An AI vision-guided multi-specification data line plug adaptive assembly method

Abstract

The application discloses a multi-specification data line plug self-adaptive assembly method based on AI visual guidance, and relates to the technical field of data line plug automatic assembly. The method comprises the following steps: an AI vision system identifies plug specifications and matches assembly parameters; a visual guidance execution mechanism performs a buckling operation according to a planned trajectory; a high-speed infrared thermal imaging is used to collect a temperature field distribution image at the moment of buckling, and a schlieren imaging is used to collect an air refractive index change image to represent local air pressure changes; double comprehensive judgment is performed by combining physical field judgment results and gap detection results to distinguish between two defect reasons, i.e., 'incomplete buckling' and 'part manufacturing tolerance deviation', and corresponding treatments are performed. The application significantly improves the accuracy and reliability of buckling quality detection through a double insurance detection mechanism, and has multi-specification self-adaptive assembly capability, and is suitable for the automatic assembly production line of various data line plugs.

Description

Technical Field

[0001] This invention relates to the field of automated assembly technology for data cable plugs, and more specifically to an adaptive assembly method for multi-specification data cable plugs based on AI vision guidance. Background Technology

[0002] As one of the most widely used connection accessories in electronic products, the plug assembly process in the manufacturing of data cables is a key step that determines product quality. Currently, data cable plugs generally use built-in snap-fit ​​connections, that is, the elastic clips on the insulating shell engage with the corresponding slots to reliably fix the plug shell to the internal components.

[0003] In existing automated assembly technologies for data cable plugs, mechanical pressing is primarily used for the snap-fit ​​operation. For example, a pressing device is used to press the plug shell and internal components into place. For instance, existing automated production and assembly devices for power cord plugs use two wire-end separating plates to separate the wire ends at a certain angle before the stripped ends are inserted into the pressing device. This ensures accurate insertion of the metal wire into the pressing mold, avoiding poor contact or incomplete pressing. In addition, there are intelligent USB production and assembly methods that employ magnetic field pre-positioning, elastic deformation pressing, and pulse vibration-assisted curing to achieve high-precision assembly of USB components.

[0004] However, the aforementioned existing technologies still have the following shortcomings in practical applications:

[0005] First, the methods for detecting the engagement quality are limited. Existing technologies mainly rely on mechanical force feedback or visual inspection alone to determine whether the engagement is complete. When the latch head and the latch engage, incomplete engagement may occur, meaning the elastic latch head may not fully engage at the bottom of the latch, or there may be tiny gaps between the engaging surfaces. This can lead to problems such as loosening and poor contact during use. Existing visual inspection methods can usually only judge the external appearance and are difficult to accurately identify the actual engagement status of the internal latching structure.

[0006] Second, the cause of the defect cannot be distinguished. Even if excessive gaps are detected after assembly, current technology makes it difficult to determine whether the root cause is incomplete fastening or manufacturing tolerance deviations in the plug components themselves. This prevents the assembly line from making targeted process adjustments or incoming material screening, affecting yield and production efficiency.

[0007] Third, it lacks multi-specification adaptability. Different specifications of data cable plugs differ in their external dimensions and snap-fit ​​structure parameters. Existing assembly equipment usually requires mechanical adjustments and parameter resetting for each specification, resulting in low cable replacement efficiency and insufficient flexibility.

[0008] Therefore, there is an urgent need for an intelligent method that can accurately determine the completion status of the latch engagement, distinguish the cause of defects, and adapt to the assembly of various specifications of data cable plugs. Summary of the Invention

[0009] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an AI-guided, vision-guided, multi-specification data cable plug adaptive assembly method to address the problems mentioned in the background art.

[0010] To achieve the above objectives, the present invention provides the following technical solution:

[0011] An AI-guided vision-based adaptive assembly method for multi-specification data cable plugs includes the following steps:

[0012] S1: Use an AI vision system to collect multi-view images of the data cable plug to be assembled, identify the plug specifications and buckle structure features through a pre-trained deep learning model, and match the corresponding assembly parameters from the assembly parameter database. The assembly parameters include at least the allowable error range of the gap.

[0013] S2: Based on the identified plug specifications and buckle spatial positions, the AI ​​vision system guides the actuator to perform the fastening operation according to the planned assembly trajectory;

[0014] S3: At the instant the card head and the bayonet engage, the high-speed infrared thermal imaging module of the AI ​​vision system acquires the temperature field distribution image of the engagement area, and at the same time, the schlieren imaging module of the AI ​​vision system acquires the air refractive index change image around the engagement area. The air refractive index change image is used to characterize the local air pressure change caused by the card head engaging the bayonet at the moment of engagement.

[0015] S4: Extract the instantaneous temperature rise amplitude and temperature gradient features from the temperature field distribution image, extract the local air pressure pulsation features from the air refractive index change image, input the instantaneous temperature rise amplitude, temperature gradient features and local air pressure pulsation features into the fastening state determination model, determine whether the card head has been completely inserted into the card slot, and output the preliminary determination result of the fastening state.

[0016] S5: Use the high-resolution industrial camera of the AI ​​vision system to acquire images of the gap in the buckle area after fastening, measure the gap width between the buckle head and the buckle mating surface through image processing algorithms, compare the measured gap width with the gap allowable error range, and output the gap detection result.

[0017] S6: Combine the preliminary judgment result of the fastening state and the gap detection result for a comprehensive judgment: if the preliminary judgment result of the fastening state is that the fastening is completed and the gap detection result is that the gap meets the standard, then the assembly is deemed qualified; if the preliminary judgment result of the fastening state is that the fastening is not completed and the gap detection result is that the gap does not meet the standard, then the fastening operation is deemed incomplete and a re-fastening command is triggered; if the preliminary judgment result of the fastening state is that the fastening is completed but the gap detection result is that the gap does not meet the standard, then the plug part is deemed to have a manufacturing tolerance deviation and a non-conforming product marking command is triggered.

[0018] S7: Record the inspection data and judgment results into the production quality database based on the comprehensive judgment results, and perform a second fastening for plugs that are judged to be incompletely fastened, and sort out defective plugs that are judged to be due to part deviation.

[0019] Preferably, the pre-trained deep learning model in step S1 uses ResNet-50 or EfficientNet as the backbone network, and performs transfer learning training on the basis of the pre-trained weights using a dataset containing labeled data cable plugs of various specifications to identify plug type, shape and size, buckle type and key geometric parameters.

[0020] Preferably, the frame rate of the high-speed infrared thermal imaging module in step S3 is not less than 500fps and the temperature resolution is not less than 0.02℃; the schlieren imaging module uses the change in light refractive index caused by the change in air density to visualize the air flow and pressure change at the moment of engagement.

[0021] Preferably, in step S4:

[0022] The instantaneous temperature rise amplitude is the difference between the peak temperature of the friction contact area between the clasp and the bayonet at the moment of engagement and the ambient temperature before engagement.

[0023] The temperature gradient feature characterizes the rate and direction of temperature change in space;

[0024] The local pressure pulsation features include the amplitude, duration, and spatial diffusion range of the pressure pulsation, which are extracted by optical flow analysis or image difference algorithm of the air refractive index change image, respectively.

[0025] The locking state determination model is a binary classification model based on a multilayer perceptron or support vector machine, which is trained from sample data with known locking states.

[0026] Preferably, in step S5, a high-resolution industrial camera is used in conjunction with a telecentric lens to acquire images of the buckle area. The image processing algorithm includes: adaptive histogram equalization to enhance contrast, Gaussian filtering to reduce noise, edge detection based on the Canny operator to extract the edge lines of the buckle head and the buckle opening, and at least 10 measurement points are selected in the length direction of the buckling surface to calculate the gap width, and the average value and the maximum value are taken as representative values ​​for comparison.

[0027] Preferably, in step S6, when the preliminary judgment result of the fastening state is that the fastening is not completed but the gap detection result is that the gap meets the standard, the system abnormal alarm is triggered to prompt manual intervention.

[0028] Preferably, step S7 further includes: when multiple plugs are determined to have failed to complete the fastening operation, the system automatically issues an alarm and suspends production; when a plug is determined to have a manufacturing tolerance deviation, the defective product information is fed back to the incoming material quality management system.

[0029] Preferably, the engagement state determination model in step S4 incorporates an online learning mechanism, using newly added labeled data during the production process to incrementally train the model, thereby gradually improving the accuracy of the determination and the adaptability to new models.

[0030] Preferably, step S7 also includes multi-station collaborative feedback: the detection data of multiple assembly stations are aggregated to the central quality management system through the industrial network, and statistical analysis is performed on the data of each station. When the failure rate of a certain station to complete the fastening or the part deviation rate deviates significantly from the statistical distribution of other stations, targeted early warning information is automatically generated.

[0031] Preferably, the temperature field acquisition in step S3 can be replaced by multi-point contact temperature measurement using an array of miniature thermocouples or thermistor sensors; the air pressure field acquisition in step S3 can be replaced by directly acquiring air pressure change signals using a miniature MEMS air pressure sensor array; and the gap detection in step S5 can be replaced by acquiring three-dimensional point cloud data using a 3D structured light camera or laser profilometer to calculate the three-dimensional gap volume between the card head and the card slot.

[0032] The technical effects and advantages of the AI ​​vision-guided adaptive assembly method for multi-specification data cable plugs of this invention are as follows:

[0033] By combining high-speed infrared thermal imaging and schlieren imaging technologies, the system simultaneously acquires images of the temperature field distribution and air refractive index changes at the moment of engagement. It extracts instantaneous temperature rise amplitude, temperature gradient features, and local air pressure pulsation features, inputting these into a binary classification model based on a multilayer perceptron or support vector machine for engagement status determination. This achieves real-time and accurate identification of the completed engagement state of the latches, effectively solving the problems of inaccurate judgment of the engagement state of internal latches by existing single mechanical force feedback or visual inspection, which are prone to incomplete engagement misjudgment and missed detection. A high-resolution industrial camera is introduced in conjunction with a telecentric lens to acquire images of the gap after engagement, and edge detection and multi-sensor imaging are used to further refine these images. The point gap measurement algorithm accurately quantifies the gap width between the clamp head and the clamping surface, improving the detection accuracy of static gaps after fastening, and forming a dual insurance mechanism of physical process detection and static result detection. Through the comprehensive judgment logic of the preliminary judgment result of the fastening state and the gap detection result, the system automatically distinguishes the situation where the physical field characteristics are normal but the gap exceeds the standard as the manufacturing tolerance deviation of the part, and automatically judges the situation where the physical field characteristics are abnormal and the gap exceeds the standard as the fastening operation is not completed and triggers re-fastening. This realizes the automatic differentiation and targeted handling of the causes of assembly defects, and improves the timeliness, accuracy and reliability of the assembly quality inspection of data cable plugs. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of an AI vision-guided adaptive assembly method for multi-specification data cable plugs according to the present invention. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0036] Example 1

[0037] Figure 1 A schematic diagram of an AI vision-guided adaptive assembly method for multi-specification data cable plugs according to the present invention is provided, which includes the following steps:

[0038] Step S1: Plug specification identification and assembly parameter matching.

[0039] The AI ​​vision system first uses multi-angle industrial cameras placed at the assembly station to capture images of the appearance of the data cable plugs to be assembled, including front views, side views, and magnified views of the snap-fit ​​area.

[0040] A pre-trained convolutional neural network (CNN) model is used to extract features and classify acquired multi-view images. This CNN model employs ResNet-50 or EfficientNet as its backbone network and is trained using transfer learning on top of ImageNet pre-trained weights, employing a labeled dataset containing at least 20 common data cable plug specifications. The trained model can identify the plug type, dimensions, latch type, and key geometric parameters of the latch.

[0041] Based on the identification results, the corresponding process parameters are automatically matched from a pre-established assembly parameter database, including at least the allowable error range for gaps. And the threshold for determining temperature changes detected by infrared thermal imaging. Barometric pressure pulsation detection threshold for schlieren imaging .

[0042] Step S2: Assembly trajectory planning and adaptive fastening execution.

[0043] Based on the plug specification parameters and buckle spatial position coordinates identified in step S1, the spatial positioning module of the AI ​​vision system calculates the relative positional relationship between the buckle head and the buckle opening and the snapping direction vector, generating the optimal assembly trajectory. The system controls the servo-driven robotic arm or pneumatic pressing mechanism to perform the snapping operation according to the planned assembly trajectory, monitoring force and displacement in real time. When the snapping displacement reaches the preset stroke tolerance range, the system enters the quality inspection stage.

[0044] Step S3: Acquire physical field parameters at the moment of engagement.

[0045] This step uses an AI vision system to simultaneously collect two types of physical field parameters at the moment of engagement.

[0046] For temperature field acquisition, a high-speed infrared thermal imaging camera with a frame rate of no less than 500fps and a temperature resolution of no less than 0.02℃ is used. The camera is aligned with the latching area, and high-speed continuous acquisition is started at the same time as the latching action is triggered to obtain a time series image of the temperature field distribution of the contact area between the latch head and the latch at the moment of latching.

[0047] For acquiring air pressure field data, a schlieren imaging system is used. This system utilizes the principle of changes in light refractive index caused by variations in air density around the latch to visualize airflow and pressure changes at the moment of engagement, capturing images of air refractive index changes. When the latch head engages at high speed, compressed air is rapidly expelled from the latch gap, creating characteristic pressure pulsations and airflow diffusion patterns. The schlieren imaging system captures this process visually.

[0048] Step S4: Physical field feature extraction and preliminary determination of the snapping state.

[0049] Feature extraction is performed on the image acquired in step S3.

[0050] From the time series image of temperature field distribution, frames within a 100ms time window before and after the snap-fit ​​instant are selected. The maximum temperature of the snap-fit ​​contact area is calculated frame by frame, and the instantaneous temperature rise value is extracted. Simultaneously calculate the temperature gradient eigenvector. From schlieren imaging sequences, dynamic features of regions exhibiting changes in air refractive index are extracted using optical flow or image difference algorithms, and the amplitude of air pressure fluctuations is calculated. pulsation duration and spatial diffusion range radius .

[0051] physical field eigenvectors Input the engagement state determination model. This determination model, as described in claim 4, is a binary classification model based on a multilayer perceptron or support vector machine, and outputs a preliminary determination result of the engagement state. ∈{“Matching complete”, “Matching incomplete”}.

[0052] Step S5: Visual inspection and secondary verification of gaps.

[0053] After the fastening action is completed and a preset stabilization delay has elapsed, the high-resolution industrial camera of the AI ​​vision system, in conjunction with the telecentric lens, captures static high-definition images of the fastening area.

[0054] The image is sequentially subjected to adaptive histogram equalization, Gaussian filtering for noise reduction, and Canny edge detection to extract the card head edge and the card slot edge. At least 10 measurement points are uniformly selected along the length of the mating surface, and the gap width at each point is calculated and averaged. and maximum value This serves as a representative value. The representative value is compared with the allowable error range for the gap, and the gap detection result is output.

[0055] Step S6: Comprehensive Judgment and Defect Cause Differentiation: A comprehensive judgment is made based on the preliminary judgment result of the fastening status and the gap detection result, using three types of judgment rules. Furthermore, in the extremely rare case where the preliminary judgment result of the fastening status is "fastening incomplete" but the gap detection result is "gap meets standards," the system triggers an abnormal alarm to prompt manual intervention for inspection, preventing misjudgments caused by detection system malfunctions.

[0056] Step S7: Feedback control and data recording.

[0057] Based on the comprehensive judgment results, corresponding operations are performed, and the detection data and judgment results are recorded in the production quality database. When multiple plugs are judged to have failed to complete the snap-fit ​​operation, the system automatically issues an alarm and suspends production; when a plug is judged to have a manufacturing tolerance deviation, the defective product information is fed back to the incoming material quality management system, providing data support for upstream supplier quality control.

[0058] Example 2

[0059] Based on Example 1, this implementation method further adds online model updates and multi-station collaboration functions for large-scale continuous production scenarios.

[0060] Specifically, an online learning mechanism is introduced into the engagement state determination model in step S4. As the amount of detection data accumulated during the production process increases, the system periodically uses newly added labeled data to incrementally train the determination model, gradually enhancing the model's adaptability to new plug models and improving the accuracy of the determination.

[0061] In step S7, a multi-station collaborative feedback function is added. Inspection data from multiple assembly stations are aggregated to the central quality management system via an industrial network. The central system performs statistical analysis on the inspection data from each station. When the "incomplete engagement" or "part deviation" judgment rate of a certain station significantly deviates from the statistical distribution of other stations, the system automatically identifies an equipment malfunction at that station or a quality fluctuation in the batch of incoming materials, and generates targeted early warning information. This spatiotemporal dimension-based verification inspection mechanism effectively suppresses the interference of environmental factors on the inspection results, improving the stability and reliability of the system.

[0062] Example 3

[0063] Based on Example 1, this implementation provides alternative solutions for the physical field detection module and the gap detection module.

[0064] In terms of temperature field acquisition, in addition to using a high-speed infrared thermal imaging camera, an array of miniature thermocouples or thermistor sensors can be used as alternatives to obtain temperature change data of the buckle area through multi-point contact temperature measurement, which is suitable for cost-sensitive application scenarios.

[0065] For acquiring air pressure fields, in addition to using a schlieren imaging system, a miniature MEMS pressure sensor array can be used as an alternative, arranged around the snap-fit ​​area to directly acquire the air pressure change signal at the moment of snapping. The MEMS pressure sensor solution is less expensive, but its spatial resolution is lower than that of the schlieren imaging solution.

[0066] In terms of gap detection, in addition to high-resolution industrial cameras with image processing, 3D structured light cameras or laser profilometers can be used to obtain three-dimensional point cloud data of the buckle area. Point cloud processing algorithms can then be used to calculate the three-dimensional gap volume between the buckle head and the buckle opening, achieving higher precision gap assessment.

[0067] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0068] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid-state drive.

[0069] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0070] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0071] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0072] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0073] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0074] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0075] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0076] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-specification data cable plug adaptive assembly method based on AI vision guidance, characterized in that, Includes the following steps: S1: Use an AI vision system to collect multi-view images of the data cable plug to be assembled, identify the plug specifications and buckle structure features through a pre-trained deep learning model, and match the corresponding assembly parameters from the assembly parameter database. The assembly parameters include at least the allowable error range of the gap. S2: Based on the identified plug specifications and buckle spatial positions, the AI ​​vision system guides the actuator to perform the fastening operation according to the planned assembly trajectory; S3: At the instant the card head and the bayonet engage, the high-speed infrared thermal imaging module of the AI ​​vision system acquires the temperature field distribution image of the engagement area, and at the same time, the schlieren imaging module of the AI ​​vision system acquires the air refractive index change image around the engagement area. The air refractive index change image is used to characterize the local air pressure change caused by the card head engaging the bayonet at the moment of engagement. S4: Extract the instantaneous temperature rise amplitude and temperature gradient features from the temperature field distribution image, extract the local air pressure pulsation features from the air refractive index change image, input the instantaneous temperature rise amplitude, temperature gradient features and local air pressure pulsation features into the fastening state determination model, determine whether the card head has been completely inserted into the card slot, and output the preliminary determination result of the fastening state. S5: Use the high-resolution industrial camera of the AI ​​vision system to acquire images of the gap in the buckle area after fastening, measure the gap width between the buckle head and the buckle mating surface through image processing algorithms, compare the measured gap width with the gap allowable error range, and output the gap detection result. S6: Combine the preliminary judgment result of the fastening state and the gap detection result to make a comprehensive judgment: if the preliminary judgment result of the fastening state is that the fastening is completed and the gap detection result is that the gap meets the standard, then the assembly is deemed qualified. If the initial assessment of the fastening status indicates that the fastening is incomplete and the gap detection result indicates that the gap does not meet the standard, then the fastening operation is determined to be incomplete, triggering a re-fastening command. If the initial assessment of the engagement status indicates that the engagement is complete, but the gap inspection result indicates that the gap does not meet the standard, then it is determined that the plug part has a manufacturing tolerance deviation, triggering a non-conforming product marking instruction. S7: Record the inspection data and judgment results into the production quality database based on the comprehensive judgment results, and perform a second fastening for plugs that are judged to be incompletely fastened, and sort out defective plugs that are judged to be due to part deviation.

2. The method according to claim 1, characterized in that, The pre-trained deep learning model described in step S1 uses ResNet-50 or EfficientNet as the backbone network. Based on the pre-trained weights, it uses a dataset containing labeled data cable plugs of various specifications for transfer learning training to identify plug type, size, buckle type and key geometric parameters.

3. The method according to claim 1, characterized in that, In step S3, the frame rate of the high-speed infrared thermal imaging module is not less than 500fps and the temperature resolution is not less than 0.02℃; the schlieren imaging module uses the change in light refractive index caused by the change in air density to visualize the air flow and pressure change at the moment of engagement.

4. The method according to claim 1, characterized in that, In step S4: The instantaneous temperature rise amplitude is the difference between the peak temperature of the friction contact area between the clasp and the bayonet at the moment of engagement and the ambient temperature before engagement. The temperature gradient feature characterizes the rate and direction of temperature change in space; The local pressure pulsation features include the amplitude, duration, and spatial diffusion range of the pressure pulsation, which are extracted by optical flow analysis or image difference algorithm of the air refractive index change image, respectively. The locking state determination model is a binary classification model based on a multilayer perceptron or support vector machine, which is trained from sample data with known locking states.

5. The method according to claim 1, characterized in that, In step S5, a high-resolution industrial camera with a telecentric lens is used to acquire images of the buckle area. The image processing algorithm includes: adaptive histogram equalization to enhance contrast, Gaussian filtering to reduce noise, edge detection based on the Canny operator to extract the edge lines of the buckle head and the buckle opening, and at least 10 measurement points are selected in the length direction of the buckling surface to calculate the gap width. The average value and the maximum value are taken as representative values ​​for comparison.

6. The method according to claim 1, characterized in that, In step S6, when the preliminary judgment result of the fastening status is that the fastening is not completed but the gap detection result is that the gap meets the standard, the system abnormal alarm is triggered to prompt manual intervention.

7. The method according to claim 1, characterized in that, Step S7 also includes: when multiple plugs are determined to have failed to complete the engagement operation, the system automatically issues an alarm and suspends production; when a plug is determined to have a manufacturing tolerance deviation, the defective product information is fed back to the incoming material quality management system.

8. The method according to claim 1, characterized in that, In step S4, the engagement state determination model introduces an online learning mechanism, using newly added labeled data during the production process to incrementally train the model, gradually improving the accuracy of the determination and the adaptability to new models.

9. The method according to any one of claims 1 to 8, characterized in that, Step S7 also includes multi-station collaborative feedback: the detection data of multiple assembly stations are aggregated to the central quality management system through the industrial network, and statistical analysis is performed on the data of each station. When the failure rate of a certain station to complete the fastening or the part deviation rate deviates significantly from the statistical distribution of other stations, targeted early warning information is automatically generated.

10. The method according to claim 1, characterized in that, The temperature field acquisition in step S3 can be replaced by multi-point contact temperature measurement using an array of miniature thermocouples or thermistor sensors; the air pressure field acquisition in step S3 can be replaced by directly acquiring air pressure change signals using a miniature MEMS air pressure sensor array; the gap detection in step S5 can be replaced by acquiring three-dimensional point cloud data using a 3D structured light camera or laser profilometer to calculate the three-dimensional gap volume between the card head and the card slot.

Images