An electrical connector fault detection method and system

Abstract

The application discloses a kind of electric connector fault detection method and system, comprising: scanning electric connector obtains three-dimensional body data, constructs electric connector model, obtains electric connector cross-sectional image;To the image segmentation of electric connector cross section obtains binary electric connector cross-sectional image, identifies the part of electric connector, according to the part of electric connector extraction structure shape, using detection rule to identify whether there is fault defect, if there is fault defect, continue to execute step S103, if there is no fault defect, system output does not exist fault defect;According to fault defect extraction candidate defect point, candidate defect point is connected to form defect point line, identifies the state change of defect point line and determines fault defect type, directly exposes internal structure by cross-sectional imaging, determines the type of fault defect, realizes the nondestructive testing of electric connector internal structure, accurately identifies internal defect, improves the accuracy and comprehensiveness of defect identification.

Description

Technical Field

[0001] This invention relates to the field of electrical connector technology, and in particular to a method and system for detecting electrical connector faults. Background Technology

[0002] In electrical systems and electronic equipment, electrical connectors, as key components, play a crucial role in transmitting electrical energy and signals. Their stability and reliability directly impact the overall system's operational efficiency and safety. However, due to the frequent exposure to external factors such as thermal cycling, mechanical vibration, and current loads during operation, electrical connectors are prone to minor deformations and damage to their internal structure, resulting in defects such as cracks, bubbles, and pores. These internal defects are not only difficult to observe with the naked eye, but traditional testing methods often struggle to accurately identify and assess them, thus posing potential risks to the long-term stable operation of the system.

[0003] Traditional methods for detecting faults in electrical connectors primarily rely on visual inspection, resistance measurement, and simple functional tests. While these methods can detect external damage and obvious functional abnormalities to some extent, they fall short in detecting minute internal defects. For example, internal defects such as cracks and bubbles often do not significantly affect the overall resistance of the connector in their early stages, making them difficult to detect through resistance measurement. Furthermore, visual inspection can only detect surface damage and is ineffective against structural defects hidden within. Summary of the Invention

[0004] This application discloses a method and system for detecting faults in electrical connectors. By directly exposing the internal structure through cross-sectional imaging, the type of fault defect can be determined, achieving non-destructive testing of the internal structure of the electrical connector, accurately identifying internal defects, and improving the accuracy and comprehensiveness of defect identification.

[0005] This application provides a method for detecting faults in electrical connectors, including: S101, Scan the electrical connector to obtain three-dimensional volume data, use three-dimensional software to construct an electrical connector model based on the three-dimensional volume data, and obtain a cross-sectional image of the electrical connector based on the electrical connector model; S102: The cross-section of the electrical connector is segmented to obtain a binary image of the electrical connector cross-section. The parts of the electrical connector are identified using the binary image of the electrical connector cross-section. The structural shape is extracted using an algorithm based on the parts of the electrical connector. Based on the structural shape, detection rules are used to identify whether there are any faults or defects. If faults or defects are found, step S103 is executed. If no faults or defects are found, the system outputs that no faults or defects exist. The parts include conductors, insulators, and gaps. The structural shape includes conductor outlines, insulator thickness, and connection status. The detection rules refer to using different detection methods for different structural shapes. S103, extract candidate defect points based on existing faults and defects, connect the obtained candidate defect points to form defect point lines, and determine the fault and defect type by identifying the state changes of the defect point lines.

[0006] Preferably, the detection rules are as follows: For the structural shape of the conductor contour, a roundness comparison method is used for detection. The roundness of the conductor is compared with a threshold. When the roundness is greater than or equal to the threshold, it indicates that the conductor has no defects; when the roundness is less than the threshold, it indicates that the conductor has defects. For the structural shape of the insulator thickness, a thickness variance comparison method is used for detection. If the calculated thickness variance is greater than a distance threshold, it indicates that there are defects; if the calculated thickness variance is less than or equal to the distance threshold, it indicates that the conductor has no defects. For the structural shape of the connection state, connectivity or contact area ratio is used for detection.

[0007] Preferably, the step of determining the type of fault by identifying the state changes of the connection lines of the fault points is as follows: if the connection lines exhibit linear characteristics and the actual length of the connection lines is greater than or equal to a set threshold, then it is determined to be a crack-type defect; if the connection lines form a closed triangular mesh and the area of ​​the mesh unit is less than the set area threshold and the mesh density is greater than the set quantity threshold, then it is determined to be a bubble-type defect.

[0008] Preferably, the fault detection method further includes: S201, Obtain the volume of the electrical connector, which includes a reference volume and a target volume. Divide the obtained reference volume into voxel subsets. Find the subset in the target volume that best matches the voxel subset. Calculate the displacement between the matching subset and the voxel subset. Represent the displacement as a displacement vector field using components. Calculate the strain tensor field based on the displacement vector field. The reference volume is the initial volume of the electrical connector when it is not subjected to external forces, and the target volume is the volume of the electrical connector when it is subjected to external forces. S202, acquire four-dimensional data of the electrical connector, which consists of three-dimensional volume data and a time dimension. According to the time dimension in the four-dimensional data, track the crack movement trajectory based on the displacement vector field, construct a velocity field, construct a defect dynamics model based on the velocity field and stress factor, output the defect evolution state of the defect dynamics model, input the defect evolution state into the neural network, and the neural network predicts the remaining life of the electrical connector.

[0009] Preferably, the method for finding the subset that best matches the voxel subset in the target volume is as follows: a global search algorithm is used to find the subset that best matches each voxel subset in the target volume according to the matching rules. The matching rules are to calculate the cross-correlation coefficient between the subset in the target volume and the voxel subset in the reference volume, and find the subset with the highest cross-correlation coefficient as the best matching subset.

[0010] Preferably, the steps for tracking the movement trajectory of the crack according to the time dimension of the four-dimensional data based on the displacement vector field are as follows: Based on the obtained displacement vector field, starting from the initial time point, the position change of the crack is tracked sequentially at different time points. For each time point, key points that can represent the shape and position characteristics of the crack at the current moment are selected. Then, based on the displacement vector field, the displacement vector corresponding to the voxel where each key point is located is determined. This vector reflects the displacement direction and magnitude of the voxel from the current time point to the next time point. Finally, based on the current position coordinates of the key point, each component of its corresponding displacement vector is added to the corresponding component of the current coordinate to obtain the position coordinates of the key point at the next time point. The positions of many key points together constitute the position of the crack at the next moment, thus obtaining the movement trajectory of the crack as time changes.

[0011] Preferably, the fault detection method further includes: S301, extract defect feature vectors based on the identified fault defect types. The defect feature vectors include geometric features and spatial features. Calculate the interaction strength factor based on the defect feature vectors. Establish a defect network when the interaction strength factor is greater than a threshold. S302, based on the defect network and the current state of the defects, use the failure criteria to obtain the failure probability, and plot the failure probability change curve with time as the horizontal axis and failure probability as the vertical axis; based on the edge weights of the defect network and the failure probability, identify the critical defect chain using the tracing method.

[0012] This application also provides an electrical connector fault detection system, including a data acquisition module, a three-dimensional reconstruction module, a cross-section extraction module, an image segmentation and recognition module, a structural shape analysis module, a defect detection module, and a decision module; The data acquisition module is used to acquire three-dimensional volume data of the electrical connector; The 3D reconstruction module is used to construct a 3D model of the electrical connector; The cross-section extraction module is used to extract axial and radial cross-sectional images of the electrical connector; The image segmentation and recognition module is used to generate a binarized image; The structural shape analysis module is used to identify the structural shape of the electrical connector; The defect detection module is used to detect the types of faults and defects in the electrical connectors. The decision module is used to determine whether the electrical connector is faulty based on the defect detection results. The data acquisition module is connected to the 3D reconstruction module, the 3D reconstruction module is connected to the cross-section extraction module, the cross-section extraction module is connected to the image segmentation and recognition module, the image segmentation and recognition module is connected to the structural shape analysis module, the structural shape analysis module is connected to the defect detection module, and the defect detection module is connected to the decision-making module.

[0013] One or more technical solutions provided in this application have at least the following technical effects or advantages: the internal structure is directly exposed by micro-CT cross-sectional imaging to achieve non-destructive testing; the connection change analysis transforms discrete defects into topological features to determine the type of fault defects, which is superior to traditional testing; the internal structure of electrical connectors is realized through non-destructive testing; internal defects (such as cracks and bubbles) are accurately identified; and the accuracy and comprehensiveness of defect identification are improved. Four-dimensional data can accurately record the changes in the shape, location, and size of defects at different points in time, quantitatively describe the rate and direction of defect expansion, and by establishing defect dynamics models and data-driven prediction models, accurate prediction of the remaining life of electrical connectors can be achieved, improving the accuracy and reliability of life prediction. By using interaction strength factors and defect networks, the synergistic amplification effect among multiple defects can be considered more accurately, significantly improving the accuracy of failure risk prediction. By identifying critical defect chains, the defect propagation path that contributes the most to system failure can be clearly identified, enabling maintenance resources to be precisely allocated to defect clusters that have the greatest impact on system reliability, and accurately assessing the failure risk of electrical connector systems. Attached Figure Description

[0014] Figure 1 This is a flowchart illustrating a fault detection method for an electrical connector according to the present invention. Figure 2 This is a schematic diagram of the process for tracking the movement trajectory of a crack according to the present invention; Figure 3 This is a schematic diagram of the process for calculating the interaction strength factor in this invention; Figure 4 This is a block diagram of an electrical connector fault detection system according to the present invention. Detailed Implementation

[0015] To facilitate understanding of the present invention, a more complete description of this application will be given below with reference to the accompanying drawings, which illustrate preferred embodiments of the invention. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to enable a more thorough and complete understanding of the disclosure of the present invention.

[0016] It should be noted that the terms "vertical," "horizontal," "up," "down," "left," "right," and similar expressions used in this article are for illustrative purposes only and do not represent the only possible implementation.

[0017] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to limit the invention; the term "and / or" as used herein includes any and all combinations of one or more of the associated listed items.

[0018] Example 1: Traditional electrical connector fault identification methods cannot detect internal defects (such as cracks and bubbles). This example achieves non-destructive testing of the internal structure of electrical connectors through cross-sectional imaging and connection status identification. The solution overcomes the shortcomings of traditional electrical connector fault identification by acquiring cross-sectional images, 3D reconstruction, analyzing structural shape and connection status, identifying defect point connections, and implementing a decision-making mechanism, thereby improving the accuracy and comprehensiveness of defect identification.

[0019] Figure 1 This is a flowchart illustrating an electrical connector fault detection method according to an embodiment of the present invention, including: S101, Scan the electrical connector to obtain three-dimensional volume data, use three-dimensional software to construct an electrical connector model based on the three-dimensional volume data, and obtain a cross-sectional image of the electrical connector based on the electrical connector model; Specifically, a microfocus X-ray computed tomography (Micro-CT) system is used to scan the electrical connector. The scanning parameters of the system are set, and the voltage is adjusted according to the material density of the connector: high voltage for metal parts and low voltage for plastic parts. A high resolution (1-5) is set. (μm / pixel) is used to detect minute faults. The exposure time is set according to the material thickness of the electrical connector, and the number of frames (continuous slice images) is set to 500-1000 to cover the entire structure of the electrical connector. The electrical connector is vertically fixed on the rotating scanning stage using a fixture. Scanning is performed in two directions: axial and radial. First, the electrical connector is scanned axially along the pin arrangement direction (Z-axis). Then, the electrical connector is scanned radially, perpendicular to the pin direction (xy plane). The scanning steps are as follows: The micro-computed tomography (Micro-CT) system is started, the X-ray source emits a cone beam of X-rays that penetrates the electrical connector, and the flat panel detector synchronously acquires projection images. The scanning stage rotates 360°, acquiring one projection image every 0.1°-0.5° of rotation, for a total of 720-3600 projections. Median filtering is used to remove electronic noise from the projections. The filtered back projection algorithm (FBP) is used to reconstruct three-dimensional volume data from the projection images. The gray value of each voxel in the three-dimensional volume data reflects the material density. VG is used. The Studio software loads 3D volume data, i.e., constructs a 3D model of the connector. The center line of the pin array is marked in the 3D model of the connector as a reference axis for cross-sectional extraction. The thickness of the connector cross-section is set to be consistent with the scanning resolution to ensure that cross-sectional details are not lost. The spacing of the cross-sections is set to 10 μm, which can be adjusted according to detection requirements; for example, it can be reduced to 5 μm for high-precision detection. The extraction direction of the cross-section is set, including axial and radial cross-sections. The axial cross-section is a plane perpendicular to the pin direction (xy plane), and the radial cross-section is a plane parallel to the pin direction (zx or zy plane). In VG Studio software, the cross-sectional image of the electrical connector is extracted according to the cross-sectional extraction reference axis, cross-sectional thickness, cross-sectional spacing, and cross-sectional extraction direction. 3D coordinate information is marked on the generated cross-sectional image of the electrical connector.

[0020] S102, perform image segmentation on the cross-section of the electrical connector to obtain a binary image of the electrical connector cross-section, identify the parts of the electrical connector through the binary image of the electrical connector cross-section, extract the structural shape based on the parts of the electrical connector using an algorithm, and identify whether there are faults or defects based on the structural shape using detection rules. If there are faults or defects, continue to step S103. If there are no faults or defects, the system outputs that there are no faults or defects. Furthermore, based on the acquired cross-sectional image of the electrical connector, an image segmentation algorithm (Otsu algorithm) is used to identify different material regions of the electrical connector. The 16-bit cross-sectional image of the electrical connector is converted to 8-bit through local region contrast enhancement (8×8 pixel blocks) and linear scaling (the cross-sectional image of the connector is divided by 256). Local region contrast enhancement can improve the visibility of conductor edges and bubble boundaries. After converting the cross-sectional image of the electrical connector from 16-bit to 8-bit, the Otsu function is called in the open-source computer vision library (OpenCV). The 8-bit cross-sectional image of the electrical connector is input into the Otsu algorithm. The algorithm traverses all possible thresholds and calculates the difference in gray-level mean (inter-class variance) of conductor, insulator, and void at each threshold. The threshold that maximizes the inter-class variance is selected as the segmentation criterion to generate a ternary image: gray value 0: void (darkest area, such as bubble or background); gray value 1: insulator (medium brightness area, such as plastic shell); gray value 2: conductor (brightest area, such as metal pin). The ternary image is then split into three independent binary images. Three types of binary masks are generated: a gap mask, an insulator mask, and a conductor mask. The gap mask sets pixels with a grayscale value of 0 in the ternary image to 255 (white), and all others to 0 (black). In the binary image, white areas represent gaps, and black areas represent non-gaps (conductors + insulators). The insulator mask sets pixels with a grayscale value of 1 to 255, and all others to 0. White represents insulators, and black represents non-insulators (conductors + gaps). The conductor mask sets pixels with a grayscale value of 2 to... The value is set to 255, with the rest set to 0. White represents conductors, and black represents non-conductors (insulators + gaps). Denoising is performed on the gap mask, insulator mask, and conductor mask using opening operations and hole filling. The three denoised binary masks are then assigned to the RGB channels: R channel: gap mask (0 = no gap, 255 = gap present); G channel: insulator mask (0 = no insulator, 255 = insulator present); B channel: conductor mask (0 = no conductor, 255 = conductor present). Conductor areas are displayed as white (R=255, ...). With G=255 and B=255, the 255 value of channel B is extended to channels R and G. The insulator is displayed as gray (R=128, G=128, B=128). The 255 value of channel G is scaled down and synchronized to channels R and B. The gap remains black (R=0, G=0, B=0). Finally, the cross-section of the electrical connector is segmented into a three-channel binary image: conductor: pure white (255, 255, 255); insulator: medium gray (128, 128, 128); gap: pure black (0, 0, 0). By segmenting the cross-section of the electrical connector into a binary image, the conductor, insulator, and gap of the electrical connector can be accurately identified.

[0021] Based on the identified conductor, insulator, and gaps in the cross-section of the electrical connector, an algorithm is used to extract the structural shape, which includes the conductor outline, insulator thickness, and connection state. For the conductor outline, boundary tracing (Moore-Neighbor algorithm) is used to search for boundary points starting from any pixel in the conductor region, following an 8-neighbor order. This 8-neighbor order refers to traversing the eight neighboring pixels centered on the current pixel in a specific (clockwise) order, generating a closed polygon. Based on the generated polygon, the roundness and area ratio are calculated. The formula for calculating roundness is: Where C represents roundness, A represents the outline area (actual area of ​​the conductor or connector pin cross-section), and P represents the outline perimeter (boundary length of the conductor or connector pin cross-section). The conductor shape (pin state, pin being the specific form of the conductor) is identified based on the calculated roundness. Specifically: a threshold is set; when the roundness is greater than or equal to the threshold, it indicates that the pin is not deformed, meaning the conductor has no defects; when the roundness is less than the threshold, it indicates that the conductor has defects. The formula for calculating the area ratio is: ,in, The actual area of ​​the conductor's current cross-section. The template area is the standard cross-sectional area of ​​the conductor. The area ratio is used to indicate the conductor's actual cross-sectional area. When the area ratio is less than 90%, it means the conductor's actual cross-sectional area is smaller than the standard template area (i.e., the conductor is thinner); when the area ratio is greater than 110%, it means the conductor's actual cross-sectional area is larger than the standard template area (i.e., the conductor is thicker). For the insulator thickness, a distance transformation is performed on the insulator region (grayscale value 128). The distance from each insulator pixel to the nearest gap (background) is calculated. In OpenCV, `distanceTransform` is called to output the distance matrix. The value of each pixel in the generated distance matrix represents the geometric distance from that point to the nearest gap (background). Pixels with a distance value of zero directly correspond to the boundary region between the insulator and the gap. The set of these pixels constitutes the inner boundary of the insulator. The conductor contour coverage is expanded through morphological dilation, so that the expanded conductor region includes its neighboring insulator pixels. Then, the intersection of the expanded conductor region and the original insulator region is extracted. The intersection closer to the conductor side is extracted from this intersection. The edge pixels constitute the boundary between the insulator and the conductor, which is the outer boundary of the insulator. 100 points are sampled radially along the insulator, and the Euclidean distance (i.e., thickness) between the inner and outer boundaries of the insulator is calculated. The thickness variance is calculated. If the calculated thickness variance is greater than the distance threshold, it indicates that the thickness is uneven, i.e., there is a fault. The connection status between the conductor and the terminal (solder joint) is judged by the connectivity and the contact area ratio. The connectivity is checked by performing a logical AND operation on the conductor and connection point regions in the binary image and counting the number of overlapping pixels. If the number of overlapping pixels is 0, or the breakpoint width is ≥2 pixels (detected by contour gap), there is a fault. The contact area ratio is obtained by the ratio of the number of overlapping pixels between the conductor and the connection point to the theoretical contact area. If the contact area ratio is less than 90%, there is a fault. If the contact area is greater than or equal to 90%, it indicates that the connection is normal. If there is a fault, continue to execute step S103. If there is no fault, the system outputs that there is no fault.

[0022] S103, extract candidate defect points based on existing faults and defects, connect the obtained candidate defect points to form defect point lines, and determine the fault and defect type by identifying the state changes of the defect point lines.

[0023] A multi-directional filter bank (such as a directional Gabor filter with a fixed wavelength of 5 pixels, covering directions from 0° to 180° at 30° intervals) is used to convolve the cross-sectional image of the electrical connector to enhance linear and blobular features and suppress background noise. Simultaneously, the LoG algorithm is applied to detect potential circular regions, and a binary image is generated through global thresholding. Defective regions (such as cracks or bubbles) are segmented into high-brightness pixels (value 1), while the background is segmented into low-brightness pixels (value 0). Subsequently, for linear structural defects (cracks), a Hough transform is applied to the binary image to detect line segments that meet a minimum length threshold, and the endpoints or center points of these line segments are extracted as candidate defect points. For closed-region defects (bubbles), the Laplacian of Gaussian (LoG) algorithm is applied. The algorithm identifies approximately circular regions in an image and extracts the centroids or boundary points of high-response regions as candidate defect points through threshold segmentation and connected component analysis. This process extracts all candidate defect points (including linear structures and closed regions). Then, based on spatial proximity (e.g., Euclidean distance < 5 pixels) and directional consistency, the candidate defect points are connected: for linearly arranged points, they are sorted by the main direction and connected sequentially to generate polyline segments; for clustered points, Delaunay triangulation is used to generate triangular mesh connections, and the geometric properties of each connection, such as length, direction, curvature, number of branch points, and mesh density, are recorded. By identifying the state changes of the connection lines at defect points, the specific type of defect is determined: if the connection line exhibits linear characteristics (long length, consistent direction, few branch points), and the actual length of the connection line is greater than or equal to a set threshold, it is identified as a crack-type defect; further, the angle between the connection line direction and the connector pin axis is calculated. If the angle is less than the angle threshold, it is marked as a crack caused by stress concentration; if the connection line exhibits a triangular or mesh structure (number of branch points > 2), it is marked as crack propagation; if the connection line forms a closed triangular mesh, and the mesh unit area is small (e.g., the side length of the triangles is all < 20 pixels) and the mesh density is high (the number of connection lines per unit area exceeds the threshold), it is identified as a bubble-type defect. The number of bubble points in the mesh is counted. If it exceeds the porosity threshold, it is marked as excessive material porosity; if the mesh is concentrated in a local area, it is marked as a large-area void.

[0024] The defect points of different cross sections are connected and mapped to a unified three-dimensional coordinate system. Projection analysis is performed along the z-axis: if the linear connection appears in ≥3 consecutive cross sections and extends along the z-axis, it is marked as a through crack; if the clustered connection is concentrated in a specific z-value range, it indicates an injection molding abnormality in combination with injection molding process parameters. Finally, based on the dynamic changes of connection characteristics (such as length, direction, density, and distribution), the defect type and its risk level are accurately identified.

[0025] The technical solutions described in the above embodiments of this application have at least the following technical effects or advantages: the internal structure is directly exposed by micro-CT cross-sectional imaging, realizing non-destructive testing; the connection change analysis transforms discrete defects into topological features, determining the type of fault defects, which is superior to traditional testing, realizing non-destructive testing of the internal structure of electrical connectors, accurately identifying internal defects (such as cracks and bubbles), and improving the accuracy and comprehensiveness of defect identification.

[0026] Example 2: In Example 1, the detection of fault defects was static. Example 1 could not determine the propagation rate and direction of defects (such as cracks) under thermal cycling or vibration loading, thus making it impossible to predict the remaining lifespan of the electrical connector. This example uses a time dimension to achieve prediction of internal fault defects and the lifespan of the electrical connector, making fault defect detection more accurate. Figure 2 As shown.

[0027] S201, Obtain the volume of the electrical connector, which includes a reference volume and a target volume. Divide the obtained reference volume into voxel subsets. Find the subset in the target volume that best matches the voxel subset. Calculate the displacement between the matching subset and the voxel subset. Represent the displacement as a displacement vector field using components. Calculate the strain tensor field based on the displacement vector field. The reference volume is the initial volume of the electrical connector when it is not subjected to external forces, and the target volume is the volume of the electrical connector when it is subjected to external forces. Specifically, the reference volume of the electrical connector is obtained by scanning it with a 3D CT scanner in its initial state, before any external forces (such as heat, force, or electrical loads). Appropriate scanning parameters are set to obtain clear and accurate 3D image data. The obtained image data contains the internal structural information of the electrical connector in its initial state, and the corresponding volume is the reference volume. This reference volume data is stored. External forces are then applied to the electrical connector, specifically thermal cycling, mechanical loading, and electrical loading. Thermal cycling involves setting a specific temperature range and number of cycles for the electrical connector. Mechanical loading refers to applying forces of different magnitudes and directions to the electrical connector. Electrical loading refers to passing a current of a certain intensity and frequency. After the external forces are applied, the electrical connector is scanned again with a 3D CT scanner to obtain 3D image data after the external forces have been applied. The corresponding volume is the target volume, and this target volume is stored. Based on the obtained reference volume and the structure of the electrical connector, voxel sizes are set, and the reference volume is divided according to the set voxel sizes to generate multiple voxel subsets. Each voxel subset contains the structural information of a local region in the initial state of the electrical connector and can be regarded as a small volume unit. A global search algorithm is used to find the subset that best matches each voxel subset in the target volume according to the matching rules. The matching rules are to calculate the cross-correlation number between the subset in the target volume and the voxel subset in the reference volume. The larger the cross-correlation number, the higher the similarity between the two. The subset with the highest cross-correlation number is the best matching subset. For each successfully matched subset of voxels and its corresponding best-matched subset, the displacement between them is calculated. Specifically, this is determined by comparing the changes in the center coordinates of the two subsets. For example, if the center coordinates of the voxel subset in the reference volume are... Given (x1, y1, z1), the center coordinates of the most matching subset in the target volume are (x2, y2, z2). Therefore, the displacement vector d = (x2 - x1, y2 - y1, z2 - z1). This displacement calculation is performed on all voxel subsets in the reference volume. The displacement vectors corresponding to each voxel subset are arranged according to their positions in the reference volume, forming a displacement vector field. Each point in the displacement vector field corresponds to a displacement vector (u, v, w), where u, v, and w represent the components of the displacement in the x, y, and z directions, respectively. The strain tensor is a physical quantity describing the degree of deformation of an object, containing information on the normal strain and shear strain in various directions. Based on the displacement vector field, the displacement gradient tensor is calculated. The components of the displacement gradient tensor can be obtained by performing partial derivative operations on the components of the displacement vector. The components are: The formula for calculating the strain tensor based on the displacement gradient tensor is: ,in, For strain tensor field, For the displacement gradient tensor, This represents the transpose of the displacement gradient tensor.

[0028] S202, acquire four-dimensional data of the electrical connector, which consists of three-dimensional volume data and a time dimension. According to the time dimension in the four-dimensional data, track the crack movement trajectory based on the displacement vector field, construct a velocity field, construct a defect dynamics model based on the velocity field and stress factor, output the defect evolution state of the defect dynamics model, input the defect evolution state into the neural network, and the neural network predicts the remaining life of the electrical connector.

[0029] Furthermore, based on the three-dimensional volume data obtained in step S101, a time dimension is introduced to obtain three-dimensional volume data at different time points. The three-dimensional volume data at different times are arranged according to timestamps to obtain time series data, i.e., four-dimensional data. The obtained four-dimensional data is stored, and image processing technology (edge ​​detection) is used to identify the position and shape of the crack in the three-dimensional volume data at the initial time point. According to the obtained displacement vector field, starting from the initial time point, the position change of the crack at different time points is tracked sequentially. For each time point, key points that can represent the shape and position characteristics of the crack at the current moment are selected. Then, based on the displacement vector field, the displacement vector corresponding to the voxel of each key point is determined. This vector reflects the voxel from the current time to the next time. The displacement direction and magnitude of the key points are determined. Finally, based on the current position coordinates of the key points, each component of the corresponding displacement vector is added to the corresponding component of the current coordinates to obtain the position coordinates of the key points at the next time point. The positions of many key points together constitute the position of the crack at the next moment. The position is verified and adjusted in the three-dimensional volume data at the next time point to obtain the movement trajectory of the crack over time. Several points are selected on the crack movement trajectory. For each point, the local instantaneous velocity of the crack at that point is calculated according to the position changes of its adjacent time points. The calculated local instantaneous velocities of each point on the crack are arranged according to their positions in three-dimensional space to form a velocity field. The velocity field describes the propagation rate of the crack at different positions.

[0030] A defect dynamics model is constructed based on the velocity field and stress factor. The expression of the defect dynamics model is as follows: ,in, This represents the crack propagation rate, i.e., the rate of change of crack length with the number of load cycles, reflecting the average length of crack growth during each load cycle. The stress intensity factor range is the difference between its maximum and minimum values, a crucial mechanical parameter describing the strength of the stress field at the crack tip. C is a material constant, reflecting the material's sensitivity to crack propagation under specific conditions. m is the exponent of the Paris formula, reflecting the influence of the stress intensity factor range on the crack propagation rate. The current stress intensity factor range, temperature, vibration, and other parameters are input into the constructed defect dynamics model. The model calculates the crack propagation rate based on the input parameters and, combined with the initial and current crack states, outputs the defect evolution state of the electrical connector over a future period, including crack length, shape, and location. A large amount of data on the defect evolution state of the electrical connector and the corresponding remaining lifetime data is collected as a neural network. The training samples for the network, and the remaining lifetime data, are obtained through experiments or actual use. A Long Short-Term Memory (LSTM) network is used. The number of nodes in the input layer, hidden layer, and output layer of the network is designed. The number of nodes in the input layer is determined according to the dimension of the defect evolution state data. The number of nodes in the output layer is 1 (representing the remaining lifetime). The number of nodes in the hidden layer is adjusted and optimized according to the actual situation. The training samples are input into the neural network to train the parameters of the neural network. The current defect evolution state data of the electrical connector output by the defect dynamics model is input into the trained neural network. Based on the input defect evolution state data, the neural network outputs the predicted value of the remaining lifetime of the electrical connector through internal calculation and inference, that is, the number of cycles or the service time that the electrical connector can withstand in the current state.

[0031] The technical solutions in the above embodiments of this application have at least the following technical effects or advantages: Four-dimensional data can accurately record the changes in the shape, location and size of defects at different time points, quantitatively describe the expansion rate and direction of defects, and by establishing a defect dynamics model and a data-driven prediction model, accurately predict the remaining life of electrical connectors, thereby improving the accuracy and reliability of life prediction.

[0032] Example 3: Examples 1 and 2 above both treat multiple detected internal defects (such as cracks, bubbles, and pores) as independent entities for evaluation. This fails to identify the synergistic amplification effect resulting from the interaction between defects within a defect cluster, thus increasing the misjudgment rate and lifespan prediction bias of electrical connectors. Figure 3 As shown.

[0033] S301, extract defect feature vectors based on the identified fault defect types. The defect feature vectors include geometric features and spatial features. Calculate the interaction strength factor based on the defect feature vectors. Establish a defect network when the interaction strength factor is greater than a threshold. Further, according to step S103, the fault defect type is identified, and the geometric and spatial features of the fault defect are measured and recorded. For extracting the geometric features of the fault defect, these features include the volume, surface area, equivalent diameter, aspect ratio, and surface curvature. The volume of the fault defect reflects its size in three-dimensional space; the surface area reflects the size of the defect surface; the equivalent diameter equates an irregularly shaped defect to the diameter of a circle with the same area; the aspect ratio, i.e., the ratio of the major axis to the minor axis, describes the elongation of the defect shape; and the surface curvature... The curvature of the defect surface is plotted. The volume, surface area, equivalent diameter, aspect ratio, and surface curvature are calculated using mathematical formulas from existing technologies, and will not be elaborated further in this embodiment. For extracting the spatial features of the fault defect, these features include centroid coordinates and principal axis directions. The centroid coordinates of the fault defect represent its center position within the internal space of the electrical connector, and the principal axis direction describes the defect's main extension direction in space. The geometric and spatial features of each defect, along with its type label, are combined to form a defect feature vector. The interaction strength factor is calculated based on the defect feature vector using the following formula: ,in, is the interaction strength factor between defect i and defect j. Its value reflects the strength of the interaction between the two defects; the larger the value, the stronger the interaction. These are distance-related weighting coefficients used to adjust the degree of influence of distance factors on the interaction strength factor. The centroid distance between defects i and j is calculated based on the centroid coordinates in the defect feature vectors. It reflects the relative spatial position of the two defects; the closer the distance, the stronger the interaction between the two defects. These are stress-related weighting coefficients used to adjust the degree of influence of stress factors on the interaction intensity factor. , These represent the stress values ​​at defects i and j, respectively, reflecting the magnitude and direction of the force borne at a specific point within the material. These are weighting coefficients related to temperature difference, used to adjust the degree of influence of temperature difference on the interaction strength factor, ΔT. ijLet the temperature difference between defect i and defect j be the threshold. Based on actual conditions and rich experience, a threshold is set. The specific value needs to be repeatedly adjusted and verified based on actual simulation and experimental data. When the calculated interaction strength factor is greater than the preset threshold, it indicates that there is a significant interaction between defect i and defect j. At this time, a directed edge is established from defect i to defect j, and the weight of the edge is set to the interaction strength factor. Defects are used as nodes in the network, and the node state is set to live or failed, which respectively indicate whether the defect has caused the electrical connector to fail. A defect network is constructed through the directed edges between nodes. According to the size, position and type in the defect feature vector, corresponding weights are assigned to the size, position and type. The failure probabilities corresponding to the size, position and type are weighted and summed. The result is the initial failure probability of each defect node, which is used as the initial state of the defect network.

[0034] S302, based on the defect network and the current state of the defects, use the failure criteria to obtain the failure probability, and plot the failure probability change curve with time as the horizontal axis and failure probability as the vertical axis; based on the edge weights of the defect network and the failure probability, identify the critical defect chain using the tracing method.

[0035] Specifically, the time step is set according to the actual working conditions of the electrical connector. The Monte Carlo simulation method is used, and the number of simulations is set. Monte Carlo simulation is performed based on the initial state of the defect node, the current state of the defect, and the interaction strength factor between defects to simulate the evolution process of the defect in the actual working environment. After the simulation of each time step, the failure criteria of the electrical connector system are used to determine whether the system has failed. The failure criteria are that the contact resistance of the electrical connector exceeds the resistance threshold or a short circuit occurs. After the simulation of each time step, the number of system failures within that time step is counted. The number of system failures is divided by the total number of simulations to obtain the failure probability of the entire electrical connector system at that time step. The time steps are arranged in order with time as the horizontal axis and the system failure probability corresponding to each time step is marked on the coordinate axis. Then, by connecting the various data points, a curve showing the change of system failure probability over time is plotted. This curve illustrates the trend of failure risk in the electrical connector system over time. Based on the edge weights (interaction strength factors) of the defect network, defect pairs with interaction strength factors greater than the weight threshold are identified, and their contribution to the node failure probability is evaluated. The evaluation method is as follows: at a given time step, the original system failure probability is P0. When a defective node fails, the system failure probability becomes P1. The contribution of this defective node to the system failure probability is P1-P0. Based on the interaction strength factor and the node failure probability contribution, critical defect chains are identified. Specifically, starting from the node with the largest contribution to the system failure probability, the chain traces along the edge with the largest interaction strength factor value. Then, it extends the edge with the second largest interaction strength factor value and finds other important nodes connected to it, gradually forming one or more critical defect chains until they extend to an interaction strength factor less than the weight threshold. These critical defect chains constitute the failure propagation path in the electrical connector system.

[0036] The technical solutions in the above embodiments of this application have at least the following technical effects or advantages: by using the interaction strength factor and defect network, the synergistic amplification effect among multiple defects can be considered more accurately, significantly improving the accuracy of failure risk prediction; by identifying the key defect chain, the defect propagation path that contributes the most to system failure can be clearly identified, enabling maintenance resources to be accurately directed to the defect cluster that has the greatest impact on system reliability, and accurately assessing the failure risk of the electrical connector system.

[0037] Example 4: Figure 4As shown, an electrical connector fault detection system includes a data acquisition module, a 3D reconstruction module, a cross-section extraction module, an image segmentation and recognition module, a structural shape analysis module, a defect detection module, and a decision module. The data acquisition module collects 3D volume data of the electrical connector; the 3D reconstruction module constructs a 3D model of the electrical connector; the cross-section extraction module extracts axial and radial cross-sectional images of the electrical connector; the image segmentation and recognition module generates binarized images; the structural shape analysis module identifies the structural shape of the electrical connector; the defect detection module detects the fault type of the electrical connector; and the decision module determines whether the electrical connector is faulty based on the defect detection results. The data acquisition module is connected to the 3D reconstruction module; the 3D reconstruction module is connected to the cross-section extraction module; the cross-section extraction module is connected to the image segmentation and recognition module; the image segmentation and recognition module is connected to the structural shape analysis module; the structural shape analysis module is connected to the defect detection module; and the defect detection module is connected to the decision module.

[0038] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for detecting faults in electrical connectors, characterized in that, include: S101, Scan the electrical connector to obtain three-dimensional volume data, use three-dimensional software to construct an electrical connector model based on the three-dimensional volume data, and obtain a cross-sectional image of the electrical connector based on the electrical connector model; S102: The cross-section of the electrical connector is segmented to obtain a binary image of the electrical connector cross-section. The parts of the electrical connector are identified using the binary image of the electrical connector cross-section. The structural shape is extracted using an algorithm based on the parts of the electrical connector. Based on the structural shape, detection rules are used to identify whether there are any faults or defects. If faults or defects are found, step S103 is executed. If no faults or defects are found, the system outputs that no faults or defects exist. The parts include conductors, insulators, and gaps. The structural shape includes conductor outlines, insulator thickness, and connection status. The detection rules refer to using different detection methods for different structural shapes. S103, extract candidate defect points based on existing faults and defects, connect the obtained candidate defect points to form defect point lines, and determine the fault and defect type by identifying the state changes of the defect point lines.

2. The method for detecting electrical connector faults as described in claim 1, characterized in that, The specific detection rules are as follows: For the structural shape of the conductor contour, a roundness comparison method is used for detection. The roundness of the conductor is compared with a threshold. When the roundness is greater than or equal to the threshold, it indicates that the conductor has no fault defects. When the roundness is less than the threshold, it indicates that the conductor has fault defects. For the structural shape of the insulator thickness, a thickness variance comparison method is used for detection. If the calculated thickness variance is greater than the distance threshold, it indicates that there are fault defects. If the calculated thickness variance is less than or equal to the distance threshold, it indicates that the conductor has no fault defects. For the structural shape of the connection state, connectivity or contact area ratio is used for detection.

3. The electrical connector fault detection method as described in claim 2, characterized in that, The steps to determine the type of fault by identifying the state changes of the connection lines of defect points are as follows: if the connection lines exhibit linear characteristics and the actual length of the connection lines is greater than or equal to the set threshold, then it is determined to be a crack-type defect; if the connection lines form a closed triangular mesh and the area of ​​the mesh unit is less than the set area threshold and the mesh density is greater than the set quantity threshold, then it is determined to be a bubble-type defect.

4. The method for detecting electrical connector faults as described in claim 1, characterized in that, Fault detection methods also include: S201, Obtain the volume of the electrical connector, which includes a reference volume and a target volume. Divide the obtained reference volume into voxel subsets. Find the subset in the target volume that best matches the voxel subset. Calculate the displacement between the matching subset and the voxel subset. Represent the displacement as a displacement vector field using components. Calculate the strain tensor field based on the displacement vector field. The reference volume is the initial volume of the electrical connector when it is not subjected to external forces, and the target volume is the volume of the electrical connector when it is subjected to external forces. S202, acquire four-dimensional data of the electrical connector, which consists of three-dimensional volume data and a time dimension. According to the time dimension in the four-dimensional data, track the crack movement trajectory based on the displacement vector field, construct a velocity field, construct a defect dynamics model based on the velocity field and stress factor, output the defect evolution state of the defect dynamics model, input the defect evolution state into the neural network, and the neural network predicts the remaining life of the electrical connector.

5. The electrical connector fault detection method as described in claim 4, characterized in that, The method for finding the subset that best matches the voxel subset in the target volume is as follows: a global search algorithm is used to find the subset that best matches each voxel subset in the target volume according to the matching rules. The matching rules are to calculate the cross-correlation coefficient between the subset in the target volume and the voxel subset in the reference volume, and find the subset with the highest cross-correlation coefficient as the best matching subset.

6. The electrical connector fault detection method as described in claim 5, characterized in that, The formula for calculating the strain tensor field is: ,in, For strain tensor field, For the displacement gradient tensor, This represents the transpose of the displacement gradient tensor.

7. The electrical connector fault detection method as described in claim 6, characterized in that, The steps for tracing the movement trajectory of a crack based on the time dimension of four-dimensional data and the displacement vector field are as follows: Based on the obtained displacement vector field, starting from the initial time point, the positional changes of the crack at different time points are tracked sequentially. For each time point, key points that can represent the shape and positional characteristics of the crack at the current moment are selected. Then, based on the displacement vector field, the displacement vector corresponding to the voxel of each key point is determined. This vector reflects the displacement direction and magnitude of the voxel from the current time point to the next time point. Finally, based on the current position coordinates of the key point, each component of its corresponding displacement vector is added to the corresponding component of the current coordinate to obtain the position coordinates of the key point at the next time point. The positions of many key points together constitute the position of the crack at the next moment, thus obtaining the movement trajectory of the crack as time changes.

8. The electrical connector fault detection method as described in claim 4, characterized in that, The fault detection method further includes: S301, extract defect feature vectors based on the identified fault defect types. The defect feature vectors include geometric features and spatial features. Calculate the interaction strength factor based on the defect feature vectors. Establish a defect network when the interaction strength factor is greater than a threshold. S302, based on the defect network and the current state of the defects, use the failure criteria to obtain the failure probability, and plot the failure probability change curve with time as the horizontal axis and failure probability as the vertical axis; based on the edge weights of the defect network and the failure probability, identify the critical defect chain using the tracing method.

9. The electrical connector fault detection method as described in claim 8, characterized in that, The formula for calculating the interaction strength factor based on the defect eigenvector is: ,in, Let i be the interaction strength factor between defect i and defect j. These are distance-related weighting coefficients used to adjust the degree of influence of distance factors on the interaction strength factor. The distance between the centroids of defects i and j reflects the relative spatial position of the two defects. These are stress-related weighting coefficients used to adjust the degree of influence of stress factors on the interaction intensity factor. , These represent the stress values ​​at defects i and j, respectively. These are weighting coefficients related to temperature difference, used to adjust the degree of influence of temperature difference on the interaction strength factor, ΔT. ij Let be the temperature difference between defect i and defect j.

10. An electrical connector fault detection system, applied to an electrical connector fault detection method as described in any one of claims 1-9, characterized in that, It includes a data acquisition module, a 3D reconstruction module, a cross-section extraction module, an image segmentation and recognition module, a structural shape analysis module, a defect detection module, and a decision-making module; The data acquisition module is used to acquire three-dimensional volume data of the electrical connector; The 3D reconstruction module is used to construct a 3D model of the electrical connector; The cross-section extraction module is used to extract axial and radial cross-sectional images of the electrical connector; The image segmentation and recognition module is used to generate a binarized image; The structural shape analysis module is used to identify the structural shape of the electrical connector; The defect detection module is used to detect the types of faults and defects in the electrical connectors. The decision module is used to determine whether the electrical connector is faulty based on the defect detection results. The data acquisition module is connected to the 3D reconstruction module, the 3D reconstruction module is connected to the cross-section extraction module, the cross-section extraction module is connected to the image segmentation and recognition module, the image segmentation and recognition module is connected to the structural shape analysis module, the structural shape analysis module is connected to the defect detection module, and the defect detection module is connected to the decision-making module.

Images