Zip-top can seaming nondestructive detection device and method based on bimodal detection

Abstract

The invention relates to a zip-top can seaming nondestructive detection device and method based on bimodal detection. The device comprises a conveying module, a cleaning module, a detection module, a control module and a collection module, wherein the conveying module continuously transfers can bodies through a mechanical arm; the cleaning module adopts a high-pressure air knife to purge and remove impurities; the detection module obtains an external three-dimensional contour through laser scanning, and an internal sealing structure is analyzed through X-ray imaging; the control module fuses the laser point cloud and the X-ray image, a machine learning model is used for identifying the seaming defect, and the classification accuracy is larger than or equal to 99.5%; and the collection module sorts qualified products and defective products according to detection results. The method comprises the steps of multi-source data space-time alignment, feature level fusion, dynamic parameter optimization and closed-loop process correction, detection parameters are adjusted in real time through quality indexes, and defects and production data are associated to achieve self-learning optimization. The method is suitable for a high-speed production line, the defect detection rate is larger than or equal to 99.9%, the false alarm rate is smaller than or equal to 0.1%, and the detection efficiency and the process controllability are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of metal packaging container detection, and in particular to a non-destructive detection device and method for the can end seaming of a beverage can based on dual-mode detection of laser and X-ray, which is applicable to the quality monitoring of end seaming in high-speed production lines in industries such as beverages and beer. Background Art

[0002] Traditional can end seaming detection mostly relies on manual visual inspection, mechanical contact detection or single optical detection, and has the following defects:

[0003] 1. Insufficient accuracy: Manual visual inspection depends on the experience of workers and cannot achieve stable and consistent detection; Optical detection is easily interfered by surface reflection and cannot detect internal structural defects (such as broken sealant and deformed hook).

[0004] 2. Low efficiency: Contact detection requires shutdown operation and is difficult to adapt to high-speed production lines (>1000 cans / minute).

[0005] 3. Isolated data: The internal and external detection data are separated and cannot comprehensively evaluate the integrity of the end seaming. Summary of the Invention

[0006] The purpose of the present invention is to provide a non-destructive detection device and method for can end seaming that integrates dual-mode detection of laser and X-ray, data fusion analysis and intelligent classification, so as to solve the problems of single detection dimension, low accuracy and insufficient efficiency in the prior art.

[0007] To achieve the above purpose, the present invention adopts the following technical solutions:

[0008] In the first aspect, the present invention provides a non-destructive detection device for can end seaming based on dual-mode detection, including:

[0009] A conveying module for continuously conveying beverage cans to be detected, including an arrangement groove and a multi-axis robotic arm. The arrangement groove is used to place continuously arranged beverage cans, and the multi-axis robotic arm is configured to sequentially transfer the beverage cans to a laser detection module and an X-ray detection module;

[0010] A dual-mode detection module, including:

[0011] A laser detection module configured to emit line laser to scan the end seaming contour outside the beverage can and obtain three-dimensional contour data through a high-precision laser camera;

[0012] An X-ray detection module configured to emit soft X-rays to penetrate the beverage can and collect internal end seaming structure images through a flat panel detector;

[0013] The control module, including a data fusion unit and a defect analysis unit, is used to synchronously control the laser detection module and the X-ray detection module, and perform multi-dimensional fusion on the detection data of both, and identify the seaming defects by combining machine learning algorithms.

[0014] Furthermore, the device is also provided with a cleaning module, including: a positioning component and a cleaning component. The positioning component is arranged below the cleaning component and is used to realize the fixed clamping of the tank body. The cleaning component is configured to use a high-pressure air knife to remove impurities on the surface of the tank lid.

[0015] Furthermore, the device is also provided with a collection module, including: a pushing component and a collection component, which are configured to transfer the tested beverage cans out of the arrangement groove.

[0016] Furthermore, the line laser wavelength of the laser detection module is 405 - 650 nm, the scanning frequency ≥ 2000 Hz, and the resolution ≤ 0.01 mm; the tube voltage of the X-ray detection module is 30 - 80 kV, and the dose rate ≤ 1 μSv / h.

[0017] Furthermore, the control module is further integrated with an automatic calibration unit and an alarm unit. The automatic calibration unit regularly performs the calibration of the laser detection module and the X-ray detection module, and the calibration deviation threshold is set to ±0.05% FS; the alarm unit emits an alarm signal when the laser detection module and / or the X-ray detection module detects a non-conforming beverage can.

[0018] Furthermore, the control module is also integrated with a wireless communication unit, which supports remote monitoring and data uploading to the cloud server; the defect analysis unit is built-in with a convolutional neural network model, and the training data set contains 100,000 groups of seaming defect samples.

[0019] Furthermore, the X-ray detection module includes a radiation source, a flat panel detector and a second tank body positioning mechanism. The second tank body positioning mechanism realizes adjustable installation in five directions through a five-axis drive mechanism.

[0020] In a second aspect, the present invention provides a method for fusing and analyzing seaming defects of beverage cans based on bimodal data, which is based on the seaming detection device described in the first aspect above, and includes the following steps:

[0021] S01. Synchronous acquisition of multi-source data:

[0022] Obtain the three-dimensional contour point cloud data of the external seaming of the beverage can through the laser detection module;

[0023] Obtain the tomographic grayscale image of the internal seaming structure of the beverage can through the X-ray detection module;

[0024] Perform spatio-temporal alignment on the point cloud data and grayscale image to generate a spatio-temporal correlation dataset;

[0025] S02. Multidimensional data fusion:

[0026] Extract contour curvature, height deviation, and surface flatness features from the point cloud data;

[0027] Extract sealant distribution density, hook continuity, and internal porosity features from the grayscale image;

[0028] Adopt a feature-level fusion strategy, input the above features into a machine learning model, and generate a joint defect probability matrix;

[0029] S03. Defect classification and decision-making:

[0030] According to the joint defect probability matrix, divide the defect levels; and trigger a hierarchical response mechanism according to the defect levels.

[0031] Furthermore, in step S01, the spatio-temporal alignment is achieved in the following manner:

[0032] Time synchronization: Add a unified timestamp to the laser and X-ray data, with an error ≤ 1 ms;

[0033] Spatial mapping: Establish a tank coordinate system, and correlate the point cloud data with the image pixel coordinates through an affine transformation matrix.

[0034] In a third aspect, the present invention provides a dynamic optimization method for detecting the can end seaming based on bimodal data, including the following steps:

[0035] S10. Real-time data feedback:

[0036] During the operation of the production line, continuously collect laser and X-ray detection data;

[0037] Calculate the comprehensive index Q_index of the seaming quality, and the formula is: Q_index = 0.6 × laser contour score + 0.4 × X-ray structure score S11. Dynamic parameter adjustment:

[0038] If Q_index of 10 consecutive cans is lower than the threshold, automatically adjust the parameters of the detection module:

[0039] Increase the laser scanning frequency by 10% - 20%;

[0040] Increase the X-ray tube voltage by 5 kV - 10 kV;

[0041] Adjust the data fusion weight to 0.8 for laser and 0.2 for X-ray;

[0042] If Q_index is continuously higher than the threshold, reduce the detection frequency to save energy consumption;

[0043] S12. Self - learning model update:

[0044] Add the real - time detection data to the training set and iteratively update the machine - learning model every 24 hours;

[0045] For new defect patterns, automatically generate adversarial samples and enhance the model's robustness.

[0046] The beneficial effects of the present invention are as follows:

[0047] 1. High precision: Double verification by laser and X - ray, the dimension detection accuracy reaches 0.01 mm, which is better than the industry standard (0.05 mm).

[0048] 2. High efficiency: Support online detection, and the efficiency is more than 3 times higher than the traditional method.

[0049] 3. Full - dimension evaluation: Simultaneously detect the external contour and internal structure to avoid missed detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 According to an embodiment of the present invention, a schematic diagram of the overall structure of a non - destructive inspection device for can end - seaming by dual - mode detection is shown;

[0051] Figure 2 According to an embodiment of the present invention, a schematic diagram of the internal structure of a non - destructive inspection device for can end - seaming is shown;

[0052] Figure 3 According to an embodiment of the present invention, a schematic diagram of the can - separating component and positioning component in the cleaning module is shown;

[0053] Figure 4 According to an embodiment of the present invention, a schematic diagram of the cleaning component in the cleaning module is shown;

[0054] Figure 5 According to an embodiment of the present invention, a schematic diagram of the laser detection module is shown;

[0055] Figure 6 According to an embodiment of the present invention, a schematic diagram of the X - ray detection module is shown;

[0056] Figure 7 According to an embodiment of the present invention, a schematic diagram of the pushing component in the collection module is shown;

[0057] Figure 8 According to an embodiment of the present invention, a schematic diagram of the collection component in the collection module is shown. DETAILED DESCRIPTION OF THE INVENTION

[0058] The following will, in conjunction with the accompanying drawings, provide a more detailed description of the technical features and advantages of this application, so that the advantages and features of this application can be more easily understood by those skilled in the art, thereby making the scope of protection of the present invention more clearly defined.

[0059] Please refer to Figures 1-8 , an embodiment of the present invention first provides a non-destructive detection device for the can end seaming based on the dual modalities of laser and X-ray. This device includes the following core modules:

[0060] Conveyor module 1: It includes an arrangement groove 101, a can separation assembly 102, and a multi-axis robotic arm 103. The arrangement groove 101 is used to place regularly arranged cans. The can separation assembly 102 is configured to push the arranged cans forward along the arrangement groove 101 in sequence. A gripper is provided at the end of the multi-axis robotic arm 103 for gripping the can 8 and transporting it between each module, with a positioning repeat accuracy reaching ±0.05 mm.

[0061] Cleaning module 2: It includes a positioning assembly 201 and a cleaning assembly 202. The positioning assembly 201 is arranged directly below the cleaning assembly 202 for achieving fixed clamping of the can body. The cleaning assembly 202 can be configured to use a high-pressure air knife 2021 to remove impurities on the surface of the can lid to improve the accuracy of subsequent seaming detection. The air pressure of the high-pressure air knife 2021 can be set to 0.3 - 0.5 MPa, and at the same time, an inclination angle of 30° - 45° can be set to optimize the purging effect. The can 8 is successively conveyed onto the positioning assembly 201 under the push of the can separation assembly 102. The positioning assembly 201 clamps and fixes the can 8. The cleaning assembly 202 is activated, and the impurities on the surface of the can lid of the can 8 are purged clean by the high-pressure air knife 2021. The positioning assembly 201 releases the purged can 8, and it continues to move forward along the arrangement groove 101 under the push of the subsequent cans until it is blocked by the limit assembly 104. To meet the cleaning requirements of cans 8 with different height dimensions, the high-pressure air knife 2021 can be drivingly connected to a first lifting cylinder 2023 to achieve up and down movement along the guide rail 2022, thereby adjusting the installation position in the Z direction.

[0062] Dual-modality detection module:

[0063] Laser detection module 3: equipped with 405-650nm line laser, scanning the sealing contour of the can body. It includes a laser camera 301 and a first can body positioning mechanism 302, which is movably mounted on the frame 7. By adjusting the installation position of the first can body positioning mechanism 302, it can meet the detection and installation requirements of cans of different specifications and sizes. Specifically, the first can body positioning mechanism 302 includes a first clamping fixture 3021, which is installed on the first mounting bracket 3022, and the first clamping fixture 3021 is connected to the second lifting cylinder 3023 to achieve Z-direction lifting adjustment, so as to meet the installation requirements of cans 8 of different heights and sizes; a slider guide mechanism 3025 is set between the first mounting bracket 3022 and the frame 7, and the first mounting bracket 3022 is connected to the first sliding cylinder 3024, and the installation position of the first clamping fixture 3021 in the X direction can be adjusted through the first sliding cylinder 3024, so as to adjust the distance between the can and the laser camera 301. In addition, the laser camera 301 can also be connected to the second sliding cylinder 3011 to achieve position adjustment in the Y direction, and the transmission effect of the above-mentioned cylinders can realize the free adjustment of the distance between the can 8 and the laser camera 301 in the X, Y, and Z directions.

[0064] X-ray detection module 4: Libra09NJ-01 micro-focus X-ray source penetrates the can body and generates an internal sealing structure image with a resolution of 5μm. It includes an X-ray source 401, a flat-panel detector 402 and a second can body positioning mechanism 403. The second can body positioning mechanism 403 is used for positioning and clamping the can 8 at the X-ray detection station. The X-ray source 401 emits soft X-rays to penetrate the can 8, and then the flat-panel detector 402 collects the sealing structure image inside the can. Specifically, the second can body positioning mechanism 403 includes a second clamping fixture 4031 and a five-axis driving mechanism 4032. The second clamping fixture 4031 can realize the installation position adjustment in the X direction, Y direction and Z direction through the five-axis driving mechanism 4032, so as to meet the detection requirements of different cans 8.

[0065] Control module (not shown): equipped with industrial computer and customized software, supporting data fusion, defect analysis, automatic calibration and real-time alarm functions. The control module has built-in data fusion unit, defect analysis unit, automatic calibration unit, alarm unit and wireless communication unit; the data fusion unit uses FPGA to achieve real-time processing, with a delay of ≤10ms; the defect analysis unit supports online model update and adapts to different can types (such as slimming cans, standard cans); the automatic calibration unit regularly (for example, monthly) performs calibration of the laser detection module 3 and the X-ray detection module 4, and the calibration deviation threshold is set to ±0.05% FS; the alarm unit sends an alarm signal when the laser detection module 3 and / or the X-ray detection module 4 detects unqualified cans.

[0066] Collection module 5: It includes a pushing component 501 and a collecting component 502, and is configured to transfer the qualified aluminum cans 8 that have completed the detection out of the arranging groove 101. The pushing component 501 includes a pushing cylinder 5011 and a pushing claw clip 5012. After the aluminum can has completed the laser detection and X-ray detection, the multi-axis robotic arm 103 transfers the aluminum can above the pushing component 501. The pushing cylinder 5011 drives the pushing claw clip 5012 to act, so that the aluminum can falls into the arranging groove 101 and moves forward one by one in the arranging groove 101 under the pushing of the subsequent aluminum cans. The collecting component 502 includes a first pushing baffle 5021, a first pushing cylinder 5022, a second pushing baffle 5023, a second pushing cylinder 5024, and a can collecting groove 5025. The first pushing cylinder 5022 drives the first pushing baffle 5021 to act once, transferring one aluminum can 8 in the arranging groove 101 to the can collecting groove 5025. The second pushing cylinder 5024 drives the second pushing baffle 5023 to act once, moving the aluminum can 8 forward by one position in the can collecting groove 5025. When the can collecting groove 5025 is full of aluminum cans 8, a row of aluminum cans 8 can be transferred out of the entire device manually or automatically.

[0067] The working principle of the embodiment of the present invention is as follows:

[0068] External contour detection: Line laser is projected onto the can body's seaming area, and the laser camera 301 captures the deformed light stripes, generating a three-dimensional point cloud through phase calculation (accuracy ±5μm);

[0069] Internal structure imaging: X-rays penetrate the can body, and the flat panel detector 402 receives the attenuation signal, and reconstructs a tomographic image through the filtered back projection algorithm (resolution 50μm);

[0070] Intelligent decision-making: The fused data is input into the ResNet-50 model, and the defect probability and type are output (such as "missing can hook" "sealant fault"), and the classification accuracy rate ≥99.5%.

[0071] The full-automatic detection process is as follows: Can body positioning → Can lid cleaning → Laser scanning → X-ray imaging → Data analysis → Can body collection, without manual intervention throughout the process.

[0072] Embodiment 1:

[0073] A certain brewery applies this device, and the parameters are set as follows:

[0074] Laser scanning speed: 5000Hz;

[0075] X-ray voltage: 60kV, current: 150μA;

[0076] Qualified standard: Can hook length ≥1.80mm, lid hook length ≥1.75mm, overlap degree ≥0.40mm.

[0077] After testing, the missed detection rate has dropped from 2.1% to 0.05%, saving about 280,000 yuan in annual rework costs.

[0078] Based on the above can end sealing detection device, an embodiment of the present invention further provides a method for fusing and analyzing can end sealing defects based on bimodal data, including the following steps:

[0079] S01. Synchronous acquisition of multi-source data:

[0080] Obtain the three-dimensional contour point cloud data of the external can end seal through the laser detection module 3;

[0081] Obtain the tomographic gray-scale image of the internal can end seal structure through the X-ray detection module 4;

[0082] Perform spatio-temporal alignment on the above point cloud data and gray-scale image to generate a spatio-temporal correlation data set;

[0083] S02. Multi-dimensional data fusion:

[0084] Extract contour curvature, height deviation, and surface flatness features from the point cloud data;

[0085] Extract sealant distribution density, hook continuity, and internal porosity features from the gray-scale image;

[0086] Adopt a feature-level fusion strategy, input the above features into a random forest model, and generate a joint defect probability matrix;

[0087] S03. Defect classification and decision-making:

[0088] According to the joint defect probability matrix, divide the defect levels:

[0089] Level 1 defect: Sealant fracture, hook missing (probability ≥ 90%);

[0090] Level 2 defect: Loose can end seal, local deformation (probability 60%-90%);

[0091] Level 3 defect: Slight burrs, uneven coating (probability 30%-60%);

[0092] Trigger a hierarchical response mechanism according to the defect level:

[0093] Level 1 defect: Immediately stop the machine and alarm;

[0094] Level 2 defect: Mark the batch and reduce the production line speed;

[0095] Level 3 defect: Record the data and prompt manual re-inspection.

[0096] In step S01, the spatio-temporal alignment is achieved in the following way:

[0097] Time synchronization: Add a unified timestamp to the laser and X-ray data, with an error ≤ 1ms;

[0098] Spatial mapping: Establish a tank coordinate system and correlate the point cloud data with the image pixel coordinates through an affine transformation matrix;

[0099] In step S02, the random forest model is trained with historical data, and the training set includes:

[0100] 100,000 groups of normal seaming samples;

[0101] 50,000 groups of artificially simulated defect samples;

[0102] 30,000 groups of on-line measured defect samples.

[0103] Embodiment 2

[0104] Taking an aluminum fiber tank as an example, the scanning frequency of the laser module is 2000Hz, and the X-ray tube voltage is 50kV;

[0105] After spatio-temporal alignment, the standard deviation of the laser contour curvature is extracted as 0.12, and the X-ray sealant coverage rate is 98.5%;

[0106] The random forest model outputs a probability of missing hook of 92.7%, triggering a first-level defect response, stopping the production line and prompting to replace the seaming die.

[0107] Based on the above seaming detection device and the seaming defect fusion analysis method, an embodiment of the present invention further provides a dynamic optimization method for detecting the seaming of aluminum cans based on bimodal data, including the following steps:

[0108] S10. Real-time data feedback:

[0109] During the operation of the production line, continuously collect laser and X-ray detection data;

[0110] Calculate the comprehensive index of seaming quality (Q_index), and the formula is: Q_index = 0.6 × laser contour score + 0.4 × X-ray structure score

[0111] The laser contour score is the weighted sum of the following indicators:

[0112] Contour smoothness (weight 0.4); seaming width deviation (weight 0.3); consistency of countersink depth (weight 0.3);

[0113] The X-ray structure score is the weighted sum of the following indicators:

[0114] Sealant coverage rate (weight 0.5); overlap degree of hook and can hook (weight 0.3); number of internal pores (weight 0.2).

[0115] S11. Dynamic parameter adjustment:

[0116] If Q_index of 10 consecutive cans is lower than the threshold, the parameters of the detection module are automatically adjusted:

[0117] The laser scanning frequency is increased by 10%-20%; the X-ray tube voltage is increased by 5 kV - 10 kV; the data fusion weight is adjusted to 0.8 for laser and 0.2 for X-ray;

[0118] If Q_index is continuously higher than the threshold, the detection frequency is reduced to save energy consumption;

[0119] S12. Self-learning model update:

[0120] The real-time detection data is added to the training set, and the random forest model is iteratively updated every 24 hours;

[0121] For new defect patterns, adversarial samples are automatically generated to enhance the robustness of the model.

[0122] Example 3

[0123] When the production line speed is 800 cans / minute, Q_index of 10 consecutive cans is 72 points (threshold 75 points);

[0124] The laser scanning frequency is automatically increased to 2200 Hz, and the X-ray tube voltage is increased to 55 kV;

[0125] After adjustment, Q_index is restored to 78 points, and the defect rate is reduced from 0.5% to 0.1%.

[0126] Furthermore, the embodiment of the present invention also provides a closed-loop control method for the seaming defects of aluminum cans, including the following steps:

[0127] S20. Defect traceability analysis:

[0128] Associate production parameters (such as seaming machine pressure, speed, temperature) according to the defect type;

[0129] Construct a defect-parameter relationship matrix to locate the source of process deviation;

[0130] S21. Automatic process correction:

[0131] If the defect is caused by insufficient seaming pressure, the control module sends an instruction to the seaming machine to increase the pressure by 5%-10%;

[0132] If the defect is caused by the fluctuation of the conveying speed, the grasping interval time of the multi-axis robotic arm 103 is adjusted;

[0133] S22. Verification and feedback:

[0134] After 50 tanks are continuously detected after correction, if the defect rate drops by ≥ 30%, the correction parameters are locked;

[0135] Otherwise, trigger the intervention of the expert system to generate a list of manual debugging suggestions.

[0136] In the description of the embodiments of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "center", "top", "bottom", "top part", "bottom part", "inner", "outer", "inner side", "outer side", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation to the present invention. Among them, the "inner side" refers to the internal or enclosed area or space. The "periphery" refers to the area around a specific component or a specific area.

[0137] In the description of the embodiments of the present invention, the terms "first", "second", "third", "fourth" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first", "second", "third", "fourth" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, the meaning of "a plurality" is two or more.

[0138] In the description of the embodiments of the present invention, it should be noted that, unless otherwise clearly specified and limited, the terms "installed", "connected", "joined", "assembled" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the internal communication of two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0139] In the description of the embodiments of the present invention, specific features, structures, materials or characteristics may be combined in a suitable manner in any one or more embodiments or examples.

[0140] In the description of the embodiments of the present invention, it should be understood that "-" and "~" represent the range between two numerical values, and this range includes the endpoints. For example: "A - B" represents the range greater than or equal to A and less than or equal to B. "A ~ B" represents the range greater than or equal to A and less than or equal to B.

[0141] In the description of the embodiments of the present invention, the term "and / or" herein is merely a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B may represent: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " herein generally represents an "or" relationship between the associated objects before and after.

[0142] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. An undamaged detection device for the can end seaming of a can based on bimodal detection, characterized in that, Including: A conveying module (1) for continuously conveying cans to be detected, including an arrangement groove (101) and a multi-axis robotic arm (103). The arrangement groove (101) is used to place continuously arranged cans, and the multi-axis robotic arm (103) is configured to sequentially transfer the cans to a laser detection module (3) and an X-ray detection module (4); A dual-modal detection module, including: A laser detection module (3) configured to emit line laser to scan the sealing profile outside the can and obtain three-dimensional profile data through a laser camera (301); An X-ray detection module (4) configured to emit soft X-rays to penetrate the can and collect an internal sealing structure image through a flat panel detector (402); A control module, including a data fusion unit and a defect analysis unit, for synchronously controlling the laser detection module (3) and the X-ray detection module (4), and performing multi-dimensional fusion on the detection data of the two, and identifying sealing defects in combination with a machine learning algorithm.

2. The hermetic seam nondestructive testing device according to claim 1, characterized in that, Further including: A cleaning module (2), including a positioning component (201) and a cleaning component (202). The positioning component (201) is arranged below the cleaning component (202) for realizing fixed clamping of the can body, and the cleaning component (202) is configured to use a high-pressure air knife (2021) to remove impurities on the can lid surface.

3. The hermetic seam non-destructive testing device according to claim 1, characterized in that, Further including: A collection module (5), including a pushing component (501) and a collection component (502), configured to transfer the cans that have completed detection out of the arrangement groove (101).

4. The sealing non-destructive detection device according to claim 1, characterized in that: The line laser wavelength of the laser detection module (3) is 405 - 650 nm, the scanning frequency ≥ 2000 Hz, and the resolution ≤ 0.01 mm; The tube voltage of the X-ray detection module (4) is 30 - 80 kV, and the dose rate ≤ 1 μSv / h.

5. The sealing non-destructive detection device according to claim 1, characterized in that: The control module is further integrated with an automatic calibration unit and an alarm unit. The automatic calibration unit regularly performs calibration of the laser detection module (3) and the X-ray detection module (4), and the calibration deviation threshold is set to ±0.05% FS; the alarm unit emits an alarm signal when the laser detection module (3) and / or the X-ray detection module (4) detects a non-conforming can.

6. The sealing non-destructive detection device according to claim 1, characterized in that: The control module is further integrated with a wireless communication unit, supporting remote monitoring and data uploading to a cloud server; The defect analysis unit is built-in with a convolutional neural network model, and the training data set contains 100,000 groups of sealing defect samples.

7. The sealing non-destructive detection device according to claim 1, characterized in that: The X-ray detection module (4) includes a radiation source (401), a flat panel detector (402), and a second can body positioning mechanism (403). The second can body positioning mechanism (403) realizes adjustable installation in five directions through a five-axis drive mechanism (4032).

8. A method for fusion analysis of defects in the curling and sealing of aluminum cans based on bimodal data, which is based on the curling and sealing detection device described in any one of claims 1-7, characterized in that, Including the following steps: S01. Multi-source data synchronous acquisition: Obtain the three-dimensional contour point cloud data of the external seam of the aluminum can through the laser detection module (3); Obtain the tomographic grayscale image of the internal seam structure of the aluminum can through the X-ray detection module (4); Perform spatio-temporal alignment on the point cloud data and the grayscale image to generate a spatio-temporal correlation data set; S02. Multidimensional data fusion: Extract the contour curvature, height deviation, and surface flatness features from the point cloud data; Extract the sealant distribution density, hook continuity, and internal porosity features from the grayscale image; Adopt a feature-level fusion strategy, input the above features into a machine learning model, and generate a joint defect probability matrix; S03. Defect classification and decision-making: According to the joint defect probability matrix, divide the defect levels; and trigger a hierarchical response mechanism according to the defect levels.

9. The method according to claim 8, wherein: In step S01, the spatio-temporal alignment is achieved in the following manner: Time synchronization: Add a unified timestamp to the laser and X-ray data, with an error ≤ 1 ms; Spatial mapping: Establish a tank coordinate system, and associate the point cloud data with the image pixel coordinates through an affine transformation matrix.

10. A dynamic optimization method for detecting the curling and sealing of aluminum cans based on bimodal data, characterized in that, It includes the following steps: S10. Real-time data feedback: During the operation of the production line, continuously collect laser and X-ray detection data; Calculate the comprehensive index of the seam quality Q_index, and the formula is: Q_index = 0.6 × laser contour score + 0.4 × X-ray structure score S11. Dynamic parameter adjustment: If Q_index of 10 consecutive cans is lower than the threshold, automatically adjust the parameters of the detection module: Increase the laser scanning frequency by 10% - 20%; Increase the X-ray tube voltage by 5 kV - 10 kV; Adjust the data fusion weight to 0.8 for laser and 0.2 for X-ray; If Q_index is continuously higher than the threshold, reduce the detection frequency to save energy consumption; S12. Self-learning model update: Add the real-time detection data to the training set, and iteratively update the machine learning model every 24 hours; For new defect patterns, automatically generate adversarial samples and enhance the robustness of the model.

Images