Method and device for identifying broken strand defects of steel-cored aluminum stranded wire based on eddy current detection

Abstract

The application discloses a steel-cored aluminum stranded conductor broken strand defect identification method and device based on eddy current detection, relates to the technical field of steel-cored aluminum stranded conductor broken strand defect identification, and comprises the following steps: installing an annular vector coil array probe composed of a plurality of excitation coils on a steel-cored aluminum stranded conductor to be measured, constructing a vectorized excitation magnetic field on the surface of the steel-cored aluminum stranded conductor by controlling the driving parameters of the excitation coils; based on the vectorized excitation magnetic field, a multi-layer selective penetration strategy is constructed, and a set of excitation modes for detecting surface aluminum strands, inner layer aluminum strands and the surrounding area of the steel core is sequentially generated; eddy current signals are collected under each excitation mode and layered feature extraction is performed to form corresponding hierarchical feature vectors; fusion discrimination is performed based on the hierarchical feature vectors to generate an identification result of the broken strand defect of the steel-cored aluminum stranded conductor; and the application solves the problems of excitation magnetic field distortion, signal aliasing and insufficient recognition sensitivity of deep defects in the traditional eddy current detection in a multi-layer stranded structure.

Description

A Method and Equipment for Identifying Stranded Defects in Steel-Cored Aluminum Stranded Wire Based on Eddy Current Detection Technical Field

[0001] This invention relates to the field of steel-cored aluminum stranded wire strand breakage defect identification technology, and more specifically, to a method and device for steel-cored aluminum stranded wire strand breakage defect identification based on eddy current detection. Background Technology

[0002] Aluminum Conductor Steel Reinforced (ACSR) is the most widely used conductor structure in overhead transmission lines. It consists of multiple layers of aluminum strands twisted together with a central steel core, possessing excellent conductivity, high mechanical strength, and strong weather resistance. However, during long-term operation, ACSR is inevitably affected by factors such as wind vibration, icing, temperature cycling, electromagnetic forces, and corrosion, making it prone to structural damage such as surface aluminum strand breakage, fatigue cracks in the inner aluminum strands, weakened steel core corrosion, and localized loosening of strands. Among these, strand breakage defects can significantly reduce the conductor's mechanical properties, potentially leading to safety accidents such as line tripping and conductor breakage. Therefore, conducting highly sensitive and reliable online or offline strand breakage detection is of significant engineering importance.

[0003] Existing technologies for defect detection in steel-cored aluminum stranded wire mainly include visual recognition, acoustic emission detection, magnetic memory detection, and eddy current detection. Among these, eddy current detection is considered an important method for identifying broken strand defects due to its advantages such as not requiring wire disassembly, sensitivity to conductive structures, and suitability for non-contact rapid scanning. However, when eddy current detection is applied to conductor structures like steel-cored aluminum stranded wire, which involves multi-layer stranding, dissimilar materials, and complex surface geometry, several technical bottlenecks remain.

[0004] First, the multi-layered stranded structure of steel-cored aluminum stranded wire results in a periodic, irregular geometric shape on its surface. The conductor cross-section position changes continuously with the strand pitch, causing varying degrees of disturbance to the excitation magnetic field on the stranded wire surface, leading to uneven eddy current density distribution. Traditional eddy current probes typically use single-coil excitation, with a fixed magnetic field direction and lacking adaptive capability. This makes them highly susceptible to electromagnetic field distortion when dealing with complex stranded structures. This distortion not only causes significant differences in detection signals for the same defect at different angles but also introduces a large number of geometrically related spurious signals, severely impacting the stability and repeatability of the detection results.

[0005] Secondly, due to the significant differences in material properties between aluminum strands and steel cores (aluminum has high electrical conductivity, while steel has high magnetic permeability), the eddy current field exhibits complex coupling and scattering effects in multilayer media. The inner aluminum strand and steel core regions are shielded by the surface layer, making it difficult for the excitation magnetic field to fully penetrate to the deeper layers. This results in traditional eddy current testing being significantly less sensitive to inner strand breaks, hidden cracks, and steel core damage. To improve deep penetration, the excitation frequency generally needs to be reduced, but this leads to decreased surface sensitivity and increased signal noise, making it difficult to simultaneously detect shallow and deep defects. Summary of the Invention

[0006] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide a method and device for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection. This method achieves highly sensitive, layered identification of multi-layer broken strand defects in steel-cored aluminum stranded wire, thereby solving the problems of insufficient sensitivity in traditional eddy current detection for identifying excitation magnetic field distortion, signal aliasing, and deep defects in multi-layer stranded structures.

[0007] To achieve the above objectives, the present invention provides the following technical solution:

[0008] In a first aspect, this application provides a method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection. The method includes: installing a ring-shaped vector coil array probe on the surface of the steel-cored aluminum stranded wire to be tested; generating a vectorized excitation magnetic field with adjustable direction and distribution on the stranded wire surface by controlling the driving parameters of each excitation coil in the probe; acquiring a set of excitation modes for the surface aluminum strands, inner aluminum strands, and the surrounding area of ​​the steel core based on the vectorized excitation magnetic field; acquiring eddy current signals under each excitation mode and extracting corresponding layered features to form a hierarchical feature vector; and performing fusion discrimination based on the hierarchical feature vector to generate an identification result for broken strand defects in the steel-cored aluminum stranded wire.

[0009] In one embodiment, based on a vectorized excitation magnetic field, a set of excitation modes for the surface aluminum strands, inner aluminum strands, and the surrounding area of ​​the steel core is obtained through a multi-layer selective penetration strategy; the multi-layer selective penetration strategy adjusts the frequency, phase, and current amplitude of the excitation magnetic field to selectively penetrate magnetic field energy into strand layers of different depths.

[0010] In one embodiment, a vectorized excitation magnetic field with adjustable direction and distribution is generated on the surface of the stranded wire by controlling the driving parameters of each excitation coil in the probe. This includes: determining the driving parameters of the amplitude and phase of each excitation coil in the annular vector coil array; performing amplitude modulation and phase calibration on each excitation coil based on the driving parameters; synchronously exciting each excitation coil with the driving current of the calibrated parameters, so that the magnetic field generated by it is superimposed and synthesized in the steel-cored aluminum stranded wire to form a directional excitation magnetic field with a predetermined direction; acquiring the actual distribution data of the directional excitation magnetic field, and adjusting the driving parameters of the excitation coil in real time accordingly to generate the vectorized excitation magnetic field.

[0011] In one embodiment, generating a vectorized excitation magnetic field further includes: acquiring the actual spatial distribution of the magnetic field generated by the excitation coil; comparing the spatial distribution of the actual magnetic field with the preset target magnetic field spatial distribution to obtain distribution deviation data; based on the distribution deviation data, calculating the required adjustment amount of the driving parameters for each excitation coil through inverse solving and constraint optimization; updating the driving parameters of each excitation coil according to the adjustment amount and performing synchronous excitation to generate a vectorized excitation magnetic field that conforms to the target distribution.

[0012] In one embodiment, the construction steps of the multilayer selective penetration strategy include: acquiring the structural characteristics of the steel-cored aluminum stranded wire to be tested and establishing a parameterized geometric model of the steel-cored aluminum stranded wire; based on the parameterized geometric model, performing pre-simulation to calculate the multilayer magnetic field distribution under different excitation parameter combinations and establishing a mapping relationship between excitation parameters and magnetic field distribution; and according to the mapping relationship, matching and generating the corresponding optimal excitation parameter combination for the steel-cored aluminum stranded wire with specific structural characteristics to constitute its multilayer selective penetration strategy.

[0013] In one embodiment, the excitation mode set includes a circumferential concentrated field mode, a radial gradient field mode, and an axial penetration field mode: the circumferential concentrated field mode is generated by controlling the drive current of each excitation coil to have a specific amplitude distribution and phase synchronization in the circumferential direction; the radial gradient field mode is generated by making the amplitude of the drive current of each excitation coil vary in a gradient according to its radial position in the array; the axial penetration field mode is generated by adjusting the amplitude and phase parameters of the drive current of each excitation coil along the axial direction of the steel-cored aluminum stranded wire.

[0014] In one embodiment, eddy current signals are acquired under various excitation modes, and corresponding hierarchical features are extracted to form hierarchical feature vectors. This includes: acquiring eddy current response signals generated by the steel-cored aluminum stranded wire under various excitation modes; preprocessing the eddy current response signals and constructing a signal matrix; performing signal separation on the signal matrix based on maximizing statistical independence to obtain multiple independent components; matching the feature information of the multiple independent components with surface, inner, and deep signal features respectively to form hierarchical signal sets; and extracting features based on each hierarchical signal set to form hierarchical feature vectors reflecting the different depth structural states of the steel-cored aluminum stranded wire.

[0015] In one embodiment, the identification result of broken strand defects in steel-cored aluminum stranded wire is generated by fusion discrimination based on hierarchical feature vectors, including: normalizing the feature vectors at each level; obtaining the optimal projection matrix through linear discriminant analysis based on the normalized feature vectors; fusing the feature vectors at each level into a comprehensive feature vector using the optimal projection matrix; and determining the defect based on the comprehensive feature vector to output the identification result of broken strand defects in steel-cored aluminum stranded wire.

[0016] In one embodiment, defect determination is performed based on the comprehensive feature vector, and the result of identifying broken strand defects in steel-cored aluminum stranded wire is output. This includes: dividing the comprehensive feature vector into multiple feature blocks according to the detection level from which it originates and performing hierarchical encoding to form hierarchical embedding vectors of different depths; extracting intra-layer discriminant features from each hierarchical embedding vector through the corresponding pathway in a multi-path neural network; weighting and fusing the intra-layer discriminant features output from each pathway in the fusion layer of the multi-path neural network to generate a weighted comprehensive feature vector; and inputting the weighted comprehensive feature vector into a classifier to output the identification result of the broken strand defect.

[0017] Secondly, this application provides an electronic device, comprising:

[0018] Memory, used to store computer programs;

[0019] A processor is used to execute the computer program to implement the steps of the aforementioned method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection.

[0020] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0021] A vectorized excitation magnetic field with controllable direction and adjustable distribution is constructed by using a ring vector coil array. Combined with a multi-layer selective penetration strategy, precise magnetic field coverage of the surface, inner layer, and surrounding area of ​​the steel core of the aluminum stranded wire is achieved. By utilizing ICA hierarchical extraction and hierarchical feature vector construction, the surface, inner, and deep layer signals are effectively separated. Through the fusion of normalization, discriminant subspace projection, multi-path neural network, and attention mechanism, adaptive weighted discrimination of multi-layer features is achieved. Finally, a multi-task classifier outputs a score for defect presence, hierarchical location, and severity. This enables the system to accurately and reliably identify deep and broken strand defects in complex multi-layer structures, significantly improving detection sensitivity, hierarchical resolution, and interpretability. It also has adaptability to different specifications of aluminum stranded wire and engineering feasibility. Attached Figure Description

[0022] Figure 1 is a schematic flowchart of the method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection provided in the embodiments of this application.

[0023] Figure 2 is a structural diagram of an electronic device provided in an embodiment of this application.

[0024] Figure 3 is a schematic diagram of the circumferential concentrated field mode structure provided in the embodiment of this application.

[0025] Figure 4 is a schematic diagram of the radial gradient field mode structure provided in the embodiment of this application.

[0026] Figure 5 is a schematic diagram of the axial penetration field mode structure provided in the embodiment of this application. Detailed Implementation

[0027] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0028] Referring to Figure 1, the flowchart of the method for identifying broken strands in steel-cored aluminum stranded wire based on eddy current detection provided by the present invention includes the following steps:

[0029] S1. A ring vector coil array probe consisting of several excitation coils is installed on the steel-cored aluminum stranded wire to be tested. By controlling the driving parameters of each excitation coil, a vectorized excitation magnetic field with controllable direction and adjustable distribution is constructed on the surface of the steel-cored aluminum stranded wire, so that the excitation magnetic field initially covers the surface and inner layer areas of the steel-cored aluminum stranded wire.

[0030] The ring probe structure with multiple independent excitation coils is fitted around the steel-cored aluminum stranded wire to be tested, so that the multiple excitation coils are distributed at a uniform angle to form a closed coil array around the steel-cored aluminum stranded wire. This ensures that each excitation coil has an independent electromagnetic drive path, providing a multi-dimensional adjustment basis for the subsequent construction of a controllable excitation magnetic field. Each excitation coil can independently adjust the excitation amplitude, phase and frequency.

[0031] In this embodiment, a ring-shaped vector coil array probe consisting of several excitation coils is installed on the steel-cored aluminum stranded wire to be tested. By controlling the driving parameters of each excitation coil, a vectorized excitation magnetic field with controllable direction and adjustable distribution is constructed on the surface of the steel-cored aluminum stranded wire, including:

[0032] In the ring vector coil array probe, a driving model is established for each excitation coil. The driving model includes the coil resistance, inductance, and mutual inductance coupling coefficient with neighboring coils. The driving model parameters are calibrated by applying a preset current and acquiring the actual magnetic field.

[0033] The driving model is specifically as follows:

[0034] ;

[0035] In the formula, The magnetic field generated by the i-th coil, To excite the current, For phase, For inductance, For resistance, The mutual inductance coupling coefficient is used to describe the magnetic coupling relationship between the i-th coil and the j-th coil. This refers to the steel core effect, which indicates the shielding effect of the steel core on the magnetic field.

[0036] Among them, the driving model It provides a mapping of "each coil input → magnetic field output".

[0037] It's important to note that the advantage of constructing a driving model lies in its transformation of the previously empirical, trial-and-error multi-coil driving method into a calculable, predictable, and controllable mathematical model. By establishing independent amplitude-phase parameter representations for each excitation coil, the system can clearly define the "contribution relationship" of each coil to the spatial magnetic field, achieving a traceable mapping from "what magnetic field to obtain" to "how each coil should be driven." Thus, in subsequent calculations of amplitude ratios and phase distributions, all solution processes are built upon this model, ensuring that the calculated parameter sets have physical meaning, are executable, and stable. This guarantees that the final synthesized excitation magnetic field has the correct direction, precise amplitude, and controllable distribution, improving the adjustability, accuracy, and engineering reliability of the entire eddy current detection system.

[0038] Based on the driving model, and according to the direction and spatial distribution of the preset target magnetic field, the driving parameters, including the amplitude and phase of each coil, are calculated by an optimization algorithm to ensure that a structured and controllable magnetic field close to the target magnetic field is formed through vector superposition.

[0039] The amplitude and phase of each coil are calculated using an optimized algorithm. The specific calculation formula is as follows:

[0040] ;

[0041] In the formula, The target magnetic field, including its direction, amplitude, and spatial distribution, is the objective of magnetic field construction. To obtain the total magnetic field by superimposing the actual magnetic field vectors generated by each coil, the adjustment is made... and Make it close , To adjust the amplitude of each coil and phase To minimize the difference between the target magnetic field and the actual synthesized magnetic field.

[0042] The driving parameters are loaded into each excitation coil, and the amplitude is modulated and the phase is calibrated for each excitation coil so that each coil can output a stable current according to the preset amplitude and start excitation under the preset phase.

[0043] After amplitude modulation and phase calibration, the calibrated drive current is simultaneously output to all excitation coils, so that multiple excitation coils are synchronously excited under the same time reference. Each excitation coil generates a corresponding local magnetic field component in space with its independent amplitude and phase, and electromagnetic coupling superposition occurs on the surface and inside of the steel-cored aluminum stranded wire to form a directional excitation magnetic field with a predetermined direction.

[0044] The actual distribution data of the directional excitation magnetic field is obtained and the excitation coil is adjusted in real time to generate a vectorized excitation magnetic field.

[0045] Furthermore, the actual distribution data of the directional excitation magnetic field is acquired and the excitation coil is adjusted in real time to generate a vectorized excitation magnetic field, including:

[0046] Based on the directional excitation magnetic field, the spatial distribution of the actual magnetic field is collected. The spatial distribution of the actual magnetic field includes the magnetic field amplitude, direction and gradient data at different spatial locations of the steel-cored aluminum stranded wire.

[0047] The spatial distribution of the target magnetic field is compared with the spatial distribution of the actual magnetic field to obtain the distribution deviation data, including magnetic field amplitude error, direction error and spatial gradient error.

[0048] A sensitivity matrix J is established based on the driving model. The sensitivity matrix describes the contribution mapping relationship between the amplitude and phase changes of each excitation coil to the distortion data of the spatial magnetic field.

[0049] The distribution deviation data is combined with the sensitivity matrix and input into a preset regularized inverse solution model to construct a weighted least squares optimization problem. By solving the weighted least squares optimization problem, the driving parameter adjustment amount of each excitation coil is obtained. The driving parameter adjustment amount includes the adjustment amount of amplitude and phase.

[0050] The regularized inverse solution model is as follows:

[0051] ;

[0052] In the formula, Let W represent the adjustment amount for the amplitude and phase of each excitation coil, and let W be the spatial weighting matrix. As a regularization factor, To optimize the variable vector, used to solve for the adjustment amount, This is the sensitivity matrix. This is distorted data.

[0053] The solution can be obtained by closed-loop analytical, pseudo-inverse, or iterative numerical optimization algorithms to ensure a stable and implementable adjustment amount under high coupling of multiple coils.

[0054] The amplitude and phase adjustments are superimposed on the original driving parameters to obtain the updated excitation current amplitude and phase, which are then synchronously excited to each excitation coil to generate a vectorized excitation magnetic field that conforms to the target distribution.

[0055] It should be noted that by transforming the multi-coil drive from an empirical, trial-and-error approach into a modelable, calculable, and controllable process, a clear mapping is established between the amplitude and phase adjustment of each coil and the spatial magnetic field distribution. Through regularized inverse solving combined with a sensitivity matrix, precise compensation for the amplitude, direction, and gradient of the spatial magnetic field is achieved. Even under complex structures of multi-layer steel-cored aluminum stranded wire and high coil coupling, a magnetic field with controllable direction and uniform distribution can be stably generated. This enables the system to achieve high-precision, controllable magnetic field superposition, providing a reliable excitation basis for subsequent eddy current detection, improving the sensitivity, accuracy, and engineering feasibility of defect identification, while ensuring the smoothness and closed-loop stability of the dynamic compensation process.

[0056] S2. Based on the vectorized excitation magnetic field, a multi-layer selective penetration strategy is constructed, and an excitation mode set for detecting the surface aluminum strands, inner aluminum strands and the surrounding area of ​​the steel core is generated sequentially according to the multi-layer selective penetration strategy. The excitation mode set includes a circumferential concentrated field mode, a radial gradient field mode and an axial penetration field mode.

[0057] In this embodiment, a multi-layer selective penetration strategy is constructed based on a vectorized excitation magnetic field, including:

[0058] The structural characteristics of the steel-cored aluminum stranded wire to be tested are obtained, and a parametric geometric model of the steel-cored aluminum stranded wire is established. The structural characteristics of the steel-cored aluminum stranded wire to be tested include the diameter of the outer aluminum strand, the diameter of the inner aluminum strand, the diameter of the steel core, the number of strands, the spacing between the aluminum strands in each layer, and the magnetic permeability and electrical conductivity of the material in each layer.

[0059] Among them, the parametric geometric model generates adjustable parameter sets for different structural types by changing geometric and material parameters, providing a model basis for steel-cored aluminum stranded wires of different specifications. The parametric geometric model is a mathematical and adjustable expression of the physical structure and material properties of the steel-cored aluminum stranded wire under test. It constructs a unified three-dimensional geometric model by parametrically defining key structural and material parameters such as the diameter of the outer aluminum strands, the diameter of the inner aluminum strands, the diameter of the steel core, the number of strands, the spacing between each aluminum strand, and the permeability and conductivity of each layer. This model allows for the generation of steel-cored aluminum stranded wire models of different specifications and structural forms simply by adjusting parameter values ​​without changing the modeling method and solution process. This provides a clear, reproducible, and scalable geometric and physical basis for subsequent excitation parameter design, finite element electromagnetic field simulation, and multi-layer magnetic field distribution analysis, avoiding the difficulty in implementing technical solutions due to model reliance on implicit assumptions.

[0060] Based on the parametric geometric model, the driving current amplitude, phase and frequency combination of each excitation coil in the ring vector coil array is defined, and each combination is associated with the corresponding parametric geometric model structure to form an excitation parameter set;

[0061] The parametric geometric model is simulated using the finite element analysis method. The electromagnetic field solver is used to calculate the multi-layer magnetic field distribution under different excitation parameter combinations. The multi-layer magnetic field distribution includes the surface aluminum strand magnetic field, the inner aluminum strand magnetic field, and the magnetic field of the surrounding area of ​​the steel core. The magnetic field strength, gradient, and direction information of each layer are recorded.

[0062] Three-dimensional magnetic field data are generated for each combination of excitation parameters to evaluate coverage, gradient uniformity, and penetration capability.

[0063] The three-dimensional magnetic field data is processed and parameterized to establish a mapping relationship between excitation parameters and magnetic field distribution. Each set of excitation parameters in the mapping relationship corresponds to the magnetic field characteristics and coverage effect of different layer regions.

[0064] Based on the mapping relationship between excitation parameters and magnetic field distribution, and combined with the structural characteristics of the steel-cored aluminum strands under test, the optimal combination of excitation parameters is matched to form a multi-layer selective penetration strategy targeting the surface aluminum strands, inner aluminum strands, and the area surrounding the steel core.

[0065] The multi-layer selective penetration strategy refers to a scheme that, by controlling the current amplitude, phase, and frequency parameters of the excitation coil, allows the vectorized excitation magnetic field to selectively and layerwise penetrate different regions of the steel-cored aluminum stranded wire (outer aluminum strands, inner aluminum strands, and the periphery of the steel core), ensuring effective coverage of each layer of magnetic field without interference. Three-dimensional magnetic field data refers to the magnetic field vector information at each point in the space of the steel-cored aluminum stranded wire, obtained through finite element simulation calculations. This includes magnetic field strength, magnetic field direction, gradient information, coverage effect indicators (whether each layer of magnetic field reaches the designed target strength range), and penetration capability indicators (the attenuation of the magnetic field in deeper layers or the steel core region). It is the core data for evaluating the effectiveness of the multi-layer selective penetration strategy.

[0066] It should be noted that by constructing a multi-layer selective penetration strategy, precise control of the magnetic field in different layers of aluminum steel-cored stranded wire (surface aluminum strands, inner aluminum strands, and the periphery of the steel core) can be achieved, ensuring that the excitation magnetic field achieves optimal coverage in each layer without interference. This significantly improves the sensitivity and accuracy of multi-layer eddy current detection, while reducing mutual interference between deep or surface magnetic field signals. This provides a reliable basis for the layered identification of strand breakage defects and has adaptability and repeatability for aluminum steel-cored stranded wires of different specifications and structures.

[0067] Furthermore, based on the multi-layer selective penetration strategy, a set of incentive patterns is sequentially generated for detecting the surface aluminum strands, inner aluminum strands, and the area surrounding the steel core, including:

[0068] As shown in Figure 3, 1 represents the surface aluminum strand, 2 represents the steel core, 3 represents the annular vector coil, and 4 represents the direction of the circumferential concentrated magnetic field. Based on the multi-layer selective penetration strategy, by controlling the current amplitude of each excitation coil in the annular vector coil array to be distributed circumferentially and synchronized in phase, the magnetic field forms a circumferential concentrated effect on the surface aluminum strand, thereby improving the sensitivity to surface defects and generating a circumferential concentrated field mode for surface aluminum strand detection.

[0069] In this method, based on the multi-layer selective penetration strategy, for the detection of surface aluminum strands, the excitation parameters of the annular vector coil array can be determined sequentially. By adjusting the current amplitude and phase of each excitation coil, the magnetic field can be concentrated in a ring on the surface of the aluminum strands.

[0070] As shown in Figure 4, 1 represents the surface aluminum strand, 5 represents the inner aluminum strand, 2 represents the steel core, 6 represents the radial gradient coil, and 7 represents the radial gradient magnetic field direction. Based on the circumferential concentrated field mode, the radial gradient field mode is generated by making the driving current amplitude of each excitation coil change in a gradient according to its radial position in the array. That is, by adjusting the current amplitude of each excitation coil to decrease or increase with the radius, a radial magnetic field gradient from the surface to the inner layer is formed to achieve selective penetration of defects in the inner aluminum strand and generate a radial gradient field mode for detecting inner aluminum strands.

[0071] As shown in Figure 5, 8 represents the aluminum strand layer, 2 represents the steel core, 9 represents the axial vector coil, and 10 represents the axial penetrating magnetic field. Based on the radial gradient field mode, by adjusting the amplitude and phase of the coil driving current along the steel core axis, an axially uniformly penetrating excitation magnetic field is formed, enabling the magnetic flux to effectively cover the surrounding area of ​​the steel core, thereby improving the penetration capability of steel core strand breakage defect detection and generating an axial penetrating field mode for steel core surrounding area detection.

[0072] The circumferential concentrated field mode, radial gradient field mode, and axial penetration field mode are switched sequentially according to a preset order to generate an excitation mode set.

[0073] It should be noted that, through a multi-layer selective penetration strategy, three excitation modes—circumferential concentrated field, radial gradient field, and axial penetration field—are generated sequentially and switched in a preset order to form a set of excitation modes. This enables the magnetic field to selectively, controllably, and efficiently cover the surface, inner layer, and the area surrounding the steel core. This not only significantly improves the detection sensitivity and penetration capability of defects in each layer but also reduces interference between magnetic fields in different layers. It achieves accurate layered identification of multi-layer broken strand defects in steel-cored aluminum stranded wire and also has adaptability and repeatability to different structural specifications.

[0074] S3. Under each excitation mode, eddy current signals are acquired and layered feature extraction is performed to form corresponding layered feature vectors. The layered feature vectors include surface feature vectors, inner feature vectors, and deep feature vectors.

[0075] In this embodiment, eddy current signals are acquired under each excitation mode and hierarchical feature extraction is performed to form corresponding hierarchical feature vectors, including:

[0076] Eddy current response signals generated by steel-cored aluminum stranded wire under various excitation modes are collected using eddy current sensors.

[0077] The acquired eddy current signal is preprocessed, including noise reduction, filtering, normalization and sampling synchronization processing.

[0078] The preprocessed eddy current signal is constructed into a signal matrix based on time or spatial sampling points;

[0079] The covariance matrix of the signal matrix is ​​obtained and its eigenvalues ​​are decomposed. The signal matrix is ​​then transformed into a whitening matrix with equal and uncorrelated variances for each channel through whitening transformation, providing a basis for ICA decomposition.

[0080] The unmixing matrix is ​​initialized, and the fast independent component analysis (ICA) algorithm is used to maximize statistical independence through iterative optimization, so that the unmixing matrix gradually converges.

[0081] After updating the unmixing matrix in each iteration, the independent component matrix is ​​calculated in conjunction with the whitening matrix. ,in, For the iteratively updated unmixing matrix, This is the whitening matrix;

[0082] For each row or column in the independent component matrix, analyze its amplitude, spectral distribution, and spatial location corresponding to the excitation mode to obtain its characteristic information;

[0083] Based on the design objectives of different excitation modes, the feature information of multiple independent components is matched with the signal features of the surface, inner and deep layers respectively to form a hierarchical signal set;

[0084] Feature extraction is performed on signal sets at each level. The feature extraction includes time-domain features, frequency-domain features, and time-frequency-domain features. These features are then organized into surface feature vectors, inner feature vectors, and deep feature vectors to obtain hierarchical feature vectors.

[0085] It should be noted that by combining the ICA method with various excitation modes to perform layered processing of eddy current signals, effective separation of surface, inner, and deep signals is achieved. Even when the actual magnetic field has a non-ideal distribution, severe signal aliasing, or deviations in the multi-layer selective penetration strategy, the independent signal components of each layer can still be extracted through statistical independence, reducing layer aliasing and improving the identifiability of deep and steel core signals. This forms reliable surface, inner, and deep feature vectors, providing accurate, stable, and repeatable basic data for the layered identification of broken strand defects in steel-cored aluminum stranded wire.

[0086] S4. Based on the hierarchical feature vector, perform fusion discrimination to generate the identification result of broken strand defects in steel-cored aluminum stranded wire.

[0087] In this embodiment, based on the hierarchical feature vector, a fusion discrimination is performed to generate an identification result for broken strand defects in steel-cored aluminum stranded wire, including:

[0088] Normalize the feature vectors at each level to ensure that features at different levels are comparable on a numerical scale.

[0089] Based on the normalized feature vectors at each level and the corresponding preset defect category labels, the feature vectors of samples of the same category are stacked in columns, and the intra-class scatter matrix and inter-class scatter matrix are calculated. The intra-class scatter matrix reflects the feature variance of samples of the same category, and the inter-class scatter matrix reflects the difference between different category centers, providing a mathematical basis for linear discriminant analysis.

[0090] The specific formula for calculating the intra-class scatter matrix is ​​as follows:

[0091] ;

[0092] The specific formula for calculating the inter-class scatter matrix is ​​as follows:

[0093] ;

[0094] In the formula, The scatter matrix is ​​within the class. Let c be the inter-class scatter matrix, and c be the number of defect categories. Let i be the set of samples of class i. For sample feature vectors, Let i be the mean vector of the features of the i-th class. Let be the number of samples in class i. This is the vector of the population mean for all samples.

[0095] Constructing generalized eigenvalues By solving for the projection direction vector v and eigenvalues The optimal projection matrix is ​​obtained, which maximizes the class spacing and minimizes the intra-class variance after projection, thereby enhancing the contribution of different levels of features to defect discrimination.

[0096] Here, the projection direction vector represents the direction of projecting the original multidimensional features onto the discriminant subspace, and the eigenvalue represents the direction of projecting the original multidimensional features onto the discriminant subspace.

[0097] The normalized hierarchical feature vectors are projected onto the discriminant subspace defined by the optimal projection matrix to obtain the fused comprehensive feature vector, which reflects surface, inner and deep information simultaneously.

[0098] Defects are determined based on the comprehensive feature vector, and the results of identifying broken strands in steel-cored aluminum stranded wire are generated.

[0099] It should be noted that by using the aforementioned technical solution for hierarchical feature fusion, the complementary information of surface, inner, and deep signals can be fully utilized in the detection of broken strands in steel-cored aluminum stranded wire. This solves the problems of limited discrimination capability of single-layer features and the ease with which deep signals are masked by surface signals. By using normalization processing and a discrimination subspace constructed based on intra-class and inter-class scatter matrices, features at different levels are projected onto the optimal direction, effectively enhancing class discrimination. At the same time, by using comprehensive feature vectors, multi-dimensional outputs of defect existence, hierarchical location, and confidence are achieved, thereby significantly improving the accuracy, reliability, and interpretability of identifying deep and steel core periphery defects, meeting the precise detection requirements of complex multi-layered structures.

[0100] Furthermore, defect determination is performed based on the comprehensive feature vector, and the results of steel-cored aluminum stranded wire breakage defect identification are generated, including:

[0101] The comprehensive feature vector is divided into blocks according to the detection level of the source, resulting in surface feature blocks, inner feature blocks, and deep feature blocks, which are then labeled with surface, inner, and deep feature information respectively.

[0102] Each feature block is hierarchically encoded to form a hierarchical embedding vector, which is used to maintain the distinguishability of features at different levels in subsequent judgments.

[0103] Construct a multi-path neural network, where each path is input with hierarchical embedding vectors from the surface, inner, and deep layers respectively, and extract intra-layer discriminative features through independent convolutional layers, pooling layers, and fully connected layers;

[0104] The multi-path neural network comprises three pathways with identical or similar structures but independent parameters. Each pathway includes an independent convolutional layer: using a one-dimensional convolutional kernel with a kernel size of 3 and a number of 32, and the activation function is ReLU; a pooling layer: following the convolutional layer, employing max pooling or average pooling, with a pooling window size of 2; and a fully connected layer: flattening the pooled features before inputting them, which may contain 1-2 hidden layers, for example, with 128 and 64 neurons respectively.

[0105] The intra-layer discriminative features output from each pathway are aggregated in the fusion layer. Preset attention weights are introduced in the fusion layer for normalized weighted fusion to obtain a weighted comprehensive feature vector, which is used for subsequent judgment to improve the ability to identify sparse, low-amplitude or deep defects.

[0106] The weighted composite feature vector is input into a multi-task classifier, which includes a fully connected layer and a Softmax layer. The fully connected layer receives the weighted composite feature vector, and its output branches are three independent task heads: a defect existence probability head: connected to a fully connected layer with 2 neurons, followed by a Softmax function, outputting the probabilities of 'no defect' and 'defective'; a defect level head: connected to a fully connected layer with 3 neurons (corresponding to surface, inner, and deep layers), followed by a Softmax function, outputting the probability of the defect being located in each level (this output is only considered valid when the 'defective' probability exceeds a set threshold); and a defect severity score head: connected to a fully connected layer with 1 neuron, whose output is mapped to the (0,1) interval through a Sigmoid function and directly used as the score; the output is the steel-cored aluminum stranded wire broken strand defect identification result, which includes the broken strand defect existence probability, the defect level (surface, inner, or deep), and the defect severity score.

[0107] The training process utilizes steel-cored aluminum stranded wire test data with manual annotations or experimental verification. The annotation information includes at least: whether there is a strand breakage defect, the level of the defect, and a label or score indicating the severity of the defect. Multi-task training employs a weighted joint loss function, including a defect existence task (cross-entropy loss), a defect level task (multi-class cross-entropy loss), and a severity task (mean squared error loss). Each loss term is weighted and summed according to preset weights to obtain the total loss. The Adam optimizer is used to optimize the loss and minimize the total loss. The parameters are then updated using the backpropagation algorithm.

[0108] It should be noted that by performing feature fusion, the complementary information of surface, inner and deep features can be fully utilized to maintain the distinguishability of multi-level signals in the discriminative subspace. Intra-layer discriminative features are extracted through a multi-path neural network, and adaptive weighted fusion is achieved by combining an attention mechanism, thereby enhancing the sensitivity to low-amplitude, deep or sparse defects. The multi-task classifier simultaneously outputs the presence, hierarchical location and severity of defects, realizing multi-dimensional judgment, making the identification of broken strand defects in steel-cored aluminum stranded wire more accurate, reliable and interpretable, meeting the high-precision detection requirements of complex multi-layer structures.

[0109] Furthermore, this application also discloses an electronic device. FIG2 is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the figure should not be considered as any limitation on the scope of use of this application.

[0110] Figure 2 is a schematic diagram of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the eddy current detection-based method for identifying broken strands in steel-cored aluminum stranded wire disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be a computer.

[0111] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0112] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field-Programmable Gate Array), and PLA (Programmable Logic Array). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0113] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk or optical disk, etc. The resources stored thereon can include operating system 221, computer program 222, etc., and the storage method can be temporary storage or permanent storage.

[0114] The operating system 221 manages and controls the various hardware devices and computer programs 222 on the electronic device 20 to enable the processor 21 to perform calculations and processing on the massive amounts of data 223 in the memory 22. It can be Windows Server, Netware, Unix, Linux, etc. The computer program 222, in addition to including a computer program capable of performing the eddy current detection-based steel-cored aluminum stranded wire breakage defect identification method disclosed in any of the foregoing embodiments, may further include computer programs capable of performing other specific tasks. The data 223 may include data received by the electronic device from external devices, as well as data collected by its own input / output interface 25.

[0115] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.

[0116] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.

[0117] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0118] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0119] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0120] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying strand breakage defects in steel-cored aluminum stranded wire based on eddy current detection, characterized in that, The process includes the following steps: Installing a ring-shaped vector coil array probe on the surface of the aluminum core stranded wire to be tested; generating a vectorized excitation magnetic field with adjustable direction and distribution on the stranded wire surface by controlling the driving parameters of each excitation coil in the probe, wherein the driving parameters include amplitude and phase; based on the vectorized excitation magnetic field, obtaining a set of excitation modes for the surface aluminum strands, inner aluminum strands, and the surrounding area of ​​the steel core through a multi-layer selective penetration strategy, wherein the multi-layer selective penetration strategy selectively penetrates the magnetic field energy to strand layers of different depths by adjusting the frequency, phase, and current amplitude of the excitation magnetic field; the set of excitation modes includes a ring-shaped concentrated field mode generated by controlling the driving current of each excitation coil to have a specific amplitude distribution and synchronous phase in the ring direction, a radial gradient field mode generated by making the amplitude of the driving current of each excitation coil change in a gradient according to its radial position in the array, and an axial penetration field mode generated by adjusting the amplitude and phase parameters of the driving current of each excitation coil along the axial direction of the aluminum core stranded wire. Eddy current signals are acquired under various excitation modes, and corresponding hierarchical features are extracted to form hierarchical feature vectors. This process includes: acquiring eddy current response signals, preprocessing them, and constructing a signal matrix; performing signal separation based on maximizing statistical independence to obtain multiple independent components; matching the feature information of multiple independent components with surface, inner, and deep signal features to form hierarchical signal sets; extracting features based on each hierarchical signal set to form hierarchical feature vectors; and performing fusion discrimination based on the hierarchical feature vectors to generate identification results for broken strand defects in steel-cored aluminum stranded wires. This process includes: normalizing each level of feature vectors and obtaining the optimal projection matrix through linear discriminant analysis; fusing each level of feature vectors into a comprehensive feature vector using the optimal projection matrix; and determining defects based on the comprehensive feature vectors to output the identification results for broken strand defects in steel-cored aluminum stranded wires.

2. The method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection according to claim 1, characterized in that, The method of generating a vectorized excitation magnetic field with adjustable direction and distribution on the surface of the stranded wire by controlling the driving parameters of each excitation coil in the probe includes: determining the driving parameters of the amplitude and phase of each excitation coil in the annular vector coil array; performing amplitude modulation and phase calibration on each excitation coil based on the driving parameters; synchronously exciting each excitation coil with the driving current of the calibrated parameters, so that the magnetic field generated by it is superimposed and synthesized in the steel-cored aluminum stranded wire to form a directional excitation magnetic field with a predetermined direction; acquiring the actual distribution data of the directional excitation magnetic field, and adjusting the driving parameters of the excitation coil in real time accordingly to generate the vectorized excitation magnetic field.

3. The method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection according to claim 2, characterized in that, The generation of the vectorized excitation magnetic field further includes: acquiring the actual spatial distribution of the magnetic field generated by the excitation coil; comparing the spatial distribution of the actual magnetic field with the preset target magnetic field spatial distribution to obtain distribution deviation data; based on the distribution deviation data, calculating the required adjustment amount of the driving parameters for each excitation coil through inverse solving and constraint optimization; updating the driving parameters of each excitation coil according to the adjustment amount and performing synchronous excitation to generate a vectorized excitation magnetic field that conforms to the target distribution.

4. The method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection according to claim 1, characterized in that, The construction steps of the multi-layer selective penetration strategy include: acquiring the structural characteristics of the steel-cored aluminum stranded wire to be tested and establishing a parameterized geometric model of the steel-cored aluminum stranded wire; based on the parameterized geometric model, performing pre-simulation to calculate the multi-layer magnetic field distribution under different excitation parameter combinations and establishing a mapping relationship between excitation parameters and magnetic field distribution; and according to the mapping relationship, matching and generating the corresponding optimal excitation parameter combination for steel-cored aluminum stranded wires with specific structural characteristics to constitute their multi-layer selective penetration strategy.

5. The method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection according to claim 1, characterized in that, The step of determining defects based on comprehensive feature vectors and outputting the identification result of broken strand defects in steel-cored aluminum stranded wire includes: dividing the comprehensive feature vector into multiple feature blocks according to the detection level from which it originates and performing hierarchical encoding to form hierarchical embedding vectors of different depths; extracting intra-layer discriminative features from each hierarchical embedding vector through the corresponding pathway in a multi-path neural network; weighting and fusing the intra-layer discriminative features output by each pathway in the fusion layer of the multi-path neural network to generate a weighted comprehensive feature vector; and inputting the weighted comprehensive feature vector into a classifier to output the identification result of broken strand defects.

6. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor is configured to execute the computer program to implement the steps of the method for identifying broken strand defects in steel-cored aluminum stranded wire based on eddy current detection as described in any one of claims 1 to 5.

Images