Gas leakage detection method and device, electronic equipment and storage medium

Abstract

The application provides a gas leakage detection method and device, electronic equipment and storage medium, and belongs to the technical field of intelligent manufacturing. The method comprises the following steps: acquiring an appearance image of a product to be detected; identifying whether a physical inkjet mark exists on the surface of the product according to the appearance image; searching and matching the physical inkjet mark with historical inkjet marks in a preset detection database to determine a target NG product record; and sending a leakage position detection image in the target NG product record to a client. The application takes the physical inkjet mark applied in a previous process as a visual feature index, constructs a mapping association between a repair physical object and historical detection data, thereby realizing accurate traceability and identity verification of NG products in a product sequence code-free scene, effectively avoiding material mixing and significantly improving repair positioning efficiency.

Description

Technical Field

[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a gas leak detection method, device, electronic equipment, and storage medium. Background Technology

[0002] In large-scale manufacturing industries such as home appliances and auto parts, airtightness testing of pressure-bearing components (such as containers and pipes) is a necessary process to ensure product quality.

[0003] To improve detection efficiency and accuracy, existing production lines widely employ advanced non-contact detection technologies, such as acoustic localization solutions based on microphone arrays. This solution, after a product is filled with a specific gas, captures and analyzes the acoustic signals generated by the leak, and then accurately generates the spatial location information of the leak point digitally on the control system. The application of this technology automates the leak location process, significantly improving both efficiency and accuracy compared to traditional manual immersion testing.

[0004] However, in current production practices, the high efficiency advantages brought by the aforementioned automated inspection process are often difficult to fully transfer to subsequent rework processes. Although the upstream inspection system has generated precise leak location data for non-conforming (NG) products, once these NG products are transferred to the rework station, workers typically still need to spend a significant amount of time performing repetitive leak detection checks to re-identify the leaks that should have been known. This lack of seamless integration between automated inspection and manual rework processes creates an efficiency bottleneck in the rework process, significantly extending the overall product manufacturing cycle and greatly diminishing the economic benefits of introducing automated inspection equipment in the early stages. Summary of the Invention

[0005] This invention provides a gas leak detection method, device, electronic equipment, and storage medium to solve the problem in the prior art that there is no effective means of correlation between digital detection results and physical products in the context of codeless traceability. This leads to the loss of accurate defect location information of non-conforming products when they are transferred to the rework station, resulting in low rework efficiency. The invention realizes the direct mapping and visualization of defect location information from the digital domain to the physical domain, enabling staff to intuitively and quickly obtain defect location guidance, thereby significantly improving rework efficiency and the overall synergy of the production line process.

[0006] This invention provides a gas leak detection method, comprising: Obtain an image of the product to be inspected located at the rework station; Based on the appearance image, identify whether there are physical inkjet marks on the surface of the product to be tested that were triggered by the previous airtightness testing process when a leak was determined; If the physical inkjet mark is detected, the physical inkjet mark is searched and matched with historical inkjet marks in the preset detection database to determine the target NG product record that uniquely corresponds to the product to be tested; the detection database stores leakage location detection images of historical NG products, and each leakage location detection image is associated with the historical inkjet mark that was triggered when a leakage was determined; The leak location detection image from the target NG product record is sent to the client.

[0007] According to a gas leak detection method provided by the present invention, the step of using the physical inkjet marker to search and match with historical inkjet markers in a preset detection database to determine a target NG (non-compliant) product record uniquely corresponding to the product to be detected includes: Obtain the current system time when the appearance image was acquired, and set a time retrieval window for backward lookup based on the current system time; Traverse the detection database, filter out multiple historical NG product records whose material feeding timestamps fall within the time retrieval window, and construct a candidate record set; Extract the retrieval feature data from the appearance image, compare it with the historical feature data associated with each historical NG product record in the candidate record set, and determine the target NG product record by combining the time factor.

[0008] According to a gas leak detection method provided by the present invention, the step of extracting retrieval feature data from the appearance image includes: The appearance image is segmented to separate the target area containing the physical inkjet mark and the background area after removing the target area; Extract the retrieval feature data, which includes at least one of the following: the outline shape data, color distribution data, and relative position coordinate data of the physical inkjet mark in the target area, or the weld morphology data, structural texture data, and product model feature data reflecting the characteristics of the product body in the background area.

[0009] According to a gas leak detection method provided by the present invention, the step of comparing the historical feature data associated with each historical NG (Not From Good) record in the candidate record set with the target NG record, and determining the target NG record by combining the time factor, includes: The extracted retrieval feature data is compared with the historical feature data of each historical NG product record in the candidate record set to calculate the difference, and the original similarity value of each historical NG product record is obtained. Calculate the time difference between the material unloading timestamp of each historical NG product record in the candidate record set and the current system time; The time difference is converted using a preset time decay logic to obtain the time weight coefficient for each historical NG product record. The smaller the time difference, the larger the corresponding time weight coefficient. The original similarity value is weighted by the time weight coefficient to generate the final matching score for each historical NG product record; The historical NG record with the highest final matching score is determined as the target NG record.

[0010] According to a gas leak detection method provided by the present invention, before acquiring the appearance image of the product to be inspected located at the rework station, the method further includes: In the airtightness testing stage, each product undergoes an airtightness test; When a gas leak is detected in any product, the coordinates of the leak location are obtained and an inkjet command is generated. In response to the inkjet command, a pneumatic nozzle is driven to apply the historical inkjet mark at a preset position on any of the product surfaces; During the product unloading process, a photograph is taken of any product to obtain a product image containing the historical inkjet markings, and the product image is stored in the detection database as a leak location detection image in the historical NG product record.

[0011] According to a gas leak detection method provided by the present invention, the step of driving a pneumatic nozzle to apply the historical inkjet mark at a preset position on any product surface in response to the inkjet command includes: Based on the obtained coordinates of the leak location, the regional attributes of the leak source on the product surface are analyzed; The motion trajectory parameters corresponding to the area attributes are called as the preset position, and the pneumatic nozzle is controlled to move and spray the historical inkjet mark.

[0012] According to a gas leak detection method provided by the present invention, the step of driving a pneumatic nozzle to apply the historical inkjet mark at a preset position on any product surface in response to the inkjet command further includes: Retrieve pre-stored fixed coordinate parameters as the preset position; The pneumatic nozzle is controlled to move to the product part corresponding to the fixed coordinate parameters to perform fixed-point inkjet printing, so as to generate the historical inkjet mark with a fixed position.

[0013] According to a gas leak detection method provided by the present invention, before performing airtightness testing on each product in the airtightness testing stage, the method further includes: The loading images of the product are captured using a camera device located at the loading station; Based on the relationship between the structural dimensions of the loading station and the pixel ratio of the loading image, the key structural dimensions of the product are determined. Read the preset production model parameters in the current programmable logic controller, and compare the measured key structural dimensions with the standard dimensions corresponding to the production model parameters; When the comparison result exceeds the preset matching threshold, an alarm is triggered and the product is blocked from entering the airtightness testing stage.

[0014] According to a gas leak detection method provided by the present invention, the step of traversing the detection database, filtering out multiple historical NG (non-compliant) records whose discharge timestamps fall within the time retrieval window, and constructing a candidate record set includes: Check the repair status attributes of the historical NG product records and filter only the records whose repair status attribute is "not returned for repair". And / or, Identify product model features in the appearance image and filter only records that belong to the same product model category as the product to be detected; The records filtered by the maintenance status attribute and / or the product model category are used as the candidate record set.

[0015] According to a gas leak detection method provided by the present invention, after sending the leak location detection image from the target NG product record to the client, the method further includes: Receive the welding completion signal sent by the client; In response to the welding completion signal, a reflow notification is generated. The reflow notification prompts or controls the transmission of the welded product to be tested to the preceding airtightness testing stage, so as to control the re-performance of airtightness testing on the product to be tested.

[0016] The present invention also provides a gas leak detection device, comprising the following components: Image acquisition unit: The user acquires an image of the appearance of the product to be inspected located at the rework station; An image processing unit is used to identify, based on the appearance image, whether there is a physical inkjet mark on the surface of the product to be inspected that was triggered by the preceding airtightness inspection process when a leak was determined; An image recognition unit is used to identify the physical inkjet mark, and then use the physical inkjet mark to search and match it with historical inkjet marks in a preset detection database to determine the target NG product record that uniquely corresponds to the product to be inspected; the detection database stores leakage location detection images of historical NG products, and each leakage location detection image is associated with the historical inkjet mark that was triggered when a leakage was determined; A leak location unit is used to send the leak location detection image from the target NG product record to the client.

[0017] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the gas leak detection method as described above.

[0018] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the gas leak detection method as described above.

[0019] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the gas leak detection method as described above.

[0020] The gas leak detection method, device, electronic equipment, and storage medium provided by this invention construct a mapping relationship between the reworked physical product and historical detection data by using the physical inkjet markings applied in the preceding process as visual feature indexes. This enables accurate traceability and identity verification of NG products in scenarios without product serial numbers, effectively avoiding material mixing and significantly improving rework location efficiency. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0022] Figure 1 This is one of the flowcharts of the gas leak detection method provided by the present invention.

[0023] Figure 2 This is a schematic diagram of the system structure for acquiring appearance images provided by the present invention.

[0024] Figure 3 This is a schematic diagram of the NG automatic inkjet system structure provided by the present invention.

[0025] Figure 4 This is one of the flowcharts for determining the target NG item record provided by the present invention.

[0026] Figure 5 This is the second schematic diagram of the process for determining the target NG item record provided by the present invention.

[0027] Figure 6This is a schematic diagram of the construction process of the detection database provided by the present invention.

[0028] Figure 7 This is one of the schematic diagrams of the inkjet marking process provided by the present invention.

[0029] Figure 8 This is the second schematic diagram of the historical inkjet marking process provided by the present invention.

[0030] Figure 9 This is a schematic diagram of product model matching test before airtightness testing provided by the present invention.

[0031] Figure 10 This is the second schematic diagram of the gas leak detection method provided by the present invention.

[0032] Figure 11 This is a schematic diagram of the gas leak detection device provided by the present invention.

[0033] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0034] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0035] It should be noted that, in the description of this invention, the terms "comprising," "including," or any other variations thereof are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element. Those skilled in the art will understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0036] In addition, "and / or" indicates that at least one of the connected objects is connected, and the character " / " generally indicates that the objects before and after are in an "or" relationship.

[0037] In the manufacturing process of home appliances, containers that withstand pressure, such as water heater inner tanks and compressors, have very high airtightness requirements and need to undergo airtightness testing before leaving the production line. For scenarios with high airtightness requirements, a more accurate testing solution is helium or hydrogen-nitrogen leak detection. For situations where airtightness requirements are relatively less stringent, microphone array detection technology is a promising alternative to traditional water immersion bubble detection. The principle of microphone array airtightness detection is to locate the leak source by using ultrasonic signals generated by the gas leak.

[0038] Current microphone array airtightness testing technology has the following problems: 1) Automated airtightness testing was achieved by using a microphone array instead of manual water testing. However, during the isolation and repair process of NG products, there was a situation where NG products and OK products were confused due to the lack of special markings.

[0039] 2) Since not all production lines in the home appliance manufacturing process have implemented code tracking, the method of checking the leak location and repairing the weld based on the SN code at the rework station cannot be implemented in the actual production line. This results in a long time to locate the leak at the rework station and cannot make good use of the airtightness detection to locate the leak location information.

[0040] 3) The testing equipment can be compatible with the testing of multiple product models by the size and position of the product structure's hanging ears. However, in actual operation, it was found that the size and position of the hanging ears alone cannot completely distinguish the product models. This can lead to malfunctions such as impacts on the sealing and inflation mechanisms after the product arrives at the testing station, which may damage the testing equipment.

[0041] In view of this, it is urgent to improve the existing gas leak detection equipment and detection methods in order to solve or alleviate the above-mentioned defects to some extent.

[0042] The following is combined with Figures 1-12 The present invention describes a gas leak detection method, apparatus, electronic device, and storage medium. By adding a pneumatic nozzle to existing automated leak detection equipment, the nozzle is driven to perform inkjet processing at a preset position on the product after an NG product is detected, thus preventing NG and OK products from being confused.

[0043] Especially for scenarios where products on the production line lack serial number traceability such as QR codes, barcodes, or engraving codes, cameras are installed at the material handling station to capture images and store them in the inspection database. Similarly, cameras are installed above the rework station to photograph defective products requiring rework. These images are then compared against historical leak location images in the inspection database, focusing on extracting key features such as product model and the ink color, shape, and location characteristics of each leak. Searching for matching leak location images clearly displays the specific leak point, facilitating rapid leak location and rework even without traceability codes.

[0044] Optionally, the present invention adds a camera device at the feeding position, and uses the known information of the size of the feeding position mechanism and the pixel relationship to measure the key features such as the size and shape of the current test product, and then compares it with the current model set in the current PLC (the key features such as the size and shape of the product of this model are known). If there is a mismatch, an alarm is triggered to prevent malfunctions such as machine collisions caused by the inconsistency between the detected product model and the equipment settings.

[0045] The gas leak detection method, device, electronic equipment, and storage medium provided by this invention can be widely applied to various industrial product production lines that require airtightness testing, such as water heater inner tanks, air conditioner compressors, and refrigerator piping in the home appliance industry, or fuel tanks, engine components, and brake lines in the automotive industry. For ease of understanding, the following explanation will use a rework scenario from an automated water heater inner tank production line as an example.

[0046] It should be noted that the gas leak detection method provided by this invention can be executed by an integrated industrial control system, such as a programmable logic controller (PLC) system, a personal computer (PC)-based control system, a distributed control system (DCS), or an embedded controller. In the following embodiments, an industrial control computer deployed next to the production line will be used as the execution subject, but the invention is not limited thereto.

[0047] Figure 1 This is one of the flowcharts of the gas leak detection method provided by the present invention, such as... Figure 1 As shown, including but not limited to the following steps: Step 11: Obtain an image of the product to be inspected located at the rework station.

[0048] In this embodiment, the rework station refers to a designated physical area or workstation on the production line specifically used for processing and repairing non-conforming (NG) products. When a product is determined to be NG after passing through the upstream quality inspection process, it will be transferred to this station by the production line's conveying system (such as a conveyor belt or AGV) to await processing by the staff.

[0049] Once a water heater inner tank is transported to the repair station and comes to a stop, it becomes a product to be inspected from the system's perspective. "Product to be inspected" does not refer to a complete leak test, but rather the system needs to identify the product through visual inspection to determine if it is a defective (NG) product requiring repair and to ascertain its specific repair information.

[0050] Figure 2 This is a schematic diagram of the system structure for acquiring appearance images provided by the present invention. To achieve this objective, as shown below... Figure 2 As shown, in this embodiment, a camera device 1, such as one or more fixed industrial cameras (which can be area scan cameras or line scan cameras), will be deployed at the rework station. Its field of view covers the surface area of ​​the product to be inspected where marks may appear.

[0051] Once the sensors (such as photoelectric sensors) at the inspection station detect the product to be inspected in place, they trigger an industrial camera to photograph its surface, thereby acquiring one or more high-resolution appearance images. These appearance images are digital image files containing all visual information about the surface of the product to be inspected; for example, they can be in JPEG, BMP, or RAW format. Besides fixed-camera shooting, acquisition can also be achieved by a multi-axis robot equipped with a camera, which plans its path based on the product's 3D model and scans the key surfaces of the product to be inspected to ensure that all possible marking information is captured.

[0052] Step 12: Based on the appearance image, identify whether there is a physical inkjet mark on the surface of the product to be tested that was triggered by the previous airtightness test process when a leak was determined.

[0053] After acquiring an image of the product to be inspected, the industrial control computer will use image processing algorithms to analyze the image. The core purpose of this step is to first confirm whether a physical inkjet mark, serving as an information carrier, exists on the product surface.

[0054] The pre-production airtightness inspection process refers to the automated inspection step that a product undergoes before it is transferred to the rework station to determine whether it has a gas leak. When the inspection system used to perform the pre-production airtightness inspection process determines that a product is NG (not good), in addition to storing the result in the inspection database, it will also activate a physical marking application device to apply a physical inkjet mark to the surface of the product.

[0055] Figure 3 This is a schematic diagram of the NG automatic inkjet system structure provided by the present invention, as shown below. Figure 3 As shown, the physical marking application device 2 can be selected from an industrial inkjet printhead, a laser marking device, or an activation nozzle, and is set at the end of the detection system that performs the preceding airtightness testing process.

[0056] In one specific embodiment of the present invention, the physical inkjet mark 3 can be formed by spraying special ink onto an industrial inkjet printhead. This ink should possess several key industrial application characteristics: first, it should be able to cure or dry quickly to avoid tailing or contamination when the product flows at high speed on a conveyor belt; second, it needs to have good adhesion to the product surface (e.g., metal or coated surfaces) to ensure that the mark does not easily fall off or become blurred during transportation; finally, the physical inkjet mark 3 should be temporary, meaning that after product repair, it can be easily and non-destructively removed by simple physical or chemical methods (e.g., wiping with a specific solvent, cleaning with alcohol, or heating), thus not affecting the final appearance quality of the product. Although this embodiment uses ink spraying as an example, those skilled in the art will understand that any other technical means capable of forming temporary physical marks on the product surface that satisfy the above-mentioned characteristics of fast drying, adhesion, and erasure fall within the scope of the present invention, such as using a laser to perform low-temperature, shallow etching on a surface coating to form a color-changing mark, or using thermal materials for printing, etc.

[0057] Crucially, the core value of the physical inkjet markings used in this embodiment lies not merely in the macroscopic shape or character content they represent. More importantly, this embodiment leverages the inherent uniqueness of physical inkjet markings at the physical level. Even when two identical target patterns (e.g., two dots) are sprayed, due to the microscopic randomness inherent in the spraying process, such as the scattering, wetting, and solidification of ink droplets, the two physical inkjet markings ultimately formed on the product surface will have unique, subtle differences at the pixel level, just as no two leaves in the world are exactly alike. This microscopic difference constitutes the physical fingerprint of each physical inkjet marking.

[0058] Therefore, the identification process in this embodiment mainly includes: first, preprocessing the appearance image (such as filtering and binarization) to highlight possible marked areas; then, using image recognition algorithms such as speckle analysis or connected component marking, detecting whether one or more areas in the appearance image conform to the basic physical characteristics (such as size, color, shape, and outline) of a physical inkjet mark. Once such an area is detected, it can be determined that a physical inkjet mark exists on the surface of the product to be inspected, and it can be segmented from the background to obtain an image slice of the physical inkjet mark.

[0059] Of course, as another feasible implementation, the physical inkjet marking 3 can also be designed as a macroscopic pattern with clear distinguishability. For example, if there are a total of K possible leak points in the product, K completely different patterns (such as circles, squares, triangles, etc.) or K different characters can be pre-designed. In this case, in addition to locating the marking, the identification process will also include using template matching or optical character recognition (OCR) technology to determine which specific predefined pattern or character is being applied.

[0060] Step 13: If the physical inkjet mark is identified, the physical inkjet mark is used to search and match with historical inkjet marks in the preset detection database to determine the target NG product record that uniquely corresponds to the product to be detected.

[0061] The detection database stores leak location detection images of historical defective products, and each leak location detection image is associated with a historical inkjet marker that was triggered when a leak was determined.

[0062] In an optional embodiment, corresponding to the above scheme utilizing the physical fingerprint of each physical inkjet mark, the retrieval and matching process is a matching based on the image content itself. In this case, the historical inkjet marks stored in the preset detection database are not simple strings, but rather the original image of the mark itself, captured and stored immediately after the previous airtightness detection process is applied, or feature vectors extracted from the original image (e.g., extracted using algorithms such as SIFT, SURF, ORB, or deep learning models).

[0063] Therefore, the specific method for retrieval and matching can be as follows: The industrial control computer extracts the feature vector of the physical inkjet mark on the product to be inspected obtained in step 12, and then compares this feature vector with the feature vectors of all historical marks stored in the inspection database one by one, calculating the similarity between them, for example by calculating Euclidean distance or cosine similarity. Since the physical fingerprint of each physical inkjet mark is unique, the historical NG product record with the highest similarity can be identified as the target NG product record uniquely corresponding to the current product to be inspected.

[0064] In another implementation using predefined patterns, the retrieval and matching process is relatively simplified. The historical inkjet marks stored in the detection database can be identifiers of predefined patterns (such as pattern A, pattern B). In this case, the retrieval and matching involves using the identifiers identified in step 12 as query keywords to search the detection database, thereby determining the corresponding target NG (Not From Good) record.

[0065] Through this step, regardless of the method used, the industrial control computer is able to successfully associate the product to be inspected with specific historical non-compliant products stored in the inspection database.

[0066] Step 14: Send the leak location detection image from the target NG product record to the client.

[0067] Once the target non-compliant product record is identified, the industrial control computer can extract the most critical rework guidance information, namely the leak location detection image. The leak location detection image is generated by the preceding airtightness testing system. It can be a two-dimensional unfolded view of a surface of the product to be inspected, clearly marking the precise location of the leak point using highlighted blocks, crosshairs, or arrows; or it can be a three-dimensional model view of the product, directly highlighting the leak area on the 3D model surface.

[0068] Industrial control computers can transmit images of identified leak locations to one or more client devices installed at the repair station via a local area network (LAN), such as wired Ethernet or wireless Wi-Fi. The client device can be a touchscreen display, a tablet, an industrial human-machine interface (HMI), or a display terminal connected to a large screen. Once the operator sees the leak location image on the client's screen, they can immediately identify the specific leak point in the water heater's inner tank and proceed with repair work such as welding, eliminating the need for any further manual leak detection.

[0069] The gas leak detection method provided in this invention constructs a mapping relationship between the reworked product and historical detection data by using the physical inkjet markings applied in the preceding process as visual feature indexes. This enables accurate traceability and identity verification of NG products in scenarios without product serial numbers, effectively avoiding material mixing and significantly improving rework location efficiency.

[0070] As an optional embodiment, in step 13 of the above embodiment, which involves using the physical inkjet marker to search and match with historical inkjet markers in a preset detection database to determine the target NG record that uniquely corresponds to the product to be detected, the present invention also provides a more specific and robust implementation method.

[0071] In some high-speed production or complex operating conditions, direct optical character recognition or template matching of physical inkjet marks may be affected by factors such as slight staining of the marks, changes in lighting, or inconsistent product placement.

[0072] Figure 4 This is one of the flowcharts for determining target NG (Not Good) item records provided by the present invention, such as... Figure 4 As shown, in order to improve the efficiency and accuracy of the retrieval and matching process, this embodiment optimizes step 13, which can be broken down into the following sub-steps: Step 131: Obtain the current system time when the appearance image was acquired, and set a forward backtracking time retrieval window based on the current system time.

[0073] The current system time refers to the millisecond-accurate timestamp that the industrial control computer obtains from its own operating system (such as Windows or Linux) at the moment when it executes step 11, which is to trigger the industrial camera to capture and successfully receive the appearance image of the product to be inspected.

[0074] The time retrieval window is a backward-looking time interval based on the current system time. For example, if a production line processes one product every 30 seconds, and the physical transport distance between the rework station and the preceding airtightness testing process takes a maximum of 5 minutes, then this time retrieval window can be reasonably set to 5 minutes or slightly longer, such as 6 minutes. If the current system time is "12 / 15 / 2025, 11:02:06 AM", a 6-minute time retrieval window would be the time interval [10:56:06 AM, 11:02:06 AM].

[0075] The size of this search window can be configured as a parameter, preset by staff based on the actual production line layout and cycle time. The purpose of setting this time-based search window is to significantly narrow down the scope of subsequent searches, avoiding a global search across the entire large inspection database.

[0076] Step 132: Traverse the detection database, filter out multiple historical NG product records whose material unloading timestamps fall within the time retrieval window, and construct a candidate record set.

[0077] After determining the time retrieval window, the industrial control computer will perform a preliminary screening operation on the preset detection database. As mentioned earlier, each historical NG (non-compliant) product record stored in the preset detection database contains, in addition to historical inkjet markings and leak location detection images, a crucial unloading timestamp. This unloading timestamp records the system time when the historical NG product completed the preceding airtightness testing process and was diverted and sent out.

[0078] The industrial control computer will traverse all historical NG (non-compliant) records in the pre-set inspection database, checking one by one whether the unloading timestamp of each NG record falls within the time retrieval window set in the previous step. All historical NG records that meet this time condition will be filtered out and collectively form a temporary candidate record set. Through this time-dimensional pre-screening, the number of NG records requiring complex feature comparison can be reduced from potentially tens of thousands to just a few or dozens that conform to the production line's cycle time logic, thereby significantly improving the inspection response speed.

[0079] Step 133: Extract the retrieval feature data from the appearance image, compare it with the historical feature data associated with each historical NG product record in the candidate record set, and determine the target NG product record by combining the time factor.

[0080] Retrieval feature data refers to a set of quantitative data extracted by an industrial control computer from a currently acquired image of the product to be inspected, focusing on the area containing the physical inkjet markings. This retrieval feature data can be a multi-dimensional data representation, and may include, but is not limited to: (1) Geometric features of physical inkjet marks calculated by image algorithms, such as centroid coordinates, aspect ratio, area, perimeter, etc. of the circumscribed rectangle in the appearance image.

[0081] (2) Apply the high-dimensional feature vectors output by deep learning feature extraction networks (such as ResNet, VGG, etc.) to the image slices marked by physical inkjet printing.

[0082] (3) Color characteristics of physical inkjet markings, such as histograms of their HSV color space.

[0083] Accordingly, in the pre-set inspection database, each historical NG record, in addition to storing the aforementioned fields, also pre-stores its associated historical feature data. This historical feature data is extracted and stored using the exact same algorithm after the physical inkjet mark is printed in the preceding airtightness inspection process, and it constitutes the baseline fingerprint of that physical inkjet mark.

[0084] The feature comparison process involves comparing the retrieved feature data extracted from the appearance image with the historical feature data associated with each historical NG (Not Qualified) record in the candidate record set. The comparison can be performed by calculating the similarity between the feature vectors of the two records, for example, using cosine similarity or Euclidean distance. The historical NG record with the highest similarity (or closest distance) is considered the most likely target NG record.

[0085] To further improve the reliability of the determination process, industrial control computers also incorporate a time factor. Specifically, the similarity score obtained from feature comparison can be weighted and fused with the degree of temporal proximity. For example, for a historical NG (Not Found) product record in the candidate record set, the smaller the difference between its material feeding timestamp and the current system time, the greater the likelihood that it is the same product to be inspected.

[0086] This invention significantly improves the efficiency and robustness of the retrieval and matching process by introducing time window pre-screening and multi-dimensional feature data comparison. The introduction of a time window avoids unnecessary global database queries, ensuring rapid response within production line cycles. Furthermore, using feature data instead of single character recognition allows the matching process to resist a certain degree of marker damage, deformation, or light interference, significantly improving matching accuracy and environmental adaptability, thereby ensuring the reliable operation of the entire gas leak detection method in industrial settings.

[0087] Based on the above embodiments, as an optional embodiment, the extraction of retrieval feature data from the appearance image specifically includes, but is not limited to: The appearance image is segmented to separate the target area containing the physical inkjet mark and the background area after removing the target area.

[0088] Image segmentation is a process of dividing an appearance image into semantically distinct components using image segmentation algorithms. Its purpose is to separate the target region containing key identification information from the remaining background region in the appearance image at the data level.

[0089] The target region refers to the image sub-region containing the physical inkjet marker, identified and separated from the appearance image using an image segmentation algorithm. This target region can be a minimal rectangular region tightly surrounding the physical inkjet marker, or an irregularly shaped region that perfectly matches the outline of the physical inkjet marker. The background region refers to the entire image portion remaining after removing pixels from the target region from the complete pixel set of the appearance image.

[0090] Optionally, the specific technical means for achieving image segmentation processing may include, but are not limited to, one of the following methods: Method 1, a threshold segmentation method based on color space, utilizes the difference between the color of the physical inkjet marker (e.g., black ink) and the surface color of the product (e.g., metallic primary color) to set a threshold to extract the target area.

[0091] Method 2, an adaptive threshold segmentation method based on image grayscale contrast, is suitable for scenarios where there is a significant brightness difference between the marker and the background.

[0092] Method 3 involves deploying a pre-trained deep learning semantic segmentation model, such as a neural network with a U-Net architecture. This semantic segmentation model can directly perform pixel-level classification on the appearance image, thereby accurately outputting the mask of the target region.

[0093] After completing the image segmentation process, the industrial control computer will perform the next step: extracting the retrieval feature data.

[0094] The composition of the retrieval feature data is flexible, and it may include at least one of the following types of data: First, data extracted from the target region to describe the inherent properties of the physical inkjet marker itself. This data directly reflects the inherent characteristics of the physical inkjet marker and may specifically include: (1) The contour shape data of the physical inkjet mark in the target area. These contour shape data are a quantitative representation of the geometric shape of the physical inkjet mark. A set of values ​​can be calculated by applying the contour extraction algorithm to the target area, such as the total perimeter of the contour, the area enclosed, and the aspect ratio of the circumscribed rectangle.

[0095] (2) The color distribution data formed by the quantitative characterization of the color composition of physical inkjet marking can be achieved by calculating the statistics (such as mean and variance) of all pixels in the target area in a specific color space (such as RGB or HSV), or by constructing a color histogram of the target area.

[0096] (3) To eliminate the influence of the positional offset of the product under inspection at the workstation, relative position coordinate data may also be included. These data represent the geometric center of the target area relative to one or more stable structural anchor points located in the background area. For example, the two-dimensional or three-dimensional spatial displacement vector of the center of a specific hole or the end point of a weld on the product under inspection.

[0097] Secondly, data extracted from the background area to reflect the product's inherent characteristics. This data utilizes the individual differences that may exist in each product as auxiliary information for identification, and may specifically include: (1) For welded parts, the weld is an inherent feature with individual differences in morphology. Welds can be identified in the background area and their morphological data, such as the set of weld path points, total length, average curvature, and intersection information with other structural elements, can be extracted.

[0098] (2) The structural texture data formed by the quantitative characterization of the surface material texture of the product can be obtained by applying texture analysis algorithms to the selected sample area in the background area, such as calculating its Local Binary Pattern (LBP) histogram or the response energy of Gabor filter.

[0099] (3) Product model feature data consisting of macroscopic structural features used to distinguish specific product models. For example, the specific model code of a product can be determined by detecting the quantity, size or spatial layout of specific components (such as pipe openings, flanges, etc.) in the background area.

[0100] Finally, the industrial control computer combines the extracted one or more types of data to form a structured data entity, such as a multidimensional feature vector. This data entity is the retrieval feature data used for subsequent retrieval and matching.

[0101] This invention constructs a highly redundant and robust retrieval feature data system by accurately segmenting the appearance image and combining a strategy of extracting feature data from two different sources: the target area and the background area. Even when the physical inkjet markings themselves are damaged or difficult to extract clearly, matching can still be achieved using individualized features such as weld morphology data extracted from the product's background area. This significantly enhances the fault tolerance and reliability of the entire identification and retrieval process in complex industrial environments.

[0102] Figure 5 This is the second schematic diagram of the process for determining the target NG item record provided by the present invention, as follows: Figure 5 As shown, for the headquarters 133 in the above embodiments, the present invention provides a specific implementation method.

[0103] Specifically, the step involves comparing the historical feature data associated with each historical NG record in the candidate record set with the relevant features, and determining the target NG record by combining the time factor, including but not limited to: Step 1331: The extracted retrieval feature data is compared with the historical feature data of each historical NG product record in the candidate record set to calculate the difference, and the original similarity value of each historical NG product record is obtained.

[0104] As described in the previous embodiments, the retrieval feature data extracted from the appearance image of the current product to be inspected, and the historical feature data associated with each historical NG product record in the preset inspection database, are both structured data entities, such as multidimensional feature vectors. Difference calculation refers to performing mathematical operations on these two feature vectors to quantify their similarity. Specific calculation methods that can be used include: Calculate the cosine of the angle between two feature vectors in the vector space. The value is between -1 and 1. The closer the value is to 1, the more consistent the directions of the two feature vectors are, that is, the higher the similarity.

[0105] Alternatively, calculate the linear distance between two feature vectors in multidimensional space. The smaller the distance, the closer the two feature vectors are in space, i.e., the higher their similarity. When using distance as a metric, it can be converted into a similarity value ranging from 0 to 1 using a function (such as a Gaussian function).

[0106] Alternatively, if the retrieval feature data and historical feature data are binary, such as binary codes generated by a hash algorithm, then the number of different bits between the two binary strings can be calculated, and the smaller the number, the higher the similarity.

[0107] In this embodiment, each historical NG (Not Qualified) record in the candidate record set is traversed, and its historical feature data is compared with the currently extracted search feature data using the aforementioned difference calculation. Each scalar value obtained after the calculation is the original similarity value of the corresponding historical NG record.

[0108] Step 1332: Calculate the time difference between the material unloading timestamp of each historical NG product record in the candidate record set and the current system time.

[0109] In this embodiment, the industrial control computer will traverse the candidate record set and read the stored unloading timestamp field for each historical NG product record.

[0110] Subsequently, the current system time when the appearance image was acquired is subtracted from the material unloading timestamp of each historical NG product record. The result is the time difference value corresponding to each historical NG product record. This time difference value represents the physical flow time of the NG product from the material unloading of the previous process to its current identification at the rework station.

[0111] Step 1333: Calculate the time difference using a preset time decay logic to obtain the time weight coefficient for each historical NG product record.

[0112] Given that on a continuously operating production line, physically adjacent products will inevitably be recorded at similar times in the system. Therefore, the smaller the time difference between historical NG (non-conforming) product records, the greater the likelihood that they are the same physical product as the product currently being inspected.

[0113] The preset time decay logic can be a mathematical function or rule, with the time difference as input and the time weighting coefficient as output. The core of this time decay logic is: the smaller the time difference, the larger the corresponding time weighting coefficient.

[0114] Optionally, the time decay logic is a linear decay function, for example: Time weighting coefficient = 1 - (time difference / maximum allowable time difference).

[0115] Time decay logic can also employ non-linear decay functions, such as exponential decay functions: Time weighting coefficient (Time difference); Here, k is a normal number decay factor, which can be configured by the staff according to the characteristics of the production line.

[0116] Step 134: The original similarity value is weighted and calculated with the time weight coefficient to generate the final matching score for each historical NG product record.

[0117] In this embodiment, for each historical NG (Not Qualified) record in the candidate record set, the following weighted calculation is performed. Through the above weighted calculation, each historical NG record will obtain a final matching score that combines both feature and time factors: Final match score = W feature Original similarity score + W time Time weighting coefficient; Among them, W feature and W time These are preset weighting factors, representing the importance of feature similarity and the importance of temporal correlation, respectively, and W feature +W time =1. The specific values ​​of these two weighting factors can be adjusted according to the signal-to-noise ratio of the actual application scenario. For example, in scenarios where the markings are very clear and reliable, W can be increased. feature The value of W can be increased in scenarios where markers are easily damaged but production line flow sequences are strict. time The value of .

[0118] Step 1335: The historical NG item record with the highest final matching score is determined as the target NG item record.

[0119] After calculating the final matching score of all historical NG records in the candidate record set, the industrial control computer will sort these final matching scores, select the only historical NG record with the highest score, and determine it as the target NG record that uniquely corresponds to the product to be inspected at the current rework station.

[0120] This embodiment provides a highly structured and configurable solution for determining target NG product records. It transforms the originally vague concepts of similarity and proximity into a calculable and comparable final matching score. Through a weighted mechanism, it achieves the organic integration of feature information and time information. This makes the matching decision process not only more accurate, but also flexible enough to adapt to different working conditions by adjusting the weight factors. Thus, while ensuring high accuracy, it also gives gas leak detection better generalization ability and engineering practicality.

[0121] Figure 6 This is a schematic diagram of the construction process of the detection database provided by the present invention, as shown below. Figure 6 As shown, in this embodiment of the invention, before obtaining the appearance image of the product to be inspected at the rework station, it is necessary to perform a step of building a pre-set inspection database. The process of building this inspection database is the data foundation for subsequent accurate matching and information retrieval at the rework station, and its specific steps include, but are not limited to, the following: Step 21: In the airtightness testing stage, conduct an airtightness test on each product.

[0122] The airtightness testing step refers to a separate, automated quality inspection station located upstream of the rework station in the production line process. Every product on the production line undergoes standardized airtightness testing as it passes through this quality inspection station.

[0123] One specific technical implementation is as follows: First, a robotic arm picks up the product and places it on a sealed fixture. Then, a tracer gas (such as helium or dry air) at a specified pressure is injected into the product. Subsequently, non-contact testing equipment, such as an acoustic camera or mass spectrometer, deployed at a quality inspection station, scans the product surface to determine whether there is a gas leak.

[0124] Step 22: When it is determined that there is a gas leak in any product, obtain the coordinates of the leak location and generate an inkjet command.

[0125] During actual testing, the control system of the testing equipment analyzes the collected data in real time. When it is determined that any product has a gas leak, for example, if the sound pressure level detected by the acoustic camera exceeds a preset threshold, the control system determines that the product is NG (Not Good).

[0126] Furthermore, once a defective product is identified as non-compliant, the control system will execute two related actions: First, obtain the coordinates of the leak location. The detection equipment will output the three-dimensional spatial coordinate data (X, Y, Z) of the specific leak source that caused this NG product determination in the product coordinate system.

[0127] Secondly, an inkjet printing command is generated and sent to the relevant physical marking application device. The industrial control computer generates an inkjet printing command based on the NG (Not Good) product determination result. This command can contain two core pieces of information: one is the content of the historical inkjet mark to be printed, used to uniquely identify this NG product determination, such as a timestamp string consisting of year, month, day, hour, minute, second, and millisecond to ensure its uniqueness; the other is the execution code used to control the printing action.

[0128] Step 23: In response to the inkjet command, drive the pneumatic nozzle to apply the historical inkjet mark at a preset position on any of the product surfaces.

[0129] Upon receiving an inkjet command from the industrial control computer, the system immediately responds and controls the industrial inkjet printing system integrated downstream of the airtightness testing station. The industrial inkjet printing system drives its pneumatic nozzles to move above the product to be marked to perform the inkjet printing action. Pneumatic nozzles are industrial-grade printheads that utilize compressed air as power to achieve high-speed, non-contact printing.

[0130] It should be noted that the printing operation is performed at a preset location on the product surface. This preset location can be a predefined marking area that is uniform across all products of the same type, based on the product type and process requirements, such as the flat surface at the bottom of a water heater's inner tank. A pneumatic nozzle uses fast-drying industrial ink to print historical inkjet markings onto the product surface at this preset location.

[0131] Step 24: During the product unloading process, take a picture of any product to obtain a product image containing the historical inkjet mark, and store the product image as a leak location detection image in the historical NG product record into the detection database.

[0132] The product unloading process can be a separate station adjacent to the physical marking application, or it can be integrated with the marking station. During the product unloading process, NG products that have been marked with historical inkjet marks will be documented here, including taking a photo of the product. This photo will be taken by a fixed industrial camera, clearly showing part of the product's appearance and the product image that has just been printed with historical inkjet marks.

[0133] According to the technical solution of this embodiment, the product image will subsequently be stored in the detection database as the leak location detection image in the historical NG product record. Specifically, the data storage operation can be as follows: the industrial control computer creates a new historical NG product record entry. In this record entry, the string content of the generated historical inkjet mark is strongly associated with the file or storage path of the currently captured product image. Then, this complete record entry is written into the preset detection database. Under this definition, the leak location detection image stored in the database contains the product image with the historical inkjet mark.

[0134] The detection database construction process provided in this invention generates and applies a unique historical inkjet mark when a product leak is detected, and takes photos for evidence during the unloading process. This ensures that every NG product has an immutable digital file in the pre-set detection database that completely corresponds to its physical mark when it leaves the detection area. This guarantees the accuracy and uniqueness of the one-to-one correspondence between the physical inkjet mark and the product image record from the source. It provides a solid data foundation for accurately retrieving the corresponding digital record by identifying the physical inkjet mark at the rework station, thereby eliminating the possibility of confusion in rework information due to missing or incorrect data association.

[0135] Figure 7 This is one of the schematic diagrams of the inkjet marking process provided by the present invention, as an optional embodiment, such as... Figure 7 As shown, this embodiment provides a method for adaptively selecting the optimal printing position based on the leakage location, mainly including: Step 231: Based on the obtained leak location coordinates, analyze the regional attributes of the leak source on the product surface; Step 232: Call the motion trajectory parameters corresponding to the area attributes as the preset position, and control the pneumatic nozzle to move and spray the historical inkjet mark.

[0136] In this embodiment, the preset position is a fixed area. However, on some complex products to be inspected, there may be multiple flat areas suitable for inkjet printing, or some areas may be unsuitable for printing due to structural obstruction, high temperature, or other reasons. When a gas leak is detected in any product, the industrial control computer not only generates inkjet printing instructions but also obtains the coordinates of the leak location, accurately indicating the physical location of the leak source in the three-dimensional model space of the product to be inspected.

[0137] The industrial control computer will use the leak location coordinates to perform an analysis on the product's internally stored 3D digital model. Specifically, the computer will determine which logical partition on the product surface the leak location coordinates fall on, thus obtaining the area attribute of the leak source. For example, a 3D digital model of a water heater's inner tank can be pre-divided into multiple logical areas such as "top spherical surface," "Cylinder A area," "Cylinder B area," and "bottom flange," each with a defined name and spatial range. When the industrial control computer determines that the leak location coordinates fall within the spatial range of "Cylinder A area," the analyzed area attribute is "Cylinder A area."

[0138] Optionally, the industrial control computer pre-stores a region trajectory mapping table, establishing a correspondence between each region attribute of the product and a specific set of motion trajectory parameters. These motion trajectory parameters can be a set of precise instruction data, defining a series of kinematic parameters required for the pneumatic nozzle to move from its standby position to an optimal printing point within the region, such as the spatial coordinates of the target point, attitude angle, movement speed, acceleration / deceleration curve, etc. This optimal printing point constitutes the dynamic preset position for this operation. The reason for associating these motion trajectory parameters with region attributes is that the flat area closest to the leak source is usually the best location for printing markings for subsequent observation by personnel.

[0139] After analyzing the regional attributes of the leak source, the industrial control computer retrieves the corresponding motion trajectory parameters from the regional trajectory mapping table. Subsequently, the industrial control computer sends these motion trajectory parameters to the motion controller mounted on the pneumatic nozzle, such as a multi-axis robot controller. Based on the received motion trajectory parameters, the motion controller controls the pneumatic nozzle to precisely move to the dynamically determined preset position and completes the printing operation of the historical inkjet markings.

[0140] In a specific example, if the industrial control computer determines that the leak occurs on the "top spherical surface" of the water heater's inner tank, it will call the corresponding motion trajectory parameters to control the robot to move the pneumatic nozzle to a flat area on the top of the inner tank for printing. If the next leak occurs on the "bottom flange," the industrial control computer will call another set of motion trajectory parameters to control the robot to print the historical inkjet markings near the bottom flange.

[0141] This invention establishes a dynamic correlation between the leak location and the printing location, upgrading the fixed marking printing location to an intelligent printing strategy that adaptively adjusts according to the leak area. This ensures that the applied historical inkjet markings always appear in the location closest to or most easily observable to the actual leak point. This not only greatly improves the efficiency of subsequent workers in finding and referring to the markings for repair, but also avoids the difficulty or confusion that may be caused by the leak location being too far from the fixed marking point, enhancing the intuitiveness and effectiveness of the entire repair guidance information.

[0142] Figure 8 This is a second schematic diagram of the inkjet marking process provided by the present invention, as another optional embodiment, such as... Figure 8 As shown, this embodiment provides a fixed-location implementation method with a simpler process and higher execution efficiency, specifically including: Step 331: Retrieve the pre-stored fixed coordinate parameters as the preset position; Step 332: Control the pneumatic nozzle to move to the product part corresponding to the fixed coordinate parameters to perform fixed-point inkjet printing, so as to generate the historical inkjet mark with a fixed position.

[0143] Specifically, during system initialization or process parameter setting, a set of fixed coordinate parameters can be configured and pre-stored within the industrial control computer. These fixed coordinate parameters can be a set of absolute coordinate values ​​defining a specific point in three-dimensional space, such as X, Y, Z, Rx, Ry, and Rz values ​​relative to the robot's base coordinate system. Each specific point physically corresponds precisely to a selected fixed area on the surface of the product to be inspected, used for marking. The selection criteria for this fixed area are typically a flat surface, ease of printing, resistance to wear during subsequent handling, and ease of image acquisition by the image acquisition equipment at the rework station. For all products of the same model, this set of fixed coordinate parameters is unique and unchanging; therefore, the printing target point defined by these parameters is the preset position described in this embodiment.

[0144] Once the inkjet command is generated, the industrial control computer immediately retrieves the pre-stored fixed coordinate parameters from its memory or configuration file and sends them as a motion target command to the motion controller mounted on the pneumatic nozzle. After parsing the motion target command, the motion controller controls the pneumatic nozzle to perform a standard point-to-point movement, precisely moving to the unique target point defined by the fixed coordinate parameters.

[0145] After the pneumatic nozzle reaches the designated position and stabilizes, the industrial control computer triggers the printing action. The pneumatic nozzle performs a pinpoint inkjet operation on the target point on the surface of the product to be inspected, printing the content of the historical inkjet markings onto it. Since the target position for each printing is strictly defined by the same set of fixed coordinate parameters, the historical inkjet markings generated on the surface of all products of the same model that are judged as NG (Not Good) will have fixed positions and consistent heights.

[0146] The fixed-position inkjet printing scheme provided in this invention defines a unique printing position using preset fixed coordinate parameters. This achieves standardized and highly repeatable application of historical inkjet markings on all defective products, resulting in fast execution speed and high reliability, well-suited to the cycle time requirements of high-speed production lines. More importantly, the fixed-position historical inkjet markings greatly simplify the image acquisition and recognition process at downstream rework stations. The image acquisition equipment can capture and process only a fixed region of interest, thereby reducing the complexity and computational load of the vision system and improving the efficiency and stability of the entire gas leak detection method in subsequent recognition stages.

[0147] Figure 9 This is a schematic diagram of product model matching test before airtightness testing provided by the present invention, as shown in the figure. Figure 9 As shown, in this embodiment, before performing airtightness testing on each product, the following steps are also performed: Step 31: Use a camera device located at the loading station to capture images of the product being loaded.

[0148] Step 32: Determine the key structural dimensions of the product based on the relationship between the structural dimensions of the loading station and the pixel ratio of the loading image.

[0149] Step 33: Read the preset production model parameters in the current programmable logic controller and compare the measured key structural dimensions with the standard dimensions corresponding to the production model parameters.

[0150] Step 34: When the comparison result exceeds the preset matching threshold, an alarm is triggered and the product is blocked from entering the airtightness testing stage.

[0151] The loading station is a physical workstation preceding the airtightness testing stage. It is responsible for picking up the product to be tested from the main conveyor line or material box and placing it at the entrance of the testing line. At this loading station, one or more camera devices, such as industrial cameras with a fixed field of view, are installed. Once a product is placed at the designated position at the loading station, the industrial control computer triggers the camera device to take a picture, thereby capturing a loading image that clearly shows the complete or key outlines of the product.

[0152] Furthermore, a non-contact dimensional measurement process using machine vision is executed. The industrial control computer pre-stores the dimensional data of the loading station, such as the vertical distance from the camera to the product placement plane and the focal length of the lens. This data is obtained through a pre-calibrated camera process. Based on this known dimensional data, the actual physical length represented by one pixel in the loading image on the physical plane can be calculated, thus determining the pixel ratio of the image; for example, one pixel equals 0.5 millimeters.

[0153] Subsequently, the industrial control computer uses image processing algorithms (such as edge detection, contour finding, and Hough transform) to locate specific structures of the product on the loading image and measures their pixel dimensions in the image. These measured structures are the product's critical structural dimensions. For example, for a water heater's inner tank, critical structural dimensions may include the tank's total height, cylinder diameter, and the distance between the top inlet and outlet water pipes. The industrial control computer multiplies the measured pixel dimensions by pixel ratios to determine the product's actual critical structural dimensions in the physical world.

[0154] Programmable Logic Controllers (PLCs) are the core units of production line automation control, responsible for managing the operational status of the entire production line. When a production line needs to switch product types, the operator selects a new product model on the Human Machine Interface (HMI). The relevant information for this product model is written into the PLC and stored as a preset production model parameter.

[0155] The industrial control computer reads the preset production model parameters from the programmable logic controller (PLC) via a communication interface (such as OPC UA or Modbus TCP / IP). Simultaneously, the industrial control computer's internal database stores a standard dimension table, which details the complete set of standard dimensions for each product model.

[0156] The industrial control computer can find the corresponding set of standard dimensions from the standard dimension table based on the preset production model parameters, and then compare the two to calculate the deviation value between them.

[0157] Within the industrial control computer, a preset matching threshold is set for each critical structural dimension. This represents the allowable manufacturing tolerance and measurement error range (e.g., ±2 mm). After comparison, the industrial control computer checks whether the deviation values ​​of all critical structural dimensions are within their respective preset matching thresholds.

[0158] When the absolute value of any comparison result (i.e., deviation value) exceeds its corresponding preset matching threshold, the industrial control computer determines that the currently fed product model does not match the planned production model. At this point, the industrial control computer will immediately perform two actions: First, it will trigger an alarm, such as sounding an alarm and flashing a red light on-site, and displaying a "Feeding Model Error" message on the HMI interface to notify the on-site operators. Second, the industrial control computer may also send a blocking command to the programmable logic controller (PLC). Upon receiving the blocking command, the PLC will immediately stop the conveyor belt or control the robotic arm to remove the erroneous product from the production line, effectively preventing this product from entering the airtightness testing stage. Only when the comparison results of all key structural dimensions are within the preset matching threshold will the industrial control computer allow the product to continue flowing.

[0159] This invention adds a machine vision-based size verification and model matching error prevention process before airtightness testing. This ensures that only products with the correct model can enter the subsequent testing and marking stages, eliminating a series of chain problems caused by manual loading errors, such as using incorrect testing programs, generating invalid NG records, and applying meaningless physical markings. This guarantees the data integrity and accuracy of the entire automated testing and rework information system, avoiding waste of production resources and potential quality risks.

[0160] Based on the above embodiments, as an optional embodiment, the step of traversing the detection database and filtering out multiple historical NG (non-compliant) records whose unloading timestamps fall within the time retrieval window, constructing a candidate record set, includes, but is not limited to: Check the repair status attributes of the historical NG product records and filter only the records whose repair status attribute is "not returned for repair". And / or, Identify product model features in the appearance image and filter only records that belong to the same product model category as the product to be detected; The records filtered by the maintenance status attribute and / or the product model category are used as the candidate record set.

[0161] In the above embodiments, the construction of the candidate record set relies solely on the time dimension filtering. In order to further reduce the size of the candidate record set and exclude records that should not logically be matching objects, this embodiment adds one or two layers of attribute filtering on the basis of time filtering.

[0162] In one optional embodiment, in the table structure design of the preset detection database, in addition to the aforementioned fields such as material unloading timestamp and historical inkjet marker, a maintenance status attribute field is also set for each historical NG product record to track the life cycle status of each NG product in the rework process. Its value can include at least two types: not reworked and reworked.

[0163] When a historical NG (Not Returned) product record is created, its default repair status attribute is "not returned for repair". Once the product is successfully repaired at the repair station, the industrial control computer or operators via a client will update the repair status attribute of the corresponding historical NG product record to "returned for repair".

[0164] In the screening process of this embodiment, after the industrial control computer initially finds a batch of historical NG (Not Returned for Repair) product records that meet the time criteria based on the time retrieval window, it further filters these records. For example, it checks the value of the maintenance status attribute field of each historical NG product record. Only when the value of this field is "not returned for repair" will the historical NG product record be retained; if the value is "returned for repair," it means that the product has already been transferred to the repair station and processed at some point in the past, and logically should not appear again, so the record will be discarded directly. In this way, invalid interference data caused by abnormal situations, such as products returning to the production line, can be effectively eliminated.

[0165] In another alternative embodiment, a model matching and filtering mechanism is provided for mixed production scenarios where multiple product models flow simultaneously on the production line. The industrial control computer first analyzes the appearance image of the product to be inspected to identify its product model characteristics. The identification method may include: reading the model string printed or etched on the product surface in the appearance image; or measuring the dimensions or geometric relationships of specific structures (such as flange diameter, number of pipe openings) on the product in the appearance image and comparing them with an internally stored model size database to determine its product model.

[0166] After identifying the specific model of the product to be inspected, the industrial control computer adds an extra filtering condition when screening historical NG product records. In addition to the timestamp falling within the time retrieval window, the product model field associated with the historical NG product record must also be completely consistent with the currently identified product model. Only those historical NG product records that belong to the same product model category as the product to be inspected will be retained.

[0167] Finally, the records filtered by the maintenance status attribute and / or the product model category are used as the candidate record set.

[0168] Industrial control computers can flexibly employ one or two filtering logics depending on the configuration. On production lines producing a single model, only maintenance status attribute filtering can be enabled; on production lines producing mixed models, both maintenance status attribute filtering and product model category filtering can be enabled simultaneously, resulting in a smaller, highly relevant set of candidate records.

[0169] This invention, by adding maintenance status attribute filtering and / or product model category filtering on the basis of time-based filtering, constructs a multi-dimensional and refined candidate record set filtering process. By eliminating logically irrelevant and physically mismatched interference records, the size of the candidate record set can be significantly reduced, thereby greatly reducing the amount of computation required for subsequent complex feature comparisons and improving the computational efficiency of the entire matching process. This precise pre-screening also fundamentally reduces the probability of false matching.

[0170] Figure 10 This is the second schematic flowchart of the gas leak detection method provided by the present invention, as shown below. Figure 10 As shown, in a specific embodiment of the present invention, after sending the leak location detection image from the target NG product record to the client, the method further includes a step of constructing a closed-loop management system for rework quality, specifically including: Receive the welding completion signal sent by the client; In response to the welding completion signal, a reflow notification is generated. The reflow notification prompts or controls the transmission of the welded product to be tested to the preceding airtightness testing stage, so as to control the re-performance of airtightness testing on the product to be tested.

[0171] In addition to displaying the leak location detection image, the client's user interface can also include an interactive element, such as a virtual button, to confirm the completion of the welding repair. When the worker completes the physical repair, they will press this button. At this time, the client application will generate a welding repair completion signal in a specific format. The client sends this welding repair completion signal to the industrial control computer via the local area network, and the industrial control computer receives the welding repair completion signal through its network monitoring service.

[0172] After receiving the welding repair completion signal, the industrial control computer will be triggered to execute a series of subsequent actions. One optional internal operation is that the industrial control computer will first update the preset inspection database, changing the repair status attribute field in the target NG product record corresponding to the current product to be inspected from "not repaired" to "repaired," thus completing the recording of the repair lifecycle.

[0173] Subsequently, the industrial control computer executes the core action of generating a return flow notification. This return flow notification is a data message containing specific instruction codes, its function being to initiate the return flow process of the product to the inspection station. The return flow notification can function in two ways: as a prompt or as a control. (1) The industrial control computer can send a signal to the indicator device at the rework station, for example, by lighting up a green indicator light, or by displaying a text message on the client screen indicating that the rework is complete and the product should be returned to the main conveyor line. This method is used to prompt the on-site operator to manually move or push the product to be inspected that has been repaired to the designated physical return inlet.

[0174] (2) In a more automated production line, the industrial control computer will send the return notification directly to the lower-level controller of the production line, such as a programmable logic controller. The return notification, as an execution instruction, will control the physical conveying equipment to start automatically, such as an independent conveyor belt, a lifting platform, or a pneumatic push rod, to transport the product to be tested after the welding is completed to the entrance of the previous airtightness testing stage.

[0175] Whether through prompts or direct control, the end result is always the same: the product is sent back to the station where leak detection was initially performed. When the product arrives at the airtightness testing stage again, the relevant sensors will identify it as a new product to be tested, thus controlling the complete airtightness testing process to be repeated. If the re-inspection passes, the product will continue to flow to subsequent processes as a qualified product; if the re-inspection finds that a leak still exists (whether the original leak point is not repaired properly or a new leak point is created due to welding), the product will be judged as a NG product again, and the new marking and filing process mentioned in the previous embodiment will be repeated, entering the rework cycle again.

[0176] This invention establishes a complete "inspection-rework-re-inspection" quality assurance system by adding a closed-loop process of product re-inspection after the rework process. This ensures that every reworked product must undergo rigorous automated inspection to verify its repair quality, fundamentally eliminating the risk of defective products flowing into the next stage due to poor rework quality or the introduction of new defects. Furthermore, it strictly achieves full traceability and closed-loop management of the NG product handling lifecycle, significantly improving the quality reliability of the final product and the control level of the production process.

[0177] Figure 11 This is a schematic diagram of the gas leak detection device provided by the present invention, as shown below. Figure 11 As shown, it mainly includes, but is not limited to: Image acquisition unit 1: The user acquires an image of the appearance of the product to be inspected located at the rework station; Image processing unit 2 is used to identify, based on the appearance image, whether there is a physical inkjet mark on the surface of the product to be inspected that was triggered by the previous airtightness inspection process when determining a leak. Image recognition unit 3 is used to identify the physical inkjet mark, and then use the physical inkjet mark to search and match with historical inkjet marks in a preset detection database to determine the target NG product record that uniquely corresponds to the product to be inspected; the detection database stores leakage location detection images of historical NG products, and each leakage location detection image is associated with the historical inkjet mark that was triggered when a leakage was determined. Leakage location unit 4 is used to send the leak location detection image from the target NG product record to the client.

[0178] It should be noted that the gas leak detection device provided by the present invention can execute the gas leak detection method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0179] The gas leak detection device provided in this invention constructs a mapping relationship between the reworked product and historical detection data by using the physical inkjet markings applied in the preceding process as visual feature indexes. This enables accurate traceability and identity verification of NG products in scenarios without product serial numbers, effectively avoiding material mixing and significantly improving rework positioning efficiency.

[0180] Figure 12 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 12 As shown, the electronic device may include: a processor 1210, a communications interface 1220, a memory 1230, and a communication bus 1240, wherein the processor 1210, the communications interface 1220, and the memory 1230 communicate with each other via the communication bus 1240. The processor 1210 can call logic instructions in the memory 1230 to execute a gas leak detection method, which includes: acquiring an appearance image of a product to be inspected located at a rework station; identifying, based on the appearance image, whether a physical inkjet mark is applied on the surface of the product to be inspected, triggered by a previous airtightness inspection process when a leak is determined; if the physical inkjet mark is identified, using the physical inkjet mark to search and match with historical inkjet marks in a preset detection database to determine a target NG product record uniquely corresponding to the product to be inspected; the detection database stores leak location detection images of historical NG products, each leak location detection image being associated with the historical inkjet mark applied when a leak is determined; and sending the leak location detection image in the target NG product record to a client.

[0181] Furthermore, the logical instructions in the aforementioned memory 1230 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, essentially, or the part that contributes to the prior art, or a part 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 the present invention. 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.

[0182] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, the computer is able to execute the gas leak detection method provided in the above embodiments, the method including: acquiring an appearance image of a product to be tested located at a rework station; identifying, based on the appearance image, whether there is a physical inkjet mark applied on the surface of the product to be tested triggered by a previous airtightness testing process when a leak is determined; if the physical inkjet mark is identified, using the physical inkjet mark to search and match with historical inkjet marks in a preset detection database, determining a target NG product record uniquely corresponding to the product to be tested; the detection database stores leak location detection images of historical NG products, each leak location detection image being associated with the historical inkjet mark applied when a leak is determined; sending the leak location detection image in the target NG product record to a client.

[0183] In another aspect, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon. When executed by a processor, the computer program is implemented to perform the gas leak detection method provided in the above embodiments. The method includes: acquiring an appearance image of a product to be inspected located at a rework station; identifying, based on the appearance image, whether a physical inkjet mark is applied on the surface of the product to be inspected, triggered by a previous airtightness inspection process when a leak is determined; if the physical inkjet mark is identified, using the physical inkjet mark to perform a search and match with historical inkjet marks in a preset detection database to determine a target NG product record uniquely corresponding to the product to be inspected; the detection database stores leak location detection images of historical NG products, each leak location detection image being associated with the historical inkjet mark applied when a leak is determined; and sending the leak location detection image in the target NG product record to a client.

[0184] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0185] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0186] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting gas leaks, characterized in that, include: Obtain an image of the product to be inspected located at the rework station; Based on the appearance image, identify whether there are physical inkjet marks on the surface of the product to be tested that were triggered by the previous airtightness testing process when a leak was determined; If the physical inkjet mark is detected, the physical inkjet mark is searched and matched with historical inkjet marks in the preset detection database to determine the target NG product record that uniquely corresponds to the product to be tested; the detection database stores leakage location detection images of historical NG products, and each leakage location detection image is associated with the historical inkjet mark that was triggered when a leakage was determined; Send the leak location detection image from the target NG product record to the client; The step of using the physical inkjet marker to search and match with historical inkjet markers in a preset inspection database to determine the target NG (non-compliant) record uniquely corresponding to the product to be inspected includes: Obtain the current system time when the appearance image was acquired, and set a time retrieval window for backward lookup based on the current system time; Traverse the detection database, filter out multiple historical NG product records whose material feeding timestamps fall within the time retrieval window, and construct a candidate record set; Extract the retrieval feature data from the appearance image, compare it with the historical feature data associated with each historical NG product record in the candidate record set, and determine the target NG product record by combining the time factor.

2. The gas leak detection method according to claim 1, characterized in that, The extraction of retrieval feature data from the appearance image includes: The appearance image is segmented to separate the target area containing the physical inkjet mark and the background area after removing the target area; Extract the retrieval feature data, which includes at least one of the following: the outline shape data, color distribution data, and relative position coordinate data of the physical inkjet mark in the target area, or the weld morphology data, structural texture data, and product model feature data reflecting the characteristics of the product body in the background area.

3. The gas leak detection method according to any one of claims 1-2, characterized in that, The step of comparing the historical feature data associated with each historical NG record in the candidate record set with the target NG record, and combining this with the time factor, includes: The extracted retrieval feature data is compared with the historical feature data of each historical NG product record in the candidate record set to calculate the difference, and the original similarity value of each historical NG product record is obtained. Calculate the time difference between the material unloading timestamp of each historical NG product record in the candidate record set and the current system time; The time difference is converted using a preset time decay logic to obtain the time weight coefficient for each historical NG product record. The smaller the time difference, the larger the corresponding time weight coefficient. The original similarity value is weighted by the time weight coefficient to generate the final matching score for each historical NG product record; The historical NG record with the highest final matching score is determined as the target NG record.

4. The gas leak detection method according to claim 1, characterized in that, Before acquiring the appearance image of the product to be inspected at the rework station, the procedure also includes: In the airtightness testing stage, each product undergoes an airtightness test; When a gas leak is detected in any product, the coordinates of the leak location are obtained and an inkjet command is generated. In response to the inkjet command, a pneumatic nozzle is driven to apply the historical inkjet mark at a preset position on any of the product surfaces; During the product unloading process, a photograph is taken of any product to obtain a product image containing the historical inkjet markings, and the product image is stored in the detection database as a leak location detection image in the historical NG product record.

5. The gas leak detection method according to claim 4, characterized in that, The step of responding to the inkjet command by driving a pneumatic nozzle to apply the historical inkjet mark at a preset position on any product surface includes: Based on the obtained coordinates of the leak location, the regional attributes of the leak source on the product surface are analyzed; The motion trajectory parameters corresponding to the area attributes are called as the preset position, and the pneumatic nozzle is controlled to move and spray the historical inkjet mark.

6. The gas leak detection method according to claim 4, characterized in that, The step of driving a pneumatic nozzle to apply the historical inkjet mark at a preset position on any product surface in response to the inkjet command further includes: Retrieve pre-stored fixed coordinate parameters as the preset position; The pneumatic nozzle is controlled to move to the product part corresponding to the fixed coordinate parameters to perform fixed-point inkjet printing, so as to generate the historical inkjet mark with a fixed position.

7. The gas leak detection method according to claim 4, characterized in that, In the airtightness testing process, before conducting airtightness testing on each product, the following steps are also included: The loading images of the product are captured using a camera device located at the loading station; Based on the relationship between the structural dimensions of the loading station and the pixel ratio of the loading image, the key structural dimensions of the product are determined. Read the preset production model parameters in the current programmable logic controller, and compare the measured key structural dimensions with the standard dimensions corresponding to the production model parameters; When the comparison result exceeds the preset matching threshold, an alarm is triggered and the product is blocked from entering the airtightness testing stage.

8. The gas leak detection method according to claim 1, characterized in that, The process involves traversing the detection database, filtering out multiple historical NG (non-compliant) records whose unloading timestamps fall within the time retrieval window, and constructing a candidate record set, including: Check the repair status attributes of the historical NG product records and filter only the records whose repair status attribute is "not returned for repair". And / or, Identify product model features in the appearance image and filter only records that belong to the same product model category as the product to be detected; The records filtered by the maintenance status attribute and / or the product model category are used as the candidate record set.

9. The gas leak detection method according to claim 1, characterized in that, After sending the leak location detection image from the target NG item record to the client, the method further includes: Receive the welding completion signal sent by the client; In response to the welding completion signal, a reflow notification is generated. The reflow notification prompts or controls the transmission of the welded product to be tested to the preceding airtightness testing stage, so as to control the re-performance of airtightness testing on the product to be tested.

10. A gas leak detection device, characterized in that, include: Image acquisition unit: The user acquires an image of the appearance of the product to be inspected located at the rework station; An image processing unit is used to identify, based on the appearance image, whether there is a physical inkjet mark on the surface of the product to be inspected that was triggered by the preceding airtightness inspection process when a leak was determined; An image recognition unit is used to identify the physical inkjet mark, and then use the physical inkjet mark to search and match it with historical inkjet marks in a preset detection database to determine the target NG product record that uniquely corresponds to the product to be inspected; the detection database stores leakage location detection images of historical NG products, and each leakage location detection image is associated with the historical inkjet mark that was triggered when a leakage was determined; A leak location unit is used to send the leak location detection image from the target NG product record to the client; The step of using the physical inkjet marker to search and match with historical inkjet markers in a preset inspection database to determine the target NG (non-compliant) record uniquely corresponding to the product to be inspected includes: Obtain the current system time when the appearance image was acquired, and set a time retrieval window for backward lookup based on the current system time; Traverse the detection database, filter out multiple historical NG product records whose material feeding timestamps fall within the time retrieval window, and construct a candidate record set; Extract the retrieval feature data from the appearance image, compare it with the historical feature data associated with each historical NG product record in the candidate record set, and determine the target NG product record by combining the time factor.

11. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the gas leak detection method as described in any one of claims 1 to 9.

12. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the gas leak detection method as described in any one of claims 1 to 9.

13. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the gas leak detection method as described in any one of claims 1 to 9.

Images