Aluminum enameled flat wire online defect detection and self-adaptive deviation correction method and system

Abstract

The invention provides an aluminum enameled flat wire online defect detection and self-adaptive deviation correction method and system, and relates to the technical field of enameled wire defect detection.The aluminum enameled flat wire online defect detection and self-adaptive deviation correction method comprises the steps that an aluminum enameled flat wire is immersed in fluid with set viscosity, and a reference fluid resistance signal and a reference electromagnetic signal are collected; fluid resistance signals and electromagnetic induction signals are detected in real time; positioning defect positions of surface defects of the enameled wire; identifying types and distribution of internal defects of the aluminum enameled flat wire; distinguishing defect types through a machine learning algorithm; outputting a corresponding correction scheme through the pre-trained neural network model; and outputting the deviation correction control signal to the correction module for deviation correction operation. Through synchronous acquisition and analysis of fluid resistance signals and electromagnetic induction signals, synchronous detection of surface defects and internal defects is realized, and the limitation of a traditional single detection means is solved; by forming a high-precision three-stage classification system, an interpretability report is generated, and the classification accuracy of the composite defects is remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of enameled wire defect detection, and particularly relates to an online defect detection and self-adaptive deviation correction method and system for aluminum enameled flat wire. BACKGROUND

[0002] In the field of enameled wire manufacturing, as the core conductive component of electrical equipment such as motors and transformers, the quality of the surface insulation coating of enameled wire directly affects the safety and reliability of equipment operation. In the production process of enameled wire, multiple processes such as wire drawing, annealing, painting, and baking are required. The insulation coating formed after the curing of the paint layer may have microscopic defects such as bubbles, pinholes, impurity particles, uneven thickness, or paint film damage. The size of these defects is usually in the micron to sub-millimeter range, and the distribution is random. Traditional manual visual inspection is inefficient and subjective, and cannot meet the needs of large-scale production. With the improvement of industrial automation level, enameled wire defect detection technology based on machine vision has gradually become a research focus.

[0003] In the field of aluminum enameled flat wire manufacturing, with the increasing demand for lightweight and integration of conductor materials in high-end equipment such as new energy vehicle drive motors and high-efficiency transformers, aluminum enameled flat wire gradually replaces traditional copper enameled wire as a key conductive component due to its light weight, excellent conductivity, and cost advantage. In the production process of aluminum enameled flat wire, it needs to go through core processes such as drawing forming, annealing softening, multi-layer insulation paint coating, and high-temperature curing. The flat cross-sectional structure and aluminum substrate characteristics of the aluminum enameled flat wire make the paint film susceptible to mechanical stress, thermal field distribution, and differences in metal ductility during the curing process, resulting in microscopic defects such as scratches, bubbles, missed coating, edge burrs, or uneven paint film thickness on the surface. In view of the above problems, online defect detection and self-adaptive deviation correction technology has become a key link in the automation of aluminum enameled flat wire production.

[0004] In general, the online defect detection and deviation correction method for aluminum enameled flat wire usually needs to detect surface defects and internal defects in steps, which is low in efficiency and easy to miss compound defects, and is difficult to meet the high-precision quality control requirements. Existing deviation correction schemes mostly use fixed parameter adjustment, without considering the differences in defect types and dynamic deformation of the wire, resulting in lagging or excessive adjustment of the deviation correction action, which is easy to cause secondary damage or deviation correction failure, and is difficult to meet the real-time precision requirements in high-speed production scenarios.

[0005] To solve the defects in the prior art, the present technical scheme provides an online defect detection and self-adaptive deviation correction method and system for aluminum enameled flat wire. SUMMARY

[0006] The present application provides an online defect detection and self-adaptive deviation correction method and system for aluminum enameled flat wire to solve the defects in the prior art.

[0007] In one aspect, the present application provides an aluminum enameled flat wire online defect detection and adaptive correction method, comprising: S1: immerse the aluminum enameled flat wire in a fluid with a set viscosity, apply a high-frequency alternating current, and collect a reference fluid resistance signal and a reference electromagnetic signal; S2: based on the reference fluid resistance signal and the reference electromagnetic signal, pass the aluminum enameled flat wire through the detection area at a constant speed, and real-time detect the fluid resistance signal and the electromagnetic induction signal; S3: based on the fluid resistance signal, locate the defect position of the surface defect of the enameled wire in combination with the reference fluid resistance signal, output the surface defect data; based on the defect position, analyze the frequency spectrum characteristics of the electromagnetic induction signal in combination with the reference electromagnetic signal, identify the type and distribution of the internal defect of the aluminum enameled flat wire, and output the internal defect data; S4: match the surface defect data and the internal defect data in time and space, distinguish the defect type through a machine learning algorithm; and establish a defect feature matrix according to the defect type, and output a corresponding correction scheme through a pre-trained neural network model; S5: based on the correction scheme, output a correction control signal to a correction module for correction operation.

[0008] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the present application, the step S1 of collecting the reference fluid resistance signal and the reference electromagnetic signal comprises: S1.1: vertically immerse the aluminum enameled flat wire into the detection fluid in the constant temperature control box through the guide wheel, and use a distributed fluid pressure sensor array to collect the reference fluid resistance signal at each axial position; S1.2: apply an alternating current with a standard frequency through a high-frequency power generator, and the current intensity is automatically matched with a preset parameter according to the wire cross-sectional area; S1.3: based on the preset parameter, collect the reference electromagnetic signal in the unmoved state through the annular electromagnetic induction coil group.

[0009] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the present application, the step S2 of real-time detecting the fluid resistance signal and the electromagnetic induction signal comprises: S2.1: control the precise traction mechanism to move the enameled wire at a constant wire speed, real-time record the dynamic resistance signal of the fluid pressure sensor in the moving process, and use a sliding window algorithm to eliminate mechanical vibration noise; S2.2: synchronously collect the time-varying induction signal of the electromagnetic induction coil in the moving state; S2.3: perform time-domain synchronous marking processing on the dynamic resistance signal and the time-varying induction signal, establish an accurate mapping relationship with the wire physical position, and output the fluid resistance signal and the electromagnetic induction signal.

[0010] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the application, in step S3, the steps of surface defect detection include: S3.1: adopt multi-stage digital filtering to eliminate traction mechanism vibration noise, extract effective signal segment through adaptive threshold algorithm, and establish resistance position mapping model; S3.2: based on fluid dynamics model and resistance position mapping model, convert fluid resistance signal into three-dimensional surface profile; adopt moving least square method to reconstruct ideal surface model, calculate deviation matrix of three-dimensional surface profile and ideal surface model, and identify abnormality through curvature analysis to obtain abnormal area; S3.3: based on abnormal area, calculate defect characteristic parameter, and establish defect characteristic vector; S3.4: through pre-trained random forest model, map defect characteristic vector to specific defect type, and output defect list with space coordinates.

[0011] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the application, in step S3, the steps of internal defect detection include; S3.5: perform synchronous compression transformation on electromagnetic induction signal, separate skin effect and defect response in time-frequency domain, adopt singular value decomposition to eliminate baseline drift, and extract transient characteristic component synchronized with wire movement; S3.6: establish electromagnetic-thermal-force coupling model, and decompose transient characteristic component into conductor defect response and insulation defect response; S3.7: based on Maxwell equation set, construct forward model, according to conductor defect response and insulation defect response, adopt Monte Carlo optimization algorithm to inverse conductor cross-sectional area change and insulation layer thickness fluctuation parameter, and output internal defect data.

[0012] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the application, in step S4, the steps of space-time matching include: S4.1: perform space registration on coordinate data in surface defect data and distribution cloud diagram in internal defect data, and establish high-precision mapping relationship; S4.2: analyze internal area below coordinate data, compare correlation of electromagnetic characteristics and surface topography, and identify associated defects; when distance between coordinate data and distribution cloud diagram is less than set threshold, determine as associated composite defects.

[0013] According to the aluminum enameled flat wire online defect detection and adaptive correction method provided by the application, in step S4, the steps of distinguishing defect type through machine learning algorithm include: S4.3: according to coordinate data and distribution cloud diagram, construct space-time correlation feature matrix, and generate fusion feature vector; S4.4: Load a pre-trained three-stage classification model based on a machine learning algorithm, and input the fusion feature vector into the three-stage classification model for defect type classification, and output a classification result; S4.5: Based on the surface defect data and the internal defect data, a three-dimensional feature matrix is constructed according to the classification result according to the defect type; S4.6: The SHAP value is used to evaluate the contribution of each feature, and the key features are dynamically weighted; the three-dimensional feature matrix is analyzed for feature importance, and an explainability analysis report is generated.

[0014] According to the online defect detection and adaptive correction method for aluminum enameled flat wire provided by the application, in step S4.4, the step of outputting the classification result comprises: S4.4.1: Load a pre-trained random forest model to perform coarse-grained classification on the fusion feature vector, and output low-confidence samples; S4.4.2: Call a 1D-CNN sub-model to perform dynamic feature enhancement on the low-confidence samples, and output a corrected probability distribution; S4.4.3: Based on the corrected probability distribution, the low-confidence samples are processed by a graph neural network for correlation defects, and the classification result is output.

[0015] According to the online defect detection and adaptive correction method for aluminum enameled flat wire provided by the application, in step S5, the step of performing the correction operation comprises: S5.1: According to the defect type and the defect position, the required correction action parameters are calculated; S5.2: Based on the correction action parameters, the action parameters of the multi-axis actuator are adjusted by a PID controller to drive the radial pressure roller to adjust the wire posture; the torsion mechanism is controlled to compensate for the axial deformation of the wire in real time, and the deflection change of the wire is dynamically tracked; S5.3: The corrected wire is re-introduced into the detection area for verification.

[0016] The application also provides an online defect detection and adaptive correction system for aluminum enameled flat wire, comprising: The fluid detection module is used for immersing the aluminum enameled flat wire into a fluid with a set viscosity, collecting a reference fluid resistance signal, and based on the reference fluid resistance signal, passing the aluminum enameled flat wire through the detection area at a constant speed to detect the fluid resistance signal in real time; The electromagnetic detection module is used for applying a high-frequency alternating current to the aluminum enameled flat wire in the fluid to collect a reference electromagnetic signal; based on the reference electromagnetic signal, the electromagnetic induction signal is detected in real time; The defect positioning module is configured to locate a defect position of the surface defect of the enameled wire based on the fluid resistance signal in combination with a reference fluid resistance signal, and analyze spectral characteristics of the electromagnetic induction signal based on the defect position in combination with a reference electromagnetic signal, so as to identify a type and distribution of the internal defect of the aluminum enameled flat wire. The defect identification module is configured to perform space-time matching on the surface defect data and the internal defect data, and distinguish the defect type through a machine learning algorithm. The deviation correction decision module is configured to establish a defect feature matrix according to the defect type, and output a corresponding deviation correction scheme through a pre-trained neural network model. The deviation correction execution module is configured to output a deviation correction control signal to the correction module for deviation correction operation based on the deviation correction scheme.

[0017] The aluminum enameled flat wire online defect detection and adaptive deviation correction method and system provided by the application realize synchronous detection of surface defects and internal defects through synchronous acquisition and analysis of fluid resistance signals and electromagnetic induction signals, and solve the limitations of traditional single detection means. The moving least squares method is used to reconstruct an ideal surface model, and the abnormal area is identified in combination with curvature analysis. The electromagnetic signal is separated from noise through synchronous compression transformation and singular value decomposition, which significantly improves the detection sensitivity of small defects. A random forest model is used to realize coarse-grained rapid classification, a 1D-CNN dynamically enhances the features of low-confidence samples, and a graph neural network processes the associated defects, forming a high-precision three-stage classification system. A SHAP value feature importance analysis is used to generate an explainability report, which significantly improves the classification accuracy of composite defects. A deviation correction scheme is generated based on the defect type, position and spatial mapping relationship, a PID controller is used to dynamically adjust a multi-axis actuator, the axial deformation and deflection change of the wire are compensated in real time, and a re-inspection mechanism after deviation correction is introduced to form a closed-loop control, thereby ensuring the effectiveness of defect correction. BRIEF DESCRIPTION OF DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0019] Fig. 1 is a flowchart of an aluminum enameled flat wire online defect detection and adaptive deviation correction method provided by an embodiment of the application; Fig. 2 is a step diagram for distinguishing the defect type through a machine learning algorithm; Fig. 3 is a structural schematic diagram of an aluminum enameled flat wire online defect detection and adaptive deviation correction system provided by an embodiment of the application. DETAILED DESCRIPTION

[0020] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions in the present application will be described clearly and completely below with reference to the drawings in the present application. Obviously, the described embodiments are only some, but not all of the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0021] Embodiment one: The present application provides an online defect detection and self-adaptive correction method and system for aluminum enameled flat wire. Figs. 1-3 The present application provides an online defect detection and self-adaptive correction method and system for aluminum enameled flat wire.

[0022] As shown in the drawings, Figs. 1-2 The present application provides an online defect detection and self-adaptive correction method and system for aluminum enameled flat wire. S1: immerse the aluminum enameled flat wire into a fluid with a set viscosity, and apply a high-frequency alternating current, collect the reference fluid resistance signal and the reference electromagnetic signal, the step comprising: S1.1: vertically immerse the aluminum enameled flat wire into the detection fluid in the constant temperature control box through the guide wheel, the guide wheel usually adopts a high-precision ceramic wheel with surface polishing treatment, the surface roughness is less than a set threshold, to ensure that the wire is vertically immersed without lateral deviation. The concentricity error of the guide wheel is controlled in microns through laser calibration, to avoid the wire twisting caused by the deformation of the wheel body, and a distributed fluid pressure sensor array is used to collect the reference fluid resistance signal at each axial position. When collecting the fluid resistance signal, the fluid resistance of the fluid on the surface of the aluminum enameled flat wire is calculated, which is expressed as:

[0023] Wherein, F f is the fluid resistance per unit length of the wire, μ is the fluid viscosity, A c is the effective contact area between the wire and the fluid, is the average velocity gradient of the fluid on the surface of the flat wire, v0 represents the velocity of the fluid, and y represents the coordinate direction perpendicular to the surface of the aluminum enameled flat wire.

[0024] S1.2: apply an alternating current with a standard frequency through a high-frequency power generator, the current intensity is automatically matched with the preset parameters according to the cross-sectional area of the wire. The power generator automatically adjusts the current frequency and amplitude according to the real-time cross-sectional area of the wire (measured online by a laser diameter gauge), to ensure that the current density is within the effective detection range, and to avoid the risk of paint film breakdown.

[0025] S1.3: Based on the preset parameters, the baseline electromagnetic signal in the non-moving state is collected by the annular electromagnetic induction coil group. The coil group adopts a multi-turn tight winding structure, the coil spacing matches the wire width, the center is concentric with the wire movement trajectory, and the edge effect is reduced. The coil is protected by a high-permeability shield to reduce external electromagnetic interference. The formula for collecting the baseline electromagnetic signal in the non-moving state by the annular electromagnetic induction coil group is:

[0026] where ε is the induced electromotive force of a single-turn coil, N is the number of turns of the coil winding, B is the magnetic induction intensity generated by the coil, L is the effective detection length of the wire, and v1 is the constant moving speed of the wire.

[0027] S2: Based on the baseline fluid resistance signal and the baseline electromagnetic signal, the aluminum enameled flat wire is moved through the detection area at a constant speed, and the fluid resistance signal and the electromagnetic induction signal are detected in real time. The steps include: S2.1: Control the precision traction mechanism to move the enameled wire at a constant wire speed. The traction mechanism adopts closed-loop control of a servo motor and a ball screw. The encoder feedbacks the wire speed in real time, and the PID algorithm dynamically adjusts the speed to ensure that the speed fluctuation is less than the set precision. The dynamic resistance signal of the fluid pressure sensor during the movement process is recorded in real time, and the sliding window algorithm is used to eliminate mechanical vibration noise. The PID algorithm for closed-loop control of the servo motor and the ball screw of the traction mechanism is:

[0028] where u(t) is the control voltage of the servo motor, e(t) is the error between the actual speed and the set speed, reflecting the deviation between the current operating state and the target state, which is the input signal of the PID control. t is the current running time of the system, K p , K i , and K d are the proportional, integral, and derivative coefficients of the PID controller, respectively. It is used to control the accumulation of historical errors, eliminate steady-state errors through integration, and improve system accuracy. It is used to control the error change rate, reflects the error change trend, suppresses overshoot (prevents error growth trend), improves system damping, and reduces oscillation.

[0029] S2.2: Synchronously collect the time-varying induction signal of the electromagnetic induction coil in the moving state. The induction signal is connected to the data acquisition card through the preamplifier, and the wire position coordinates are recorded synchronously (obtained through the absolute encoder of the traction mechanism), ensuring that the signal and position correspond accurately.

[0030] S2.3: Time domain synchronization marking processing is performed on the dynamic resistance signal and the time-varying induction signal, a hardware trigger mode is adopted, the two types of sensors are synchronously triggered at the time of the traction mechanism pulse signal (corresponding to the fixed distance movement of the wire), microsecond-level synchronization is realized through timestamp recording, an accurate mapping relationship with the physical position of the wire is established, and the fluid resistance signal and the electromagnetic induction signal are output.

[0031] S3: Based on the fluid resistance signal, the defect position of the enameled wire surface defect is located in combination with the reference fluid resistance signal, and the surface defect data is output. Based on the defect position, the spectrum characteristics of the electromagnetic induction signal are analyzed in combination with the reference electromagnetic signal, the type and distribution of the internal defect of the aluminum enameled flat wire are recognized, and the internal defect data is output.

[0032] The steps of surface defect detection include: S3.1: Multi-stage digital filtering is adopted to eliminate the vibration noise of the traction mechanism, the three-stage filtering structure is low-pass filtering (remove high-frequency noise), band-pass filtering (extract defect characteristic frequency), and adaptive notch filtering (suppress power frequency interference) in turn, the signal-to-noise ratio is gradually improved, the effective signal segment is extracted through the adaptive threshold algorithm, the threshold is dynamically adjusted based on the mean and standard deviation of the signal in the sliding window, and the signal segment that continuously exceeds the threshold is determined as the effective defect segment, and the resistance position mapping model is established.

[0033] S3.2: Based on the fluid dynamics model and the resistance position mapping model, the fluid resistance signal is converted into a three-dimensional surface profile. The fluid dynamics model is based on the laminar flow assumption, and the relationship between resistance and surface topography is obtained by simplifying the Navier-Stokes equation, and the model parameters are calibrated by standard samples. The moving least squares method is adopted to reconstruct the ideal surface model, the moving least squares method adopts a variable window and an adaptive polynomial order, the fitting accuracy of the overall profile and the local details is balanced, the deviation matrix of the three-dimensional surface profile and the ideal surface model is calculated, and the abnormality is identified through curvature analysis, the Gaussian curvature and the average curvature are calculated, and the curvature abnormal region that exceeds the ideal surface deviation threshold and is continuously distributed is determined as the defect, and the abnormal region is obtained. The fluid dynamics three-dimensional surface profile reconstruction model is represented as:

[0034] wherein, is the local resistance at the coordinates (x, s), μ is the fluid viscosity, v1 is the constant moving speed of the wire, h(x, s) is the surface height deviation function, , are the gradients of the surface profile in the x and s directions respectively. n x1 represents the upper and lower limits of integration in the x direction, s m s1 represents the upper and lower limits of integration in the s direction. dxds is the area microelement in the two-dimensional Cartesian coordinate system, which is the cumulative load from the local to the whole.

[0035] S3.3: Based on the abnormal area, calculate the defect feature parameters (such as the maximum value of profile deviation, standard deviation, number of curvature anomaly points and distribution density), and establish a defect feature vector.

[0036] S3.4: Map the defect feature vector to a specific defect type through a pre-trained random forest model, output a defect list with spatial coordinates, and the model input is a multi-dimensional feature vector, which is trained and optimized through historical defect samples (scratches, pits, bumps, etc.) to classify the boundary.

[0037] The steps for internal defect detection include.

[0038] S3.5: Perform synchronous compression transformation on the electromagnetic induction signal. The synchronous compression transformation adjusts the time-frequency resolution, separates the skin effect (high-frequency linear correlation) and defect response (low-frequency nonlinear) signals, and uses singular value decomposition to eliminate baseline drift when separating the skin effect and defect response in the time-frequency domain. The signal time-varying matrix is decomposed into low-rank (baseline drift) and sparse (defect transient) components, the low-rank part is retained to remove noise, and the transient feature component synchronized with the wire movement is extracted. The formula for time-frequency domain signal separation is:

[0039] where, is the wavelet coefficient, is the original electromagnetic induction signal, t1 is the time window, ψ(t1) is the mother wavelet function, which is the basis function of wavelet transform, and has the characteristics of oscillation, rapid decay, and tight support or approximate tight support in both time and frequency domains. represents the complex conjugate of the mother wavelet function. a is the scale factor, which is a positive real number, used for scaling operation of the mother wavelet function. When a>1, the mother wavelet function is stretched, the width in time domain increases, and the width in frequency domain decreases; when 0

[0040] S3.6: Establish an electromagnetic-thermal-force coupling model, which is based on Maxwell's equations (electromagnetic field), heat conduction equation (eddy current loss heat production) and elastic mechanics equation (stress distribution) multi-field coupling, solve the temperature field and stress field at the defect position through finite element simulation, and decompose the transient feature component into conductor defect response and insulation defect response.

[0041] S3.7: Constructing a forward model based on Maxwell's equations, the forward model adopts a finite element simulation to establish a conductor / insulation layer geometric model, inputting parameters such as cross-sectional area variation, insulation layer thickness fluctuation, etc., simulating the electromagnetic induction signal response, according to the conductor defect response and the insulation defect response, using the Monte Carlo optimization algorithm to invert the conductor cross-sectional area variation, insulation layer thickness fluctuation parameters, outputting internal defect data, the Monte Carlo optimization generates parameter combinations randomly and calculates errors, and the parameter solution closest to the actual signal is obtained through optimization. The model can be expressed as:

[0042] wherein, is the first-order partial derivative of the electric field intensity E(j) with respect to position j, E(j) is the electric field intensity at defect position j, γ is the magnetic permeability of the conductor, σ is the electrical conductivity of the conductor, is the conductor cross-sectional area variation function caused by defects, t2 is the duration, is the conductor cross-sectional area variation function is the first-order partial derivative of the time t2.

[0043] S4: Spatiotemporal matching of surface defect data and internal defect data, distinguishing defect types through machine learning algorithms. And according to the defect type, a defect feature matrix is established, and a corresponding correction scheme is output through a pre-trained neural network model.

[0044] The steps of spatiotemporal matching include: S4.1: Spatial registration of coordinate data in surface defect data and distribution cloud in internal defect data, using a feature point-based registration method, extracting defect center points (surface) and centroids (internal cloud), and establishing a high-precision mapping relationship through affine transformation (translation, rotation, scaling).

[0045] S4.2: Analyzing the internal area below the coordinate data, comparing the correlation between electromagnetic characteristics and surface topography, quantifying the correlation degree through cross-correlation function and mutual information, and determining as associated defects when the distance between coordinate data and distribution cloud is less than the set threshold.

[0046] The steps of distinguishing defect types through machine learning algorithms include: S4.3: According to the coordinate data and the distribution cloud, a spatiotemporal correlation feature matrix is constructed, and a fusion feature vector is generated, which contains multi-dimensional information such as surface defect spatial coordinates, contour deviation, curvature anomaly, and internal defect cloud, electromagnetic characteristics, etc., and the dimension is eliminated through normalization processing.

[0047] S4.4: Load the pre-trained three-stage classification model based on machine learning algorithm, and input the fusion feature vector into the three-stage classification model for defect type classification, and output the classification result.

[0048] The step of outputting the classification result includes: S4.4.1: Load the pre-trained random forest model, perform coarse-grained classification on the fusion feature vector, and output low-confidence samples, denoted as:

[0049] Where Gini is the node Gini index, C is the total number of defect types such as scratches, dents, etc. c is the defect sample index, P c is the probability of the cth defect sample in the node.

[0050] S4.4.2: Call 1D-CNN submodel to perform dynamic feature enhancement on low-confidence samples. 1D-CNN extracts time series features (such as defect development time correlation) and enhances them to correct the probability distribution. The 1D-CNN submodel (One-Dimensional Convolutional Neural Network Submodel) is a deep learning model specifically designed for processing one-dimensional time series signals (such as electromagnetic induction signals, defect feature sequences, etc.). Its main function is to dynamically enhance the features of low-confidence samples to improve the accuracy of defect classification.

[0051] S4.4.3: Based on the corrected probability distribution, perform correlation defect processing on low-confidence samples through a graph neural network, which models the correlation between defect types (such as the probability of surface scratches and insulation shedding accompanying each other), and outputs the classification result.

[0052] S4.5: Based on the classification result, construct a three-dimensional feature matrix according to the defect type based on the surface defect data and internal defect data.

[0053] S4.6: Evaluate the contribution of each feature using SHAP values and dynamically weight the key features. Perform feature importance analysis on the three-dimensional feature matrix to generate an explainability analysis report. SHAP values quantify the contribution of each feature to the classification. Draw an importance ranking chart and mark the impact direction. Combine expert knowledge to generate a written explanation. SHAP (SHapley Additive exPlanations) is an explainability analysis method used to quantify the contribution of each feature to the prediction result in machine learning models. It is based on the Shapley value concept in game theory, which can fairly distribute the "contribution" of each feature to the model output, helping to understand the decision-making logic of the model. It calculates the influence of features in different feature subsets to finally obtain the global and local explanation of the feature to the model output. Shapley value is a mathematical method derived from game theory, used to fairly distribute cooperative benefits or quantify individual contributions to collective results. In the field of machine learning, it is used to explain model predictions and measure the contribution of each feature to the final result.

[0054] S5: Based on the correction scheme, output the correction control signal to the correction module for correction operation. The steps of the correction operation include: S5.1: Calculate the required correction action parameters according to the defect type and defect position, including radial pressure roller pressure adjustment amount (lateral deviation calculation), torsion mechanism torque adjustment amount (axial deformation calculation), and multi-axis actuator movement speed (deflection rate adjustment).

[0055] S5.2: Based on the correction action parameters, adjust the action parameters of the multi-axis actuator through the PID controller to drive the radial pressure roller to adjust the wire posture. Control the torsion mechanism to compensate for the axial deformation of the wire in real time, dynamically track the deflection change of the wire, and optimize the response speed and stability of the PID parameters through the Ziegler-Nichols method. Ziegler-Nichols method is a classic PID controller parameter tuning method that quickly determines the initial values of the proportional, integral, and derivative control parameters, making the correction system achieve better dynamic response performance. Through experimental testing of the critical gain and critical oscillation period of the system, then calculate the PID parameters according to the specific formula.

[0056] S5.3: Re-introduce the corrected wire into the detection area for verification. The same fluid and electromagnetic detection modules are used to collect signals, and the defect positioning and identification module is used to verify the effect. If it does not meet the standard, adjust the parameters and repeat the verification.

[0057] In summary, the present application provides an aluminum enameled flat wire online defect detection and adaptive correction method and system. Through synchronous acquisition and analysis of fluid resistance signals and electromagnetic induction signals, synchronous detection of surface defects and internal defects is realized, solving the limitations of traditional single detection methods. The moving least squares method is used to reconstruct the ideal surface model, and the curvature analysis is used to identify abnormal areas. The electromagnetic signal is separated from the noise through synchronous compression transformation and singular value decomposition, significantly improving the detection sensitivity of small defects. Through the random forest model, coarse-grained rapid classification is realized. The 1D-CNN dynamically enhances the low-confidence sample features, and the graph neural network processes the associated defects, forming a high-precision three-stage classification system. Combined with SHAP value feature importance analysis, an interpretable report is generated, significantly improving the classification accuracy of composite defects. By using the defect type, position and spatial mapping relationship to generate a correction scheme, the PID controller dynamically adjusts the multi-axis actuator to compensate for the axial deformation and deflection changes of the wire in real time, and a re-inspection mechanism is introduced after correction to form a closed-loop control, ensuring the effectiveness of defect correction.

[0058] As shown in Fig. 3 The present application also provides an aluminum enameled flat wire online defect detection and adaptive correction system, which comprises a fluid detection module, an electromagnetic detection module, a defect positioning module, a defect identification module, a correction decision module and a correction execution module.

[0059] The fluid detection module is used to immerse the aluminum enameled flat wire in a fluid with a set viscosity to collect a reference fluid resistance signal. Based on the reference fluid resistance signal, the aluminum enameled flat wire is passed through the detection area at a constant speed, and the fluid resistance signal is detected in real time. The fluid detection module includes a fluid circulation filtration unit (pump body + filter to maintain cleanliness) and a Teflon-coated guide wheel (to reduce the coefficient of friction).

[0060] The electromagnetic detection module is used to apply a high-frequency alternating current to the aluminum enameled flat wire in the fluid to collect a reference electromagnetic signal. Based on the reference electromagnetic signal, the electromagnetic induction signal is detected in real time. The electromagnetic detection module is provided with an electromagnetic shield (high permeability material) and a coaxial cable (to reduce transmission loss).

[0061] The defect positioning module is used to locate the defect position of the enameled wire surface defect based on the fluid resistance signal in combination with the reference fluid resistance signal. Based on the defect position, the frequency spectrum characteristics of the electromagnetic induction signal are analyzed in combination with the reference electromagnetic signal to identify the type and distribution of the internal defects of the aluminum enameled flat wire. The reference signal is stored in an industrial-grade database (supporting version management and backup), and the position mapping model is continuously optimized through historical data.

[0062] The defect identification module is used to perform space-time matching of surface defect data and internal defect data, and distinguish the defect types through machine learning algorithms.

[0063] The deviation rectification decision module is configured to establish a defect feature matrix according to a defect type, and output a corresponding deviation rectification scheme through a pre-trained neural network model.

[0064] The deviation rectification execution module is configured to output a deviation rectification control signal to a correction module for deviation rectification operation based on the deviation rectification scheme. The correction module includes a radial pressure roller group (servo motor driven), an axial torsion mechanism (step motor driven), and a laser displacement sensor (monitoring deflection change).

[0065] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software plus necessary universal hardware platforms, and of course, can also be realized by hardware. Based on such understanding, the above technical solutions, essentially or in other words, the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0066] Finally, it should be noted that: the above examples are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing examples, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing examples, or make equivalent replacement for some technical features therein; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An online defect detection and adaptive deviation correction method for aluminum enameled flat wire, characterized in that, The method comprises the following steps: S1: immerse the aluminum enameled flat wire into a fluid with a set viscosity and apply a high-frequency alternating current, collect the baseline fluid resistance signal and the baseline electromagnetic signal; S2: based on the baseline fluid resistance signal and the baseline electromagnetic signal, pass the aluminum enameled flat wire through the detection area at a constant speed, and detect the fluid resistance signal and the electromagnetic induction signal in real time; S3: based on the fluid resistance signal, combine the baseline fluid resistance signal, locate the defect position of the surface defect of the enameled wire, and output the surface defect data; based on the defect position, combine the baseline electromagnetic signal, analyze the spectral characteristics of the electromagnetic induction signal, identify the type and distribution of the internal defects of the aluminum enameled flat wire, and output the internal defect data; S4: space-time match the surface defect data and the internal defect data, distinguish the defect type through a machine learning algorithm; and establish a defect feature matrix according to the defect type, and output a corresponding correction scheme through a pre-trained neural network model; S5: based on the correction scheme, output a correction control signal to a correction module for correction operation.

2. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 1, characterized in that, In step S1, the step of collecting the baseline fluid resistance signal and the baseline electromagnetic signal comprises: S1.1: vertically immerse the aluminum enameled flat wire into the detection fluid in the constant temperature control box through the guide wheel, and use a distributed fluid pressure sensor array to collect the baseline fluid resistance signal at each axial position; S1.2: apply an alternating current with a standard frequency through a high-frequency power generator, and the current intensity is automatically matched with a preset parameter according to the wire cross-sectional area; S1.3: based on the preset parameter, collect the baseline electromagnetic signal in the unmoved state through the annular electromagnetic induction coil group.

3. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 1, characterized in that, In step S2, the step of detecting the fluid resistance signal and the electromagnetic induction signal in real time comprises: S2.1: control the precision traction mechanism to move the enameled wire at a constant linear speed, and record the dynamic resistance signal of the fluid pressure sensor in real time during the movement process, and use a sliding window algorithm to eliminate mechanical vibration noise; S2.2: synchronously collect the time-varying induction signal of the electromagnetic induction coil in the motion state; S2.3: perform time domain synchronous marking processing on the dynamic resistance signal and the time-varying induction signal, establish an accurate mapping relationship with the wire physical position, and output the fluid resistance signal and the electromagnetic induction signal.

4. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 1, characterized in that, In step S3, the step of surface defect detection comprises: S3.1: use multi-stage digital filtering to eliminate the vibration noise of the traction mechanism, extract the effective signal segment through an adaptive threshold algorithm, and establish a resistance position mapping model; S3.2: based on the fluid dynamics model and the resistance position mapping model, convert the fluid resistance signal into a three-dimensional surface profile; use a moving least squares method to reconstruct an ideal surface model, calculate the deviation matrix of the three-dimensional surface profile and the ideal surface model, and identify the abnormality through curvature analysis to obtain an abnormal area; S3.3: based on the abnormal area, calculate the defect feature parameters, and establish a defect feature vector; S3.4: map the defect feature vector to a specific defect type through a pre-trained random forest model, and output a defect list with spatial coordinates.

5. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 1, characterized in that, In step S3, the step of performing internal defect detection includes: S3.5: performing synchronous compression transformation on the electromagnetic induction signal, separating the skin effect and defect response in the time-frequency domain, eliminating baseline drift by singular value decomposition, and extracting transient characteristic components synchronized with the wire movement; S3.6: establishing an electromagnetic-thermal-force coupling model, decomposing the transient characteristic components into conductor defect response and insulation defect response; S3.7: based on the Maxwell equations, a forward model is constructed, and based on the conductor defect response and insulation defect response, a Monte Carlo optimization algorithm is used to invert the conductor cross-sectional area variation and insulation layer thickness fluctuation parameters, and the internal defect data is output.

6. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 1, characterized in that, In step S4, the step of performing space-time matching includes: S4.1: spatially registering the coordinate data in the surface defect data with the distribution cloud in the internal defect data, and establishing a high-precision mapping relationship; S4.2: analyzing the internal region below the coordinate data, comparing the correlation of electromagnetic characteristics and surface topography, and identifying associated defects; when the distance between the coordinate data and the distribution cloud is less than a set threshold, it is determined as a related composite defect.

7. The online defect detection and adaptive correction method for aluminum enameled flat wire according to claim 6, characterized in that, In step S4, the step of distinguishing defect types by machine learning algorithm includes: S4.3: constructing a space-time correlation feature matrix according to the coordinate data and the distribution cloud, and generating a fusion feature vector; S4.4: loading a pre-trained three-stage classification model based on machine learning algorithm, and inputting the fusion feature vector into the three-stage classification model for defect type classification, and outputting the classification result; S4.5: based on the surface defect data and the internal defect data, according to the classification result, constructing a three-dimensional feature matrix according to the defect type; S4.6: using SHAP value to evaluate the contribution of each feature, and dynamically weighting the key features; performing feature importance analysis on the three-dimensional feature matrix, and generating an explainability analysis report.

8. The method according to claim 7, wherein the method is characterized by, In step S4.4, the step of outputting the classification result includes: S4.4.1: loading a pre-trained random forest model, performing coarse-grained classification on the fusion feature vector, and outputting low-confidence samples; S4.4.2: calling a 1D-CNN sub-model to dynamically enhance the features of the low-confidence samples, and outputting a corrected probability distribution; S4.4.3: based on the corrected probability distribution, performing associated defect processing on the low-confidence samples by graph neural network, and outputting the classification result.

9. The online defect detection and adaptive correction method of an aluminum enameled flat wire according to claim 1, characterized in that, In step S5, the step of performing rectification operation includes: S5.1: calculating the required rectification action parameters according to the defect type and defect position; S5.2: based on the rectification action parameters, adjusting the action parameters of the multi-axis actuator through the PID controller, driving the radial pressure roller to adjust the wire posture; controlling the torsion mechanism to compensate for the axial deformation of the wire in real time, and dynamically tracking the deflection change of the wire; S5.3: re-introducing the rectified wire into the detection area for verification.

10. An aluminum enameled flat wire on-line defect detection and adaptive correction system employing an aluminum enameled flat wire on-line defect detection and adaptive correction method according to any one of claims 1 to 9, characterized in that, It includes: A fluid detection module for immersing the aluminum enameled flat wire in a fluid of a set viscosity to collect a reference fluid resistance signal; The aluminum enameled flat wire is passed through a detection area at a constant speed based on the reference fluid resistance signal, and a fluid resistance signal is detected in real time; An electromagnetic detection module is configured to apply a high-frequency alternating current to the aluminum enameled flat wire in the fluid, and collect a reference electromagnetic signal; An electromagnetic induction signal is detected in real time based on the reference electromagnetic signal; A defect positioning module is configured to locate a defect position of the surface defect of the enameled flat wire based on the fluid resistance signal in combination with the reference fluid resistance signal; Based on the defect position, a spectrum feature of the electromagnetic induction signal is analyzed in combination with the reference electromagnetic signal, and a type and distribution of an internal defect of the aluminum enameled flat wire are identified; A defect recognition module is configured to perform space-time matching on the surface defect data and the internal defect data, and distinguish the defect type through a machine learning algorithm; A rectification decision module is configured to establish a defect feature matrix according to the defect type, and output a corresponding rectification scheme through a pre-trained neural network model; A rectification execution module is configured to output a rectification control signal to a correction module for rectification operation based on the rectification scheme.

Images