Carbon fiber wire defect detection method and system

Abstract

The invention discloses a carbon fiber wire defect detection method and system, and the method comprises the steps: obtaining magnetic flux leakage data of a carbon fiber wire, and generating a detection image according to the magnetic flux leakage data; inputting the detection image into a preset carbon fiber wire classification model; performing feature extraction on the detection image through a feature extraction network to generate a lead feature map; and performing defect detection on the wire characteristic pattern by adopting a defect classification network to obtain a detection result corresponding to the carbon fiber wire. Various defect types of the carbon fiber wire are accurately identified, the defects at the crimping head are effectively identified, and the limitation that X-rays cannot penetrate through a thick fitting is overcome. By accurately detecting and timely processing the defects of the carbon fiber wire, the potential fault risk is reduced, and the safety and reliability of power transmission of a power distribution network are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon fiber wire defect detection, and in particular to a carbon fiber wire defect detection method and system. Background Art

[0002] As a representative of low-carbon and environmentally friendly new conductors, carbon fiber core conductors are often used in the capacity expansion and reconstruction of transmission corridors. The crimping reliability of carbon fiber core conductors is the focus of engineering attention. Unlike traditional steel-core aluminum stranded wires, the process technology of carbon fiber crimping is immature, and its conductors are prone to failure, which seriously affects the safety of power transmission. Therefore, it is very important to regularly inspect carbon fiber core conductors.

[0003] At present, the existing technology mainly uses X-rays to detect defects in carbon fiber cores. However, for the crimping joints of the carbon fiber core, due to the thick size of each crimping hardware, it is difficult for X-rays to penetrate, and it is impossible to identify defects at the crimping joints of the carbon fiber core conductors, which reduces the safety of power transmission in the distribution network. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the technical problem solved by the present invention is that the existing technology mainly uses X-rays to detect defects in carbon fiber cores, but is unable to identify defects at the crimping joints. Due to the thick size of the crimping hardware, it is difficult for X-rays to penetrate effectively, resulting in the inability to accurately detect defects in the carbon fiber core conductors at the crimping joints, thereby reducing the safety of power transmission in the distribution network. By adopting magnetic flux leakage detection technology and classification models, the present invention can accurately identify various defects in carbon fiber conductors, including normal, broken, under-voltage and other conditions of the conductors, thereby improving the detection accuracy and safety of the power system. In order to solve the above technical problems, the present invention provides the following technical solutions: A method for detecting defects in carbon fiber conductors, comprising:

[0006] Obtaining magnetic flux leakage data of the carbon fiber conductor and generating a detection image based on the magnetic flux leakage data;

[0007] Input the detection image into a preset carbon fiber wire classification model;

[0008] Extract features from the detection image through a feature extraction network to generate a wire feature map;

[0009] The defect classification network is used to perform defect detection on the wire feature map and obtain the corresponding detection results of the carbon fiber wire.

[0010] As a preferred embodiment of the carbon fiber conductor defect detection method of the present invention, the steps of obtaining magnetic flux leakage data of the carbon fiber conductor and generating a detection image according to the magnetic flux leakage data include: the carbon fiber conductor classification model includes a feature extraction network and a defect classification network;

[0011] Performing magnetic treatment on the carbon fiber conductor to obtain a magnetically marked section;

[0012] Using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data;

[0013] Determining whether the detection speed is within a preset speed range;

[0014] When the detection speed is within the speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and a detection image is generated according to the magnetic flux leakage data;

[0015] When the detection speed is not within the speed range, the steps of using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data are performed until the detection speed is within the speed range.

[0016] As a preferred solution of the carbon fiber wire defect detection method of the present invention, wherein: the feature extraction network includes a first extraction module and two second extraction modules, and the feature extraction network is used to extract features of the detection image, and generating a wire feature map includes extracting features of the detection image by the first extraction module to generate a first feature map, wherein the first extraction module includes a 3×3 standard convolution layer, a 2×2 maximum pooling layer, a batch normalization layer, and a Swish activation layer connected in sequence;

[0017] The first feature map is sequentially subjected to feature extraction by two second extraction modules to generate a wire feature map, wherein the second extraction module includes a 3×3 standard convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer, and a discard layer connected sequentially.

[0018] As a preferred embodiment of the carbon fiber wire defect detection method of the present invention, the defect classification network includes a third extraction module, a fourth extraction module and a PLTanh activation layer, and the step of using the defect classification network to perform defect detection on the wire feature map to obtain the detection result corresponding to the carbon fiber wire includes:

[0019] Performing feature extraction on the wire feature map by the third extraction module to obtain a second feature map, wherein the third extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, and a discard layer connected in sequence;

[0020] Using a fourth extraction module to perform feature extraction on the second feature map to obtain a third feature map, wherein the fourth extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer and a fully connected layer connected in sequence;

[0021] The PLTanh activation layer is used to perform defect detection on the third feature map to obtain the corresponding detection results of the carbon fiber wire.

[0022] As a preferred solution of the carbon fiber wire defect detection method of the present invention, wherein: the step of extracting features from the wire feature map by a third extraction module to obtain a second feature map includes sequentially performing nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer to obtain a first intermediate feature map;

[0023] A batch normalization layer is used to normalize the first intermediate feature map to obtain a second intermediate feature map;

[0024] The second intermediate feature map is randomly discarded through the discard layer to obtain the second feature map.

[0025] As a preferred solution of the carbon fiber wire defect detection method of the present invention, wherein: inputting the detection image into a preset carbon fiber wire classification model also includes obtaining training wire data, performing data preprocessing on the training wire data, and generating a wire feature set;

[0026] Input the wire feature set into a preset initial carbon fiber wire classification model for training, and output the training wire detection data;

[0027] Calculating a training loss function value of the wire feature set according to the training wire detection data;

[0028] When the training loss function value is greater than or equal to the preset standard loss function value, the network parameters of the preset initial carbon fiber wire classification model are adjusted until the training loss function value is less than the preset standard loss function value;

[0029] When the training loss function value is less than the preset standard loss function value, a carbon fiber wire classification model is generated.

[0030] A carbon fiber wire defect detection system, wherein:

[0031] An acquisition module is used to obtain magnetic flux leakage data of the carbon fiber conductor and generate a detection image according to the magnetic flux leakage data;

[0032] An analysis module, used for inputting the detection image into a preset carbon fiber wire classification model, wherein the carbon fiber wire classification model includes a feature extraction network and a defect classification network;

[0033] An extraction module, used for performing feature extraction on the detection image through the feature extraction network to generate a wire feature map;

[0034] The detection module is used to use the defect classification network to perform defect detection on the wire feature map to obtain a detection result corresponding to the carbon fiber wire.

[0035] A carbon fiber wire defect detection system, wherein:

[0036] The acquisition module includes a magnetic processing submodule for performing magnetic processing on the carbon fiber conductor to obtain a magnetically marked section;

[0037] The detection submodule is used to detect the magnetic mark section using a preset magnetic flux leakage detection device to obtain the detection speed and intermediate magnetic flux leakage data;

[0038] The first analysis submodule is used to determine whether the detection speed is within a preset speed range;

[0039] The extraction module includes a first extraction submodule, which is used to extract features from the detection image through the first extraction module to generate a first feature map;

[0040] A second extraction submodule is used to extract features from the first feature map through two second extraction modules in sequence to generate a wire feature map;

[0041] The detection module includes a third extraction submodule, which is used to extract features from the wire feature map through the third extraction module to obtain a second feature map;

[0042] A fourth extraction submodule, configured to extract features from the second feature map using the fourth extraction module to obtain a third feature map;

[0043] A defect detection submodule, used for performing defect detection on the third feature map by using a PLTanh activation layer to obtain a detection result;

[0044] The third extraction submodule includes a mapping unit, which is used to perform nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer in sequence to obtain a first intermediate feature map;

[0045] A normalization unit, used for normalizing the first intermediate feature map by using a batch normalization layer to obtain a second intermediate feature map;

[0046] The discarding unit is used to randomly discard the second intermediate feature map through the discarding layer to obtain the second feature map.

[0047] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0048] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of any one of the methods of the present invention.

[0049] Beneficial effects of the invention: The carbon fiber conductor defect detection method provided by the invention accurately identifies various defect types of carbon fiber conductors based on magnetic flux leakage detection data and classification models, effectively identifies defects at crimping joints, and overcomes the limitation that X-rays cannot penetrate thick metal fittings. By accurately detecting and promptly handling carbon fiber conductor defects, potential fault risks are reduced, and the safety and reliability of power transmission in the distribution network are improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0051] Figure 1 An overall flow chart of a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0052] Figure 2 A schematic diagram of the structure of a classification model of identical carbon fiber wires in a method for detecting defects in carbon fiber wires provided in the first embodiment of the present invention;

[0053] Figure 3 A schematic diagram of the structure of a carbon fiber wire classification model for a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0054] Figure 4 A waveform diagram of magnetic flux leakage detection of a carbon fiber conductor defect detection method provided by the first embodiment of the present invention;

[0055] Figure 5 A schematic structural diagram of a magnetic flux leakage detection device for a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0056] Figure 6 A schematic diagram of a fracture defect of a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0057] Figure 7 A schematic diagram of a low-pressure defect of a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0058] Figure 8A schematic diagram of inductance signal element detection of a carbon fiber wire defect detection method provided by the first embodiment of the present invention;

[0059] Fig. 9 A structural block diagram of a carbon fiber wire defect detection method and a carbon fiber wire defect detection system provided by the first embodiment of the present invention;

[0060] Fig.10 A structural block diagram of a computer device for a carbon fiber wire defect detection method provided by the first embodiment of the present invention. DETAILED DESCRIPTION

[0061] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0062] Example 1

[0063] Reference Figure 1 , as an embodiment of the present invention, provides a carbon fiber wire defect detection method, comprising:

[0064] S1: Obtain the magnetic flux leakage data of the carbon fiber conductor and generate a detection image based on the magnetic flux leakage data.

[0065] The carbon fiber conductor classification model includes a feature extraction network and a defect classification network;

[0066] Performing magnetic treatment on the carbon fiber conductor to obtain a magnetically marked section;

[0067] Using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data;

[0068] Determining whether the detection speed is within a preset speed range;

[0069] When the detection speed is within the speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and a detection image is generated according to the magnetic flux leakage data;

[0070] When the detection speed is not within the speed range, the steps of using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data are performed until the detection speed is within the speed range.

[0071] The feature extraction network includes a first extraction module and two second extraction modules. The feature extraction network is used to extract features from the detection image. Generating a wire feature map includes extracting features from the detection image through the first extraction module to generate a first feature map, wherein the first extraction module includes a 3×3 standard convolution layer, a 2×2 maximum pooling layer, a batch normalization layer, and a Swish activation layer connected in sequence.

[0072] The first feature map is sequentially subjected to feature extraction by two second extraction modules to generate a wire feature map, wherein the second extraction module includes a 3×3 standard convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer, and a discard layer connected sequentially.

[0073] Use the preset magnetic flux leakage detection equipment to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data;

[0074] In the embodiment of the present invention, refer to Figure 5 As shown, the magnetic flux leakage detection equipment includes a magnetic flux leakage detection device 2, an inductor 3, a handle 4 and an encoder wheel 5. During the detection, the operator holds the handle 4 to control the magnetic flux leakage detection device 2 to slide along the surface of the magnetic mark section (crimping hardware 1). At the starting point, the encoder 5 will record the starting time t1; at the ending point, the encoder will record the ending time t2; and record the sliding distance L, so as to calculate the average speed (L / (t1+t2)) as the detection speed, and collect the intermediate magnetic flux leakage data during the movement process through the magnetic flux leakage detection device.

[0075] Determine whether the detection speed is within a preset speed range;

[0076] The speed range refers to the speed range that does not affect the acquisition of intermediate magnetic flux leakage data. The general value is 0.4m / s-0.6m / s.

[0077] Determine whether the detection speed is between [0.4-0.6].

[0078] It should be noted that if Figure 4 As shown, the detection speed will affect the inconsistency of the intermediate magnetic flux leakage data of the magnetic flux leakage detection device.

[0079] When the detection speed is within the speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and a detection image is generated according to the magnetic flux leakage data;

[0080] In the embodiment of the present invention, when the detection speed is in [0.4-0.6], it means that the current measurement is in the normal detection speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and the detection image is constructed according to the magnetic flux leakage data.

[0081] When the detection speed is not within the speed range, the process jumps to the step of using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data until the detection speed is within the speed range.

[0082] In the embodiment of the present invention, when the detection speed is not in [0.4-0.6], it means that the current measurement is not in the normal detection speed range, and the intermediate magnetic flux leakage data cannot be used as identification data for this defect detection. The magnetic flux leakage detection equipment performs a voice broadcast to prompt the technician to re-perform the detection through the magnetic flux leakage detection equipment.

[0083] It is worth mentioning that by limiting the detection speed, the influence of the detection speed on the detection results during the detection process is avoided, and the accuracy of carbon fiber wire defect detection is further improved.

[0084] Step 206: input the inspection image into a preset carbon fiber wire classification model, wherein the carbon fiber wire classification model includes a feature extraction network and a defect classification network;

[0085] Acquire training conductor data, perform data preprocessing on the training conductor data, and generate a conductor feature set;

[0086] In the embodiment of the present invention, a preset detection device is used to detect the carbon fiber conductor fault simulation model to obtain training conductor data, and the training conductor data is preprocessed to generate a conductor feature set.

[0087] It should be noted that, see Figure 8 As shown, a circle of inductive signal elements is arranged on the inner diameter of the detection device, with a total of 24 magnetic sensitive elements along the circumferential direction, wherein the circumferential direction is the AB direction.

[0088] It should be noted that, see Figure 6-7 As shown in the figure, during the data preprocessing process of the training conductor data, the training conductor data is marked according to the defect type. The fracture size is 0-3mm for mild, 3-6mm for moderate, and 6-9mm for severe, among which 3mm is mild and 6mm is moderate. The degree of underpressure is 0-30mm for mild, 30-60mm for moderate, and 60-90mm for severe, among which 30mm is mild and 60mm is moderate. The composite defect of fracture and underpressure can be divided into 9 types according to the size of the above fracture defect and underpressure defect.

[0089] The defect classification network includes a third extraction module, a fourth extraction module and a PLTanh activation layer.

[0090] S2: Input the detection image into a preset carbon fiber wire classification model.

[0091] Performing feature extraction on the wire feature map by the third extraction module to obtain a second feature map, wherein the third extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, and a discard layer connected in sequence;

[0092] Using a fourth extraction module to perform feature extraction on the second feature map to obtain a third feature map, wherein the fourth extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer and a fully connected layer connected in sequence;

[0093] The PLTanh activation layer is used to perform defect detection on the third feature map to obtain the corresponding detection results of the carbon fiber wire.

[0094] The wire feature map is nonlinearly mapped through the fully connected layer and the Swish activation layer in sequence to obtain the first intermediate feature map;

[0095] In the embodiment of the present invention, refer to Figure 3 As shown, the wire feature map is nonlinearly mapped through the fully connected layer and the Swish activation layer in sequence to obtain the first intermediate feature map.

[0096] A batch normalization layer is used to normalize the first intermediate feature map to obtain a second intermediate feature map;

[0097] In an embodiment of the present invention, the first intermediate feature map is normalized by a batch normalization layer to obtain a second intermediate feature map.

[0098] The second intermediate feature map is randomly discarded through the discard layer to obtain the second feature map.

[0099] In order to prevent overfitting, the second intermediate feature map is randomly discarded through the discard layer to obtain the second feature map.

[0100] Using a fourth extraction module to perform feature extraction on the second feature map to obtain a third feature map, wherein the fourth extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, and a fully connected layer connected in sequence;

[0101] In the embodiment of the present invention, feature extraction is performed on the second feature map in sequence through a fully connected layer, a Swish activation layer, a batch normalization layer and a fully connected layer to obtain a third feature map.

[0102] The PLTanh activation layer is used to perform defect detection on the third feature map to obtain the corresponding detection results of the carbon fiber wire.

[0103] In the embodiment of the present invention, the PLTanh activation layer When x≥0, f(x)=x, when x<0, f(x)=α(e x -1);

[0104] Defect detection is performed on the third characteristic graph to obtain the detection results corresponding to the carbon fiber wire (normal wire, slight break, moderate break, severe break, slight less pressure, moderate less pressure, severe less pressure).

[0105] The step of extracting features from the wire feature map by a third extraction module to obtain a second feature map includes sequentially performing nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer to obtain a first intermediate feature map;

[0106] A batch normalization layer is used to normalize the first intermediate feature map to obtain a second intermediate feature map;

[0107] The second intermediate feature map is randomly discarded through the discard layer to obtain the second feature map.

[0108] S3: Extract features from the detection image through the feature extraction network to generate a wire feature map.

[0109] An extraction module 303 is used to extract features from the detection image through a feature extraction network to generate a wire feature map;

[0110] The detection module 304 is used to perform defect detection on the wire feature map using a defect classification network to obtain a detection result corresponding to the carbon fiber wire.

[0111] Furthermore, the acquisition module 301 includes:

[0112] A magnetic processing submodule, used for performing magnetic processing on the carbon fiber conductor to obtain a magnetically marked section;

[0113] The detection submodule is used to detect the magnetic mark section using a preset magnetic flux leakage detection device to obtain the detection speed and intermediate magnetic flux leakage data;

[0114] The first analysis submodule is used to determine whether the detection speed is within a preset speed range;

[0115] When the detection speed is within the speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and a detection image is constructed based on the magnetic flux leakage data;

[0116] When the detection speed is not within the speed range, the process jumps to the step of using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data.

[0117] Furthermore, the feature extraction network includes a first extraction module and two second extraction modules, and the extraction module 303 includes:

[0118] A first extraction submodule, used for performing feature extraction on the detection image through a first extraction module to generate a first feature map, wherein the first extraction module includes a 3×3 standard convolution layer, a 2×2 maximum pooling layer, a batch normalization layer, and a Swish activation layer connected in sequence;

[0119] The second extraction submodule is used to extract features from the first feature map through two second extraction modules in sequence to generate a wire feature map, wherein the second extraction module includes a 3×3 standard convolution layer, a 2×2 maximum pooling layer, a batch normalization layer and a discard layer connected in sequence.

[0120] Further, the defect classification network includes a third extraction module, a fourth extraction module and a PLTanh activation layer, and the detection module 304 includes:

[0121] A third extraction submodule, used for performing feature extraction on the wire feature map through a third extraction module to obtain a second feature map, wherein the third extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, and a discard layer connected in sequence;

[0122] A fourth extraction submodule, used for performing feature extraction on the second feature map by using a fourth extraction module to obtain a third feature map, wherein the fourth extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer and a fully connected layer connected in sequence;

[0123] The defect detection submodule is used to perform defect detection on the third feature map using the PLTanh activation layer to obtain a detection result.

[0124] Furthermore, the third extraction submodule includes:

[0125] A mapping unit, used to perform nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer in sequence to obtain a first intermediate feature map;

[0126] A normalization unit, used for normalizing the first intermediate feature map by using a batch normalization layer to obtain a second intermediate feature map;

[0127] The discarding unit is used to randomly discard the second intermediate feature map through the discarding layer to obtain the second feature map.

[0128] Furthermore, it also includes:

[0129] A training module is used to obtain training conductor data, perform data preprocessing on the training conductor data, and generate a conductor feature set;

[0130] Input the wire feature set into a preset initial carbon fiber wire classification model for training, and output the training wire detection data;

[0131] Calculate the training loss function value of the wire feature set according to the training wire detection data;

[0132] When the training loss function value is greater than or equal to the preset standard loss function value, a grid search method or a random search method is used to adjust the network parameters of the preset initial carbon fiber wire classification model until the training loss function value is less than the preset standard loss function value;

[0133] When the training loss function value is less than the preset standard loss function value, a carbon fiber wire classification model is generated.

[0134] S4: Use the defect classification network to perform defect detection on the wire feature map to obtain the detection result corresponding to the carbon fiber wire. Obtain training wire data, perform data preprocessing on the training wire data, and generate a wire feature set; input the wire feature set into a preset initial carbon fiber wire classification model for training, and output the training wire detection data; calculate the training loss function value of the wire feature set according to the training wire detection data; when the training loss function value is greater than or equal to the preset standard loss function value, adjust the network parameters of the preset initial carbon fiber wire classification model until the training loss function value is less than the preset standard loss function value; when the training loss function value is less than the preset standard loss function value, generate a carbon fiber wire classification model.

[0135] The accuracy of the model can be significantly improved by training the preset initial carbon fiber wire classification model with the training wire data and continuously adjusting the network parameters according to the training loss function value. The resulting carbon fiber wire classification model can better identify and classify the defect types of the wire. The model parameters are adjusted until the training loss function value is less than the preset standard loss function value. This process ensures that the model has high accuracy and stability in practical applications. False positives and false negatives are effectively reduced, and the reliability of the detection results is improved. By processing a large amount of training data, the generated classification model can adapt to defects of different types and severity. This makes the system have a high generalization ability and can cope with the defect detection needs in different working environments. The design scheme realizes the automation of the detection process. From data collection, preprocessing to model training and adjustment, the entire process can be automatically executed, which greatly improves the detection efficiency and reduces manual intervention and operation errors. By dynamically adjusting the network parameters and continuously optimizing the model performance, it is ensured that the detection effect can be maintained efficiently under different environments and conditions. This design enables the system to adapt to new data and changing detection requirements at any time. Using actual training data to optimize the model makes the classification model more practical in practical applications. The performance of the model is trained and verified based on real data, which can reflect the problems and needs in actual work. By accurately identifying and classifying the defects of carbon fiber conductors, potential problems can be discovered and handled in a timely manner, ensuring the stable operation of the power system, reducing safety accidents caused by conductor defects, and improving the safety and reliability of the entire distribution network.

[0136] On the other hand, this embodiment also provides a carbon fiber wire defect detection system, which includes:

[0137] An acquisition module is used to obtain magnetic flux leakage data of the carbon fiber conductor and generate a detection image according to the magnetic flux leakage data;

[0138] An analysis module, used for inputting the detection image into a preset carbon fiber wire classification model, wherein the carbon fiber wire classification model includes a feature extraction network and a defect classification network;

[0139] An extraction module, used for performing feature extraction on the detection image through the feature extraction network to generate a wire feature map;

[0140] The detection module is used to use the defect classification network to perform defect detection on the wire feature map to obtain a detection result corresponding to the carbon fiber wire.

[0141] The acquisition module includes a magnetic processing submodule, which is used for performing magnetic processing on the carbon fiber wire to obtain a magnetically marked section.

[0142] The detection submodule is used to detect the magnetic mark section using a preset magnetic flux leakage detection device to obtain the detection speed and intermediate magnetic flux leakage data.

[0143] The first analysis submodule is used to determine whether the detection speed is within a preset speed range;

[0144] The extraction module includes a first extraction submodule, which is used to extract features from the detection image through the first extraction module to generate a first feature map.

[0145] The second extraction submodule is used to extract features from the first feature map through two second extraction modules in sequence to generate a wire feature map.

[0146] The detection module includes a third extraction submodule, which is used to extract features from the wire feature map through the third extraction module to obtain a second feature map.

[0147] The fourth extraction submodule is used to use the fourth extraction module to perform feature extraction on the second feature map to obtain a third feature map.

[0148] The defect detection submodule is used to perform defect detection on the third feature map using the PLTanh activation layer to obtain a detection result.

[0149] The third extraction submodule includes a mapping unit, which is used to perform nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer in sequence to obtain a first intermediate feature map.

[0150] The normalization unit is used to normalize the first intermediate feature map by using a batch normalization layer to obtain a second intermediate feature map.

[0151] The discarding unit is used to randomly discard the second intermediate feature map through the discarding layer to obtain the second feature map.

[0152] A computer device comprises: a memory and a processor; the memory stores a computer program, wherein the processor implements the steps of any one of the methods of the present invention when executing the computer program.

[0153] If the above functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0154] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0155] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0156] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0157] Example 2

[0158] Experiments were conducted on carbon fiber conductors of different types and states. Six carbon fiber conductors of different states were selected, including normal conductors, slight fractures, moderate fractures, severe fractures, slight underpressure and severe underpressure. Ten samples of conductors of each state were prepared. The magnetic flux leakage detection equipment was calibrated to ensure the detection accuracy. The image processing software and the preset carbon fiber conductor classification model have been loaded into the computer system.

[0159] Place the carbon fiber conductor samples one by one in the magnetic flux leakage detection equipment for magnetic treatment. The magnetic flux leakage detection equipment moves within the preset speed range (0.4m / s-0.6m / s) to obtain magnetic flux leakage data. Generate a magnetic flux leakage data map for each sample. Based on the acquired magnetic flux leakage data, use image processing software to generate the corresponding detection image. The image resolution is set to 24x24 pixels. Collect the training data of the conductor, including the data of normal conductors and various defective conductors. Preprocess the training data to generate a conductor feature set. Input the feature set into the preset initial carbon fiber conductor classification model for training, and output the training detection data.

[0160] The training loss function value is calculated based on the training test data. If it is greater than or equal to the preset standard loss function value, the model parameters are adjusted until the loss function value is less than the preset standard.

[0161] The generated detection image is input into the trained feature extraction network, and the wire feature map is extracted through the first extraction module (including 3x3 standard convolution layer, 2x2 maximum pooling layer, batch normalization layer and Swish activation layer) and two second extraction modules (including 3x3 standard convolution layer, 2x2 maximum pooling layer, batch normalization layer and discard layer).

[0162] The defect classification network is used to detect the extracted wire feature map. The defect classification network includes the third extraction module (fully connected layer, Swish activation layer, batch normalization layer and discard layer) and the fourth extraction module (fully connected layer, Swish activation layer, batch normalization layer and fully connected layer). Finally, the feature map is classified through the PLTanh activation layer to obtain the detection result. The experimental results are shown in Table 1.

[0163] Table 1 Experimental results

[0164]

[0165] It can be seen from the experimental data that high accuracy and consistency are shown when detecting defects in carbon fiber conductors in different states. The detection speed of all samples is within the preset range of 0.4m / s to 0.6m / s, ensuring the reliability of the leakage magnetic data. The results of the mean and variance of the leakage magnetic data show that there are significant differences in the leakage magnetic data of conductors in various states, which provides an effective basis for subsequent feature extraction and defect detection.

[0166] The feature map of the normal sample has the highest similarity, reaching 0.98, which shows that the feature extraction network can extract the features of normal wires well.

[0167] The similarities of the characteristic graphs of samples with slight fracture, moderate fracture and severe fracture are 0.95, 0.92 and 0.88 respectively, showing that the characteristics of wires with different fracture degrees gradually change. The defect classification network can accurately distinguish wires with different fracture degrees, and the detection results are completely consistent with the actual defect type.

[0168] The similarities of the feature maps for the mildly under-stressed and severely under-stressed samples are 0.96 and 0.93, respectively, which also shows the extent to which the classification network can accurately identify under-stress defects.

[0169] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for detecting defects in a carbon fiber conductor, characterized in that: include: Obtaining magnetic flux leakage data of the carbon fiber conductor and generating a detection image based on the magnetic flux leakage data; Input the detection image into a preset carbon fiber wire classification model; Extract features from the detection image through a feature extraction network to generate a wire feature map; The defect classification network is used to perform defect detection on the wire feature map and obtain the corresponding detection results of the carbon fiber wire.

2. The carbon fiber wire defect detection method according to claim 1, characterized in that: The steps of obtaining magnetic flux leakage data of the carbon fiber conductor and generating a detection image according to the magnetic flux leakage data include: the carbon fiber conductor classification model includes a feature extraction network and a defect classification network; Performing magnetic treatment on the carbon fiber conductor to obtain a magnetically marked section; Using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data; Determining whether the detection speed is within a preset speed range; When the detection speed is within the speed range, the intermediate magnetic flux leakage data is determined as the magnetic flux leakage data of the carbon fiber conductor, and a detection image is generated according to the magnetic flux leakage data; When the detection speed is not within the speed range, the steps of using a preset magnetic flux leakage detection device to detect the magnetic mark section to obtain the detection speed and intermediate magnetic flux leakage data are performed until the detection speed is within the speed range.

3. The carbon fiber wire defect detection method according to claim 2, characterized in that: The feature extraction network includes a first extraction module and two second extraction modules. The feature extraction network is used to extract features from the detection image to generate a wire feature map, which includes extracting features from the detection image through the first extraction module to generate a first feature map, wherein the first extraction module includes a 3×3 standard convolution layer, a 2×2 maximum pooling layer, a batch normalization layer, and a Swish activation layer connected in sequence; The first feature map is sequentially subjected to feature extraction by two second extraction modules to generate a wire feature map, wherein the second extraction module includes a 3×3 standard convolutional layer, a 2×2 maximum pooling layer, a batch normalization layer, and a discard layer connected sequentially.

4. The carbon fiber wire defect detection method according to claim 3, characterized in that: The defect classification network includes a third extraction module, a fourth extraction module and a PLTanh activation layer. The step of using the defect classification network to perform defect detection on the wire feature map to obtain a detection result corresponding to the carbon fiber wire includes: Performing feature extraction on the wire feature map by the third extraction module to obtain a second feature map, wherein the third extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer, and a discard layer connected in sequence; Using a fourth extraction module to perform feature extraction on the second feature map to obtain a third feature map, wherein the fourth extraction module includes a fully connected layer, a Swish activation layer, a batch normalization layer and a fully connected layer connected in sequence; The PLTanh activation layer is used to perform defect detection on the third feature map to obtain the corresponding detection results of the carbon fiber wire.

5. The carbon fiber wire defect detection method according to claim 4, characterized in that: The step of extracting features from the wire feature map by a third extraction module to obtain a second feature map includes sequentially performing nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer to obtain a first intermediate feature map; A batch normalization layer is used to normalize the first intermediate feature map to obtain a second intermediate feature map; The second intermediate feature map is randomly discarded through the discard layer to obtain the second feature map.

6. The carbon fiber wire defect detection method according to claim 5, characterized in that: Inputting the detection image into a preset carbon fiber wire classification model also includes acquiring training wire data, performing data preprocessing on the training wire data, and generating a wire feature set; Input the wire feature set into a preset initial carbon fiber wire classification model for training, and output the training wire detection data; Calculating a training loss function value of the wire feature set according to the training wire detection data; When the training loss function value is greater than or equal to the preset standard loss function value, the network parameters of the preset initial carbon fiber wire classification model are adjusted until the training loss function value is less than the preset standard loss function value; When the training loss function value is less than the preset standard loss function value, a carbon fiber wire classification model is generated.

7. A computing platform load balancing system based on a particle swarm genetic algorithm using the method according to any one of claims 1 to 6, characterized in that: An acquisition module is used to obtain magnetic flux leakage data of the carbon fiber conductor and generate a detection image according to the magnetic flux leakage data; An analysis module, used for inputting the detection image into a preset carbon fiber wire classification model, wherein the carbon fiber wire classification model includes a feature extraction network and a defect classification network; An extraction module, used for performing feature extraction on the detection image through the feature extraction network to generate a wire feature map; The detection module is used to use the defect classification network to perform defect detection on the wire feature map to obtain a detection result corresponding to the carbon fiber wire.

8. A computing platform load balancing system based on a particle swarm genetic algorithm using the method as claimed in any one of claims 7, characterized in that: The acquisition module includes a magnetic processing submodule for performing magnetic processing on the carbon fiber conductor to obtain a magnetically marked section; The detection submodule is used to detect the magnetic mark section using a preset magnetic flux leakage detection device to obtain the detection speed and intermediate magnetic flux leakage data; The first analysis submodule is used to determine whether the detection speed is within a preset speed range; The extraction module includes a first extraction submodule, which is used to extract features from the detection image through the first extraction module to generate a first feature map; A second extraction submodule is used to extract features from the first feature map through two second extraction modules in sequence to generate a wire feature map; The detection module includes a third extraction submodule, which is used to extract features from the wire feature map through the third extraction module to obtain a second feature map; A fourth extraction submodule, configured to extract features from the second feature map using the fourth extraction module to obtain a third feature map; A defect detection submodule, used for performing defect detection on the third feature map by using a PLTanh activation layer to obtain a detection result; The third extraction submodule includes a mapping unit, which is used to perform nonlinear mapping on the wire feature map through a fully connected layer and a Swish activation layer in sequence to obtain a first intermediate feature map; A normalization unit, used for normalizing the first intermediate feature map by using a batch normalization layer to obtain a second intermediate feature map; The discarding unit is used to randomly discard the second intermediate feature map through the discarding layer to obtain the second feature map.

9. A computer device comprising: Memory and processor; The memory stores a computer program, and is characterized in that the processor implements the steps of the carbon fiber wire defect detection method when executing the computer program.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the carbon fiber wire defect detection method are implemented.

Images