Nondestructive testing method and device for bottle cap sealing structure based on X-ray imaging

Abstract

The invention discloses a nondestructive testing method and device for a bottle cap sealing structure based on X-ray imaging, and belongs to the technical field of industrial nondestructive testing and image recognition. The device comprises a rack, a driving unit, a jig disc, a detection unit, a PLC control unit, an X-ray imaging unit and an image processing unit. The method comprises the following steps: accurately positioning a bottle body to an imaging area through a driving unit; an X-ray imaging unit is used for obtaining an internal structure image of the bottle cap; inputting the image into an image processing unit integrated with a defect identification model based on a YOLOv8 algorithm for intelligent analysis, and identifying defects such as missing, dislocation and deformation of the sealing washer; and finally, whether the sealing performance is qualified or not is judged according to an identification result, and a sorting mechanism can be linked. The limitation that visible light cannot penetrate through a non-transparent bottle cap material is solved, rapid, accurate, full-automatic and non-destructive detection of a completely hidden bottle cap internal sealing structure is achieved, and the quality control efficiency and reliability of a production line are improved.

Description

Technical Field

[0001] This invention relates to image recognition technology for detecting the sealing status of bottle caps, and more particularly to a non-destructive testing method and apparatus for bottle cap sealing structures based on X-ray imaging. Background Technology

[0002] In the modern food, pharmaceutical, and beverage packaging industry, the sealing integrity of bottle caps is a key factor in ensuring product safety, preventing leakage and spoilage, and maintaining shelf life. Bottle caps typically contain complex and sophisticated components such as sealing gaskets, tamper-evident rings, inner plugs, or pressure valves. Once sealed, these structures are completely concealed within the cap shell or the bottle neck joint, forming a visually impenetrable enclosed space.

[0003] Currently, the industry's conventional methods for testing the sealing quality of bottle caps mainly rely on the following approaches:

[0004] Sampling inspection is performed manually by opening, cutting, or using simple tools. This method is inefficient, subjective, cannot achieve full inspection, and carries the risk of missed inspections, making it unsuitable for the quality control requirements of modern high-speed production lines.

[0005] Visible light-based optical vision inspection: This method uses industrial cameras to automatically inspect the appearance, printing, and dimensions of bottle caps. However, the physical properties of light mean it cannot penetrate opaque plastic or metal bottle cap materials. Therefore, this method cannot obtain any information about the internal sealing structure and is powerless to detect internal defects such as missing, misaligned, deformed, or unevenly distributed gaskets.

[0006] For example, the sealing performance can be indirectly inferred by detecting the tightening torque of the bottle cap or by online weighing (leak detection). However, these methods cannot pinpoint the specific type and location of defects and are easily affected by other production factors, resulting in limited accuracy and reliability.

[0007] Therefore, how to achieve rapid, accurate, and fully automated detection of the concealed sealing structure inside the bottle cap without damaging the product has always been a core technological bottleneck that needs to be overcome in the field of intelligent packaging quality control.

[0008] X-ray imaging technology has been widely used in industrial non-destructive testing due to its powerful material penetrating ability. Its principle lies in the fact that when X-rays penetrate an object, materials of different densities and thicknesses absorb the rays to varying degrees, thus forming a grayscale image on the detector that reflects information about the internal structure. Summary of the Invention

[0009] To address the shortcomings of the existing technology, this invention proposes a non-destructive testing method and apparatus for bottle cap sealing structures based on X-ray imaging.

[0010] The technical solution of this invention is implemented as follows:

[0011] A non-destructive testing device for bottle cap sealing structures based on X-ray imaging, characterized in that it comprises:

[0012] frame;

[0013] The drive unit is mounted on the frame;

[0014] A fixture tray, mounted on the drive unit, is used to support and position the bottle.

[0015] The detection unit is used to emit X-rays and acquire images of the inside of the bottle cap;

[0016] The PLC control unit is signal-connected to the detection unit and the drive unit, and is used to control the movement of the drive unit according to the position signal;

[0017] It also includes an image processing unit for defect identification of the acquired images.

[0018] Preferably, the drive unit includes a first lead screw drive device, a second lead screw drive device, and a third lead screw drive device arranged perpendicularly to each other in space, for driving the fixture disk to perform three-dimensional linear motion.

[0019] The first, second, and third lead screw transmission devices respectively include a motor, a ball screw, and a belt transmission device.

[0020] Preferably, the image processing unit integrates a bottle cap sealing defect recognition model trained based on the YOLOv8 algorithm.

[0021] Preferably, it also includes a sorting mechanism connected to the PLC control unit for automatically rejecting defective products based on the identification results.

[0022] A non-destructive testing method for bottle cap sealing structures based on X-ray imaging, characterized by comprising:

[0023] The drive unit positions the bottle, which is supported on the fixture plate, to the imaging area of ​​the X-ray imaging unit.

[0024] The X-ray imaging unit is triggered to work and acquire X-ray images of the inside of the bottle cap;

[0025] The acquired X-ray images are input into the image processing unit for defect identification;

[0026] The identification results determine whether the bottle cap sealing structure is up to standard.

[0027] Preferably, the step of the driving unit positioning the bottle to the imaging area specifically includes:

[0028] After the detection unit detects that the bottle is in place, it sends a trigger signal to the PLC control unit.

[0029] The PLC control unit controls the drive unit to drive the fixture disk and the bottle to perform three-dimensional linear motion until the bottle reaches the preset imaging position and locks.

[0030] Preferably, before inputting the image into the image processing unit, a step of preprocessing the X-ray image is further included, wherein the preprocessing includes at least one of denoising, enhancement, and normalization.

[0031] Preferably, the defect identification is specifically implemented through a model trained based on the YOLOv8 algorithm. The model uses the CSPDarknet backbone network to extract features, combines PAN-FPN for multi-scale feature fusion, and directly predicts the defect bounding box through an anchorless mechanism.

[0032] Preferably, the training steps of the model include:

[0033] Multiple X-ray images were acquired showing different states of the bottle cap sealing structure;

[0034] Use the annotation tool to mark the locations of defects in the image;

[0035] The labeled images are divided into training set, validation set and test set;

[0036] Model training and validation were performed using the YOLOv8 network architecture.

[0037] Preferably, the inference steps of the model include: inputting a preprocessed image, outputting recognition results containing defect type, location, and size, and filtering overlapping detection boxes using a non-maximum suppression algorithm.

[0038] The process includes, after determining whether the bottle cap sealing structure is qualified, binding and storing the identification result with the bottle ID, and controlling the sorting mechanism to reject unqualified products.

[0039] The non-destructive testing method and apparatus for bottle cap sealing structures based on X-ray imaging of the present invention have the following beneficial effects:

[0040] By utilizing the powerful penetrating ability of X-rays, the limitation that visible light cannot penetrate opaque bottle cap materials is overcome. Clear images of the internal sealing structure of the bottle cap can be obtained without damaging the product, making hidden defects such as missing, misaligned, or deformed gaskets directly visible.

[0041] By combining advanced target detection algorithms such as YOLOv8, intelligent analysis and recognition of X-ray images can accurately determine the type, location, and size of defects, overcoming the shortcomings of traditional indirect detection methods that cannot locate specific defects.

[0042] Through a PLC-controlled multi-axis motion mechanism and photoelectric sensing system, the bottle can be automatically positioned, transferred, imaged, and reset. It can be seamlessly integrated with the production line to achieve high speed and full inspection, greatly improving inspection efficiency and consistency. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of the non-destructive testing device for bottle cap sealing structure based on X-ray imaging according to the present invention.

[0044] Figure 2 This is a schematic diagram of the non-destructive testing device for bottle cap sealing structure based on X-ray imaging according to the present invention.

[0045] Figure 3 This is a flowchart of the non-destructive testing method for bottle cap sealing structure based on X-ray imaging according to the present invention.

[0046] Figure 4 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0047] Figure 5 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0048] Figure 6 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0049] Figure 7 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0050] Figure 8 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0051] Figure 9 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0052] Figure 10 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0053] Figure 11 This is a schematic diagram of the non-destructive testing method for bottle cap sealing structures based on X-ray imaging according to the present invention.

[0054] The reference numerals in the attached figures are as follows:

[0055] 10-Frame, 20-Drive unit, 21-First lead screw drive device, 22-Second lead screw drive device, 23-Third lead screw drive device, 30-Jig plate, 40-Detection unit, 50-PLC control unit. Detailed Implementation

[0056] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0057] Example 1

[0058] Reference Figures 1 to 2 As shown in the figure, the non-destructive testing device for bottle cap sealing structure based on X-ray imaging proposed in this embodiment includes:

[0059] The frame 10 serves as the structural support for the entire motion mechanism.

[0060] The drive unit 20 is installed inside the frame 10, and the drive unit 20 consists of a first lead screw drive device 21, a second lead screw drive device 22, and a third lead screw drive device 23, which are respectively installed perpendicularly to each other.

[0061] The jig disc 30 is mounted on the drive unit 20 and is used to support and position the bottle. The first lead screw drive device 21, the second lead screw drive device 22, and the third lead screw drive device 23 each include a motor, a ball screw, and a belt drive device. The motor provides power to drive the ball screw and belt to achieve precise linear motion of the jig disc 30.

[0062] The detection unit 40 uses a through-beam infrared photoelectric switch as a sensor, and its output is connected to the PLC control unit 50 to provide precise hard limit control for the motion mechanism, ensuring that each bottle can stop at the preset imaging position.

[0063] In this embodiment, the bottle to be tested is placed or transported onto the fixture tray 30. When the bottle reaches the initial position, the through-beam infrared photoelectric switch of the detection unit 40 is triggered, generating an arrival signal and sending it to the PLC control unit 50.

[0064] The PLC control unit 50 sends commands to the drive unit 20 according to a preset program. The drive unit 20 has three mutually perpendicularly mounted first lead screw drive devices 21, second lead screw drive device 22, and third lead screw drive device 23, which work collaboratively under the drive of their respective motors. Typically, the control fixture disk 30 performs sequential or synchronous linear movements in three dimensions: forward / backward (Y-axis), left / right (X-axis), and up / down (Z-axis), thereby smoothly and accurately moving the bottle from its initial position to the preset detection position, i.e., the optimal imaging field of view center between the X-ray source and the flat panel detector.

[0065] Throughout the entire movement, mechanical hard limits along each axis ensure precise movement and prevent overtravel. When the fixture disk 30 moves the bottle to the target position, the detection unit 40 reconfirms the position, and the PLC control unit 50 then locks the motion mechanism, keeping the bottle absolutely stationary during imaging.

[0066] A position-ready signal is sent from the PLC to the computer host, triggering the image acquisition device to start. The X-ray source emits X-rays, and the flat panel detector receives the rays that penetrate the bottle cap and generates an image, completing one non-destructive inspection. After imaging is completed, the PLC control drive unit 20 moves the fixture tray 30 and the bottle out of the imaging area, resets them to their initial position, or moves them to the next station to prepare for receiving the next bottle, thus forming a continuous and automatic work cycle.

[0067] Example 2

[0068] Based on the above embodiments, referring to Figures 3 to 11 As shown in the figure, this embodiment also proposes a non-destructive testing method for bottle cap sealing structures based on X-ray imaging, which specifically includes the following steps:

[0069] Step 1: Construct a bottle cap sealing defect identification model using the YOLOv8 target detection algorithm;

[0070] First, 10,000 X-ray images of different bottle openings were acquired using an image acquisition device, and the locations of sealing defects were marked using LabelMe. All images were divided into three parts in a 7:2:1 ratio: training set, validation set, and test set. Finally, a sealing defect detection model for bottle opening threads and bottle cap internal threads was trained.

[0071] The principles of YOLOv8 can be summarized into the following key steps and the core ideas behind them:

[0072] Backbone network and feature extraction: The input image first passes through a powerful CSPDarknet backbone network to progressively extract multi-scale feature maps from low to high levels.

[0073] Feature Pyramid Network: By using the Path Aggregation Network (PAN-FPN) structure, the semantic information of deep feature maps is fused with the fine positional information of shallow feature maps, thereby enhancing the model's ability to detect targets at different scales.

[0074] Anchor-free prediction: YOLOv8 abandons the pre-set anchor box design of earlier versions of the YOLO series and adopts an anchor-free mechanism instead. It directly predicts the offset of the bounding box relative to the center of the grid cell. This simplifies model training and reduces hyperparameter tuning related to anchor box size.

[0075] Decoupled Head: YOLOv8 uses a decoupled detection head to separate classification and regression tasks, allowing the network to focus more on learning the features of different tasks, thereby improving overall performance.

[0076] In this embodiment, the core detection output formula can be simplified as follows:

[0077] For each grid cell in the feature map, the model predicts N bounding boxes. The prediction output for each bounding box includes the following key components:

[0078] (t x , t y ): The offset of the center point of the bounding box relative to the top left corner of the grid cell.

[0079] (t v , t h P1: The logarithm of the width and height of the bounding box relative to a reference scale (such as the receptive field of this feature layer). P0: The objectness score of the bounding box containing the object. (P1, P2, ..., P...) k ): A probability distribution representing the conditional probability that the bounding box belongs to each of the K categories.

[0080] Finally, the complete prediction vector of a bounding box can be represented as:

[0081] B = [t] x , t y , t v , t h , P0, P1, P2, ..., P k ]

[0082] During inference, the model generates a large number of such candidate boxes. Then, through a non-maximum suppression (NMS) post-processing step, redundant boxes with high overlap and low confidence are filtered out, and finally clear and accurate detection results are output.

[0083] Step 2: After power-on, the drive unit 20, detection unit 40 and PLC control unit 50 complete self-test and initialization, load the trained YOLOv8 bottle cap sealing defect recognition model, and enter the detection state.

[0084] Step 3: The bottle to be tested is placed on the fixture plate 30 via a conveyor line or manually. The fixture plate 30 is used to stably support and position the bottle. After the through-beam infrared photoelectric switch detects that the bottle is in place, it sends a trigger signal to the PLC control unit 50.

[0085] Step 4: The PLC control unit 50 controls the drive unit 20 to drive the fixture disk to perform three-dimensional linear motion according to the preset program. The bottle is moved smoothly to the preset imaging position, which is the optimal field of view center between the X-ray source and the flat panel detector. After it is in place, the drive unit 20 locks to ensure that the bottle remains stationary during imaging.

[0086] Step 5: The PLC control unit 50 sends a position ready signal, the X-ray source emits X-rays, which penetrate the bottle cap and bottle mouth area to generate a digital image. After the image acquisition is completed, the X-ray source is turned off, and the PLC control unit 50 controls the drive unit 20 to move the bottle out of the imaging area and reset or transfer it to the next station.

[0087] Step 6: Preprocess the acquired X-ray images, typically involving denoising, enhancement, and normalization. Then, input the processed images into the pre-trained YOLOv8 model for inference and recognition, including:

[0088] The model extracts multi-scale features based on the CSPDarknet backbone network and the PAN-FPN structure.

[0089] An anchor-free mechanism is used to directly predict the location and category of defect bounding boxes.

[0090] Output the coordinates of each detection box. Confidence level and category probability ,

[0091] By filtering overlapping boxes using non-maximum suppression (NMS), a clear defect identification result is finally output, which includes defect type, location, size, etc.

[0092] Step 7: Determine whether the bottle cap sealing structure is qualified based on the identification results. The results can be linked to the bottle ID, recorded in the database, and linked to the subsequent sorting mechanism to remove unqualified products. The sorting machine can be equipped with push rods, pneumatic devices, etc.

[0093] In this embodiment, a YOLOv8-based object detection model is constructed and trained specifically for identifying defects in bottle cap sealing structures. The specific steps include: acquiring no fewer than 10,000 X-ray images of different bottle openings; labeling sealing defects in the images using an annotation tool; dividing the images into training, validation, and test sets; and training the model using the YOLOv8 network structure. YOLOv8 uses CSPDarknet as the backbone network to extract features, combines it with PAN-FPN for multi-scale feature fusion, and directly predicts the bounding box position and category through an anchor-free mechanism. The final output includes detection results containing coordinates, confidence scores, and category probabilities. After non-maximum suppression processing, clear defect identification information is obtained.

[0094] After power-on, the drive unit 20, the detection unit 40 and the PLC control unit 50 complete self-test and initialization in sequence, load the trained YOLOv8 model and enter the detection state.

[0095] After the detection process begins, the bottle to be tested is placed on the fixture tray 30 via a conveyor line or manually. Once the through-beam infrared photoelectric switch in the detection unit 40 detects the bottle's positioning, it sends a trigger signal to the PLC control unit 50. The PLC control unit 50, according to a preset program, controls the drive unit 20, which, through the first lead screw drive device 21, the second lead screw drive device 22, and the third lead screw drive device 23, drives the fixture tray 30 and the bottle in a three-dimensional linear motion, precisely moving them to the center of the imaging field of view between the X-ray source and the flat panel detector. Once in position, the drive unit 20 locks, ensuring the bottle remains stationary during imaging.

[0096] The PLC control unit 50 sends a position ready signal to the image acquisition system, triggering the X-ray source to emit X-rays. After the X-rays penetrate the bottle cap and bottle mouth area, they are received by the flat panel detector and a digital image is generated. After imaging is completed, the X-ray source is turned off, and the PLC control unit 50 controls the drive unit 20 to move the bottle out of the imaging area, resetting it or transferring it to the next station.

[0097] The acquired X-ray images are first preprocessed, and then input into a pre-trained YOLOv8 model for inference and recognition. The model outputs information on the type, location, and size of defects, which are then filtered using non-maximum suppression to obtain the final recognition result.

[0098] The system determines whether the bottle cap sealing structure is up to standard based on the identification results and records the results in the database, binding them to the bottle ID. Simultaneously, it can be linked to the subsequent sorting mechanism to automatically remove defective products, achieving full automation from detection to sorting.

[0099] 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 and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A non-destructive testing device for bottle cap sealing structures based on X-ray imaging, characterized in that, include: Rack (10); A drive unit (20) is mounted on the frame (10); A fixture plate (30) is provided on the drive unit (20) for supporting and positioning the bottle body; The detection unit (40) is used to emit X-rays and acquire images of the inside of the bottle cap; The PLC control unit (50) is signal-connected to the detection unit (40) and the drive unit (20) and is used to control the movement of the drive unit (20) according to the position signal; It also includes an image processing unit for defect identification of the acquired images.

2. The non-destructive testing device for bottle cap sealing structure according to claim 1, characterized in that, The drive unit (20) includes a first lead screw drive device (21), a second lead screw drive device (22), and a third lead screw drive device (23) arranged perpendicularly to each other in space, for driving the jig disk (30) to perform three-dimensional linear motion. The first lead screw drive device (21), the second lead screw drive device (22) and the third lead screw drive device (23) respectively include a motor, a ball screw and a belt drive device.

3. The non-destructive testing device for bottle cap sealing structure according to claim 1, characterized in that, The image processing unit integrates a bottle cap sealing defect recognition model trained based on the YOLOv8 algorithm.

4. The non-destructive testing device for bottle cap sealing structure according to claim 1, characterized in that, It also includes a sorting mechanism that is connected to the PLC control unit (50) by signal, for automatically rejecting defective products based on the identification results.

5. A non-destructive testing method for bottle cap sealing structures based on X-ray imaging, comprising the non-destructive testing device for bottle cap sealing structures as described in claim 1, characterized in that, include: The drive unit (20) positions the bottle, which is supported on the fixture plate (30), to the imaging area of ​​the X-ray imaging unit. The X-ray imaging unit is triggered to work and acquire X-ray images of the inside of the bottle cap; The acquired X-ray images are input into the image processing unit for defect identification; The identification results determine whether the bottle cap sealing structure is up to standard.

6. The non-destructive testing method for bottle cap sealing structure according to claim 5, characterized in that, The specific steps of the driving unit (20) positioning the bottle to the imaging area include: After the detection unit (40) detects that the bottle is in place, it sends a trigger signal to the PLC control unit (50); The PLC control unit (50) controls the drive unit (20) to drive the fixture disk (30) and the bottle to perform three-dimensional linear motion until the bottle reaches the preset imaging position and locks.

7. The non-destructive testing method for bottle cap sealing structure according to claim 5, characterized in that, Before inputting the image into the image processing unit, the process includes a preprocessing step for the X-ray image, which includes at least one of denoising, enhancement, and normalization.

8. The non-destructive testing method for bottle cap sealing structure according to claim 5, characterized in that, The defect identification is specifically achieved through a model trained based on the YOLOv8 algorithm. The model uses the CSPDarknet backbone network to extract features, combines PAN-FPN for multi-scale feature fusion, and directly predicts the defect bounding box through an anchorless mechanism.

9. The non-destructive testing method for bottle cap sealing structure according to claim 8, characterized in that, The training steps of the model include: Multiple X-ray images were acquired showing different states of the bottle cap sealing structure; Use the annotation tool to mark the locations of defects in the image; The labeled images are divided into training set, validation set and test set; Model training and validation were performed using the YOLOv8 network architecture.

10. The non-destructive testing method for bottle cap sealing structure according to claim 9, characterized in that, The inference steps of the model include: inputting a preprocessed image, outputting recognition results containing defect type, location, and size, and filtering overlapping detection boxes using a non-maximum suppression algorithm. The process includes, after determining whether the bottle cap sealing structure is qualified, binding and storing the identification result with the bottle ID, and controlling the sorting mechanism to reject unqualified products.

Images