0037] A preprocessing module is used to preprocess the collected original ultrasonic guided wave signals to produce training samples and test samples;

[0038] Dimensionality reduction and reconstruction module, used to merge training samples and test samples, reduce the dimensions of training samples and test samples and reconstruct them;

[0039] A model training module is used to construct a neural network model for cable aluminum sheath defect recognition, and to train the neural network model for cable aluminum sheath defect recognition using reconstructed training samples and test samples;

[0040] The model recognition module is used to identify the corrosion defects of the cable aluminum sheath by using the cable aluminum sheath defect recognition neural network model.

[0041] The third object of the present invention can be achieved by the following technical solutions:

[0042] A storage medium stores a program, and when the program is executed by a computer, the above-mentioned data-driven high-voltage cable aluminum sheath defect identification method is implemented.

[0043] The present invention has the following beneficial effects compared with the prior art:

[0044] (1) The present invention reconstructs the original ultrasonic guided wave signal by introducing a dimensionality reduction matrix to achieve data dimensionality reduction, reduce redundant information in the original ultrasonic guided wave signal, and improve the computational efficiency of high-voltage cable aluminum sheath defect identification.

[0045] (2) The present invention combines the training samples and the test samples, reduces the dimensions of the training samples and the test samples, and reconstructs them. The reconstruction method of the present invention reduces the redundant information in the original ultrasonic guided wave signal, and achieves less loss of relevant information about corrosion defects in the original ultrasonic guided wave signal, thereby improving the calculation accuracy of high-voltage cable aluminum sheath defect recognition.

[0046] (3) The present invention constructs a GRU-based recurrent neural network model, and uses the GRU-based recurrent neural network model to train and identify reconstructed samples, thereby solving problems such as long-term memory and gradients in back propagation, and further improving the computational accuracy and efficiency of high-voltage cable aluminum sheath defect identification, which is beneficial to improving the reliability and safety of power system operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] 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.

[0048] Figure 1 This is a flow chart of a data-driven high-voltage cable aluminum sheath defect identification method according to Embodiment 1 of the present invention;

[0049] Figure 2 A reconstructed network flow chart of a data-driven high-voltage cable aluminum sheath defect recognition method according to Embodiment 1 of the present invention;

[0050] Figure 3 This is a GRU input and output structure diagram of Example 1 of the present invention;

[0051] Figure 4 This is a diagram of the internal structure of the GRU of Example 1 of the present invention. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0053] Embodiment 1:

[0054] like Figure 1 As shown, this embodiment provides a data-driven high-voltage cable aluminum sheath defect recognition method, comprising the following steps:

[0055] S10, constructing a cable aluminum sheath corrosion simulation sample, and obtaining an ultrasonic guided wave original signal of the cable aluminum sheath corrosion simulation sample;

[0056] In this embodiment, the cable aluminum sheath corrosion defect samples include uniform corrosion, pitting corrosion, and filamentary corrosion defect samples, which are obtained by artificially simulating with heavy blocks or collecting actual corrosion defect samples.

[0057] S20, preprocessing the collected ultrasonic guided wave original signal to produce training samples and test samples, specifically:

[0058] In order to make the amplitudes of the collected waveguide signals have the same numerical scale and avoid the signal reconstruction network from biasing the extraction of signals with larger amplitude characteristics, the amplitude scale of each signal is normalized. The normalized expression is as follows:

[0059]

[0060] Among them, B i represents the amplitude of the i-th signal, B max Indicates the maximum value of the signal, B min Indicates the minimum value of the signal;

[0061] Get N training samples, denoted as [X 1 ,X 2 ,...,X N ]; n test samples, denoted as [Y 1 ,Y 2 ,...,Y n ].

[0062] S30, merging the training sample and the test sample, reducing the dimension of the training sample and the test sample and reconstructing them, such as Figure 2 As shown, the following steps are included:

[0063] S31, calculating the distance matrix, fusing the time-frequency information of the guided wave signals of the training sample and the test sample;

[0064] In this embodiment, the distance matrix M in step S31 is calculated by the following expression:

[0065]

[0066] Among them, M i,j represents the element in the i-th row and j-th column of the distance matrix M, represents the amplitude difference of time-frequency information between different samples, and m is the parameter weight parameter that controls the weight of the amplitude difference of time-frequency information.

[0067] S32, constructing an optimization problem objective function and solving a dimension reduction matrix;

[0068] In this embodiment, the expression of the optimization problem is:

[0069]

[0070] sTP T =E

[0071] Among them, P represents the dimension reduction matrix, X represents the training sample, Y represents the test sample, A represents the representation coefficient of the discriminant feature obtained by the matrix P, ⊙ represents the Hadamard operator, M represents the distance matrix, J represents the auxiliary variable for constructing the distance matrix M, and λ 1 and λ 2 are two parameters for adjusting the weights of training samples and test samples, E is the unit matrix, F is the original dimension, τ is the parameter for adjusting the size of the auxiliary variable, I represents the signal time-frequency atlas, I = [X; Y] = [X 1 ,X 2 ,...,X N ; Y 1 ,Y 2 ,...,Y n ]∈R d ×(N+n) .

[0072] S33. Reconstruct the training sample and the test sample using the dimension reduction matrix. Specifically, PY is the reconstructed test sample, and PX is the reconstructed training sample.

[0073] S40, constructing a neural network model for cable aluminum sheath defect recognition, and using the reconstructed training samples and test samples to train the neural network model for cable aluminum sheath defect recognition;

[0074] like Figure 3 and Figure 4 As shown, in this embodiment, the cable aluminum sheath defect recognition neural network model is a GRU-based recurrent neural network model, and the input of GRU is x at time t. t and the hidden layer state h at time t-1 t-1 The hidden layer state contains the relevant information of the previous node, and the output of GRU is composed of the output y of the hidden node at time t. t and the hidden state h passed to the next node t composition.

[0075] Through the last transmitted state h t-1 and the input x of the current node t To obtain two gating states, namely the update gate and the reset gate.

[0076] After getting the gate signal, first use the reset gate to reset h t-1 , and then let h t-1 and x t Finally, a tanh activation function is used to scale the data to the range of -1 to 1.

[0077]

[0078] Here Mainly contains the current input x t Data, targeted Adding to the current hidden state realizes the filtering of information.

[0079] For the hidden units at the current and previous time steps, the reset gate determines how to combine the new input information with the previous memory, and the update gate defines how much of the previous memory is preserved to the current time step.

[0080]

[0081] where h t-1 Contains information from the past. is the candidate hidden state, and z is the update gate. The operation of this step is to forget the h passed down t-1 Some dimensional information in the memory is added, and some dimensional information of the current node input is added. The range of z is 0 to 1. The closer the gate signal is to 1, the more past data is remembered; and the closer it is to 0, the more past data is forgotten.

[0082] In another embodiment of the present invention, the cable aluminum sheath defect recognition neural network model is a LSTM-based recurrent neural network model.

[0083] S50, using a cable aluminum sheath defect recognition neural network model to identify corrosion defects of the cable aluminum sheath, including the following steps:

[0084] S51, obtaining the original ultrasonic guided wave signal of the cable aluminum sheath;

[0085] S52, preprocessing the collected ultrasonic guided wave original signal;

[0086] S53, using the dimension reduction matrix obtained in step S32, reducing the dimension and reconstructing the original ultrasonic guided wave signal;

[0087] S54, inputting the reconstructed signal obtained in step S53 into the cable aluminum sheath defect recognition neural network model obtained in step S40 to obtain the corrosion defect recognition result of the cable aluminum sheath.

[0088] In summary, this embodiment introduces a dimensionality reduction matrix to reconstruct the original ultrasonic guided wave signal to achieve data dimensionality reduction, reduce redundant information in the original ultrasonic guided wave signal, and improve the computational efficiency of high-voltage cable aluminum sheath defect recognition; this embodiment merges the training sample and the test sample, and simultaneously reduces the dimension of the training sample and the test sample and reconstructs them. The reconstruction method of the present invention reduces the redundant information in the original ultrasonic guided wave signal while achieving less loss of relevant information of corrosion defects in the original ultrasonic guided wave signal, thereby improving the computational accuracy of high-voltage cable aluminum sheath defect recognition; this embodiment constructs a GRU-based recurrent neural network model, and uses the GRU-based recurrent neural network model to train and recognize the reconstructed samples, so that the computational accuracy and efficiency of high-voltage cable aluminum sheath defect recognition are further improved, which is beneficial to improving the reliability and safety of power system operation.

[0089] Embodiment 2:

[0090] This embodiment provides a data-driven high-voltage cable aluminum sheath defect recognition system, including:

[0091] Sample construction module, used to construct cable aluminum sheath corrosion simulation samples;

[0092] A data acquisition module, which acquires the original ultrasonic guided wave signal;

[0093] A preprocessing module is used to preprocess the collected original ultrasonic guided wave signals to produce training samples and test samples;

[0094] Dimensionality reduction and reconstruction module, used to merge training samples and test samples, reduce the dimensions of training samples and test samples and reconstruct them;

[0095] A model training module is used to construct a neural network model for cable aluminum sheath defect recognition, and to train the neural network model for cable aluminum sheath defect recognition using reconstructed training samples and test samples;

[0096] The model recognition module is used to identify the corrosion defects of the cable aluminum sheath by using the cable aluminum sheath defect recognition neural network model.

[0097] That is to say, in the above modules of this embodiment, the sample construction module and the data acquisition module cooperate to implement step S10 of embodiment 1 of the present invention; the preprocessing module is used to implement step S20 of embodiment 1, and the dimension reduction and reconstruction module is used to implement step S30 of embodiment 1; the model training module is used to implement step S40 of embodiment 1; the model identification module and the data acquisition module cooperate to implement step S50 of embodiment 1. Since steps S10-S50 have been described in detail in embodiment 1, in order to make the description of the specification concise, the detailed implementation process of the above modules in this embodiment refers to embodiment 1 and will not be repeated.

[0098] Embodiment 3:

[0099] A storage medium stores a program, and when the program is executed by a computer, the data-driven high-voltage cable aluminum sheath defect identification method of embodiment 11 of the present invention is implemented, comprising the following steps:

[0100] Construct a cable aluminum sheath corrosion simulation sample and obtain the ultrasonic guided wave original signal of the cable aluminum sheath corrosion simulation sample;

[0101] Preprocess the collected ultrasonic guided wave original signals to produce training samples and test samples;

[0102] Merge the training samples and the test samples, reduce the dimensions of the training samples and the test samples, and reconstruct them;

[0103] Construct a neural network model for cable aluminum sheath defect recognition, and use the reconstructed training samples and test samples to train the neural network model for cable aluminum sheath defect recognition;

[0104] The corrosion defects of cable aluminum sheath are identified using the neural network model for cable aluminum sheath defect recognition.

[0105] It should be noted that the computer-readable storage medium of the present embodiment may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0106] In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device, or device. In this embodiment, the computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer-readable signal medium may also be any computer-readable storage medium other than a computer-readable storage medium, which may send, propagate, or transmit a program used by or in combination with an instruction execution system, device, or device. The computer program contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0107] The computer readable storage medium can be written in one or more programming languages ​​or a combination thereof to execute the computer program of the present embodiment, and the programming language includes an object-oriented programming language, such as Java, Python, C++, and also includes a conventional procedural programming language, such as C language or a similar programming language. The program can be executed entirely on the user's computer, partially on the user's computer, as an independent software package, partially on the user's computer and partially on a remote computer, or completely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., using an Internet service provider to connect through the Internet).

[0108] Obviously, the embodiments described above are only part of the embodiments of the present invention, rather than all of the embodiments. The present invention is not limited to the details of the above embodiments, and any appropriate changes or modifications made by ordinary technicians in the relevant technical field are deemed to be within the patent scope of the present invention.

Claims

1. A data-driven method for identifying defects in aluminum sheaths of high-voltage cables. It is characterized in that The following steps are involved: Construct a cable aluminum sheath corrosion simulation sample and obtain the ultrasonic guided wave original signal of the cable aluminum sheath corrosion simulation sample; The collected ultrasonic guided wave original signal is preprocessed to produce training samples and test samples, including the following steps: The amplitude scale of each signal is normalized, and the normalized expression is as follows: Among them, B i represents the amplitude of the i-th signal, B max Indicates the maximum value of the signal, B min Indicates the minimum value of the signal; Obtain N training samples, denoted as [X 1 , X 2 ,..., X N ; n test samples, denoted as [Y 1 , Y 2 ,..., Y n ; The training samples and the test samples are merged, and the dimensions of the training samples and the test samples are reduced and reconstructed, including the following steps: Calculate the distance matrix and fuse the time-frequency information of the guided wave signals of the training samples and the test samples; Construct the objective function of the optimization problem and solve the dimension reduction matrix; Reconstruct training and test samples using the dimension reduction matrix; Calculate the distance matrix, integrate the time-frequency information of the guided wave signals of the training samples and the test samples, and the distance matrix M is calculated by the following expression: Among them, M i,j represents the element in the i-th row and j-th column of the distance matrix M, represents the amplitude difference of time-frequency information between different samples, and m is the parameter weight parameter that controls the weight of the amplitude difference of time-frequency information; The optimization problem is expressed as: stPP T =E Among them, P represents the dimension reduction matrix, X represents the training sample, Y represents the test sample, A represents the representation coefficient of the discriminant feature obtained by the matrix P, ⊙ represents the Hadamard operator, M represents the distance matrix, J represents the auxiliary variable for constructing the distance matrix M, and λ 1 and λ 2 are two parameters for adjusting the weights of training samples and test samples, E is the unit matrix, F is the original dimension, τ is the parameter for adjusting the size of the auxiliary variable, I represents the signal time-frequency atlas, I = [X; Y] = [X 1 ,X 2 ,...,X N ; Y 1 ,Y 2 ,...,Y n ]∈R d×(N+n) ; Construct a neural network model for cable aluminum sheath defect recognition, and use the reconstructed training samples and test samples to train the neural network model for cable aluminum sheath defect recognition; The corrosion defects of cable aluminum sheath are identified using the neural network model for cable aluminum sheath defect recognition.

2. According to claim 1, the data-driven high-voltage cable aluminum sheath defect identification method, It is characterized in that The cable aluminum sheath corrosion defect samples include uniform corrosion, pitting corrosion, and filamentary corrosion defect samples, which are obtained by artificially simulating with heavy blocks or collecting actual corrosion defect samples.

3. The data-driven high-voltage cable aluminum sheath defect identification method according to claim 1, It is characterized in that The cable aluminum sheath defect recognition neural network model is a GRU-based recurrent neural network model or a LSTM-based recurrent neural network model.

4. A data-driven high-voltage cable aluminum sheath defect identification system, It is characterized in that The system applies the data-driven high-voltage cable aluminum sheath defect identification method according to any one of claims 1 to 3, comprising: Sample construction module, used to construct cable aluminum sheath corrosion simulation samples; A data acquisition module, which acquires the original ultrasonic guided wave signal; A preprocessing module is used to preprocess the collected original ultrasonic guided wave signals to produce training samples and test samples; Dimensionality reduction and reconstruction module, used to merge training samples and test samples, reduce the dimensions of training samples and test samples and reconstruct them; A model training module is used to construct a neural network model for cable aluminum sheath defect recognition, and to train the neural network model for cable aluminum sheath defect recognition using reconstructed training samples and test samples; The model recognition module is used to identify the corrosion defects of the cable aluminum sheath by using the cable aluminum sheath defect recognition neural network model.

5. A storage medium storing a program, It is characterized in that When the program is executed by a computer, the data-driven high-voltage cable aluminum sheath defect identification method according to any one of claims 1 to 3 is implemented.

Images