Pulse eddy current detection method, system, terminal and medium for high voltage cable lead seal defect
By constructing and using deep learning models to process pulsed eddy current signal data, combining short-time Fourier transform and SE attention module, the single-frequency eddy current detection method has solved the shortcomings in deep defect detection capabilities and efficiency, and achieved efficient and accurate High voltage cable lead seal defect detection.
Patent Information
- Application Number
- CN202311498855.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2022-11-25
- Filing Date
- 2023-11-09
- Publication Date
- 2024-12-20
- Estimated Expiration
- 2043-11-09
AI Technical Summary
The existing single-frequency eddy current detection method has low detection capabilities and inspection efficiency in deep defects in high-voltage cable lead seals, and is sensitive to interference signals, making it difficult to meet the needs of on-site inspection.
Build a pulse eddy current signal data set for cable lead seal defects, use deep learning models to process simulated and measured pulse eddy current signal data, establish a cable lead seal defect detection model, and improve detection capabilities through short-time Fourier transform and SE attention modules , combined with the MMD field adaptive algorithm to reduce the difference between simulation and measured data.
It improves the deep defect detection capability and inspection efficiency of cable lead seal defect detection, achieves fast and accurate on-site inspection, and reduces engineering complexity and cost.
Smart Images
Figure CN118010838B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of cable seal detection, and more specifically to a pulse eddy current detection method, system, terminal and medium for high-voltage cable seal defect detection. Background Art
[0002] Lead sealing refers to the use of alloy materials such as lead and tin to seal the gap between the end of the tail pipe and the metal sleeve, forming a molded structure after sealing.
[0003] The lead seal of high-voltage cables is used to make the copper shell or tail tube of the accessories electrically connected to the aluminum sheath of the cable, and also plays a role in sealing and waterproofing. According to statistics, the main cause of high-voltage cable failure is the quality defects of the cable body and accessories. As one of the important parts of high-voltage cable accessories, the lead seal, once cracked, can easily lead to water ingress or poor electrical connection of the accessories, causing tripping, breakdown and other problems. In severe cases, it can also cause power grid shutdown accidents.
[0004] Existing lead seal detection technologies include ultrasonic detection, X-ray detection, loop resistance detection, and single-frequency eddy current detection. The ultrasonic method is limited by coupling and requires the removal of the surface protective layer for detection. It is mostly used for testing before cable installation. The X-ray detection method has high requirements for X-ray energy and penetration angle, so on-site detection is inconvenient. The loop resistance method can only detect the situation where the lead seal is completely disconnected. The detection requires the removal of the lead seal heat shrink sleeve and waterproof tape, which is inefficient. Compared with other detection methods, the single-frequency eddy current detection method has the advantages of no coupling agent, no need to contact the test piece, easy automation, and fast detection speed. However, the single-frequency eddy current detection method has problems such as small effective penetration depth, limited spectrum range, and sensitivity to interference signals, resulting in low deep defect detection capability and inspection efficiency of the single-frequency eddy current detection method. Summary of the invention
[0005] The purpose of the present invention is to provide a pulse eddy current detection method, system, terminal and medium for high-voltage cable seal defects. After processing the simulated pulse eddy current signal data and the measured pulse eddy current signal data of different cable seal defects, a cable seal defect detection model based on the simulated pulse eddy current signal data and the measured pulse eddy current signal data is constructed using a deep learning model, and the cable seal defect detection model is used to detect the cable seal defects, so as to achieve the purpose of improving the deep defect detection capability and inspection efficiency in the cable seal defect detection process.
[0006] The above technical objectives of the present invention are achieved through the following technical solutions:
[0007] A pulsed eddy current detection method for high voltage cable lead seal defects comprises the following steps:
[0008] Constructing a pulsed eddy current signal data set of different defects of cable seals, wherein the pulsed eddy current signal data in the pulsed eddy current signal data set includes both simulated pulsed eddy current signal data of different defect types of cable seals and measured pulsed eddy current signal data of different defect types of cable seals;
[0009] Processing pulsed eddy current signal data in a pulsed eddy current signal data set, and using the processed pulsed eddy current signal data as a training set;
[0010] The pulsed eddy current signal data in the training set is trained using a deep learning model to obtain a trained cable seal defect detection model;
[0011] The pulsed eddy current signal data to be detected corresponding to the cable seal to be detected is collected, and the pulsed eddy current signal data to be detected is used as the input of the cable seal defect detection model. The obtained cable seal defect detection model is used to detect the on-site cable seals to obtain the high-voltage cable seal defect detection results.
[0012] Furthermore, the process of processing the pulsed eddy current signal data in the pulsed eddy current signal data set is specifically as follows:
[0013] The pulse eddy current signal data in each period of the pulse eddy current signal data set are converted into a time-frequency image by using short-time Fourier transform to obtain a time-frequency image data set;
[0014] The time-frequency images in the time-frequency image data set are translated, rotated, cropped, enhanced, and expanded to obtain an expanded time-frequency image data set;
[0015] The expanded time-frequency image data set is standardized and used as a training set, and the time-frequency image data in the training set is randomized.
[0016] Furthermore, the deep learning model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer;
[0017] The input layer inputs the time-frequency image data in the training set;
[0018] In the convolution layer, the convolution kernel is convolved with the input time-frequency image and traverses every pixel of the entire image;
[0019] Reduce the dimension and number of parameters of convolution output features through pooling operations;
[0020] After convolution and pooling, the fully connected layer flattens the features into a one-dimensional vector:
[0021] In the output layer, the loss function is used to predict the probability of the sample in the label space.
[0022] Furthermore, the SE attention module is introduced before the fully connected layer to compress the input feature map through global averaging to obtain global features on different channels;
[0023] Use the simulated pulsed eddy current signal data in the training set to pre-train the established deep learning model and freeze the weight parameters except for the convolution layer;
[0024] The measured pulsed eddy current signal data in the training set is input into the pre-trained deep learning model, and the MMD distance between the simulated pulsed eddy current signal data and the measured pulsed eddy current signal data in the training set is calculated in the fully connected layer;
[0025] The difference distance MK-MMD between the simulated pulse eddy current signal data and the measured pulse eddy current signal data in the training set, as well as the classification loss function of the simulated pulse eddy current signal data in the training set are used as the total loss function of the deep learning model. The weight parameters in the deep learning model are continuously updated to obtain the trained cable seal defect detection model.
[0026] Furthermore, there are three convolutional layers and three pooling layers.
[0027] Furthermore, the pulsed eddy current simulation model of cable seal defects is used to obtain simulated pulsed eddy current signal data of different types of defects in cable seals;
[0028] Among them, the construction process of the pulsed eddy current simulation model of cable seal defects includes:
[0029] Use SolidWorks software to build 3D models of different types of defective cable seals;
[0030] Import the 3D models of different types of defective seals of cables into COMSOL software, and set the geometric model of the COMSOL software probe;
[0031] The material and electromagnetic parameters are set respectively to establish a pulsed eddy current simulation model for cable seal defects.
[0032] Furthermore, the detection module in the high-voltage cable seal defect pulse eddy current detection system collects the measured pulse eddy current signal data of different cable seal defects.
[0033] High voltage cable seal defect pulse eddy current detection system, including:
[0034] The detection module includes an electrically connected function generator, a data acquisition card, and a pulsed eddy current probe, wherein the excitation parameters and detection parameters of the pulsed eddy current probe are set by the function generator, the pulsed eddy current probe is used to perform real-time physical state detection on the cable seal and generate pulsed eddy current signal data in real time, and the data acquisition card is used to collect the pulsed eddy current signal data generated by the pulsed eddy current probe in real time;
[0035] The signal processing module pre-processes the pulsed eddy current signal from the detection module to ensure that its format and attributes meet the input standards of the subsequent deep learning model;
[0036] Defect detection module, which embeds a pre-trained cable seal defect detection model designed specifically for cable seal defects. When the processed pulsed eddy current signal is input into the defect detection module, the model immediately performs forward propagation. The input signal is transformed and screened in each weight layer of the model to identify features that match known defects and finally converted into specific defect identification results.
[0037] An electronic terminal comprises: a memory for storing a computer program; and a processor for executing the computer program stored in the memory so that the electronic terminal executes the pulse eddy current detection method for high-voltage cable lead seal defects.
[0038] A computer-readable storage medium stores a computer program: when the program is executed by a processor, the pulse eddy current detection method for high-voltage cable lead seal defects is implemented.
[0039] Compared with the prior art, the present invention has the following beneficial effects:
[0040] ①The problem of difficulty in acquiring samples for deep learning models was solved through the pulsed eddy current simulation model of cable seal defects.
[0041] ② The pulsed eddy current method is used as a means of signal acquisition, which has the advantages of fast detection speed, low cost, and easy automation, and improves the deep defect detection capability and inspection efficiency in the process of cable seal defect detection.
[0042] ③ The short-time Fourier transform is used to fuse the time-frequency information of the pulsed eddy current signal, and the response to defects is more obvious.
[0043] ④The SE attention mechanism enables the neural network to focus on a subset of its input (or features), greatly improving the model performance.
[0044] ⑤The MMD field adaptive algorithm can reduce the difference between simulation data and experimental data and improve detection accuracy.
[0045] ⑥ Through deep learning, end-to-end detection of cable seal defects is achieved without human intervention, reducing the complexity and cost of the project. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:
[0047] Figure 1 A schematic flow chart of a pulsed eddy current detection method for high-voltage cable seal defects provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0048] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.
[0049] like Figure 1 As shown, an embodiment of the present invention provides a pulse eddy current detection method for high-voltage cable lead seal defects, comprising the following steps:
[0050] A pulse eddy current signal data set of different defects in cable seals is constructed, wherein the pulse eddy current signal data in the pulse eddy current signal data set includes simulated pulse eddy current signal data and measured pulse eddy current signal data of different defects in cable seals; the pulse eddy current signal data in the pulse eddy current signal data set are processed, and the processed pulse eddy current signal data are used as a training set; the pulse eddy current signal data in the training set are trained using a deep learning model to obtain a trained cable seal defect detection model; and the obtained cable seal defect detection model is used to detect on-site cable seals to identify their different defects.
[0051] The process of processing the pulsed eddy current signal data in the pulsed eddy current signal data set is specifically: using short-time Fourier transform to transform the pulsed eddy current signal data in each period in the pulsed eddy current signal data set into a time-frequency image to obtain a time-frequency image data set;
[0052] In the specific implementation process, pulsed eddy current signal data with a period of l is selected and converted into l×m time-frequency image data through short-time Fourier transform. The transformation formula (1) is:
[0053]
[0054] Among them, g(ut) is the time window function, and z(u) is the pulse eddy current signal data intercepted by the window function.
[0055] The time-frequency images in the time-frequency image dataset are translated, rotated, cropped, enhanced, and expanded to obtain an expanded time-frequency image dataset; the expanded time-frequency image dataset is standardized and used as a training set, and the time-frequency image data in the training set are randomized.
[0056] The deep learning model includes an input layer, a convolution layer, a pooling layer, a fully connected layer and an output layer; the input layer inputs the time-frequency image data in the training set; in the convolution layer, the convolution kernel is convolved with the input time-frequency image and traverses each pixel of the entire image; the dimension and parameter amount of the convolution output feature are reduced by the pooling operation; after convolution pooling, the fully connected layer flattens the feature into a one-dimensional vector: in the output layer, the probability of the sample in the label space is predicted by the loss function. The convolution layer and the pooling layer are both set to 3.
[0057] In the specific implementation process, the model includes 1 input layer, 3 convolutional layers, 3 pooling layers, 2 fully connected layers and 1 output layer; the input layer inputs the time-frequency image data of size l×m. In the convolutional layer, the convolution kernel is convolved with the input image and traverses every pixel of the entire image, as shown in formula (2):
[0058] X l =f(K l *X l-1 +B l ) (2)
[0059] Among them, * represents the convolution operation, K l is the convolution kernel, X l-1 is the input of the convolutional layer, X l is the output of the convolutional layer, B l is the bias, and f represents the activation function.
[0060] The pooling operation reduces the dimension and parameter quantity of the convolution output features, as shown in formula (3):
[0061]
[0062] in, is the pooling output, For pooled input, pool represents the pooling method, and maximum pooling is used.
[0063] After three convolutions and pooling, the relevant features are extracted and flattened into a 1-dimensional vector by the fully connected layer. Since a single fully connected layer cannot solve the nonlinear problem well, a two-layer fully connected layer is used. The probability of the sample in the label space is predicted by the softmax loss function at the output layer, as shown in formula (4),
[0064]
[0065] The SE attention module is introduced before the fully connected layer, and the input feature map is compressed by global averaging to obtain the global features on different channels; the simulated pulse eddy current signal data in the training set (i.e., the target domain) is used to pre-train the established deep learning model, and the weight parameters except the convolution layer are frozen; the measured pulse eddy current signal data in the training set (i.e., the source domain) is input into the pre-trained deep learning model, and the MMD distance between the simulated pulse eddy current signal data and the measured pulse eddy current signal data in the training set is calculated in the fully connected layer; the difference distance MK-MMD between the simulated pulse eddy current signal data and the measured pulse eddy current signal data in the training set, and the classification loss function of the simulated pulse eddy current signal data in the training set are used as the total loss function of the deep learning model, and the weight parameters in the deep learning model are continuously updated to obtain the trained cable seal defect detection model.
[0066] In the specific implementation process, the SE attention module is introduced before the first fully connected layer, and the input feature map is compressed by global averaging to obtain the global features on different channels; the simulated pulsed eddy current signal data is used to pre-train the established model, and the weight parameters except the convolution layer are frozen; the measured pulsed eddy current signal data is input into the pre-trained model, and the MMD distance (maximum mean difference) between the simulated pulsed eddy current signal data features and the measured pulsed eddy current signal data features is calculated in the fully connected layer, which is expressed as formula (5):
[0067]
[0068] where φ(x) maps each instance to a i ,x j )=φ(x i ) T φ(x j ) The Hilbert space H, X associated with s and X t are the simulated pulsed eddy current signal data in the training set and the measured pulsed eddy current signal data, respectively. s and n t They are the size of the simulated pulsed eddy current signal data and the size of the measured pulsed eddy current signal data in the training set. The MMD distance of the simulated pulsed eddy current signal data and the measured pulsed eddy current signal data features and the simulated pulsed eddy current signal data classification loss function are used as the total loss function of the model as shown in formula (6). As the basis for the back propagation of the model, the weight parameters in the neural network are continuously updated until completion.
[0069] loss total =loss CNN +λloss MMD (6)
[0070] Where λ is the weight factor of the distribution adaptation optimization part, λ = 0.001.
[0071] Through the pulse eddy current simulation model of cable seal defects, simulated pulse eddy current signal data of different types of cable seal defects are obtained; wherein, the construction process of the pulse eddy current simulation model of cable seal defects includes: using solidworks software to establish a three-dimensional model of cable seals with different types of defects; importing the three-dimensional model of cable seals with different types of defects into COMSOL software, and setting the geometric model of the COMSOL software probe; setting the material and electromagnetic parameters respectively to establish the pulse eddy current simulation model of cable seal defects.
[0072] The detection module in the pulsed eddy current detection system for high-voltage cable seal defects is used to collect the measured pulsed eddy current signal data of different defects in cable seals.
[0073] The embodiment of the present invention also provides a pulse eddy current detection system for high-voltage cable lead seal defects, including: a detection module, including an electrically connected function generator, a data acquisition card, and a pulse eddy current probe, the excitation parameters and detection parameters of the pulse eddy current probe are set by the function generator, the pulse eddy current probe is used to perform real-time physical state detection on the cable lead seal and generate pulse eddy current signal data in real time, and the pulse eddy current signal data generated by the pulse eddy current probe is collected in real time by the data acquisition card; a signal processing module, which pre-processes the pulse eddy current signal from the detection module to ensure that its format and attributes meet the input standards of the subsequent deep learning model; a defect detection module, which embeds a pre-trained cable lead seal defect detection model designed specifically for cable lead seal defects. When the processed pulse eddy current signal is input into the defect detection module, the model immediately performs forward propagation. The input signal is transformed and screened in each weight layer of the model to identify features that match known defects, and finally converted into specific defect recognition results.
[0074] An embodiment of the present invention further provides an electronic terminal, comprising: a memory, the memory being used to store a computer program; a processor, the processor being used to execute the computer program stored in the memory, so that the electronic terminal executes the pulse eddy current detection method for high-voltage cable seal defects.
[0075] Specifically: the electronic terminal includes one or more processors; a memory coupled to the processor for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the steps of a power system structure optimization method described in the above embodiment. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), off-the-shelf programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, which are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in a computer storage medium to implement the corresponding method flow or corresponding functions; the processor described in the embodiment of the present invention can be used to perform the operation of a power system structure optimization method.
[0076] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored: when the program is executed by a processor, the pulse eddy current detection method for high-voltage cable lead seal defects is implemented.
[0077] Specifically: Computer-readable storage medium (Memory) is a memory device in a computer device for storing programs and data. It is understandable that the computer-readable storage medium here can include both built-in storage media in a computer device and, of course, extended storage media supported by the computer device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by a processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one disk memory. The processor can load and execute one or more instructions stored in the computer-readable storage medium to implement the corresponding steps of the power system structure optimization method in the above embodiment. Those skilled in the art should understand that the embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.
[0078] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A pulsed eddy current detection method for high voltage cable seal defects, characterized in that: The following steps are involved: Constructing a pulsed eddy current signal data set of different defects of cable seals, wherein the pulsed eddy current signal data in the pulsed eddy current signal data set includes both simulated pulsed eddy current signal data of different defect types of cable seals and measured pulsed eddy current signal data of different defect types of cable seals; Processing pulsed eddy current signal data in a pulsed eddy current signal data set, and using the processed pulsed eddy current signal data as a training set; The pulsed eddy current signal data in the training set is trained using a deep learning model to obtain a trained cable seal defect detection model; Collect the pulsed eddy current signal data to be detected corresponding to the cable seal to be detected, use the pulsed eddy current signal data to be detected as the input of the cable seal defect detection model, use the obtained cable seal defect detection model to detect the on-site cable seal, and obtain the high-voltage cable seal defect detection result; The process of processing the pulsed eddy current signal data in the pulsed eddy current signal data set is specifically: The pulse eddy current signal data in each period of the pulse eddy current signal data set are converted into a time-frequency image by using short-time Fourier transform to obtain a time-frequency image data set; The time-frequency images in the time-frequency image data set are translated, rotated, cropped, enhanced, and expanded to obtain an expanded time-frequency image data set; The expanded time-frequency image data set is standardized and used as a training set, and the order of the time-frequency image data in the training set is randomized; The deep learning model includes an input layer, a convolutional layer, a pooling layer, a fully connected layer and an output layer; The input layer inputs the time-frequency image data in the training set; In the convolution layer, the convolution kernel is convolved with the input time-frequency image and traverses every pixel of the entire image; Reduce the dimension and number of parameters of convolution output features through pooling operations; After convolution and pooling, the fully connected layer flattens the features into a one-dimensional vector: In the output layer, the probability of the sample in the label space is predicted through the loss function; The SE attention module is introduced before the fully connected layer to compress the input feature map through global averaging to obtain global features on different channels; Use the simulated pulsed eddy current signal data in the training set to pre-train the established deep learning model and freeze the weight parameters except the convolution layer; The measured pulsed eddy current signal data in the training set is input into the pre-trained deep learning model, and the MMD distance between the simulated pulsed eddy current signal data and the measured pulsed eddy current signal data in the training set is calculated in the fully connected layer; The difference distance MK-MMD between the simulated pulse eddy current signal data and the measured pulse eddy current signal data in the training set, as well as the classification loss function of the simulated pulse eddy current signal data in the training set are used as the total loss function of the deep learning model. The weight parameters in the deep learning model are continuously updated to obtain the trained cable seal defect detection model.
2. The pulse eddy current detection method for high-voltage cable seal defects according to claim 1 is characterized in that: There are three convolutional layers and three pooling layers, and two fully connected layers.
3. The pulse eddy current detection method for high-voltage cable seal defects according to claim 1 is characterized in that: Through the pulsed eddy current simulation model of cable seal defects, the simulated pulsed eddy current signal data of different types of cable seal defects are obtained; Among them, the construction process of the pulsed eddy current simulation model of cable seal defects includes: Use SolidWorks software to build 3D models of different types of defective cable seals; Import the 3D models of different types of defective seals of cables into COMSOL software, and set the geometric model of the COMSOL software probe; The material and electromagnetic parameters are set respectively to establish a pulsed eddy current simulation model for cable seal defects.
4. The pulse eddy current detection method for high-voltage cable seal defects according to claim 1 is characterized in that: The detection module in the pulsed eddy current detection system for high-voltage cable seal defects is used to collect the measured pulsed eddy current signal data of different defects in cable seals.
5. The pulse eddy current detection system for high-voltage cable seal defects according to claim 4 is characterized in that: include: The detection module includes an electrically connected function generator, a data acquisition card, and a pulsed eddy current probe, wherein the excitation parameters and detection parameters of the pulsed eddy current probe are set by the function generator, the pulsed eddy current probe is used to perform real-time physical state detection on the cable seal and generate pulsed eddy current signal data in real time, and the data acquisition card is used to collect the pulsed eddy current signal data generated by the pulsed eddy current probe in real time; The signal processing module pre-processes the pulsed eddy current signal from the detection module to ensure that its format and attributes meet the input standards of the subsequent deep learning model; Defect detection module, which embeds a pre-trained cable seal defect detection model designed specifically for cable seal defects; When the processed pulsed eddy current signal is input into the defect detection module, the model immediately performs forward propagation; The input signal is transformed and filtered in each weight layer of the model to identify features that match known defects and finally converted into specific defect identification results.
6. An electronic terminal, characterized in that: include: A memory, the memory being used to store a computer program; A processor, wherein the processor is used to execute the computer program stored in the memory so that the electronic terminal executes the pulse eddy current detection method for high-voltage cable seal defects according to any one of claims 1 to 5.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the pulse eddy current detection method for high-voltage cable seal defects described in any one of claims 1 to 5 is implemented.