A method and system for vacuum decay leak testing on a condom production line

Abstract

The present application relates to the technical field of condom testing, and particularly relates to a vacuum attenuation sealing property testing method and system on a condom production line, which first acquires latex micropore parameters and product profile data, divides a sealing risk area in combination with porosity distribution and reservoir structure; secondly, a vacuum testing unit attenuation section is decomposed according to testing accuracy requirements and risk area data; then, cavity background pressure fluctuation is monitored in real time, and pressure deviation compensation estimation is performed in combination with attenuation section data to eliminate environmental and material interference; subsequently, a defect sensitive response map is constructed based on compensation data and leakage risk is quantified; finally, a pass or fail determination is output according to risk quantification results, and online high-precision dynamic detection is realized. The method effectively distinguishes real leakage from false signals caused by material air permeability or structural deformation, and significantly improves the micro-defect detection rate and production quality control level.

Description

Technical Field

[0001] This invention relates to the field of condom testing technology, and in particular to a method and system for testing vacuum decay sealing performance on a condom production line. Background Technology

[0002] In the production of condoms, the product's airtightness is a core quality indicator determining its safety and effectiveness. Traditional airtightness testing methods often employ water testing or bubble testing, relying on manual visual inspection to determine the presence of leaks. These methods suffer from low efficiency, high subjectivity, and difficulty in detecting minute defects (such as micron-sized pinholes or abnormal material permeability), failing to meet the real-time, accuracy, and traceability requirements of modern high-speed automated production lines. With the development of polymer materials science and precision testing technology, the vacuum decay method, as a non-destructive and highly sensitive airtightness testing method, is gradually being applied in the fields of medical devices and safety protective equipment. This method determines the presence of leaks in the tested component by monitoring pressure changes in a sealed cavity under vacuum conditions, offering advantages such as fast detection speed and good repeatability.

[0003] However, condoms, as ultra-thin latex products, possess inherent polymer microporous permeability. Furthermore, under vacuum, the reservoir tip is prone to significant elastic stress deformation. This results in pressure changes within the cavity originating not only from actual leakage but also from spurious leak signals caused by the material's inherent permeability or structural deformation, leading to misjudgments. Existing vacuum detection technologies generally treat these effects as noise, lacking a collaborative modeling and dynamic compensation mechanism for latex microporous permeability and local structural stress response, thus limiting further improvements in detection sensitivity and accuracy.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] The main objective of this invention is to provide a vacuum attenuation sealing test method and system for condom production lines. This invention aims to solve the technical problem that existing vacuum attenuation detection technologies lack a collaborative modeling and dynamic compensation mechanism for the inherent microporous permeability characteristics of latex materials and the interference of stress deformation in the reservoir tip, making it difficult to effectively distinguish between real leakage signals and pseudo-leakage noise caused by material permeability or structural deformation, thus limiting the sensitivity and accuracy of condom sealing tests.

[0006] To achieve the above objectives, the present invention provides a method for testing vacuum attenuation sealing performance on a condom production line, the method comprising:

[0007] The micropore parameters and product structure contour data of the latex polymer material of the condom to be tested are obtained, and the sealing risk area is divided according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip to obtain the data of the risk area to be tested.

[0008] Obtain the accuracy requirement data for the production line sealing test; decompose the test accuracy requirement data and the data of the risk area to be inspected into the attenuation section of the vacuum test unit to obtain the test attenuation section decomposition data;

[0009] Acquire the cavity background pressure fluctuation data of the current vacuum test system, and estimate the pressure deviation compensation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain the pressure compensation estimation data;

[0010] Based on the pressure compensation estimation data, a defect sensitivity response map is constructed to obtain defect response characteristic map data; based on the defect response characteristic map data, leakage risk is quantified to obtain leakage risk quantification data.

[0011] Based on the quantitative data of leakage risk, a pass or fail judgment is output to obtain the online sealing test result, so as to realize high-precision dynamic sealing test of condom production line.

[0012] Optionally, the step of dividing the sealing risk area based on the latex micropore porosity distribution and the seminal vesicle structural parameters to obtain the risk area data to be inspected includes:

[0013] Based on the latex micropore porosity distribution data and seminal vesicle stress parameters, the defect-pressure response coupling relationship was analyzed to obtain the defect-pressure coupling matrix data.

[0014] Obtain statistical data on the distribution of defects in the condom production process, and calculate the frequency of risk occurrence based on the statistical data to obtain defect risk frequency data;

[0015] Based on the defect-pressure coupling matrix data and defect risk frequency data, a digital twin inspection domain is constructed to obtain the inspection risk area data.

[0016] Optionally, the step of constructing a digital twin inspection domain based on defect-pressure coupling matrix data and defect risk frequency data to obtain inspection risk area data includes:

[0017] Based on the defect-pressure coupling matrix data and defect risk frequency data, virtual vacuum test field coverage features and pressure attenuation transmission features are extracted, resulting in virtual test field coverage feature data and pressure attenuation transmission feature data, respectively.

[0018] Based on the virtual test field coverage feature data and pressure attenuation transmission feature data, twin defect response maps are generated from the defect-pressure coupling matrix data to obtain twin defect response map data.

[0019] The virtual-real mismatch region detection was performed on the twin defect response map data to obtain virtual-real calibration map data. The virtual-real mismatch region detection included environmental temperature fluctuation mismatch detection, cavity leakage background anomaly detection, and latex elastic creep deviation detection.

[0020] The production line cycle time constraint is injected into the virtual and real calibration diagram data to obtain the time constraint diagram data;

[0021] High-risk connected subdomains are extracted from the time-series constraint graph data to obtain the data of the risk areas to be inspected.

[0022] Optionally, the process of decomposing the test accuracy requirement data and the data of the risk area to be inspected into attenuation segment data using a vacuum test unit to obtain test attenuation segment decomposition data includes:

[0023] The topology of the vacuum attenuation test section is divided based on the test accuracy requirement data and the data of the risk area to be inspected, and the test section topology division data is obtained.

[0024] Based on the test section topology division data, gas permeation characteristics and key defect nodes are extracted, and gas permeation characteristic data and key defect node data are obtained respectively.

[0025] Based on the gas permeation characteristic data and key defect node data, the pressure decay time-series dependency processing of the test section topology partitioning data is performed to obtain time-series dependency data.

[0026] A dynamic decay response map is constructed based on time-dependent data to obtain the decomposed data of the test decay segment.

[0027] Optionally, the step of acquiring the cavity background pressure fluctuation data of the current vacuum testing system, and performing pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data includes:

[0028] Based on the cavity background pressure fluctuation data and test attenuation segment decomposition data of the current vacuum testing system, multi-cycle pressure attenuation superposition calculation is performed to obtain attenuation superposition data.

[0029] The pressure deviation kernel distribution is calculated based on the attenuation superposition data to obtain the local pressure deviation intensity data;

[0030] Based on the current background pressure data of the vacuum test chamber, the ultimate vacuum degree of the chamber and the response error of the pressure sensor are calculated, and the ultimate vacuum data and sensor error data are obtained respectively.

[0031] Based on the ultimate vacuum data and sensor error data, the dynamic pressure compensation amount is extrapolated from the local pressure deviation intensity data to obtain the pressure compensation estimate data.

[0032] Optionally, the constructing a defect-sensitive response map based on pressure compensation estimation data to obtain defect response feature map data comprises:

[0033] constructing a test attenuation segment response subgraph based on the current background pressure data of the vacuum test cavity and the decomposition data of the test attenuation segment, to obtain response subgraph data;

[0034] assigning a dynamic weight of defect sensitivity to the response subgraph data based on the pressure compensation estimation data, to obtain dynamic weight response map data;

[0035] performing acyclic optimization of abnormal responses based on the dynamic weight response map data, to obtain the defect response feature map data.

[0036] Optionally, the quantifying leakage risk based on the defect response feature map data to obtain leakage risk quantification data comprises:

[0037] performing pressure attenuation feature mapping based on the defect response feature map data, to obtain attenuation feature data;

[0038] extracting abnormal leakage tuples based on the attenuation feature data, to obtain leakage tuple data;

[0039] performing micropore permeation-stress deformation coupling quantification based on the leakage tuple data, to obtain the leakage risk quantification data.

[0040] Optionally, the performing micropore permeation-stress deformation coupling quantification based on the leakage tuple data to obtain leakage risk quantification data comprises:

[0041] calculating the unsteady pressure attenuation deviation and the deformation amplitude deviation of the seminal vesicle based on the leakage tuple data, to obtain dynamic attenuation deviation data and deformation amplitude deviation data respectively;

[0042] analyzing tiny micropore leakage features based on the dynamic attenuation deviation data and the deformation amplitude deviation data, to obtain micropore leakage analysis data;

[0043] establishing a micropore permeation-deformation response coupling model based on the micropore leakage analysis data, to obtain coupling response matrix data;

[0044] performing high-sensitivity defect risk threshold compensation regulation based on the coupling response matrix data, to obtain the leakage risk quantification data.

[0045] Optionally, the outputting a qualified or unqualified judgment based on the leakage risk quantification data to obtain an online tightness detection result, so as to realize high-precision dynamic tightness detection of a condom production line, comprises:

[0046] identifying high-risk defect clusters based on the leakage risk quantification data, to obtain defect core cluster data;

[0047] Based on the defective core cluster data, the vacuum test accuracy level judgment threshold is optimized to obtain the judgment threshold optimization data;

[0048] Based on the judgment threshold, the optimized data generates online sealing test results to drive the production line sorting mechanism to dynamically reject unqualified products, thereby achieving high-precision online dynamic sealing test of the condom production line.

[0049] Furthermore, to achieve the above objectives, the present invention also provides a vacuum attenuation sealing performance testing system for a condom production line, the system comprising:

[0050] The risk classification module is used to obtain the micropore parameters of the latex polymer material of the condom to be tested and the product structure contour data, and to classify the sealing risk area according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip, so as to obtain the data of the risk area to be tested.

[0051] The attenuation decomposition module is used to acquire the accuracy requirement data for the production line sealing test; it decomposes the test accuracy requirement data and the data of the risk area to be inspected into attenuation segments in the vacuum test unit to obtain test attenuation segment decomposition data.

[0052] The deviation compensation module is used to acquire the cavity background pressure fluctuation data of the current vacuum test system, and to perform pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data.

[0053] The quantification and judgment module is used to construct a defect sensitivity response map based on the pressure compensation estimation data to obtain defect response feature map data; and to quantify leakage risk based on the defect response feature map data to obtain leakage risk quantification data.

[0054] The results output module is used to determine whether a product is qualified or unqualified based on the leakage risk quantification data, and to obtain the online sealing test results, so as to realize high-precision dynamic sealing test of the condom production line.

[0055] This invention provides a vacuum attenuation sealing performance testing method for condom production lines. By introducing latex polymer micropore parameters and reservoir structure parameters, this method can accurately identify and quantify pressure changes caused by the inherent permeability of the material and the elastic deformation of the reservoir under vacuum. Utilizing pressure deviation compensation estimation technology, background noise generated by these non-defect factors is separated from the total pressure attenuation signal, thus avoiding misjudgments or missed detections caused by the inability of traditional vacuum attenuation methods to distinguish between actual pinhole leaks and material structure interference. This significantly improves the detection rate of minute defects. The sealing risk area division based on porosity distribution and structural parameters allows the test to focus on stress concentrations such as the reservoir, rather than a homogeneous overall scan. For critical components prone to deformation, the system, combined with the attenuation segment decomposition of the vacuum testing unit, can set differentiated detection windows and sensitivity thresholds for the physical characteristics of different areas, achieving precise capture of potential leak points. By acquiring and dynamically compensating for real-time fluctuations in the chamber's background pressure, this method effectively offsets the impact of external factors such as ambient temperature changes, equipment vibration, or vacuum pump fluctuations on the test results. This real-time dynamic load compensation mechanism ensures the stability of the testing benchmark under complex conditions of high-speed continuous production, guaranteeing the consistency and reliability of batch product testing results. By constructing a defect sensitivity response map and quantifying leakage risk, the originally abstract pressure curve is transformed into a concrete risk value. This not only provides scientific and traceable data support for determining whether a product is qualified or unqualified, but also guides the production line to conduct more refined quality control through risk classification, such as distinguishing between minor defects and serious leaks, providing data feedback for subsequent production process optimization, and realizing a shift from passive rejection to proactive quality control. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating an embodiment of the vacuum attenuation sealing test method for a condom production line according to the present invention.

[0057] Figure 2 This is a structural block diagram of an embodiment of the vacuum attenuation sealing performance testing system for condom production lines according to the present invention.

[0058] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0059] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0060] Reference Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the vacuum attenuation sealing test method for a condom production line according to the present invention.

[0061] In one embodiment, the vacuum attenuation sealing test method on the condom production line includes:

[0062] Step S100: Obtain the micropore parameters and product structure contour data of the latex polymer material of the condom to be tested, and divide the sealing risk area according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip to obtain the risk area data to be tested.

[0063] The condom under test can be an unsealed or sealed finished condom in the sealing test stage of the production line. It can be used as the test object for vacuum attenuation sealing test, and its sealing integrity needs to be determined by pressure change response. In an exemplary embodiment, the condom under test can be fed into the vacuum test chamber by an automated conveying mechanism for sealing clamping and testing. The micropore parameters of the latex polymer material can be a set of physical parameters describing the inherent micropore structure characteristics of the latex material used in the condom. It can be used to quantify the gas permeation behavior caused by the gaps between polymer chains and distinguish between real leakage and false signals caused by material permeability. Furthermore, the micropore parameters of the latex polymer material can be obtained through material batch testing, laboratory gas permeability testing, or by calling historical process databases. For example, the micropore parameters of the latex polymer material can include, but are not limited to, porosity, average pore size, and permeability coefficient. The product structural contour data can be digital characterization data of the three-dimensional geometric shape of the condom, which can be used for the identification and structural stress modeling of key parts such as the reservoir tip. In a specific embodiment, the product structural contour data can be obtained through optical scanning, laser profilometer, or CAD model import. The micropore porosity distribution of latex can be considered as the non-uniform distribution of micropore density in the spatial dimension of latex materials. It can be used to reflect the differences in the air permeability of materials in different regions and provide a basis for risk area delineation.

[0064] The reservoir tip structural parameters can be a set of parameters describing the geometry and mechanical properties of the reservoir tip of the condom. These parameters can be used to model the effect of elastic deformation in this region on the intracavity pressure under vacuum. In one embodiment, the reservoir tip structural parameters can be extracted based on product structural contour data or obtained through finite element simulation inversion. Furthermore, the reservoir tip structural parameters can include, but are not limited to, the reservoir volume, wall thickness, radius of curvature, and Young's modulus. The sealing risk area can be an area on the condom surface that is more prone to leakage or false leakage signals due to material properties or structural stress concentration. This can be used to guide the testing system to focus on key areas and improve the targeting of the test. In a specific embodiment, the sealing risk area can be combined with the latex micropore porosity distribution and the reservoir tip structural parameters to jointly determine the spatial location and weight of the risk area. The data of the risk area to be tested can be structured data output after the sealing risk area is divided, including area coordinates, risk level, and physical characteristic labels. This data can be used as input for the subsequent decomposition of the attenuation section in the vacuum testing unit.

[0065] Obtaining the micropore parameters and structural contour data of the latex polymer material of the condom under test can be achieved by reading micropore parameters from a material database or online detection module and acquiring geometric data through a contour scanning device. Furthermore, this operation can be implemented by reading batch material parameters via RFID tags and simultaneously triggering laser contour scanning, or by calling micropore parameters through a cloud-based material library API and fusing them with a vision system to obtain structural data. This provides a dual input of material and structure for subsequent risk modeling. Delineating sealing risk areas based on the latex micropore porosity distribution and the reservoir tip structural parameters can be achieved by combining the spatial distribution of porosity with a structural stress model to calculate the contribution of false leakage in each area and delineate high-risk areas. For example, this operation can be achieved by using a weighted superposition model to fuse porosity and deformation energy density to generate a risk heat map, and using a machine learning model to cluster historical leakage data and parameter combinations to delineate risk areas. This achieves a shift from overall homogeneous detection to focusing on key areas.

[0066] Step S200: Obtain the accuracy requirement data for the production line sealing test. Decompose the test accuracy requirement data and the data of the risk areas to be inspected into attenuation sections using the vacuum test unit to obtain the test attenuation section decomposition data.

[0067] The accuracy requirements for production line sealing tests can be based on indicators such as detection sensitivity, false positive rate, and false negative rate set according to product quality standards. These can be used to constrain the granularity and sensitivity threshold configuration of the test attenuation segment decomposition. In one specific embodiment, the accuracy requirements for production line sealing tests can be read from the quality control indicators of the current production line in the production management system or process configuration file. Furthermore, this data can be dynamically adjusted in real time through the MES system interface, or preset detection level parameters can be loaded from the local configuration file. The attenuation segment decomposition data of the vacuum test unit can be a data structure that divides the entire pressure attenuation process into multiple sub-segments according to time or spatial dimensions. This can be used to support differentiated detection strategies for different risk areas or different attenuation stages. For example, the attenuation segment decomposition data of the vacuum test unit can be coordinated with the data of the risk area to be inspected, so that each attenuation segment corresponds to a specific physical area or response mechanism.

[0068] Obtaining the accuracy requirements data for production line sealing tests can be achieved by reading the current production line's quality control indicators from the production management system or process configuration file. Furthermore, this operation can be implemented in real-time through the MES system interface by acquiring dynamically adjusted accuracy requirements and loading preset detection level parameters from a local configuration file, thereby ensuring that the testing strategy conforms to actual production standards. Decomposing the test accuracy requirement data and the data of the risk areas to be inspected into vacuum test unit attenuation segments can be achieved by dividing the entire pressure attenuation process into multiple sub-time periods or sub-regional segments, each associated with a specific risk area and accuracy requirements. In an exemplary embodiment, this operation can be implemented by dividing the attenuation curve into an initial deformation segment, a steady-state permeation segment, and a leakage response segment according to the time axis, and by dividing the cavity into a reservoir segment, a cylinder segment, and a sealing segment according to spatial regions, and sampling them separately, thereby achieving the technical effect of supporting the dynamic configuration of differentiated detection windows and sensitivity thresholds.

[0069] Step S300: Obtain the cavity background pressure fluctuation data of the current vacuum test system, and perform pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data.

[0070] The cavity background pressure fluctuation data of the current vacuum testing system can be the time-series data of pressure fluctuations caused by environmental disturbances within the vacuum cavity when there is no test piece or standard part present. This data can be used to identify and compensate for the impact of external interference (such as temperature drift and pump fluctuations) on the test signal. In one specific embodiment, the cavity background pressure fluctuation data of the current vacuum testing system can be acquired in real time before or between tests using a high-precision pressure sensor. The vacuum testing system cavity can be a sealed mechanical chamber that houses the condom under test and performs vacuum attenuation testing. This chamber can provide a controllable vacuum environment, isolate external interference, and ensure the accuracy of pressure signal acquisition. Furthermore, the vacuum testing system cavity can be integrated with a vacuum pump, sealing fixture, pressure sensor, and control system. The pressure compensation estimation data can be a corrected pressure attenuation signal obtained after processing by a deviation compensation algorithm. This signal can be used to eliminate non-defect pressure change components caused by material permeability, structural deformation, and environmental disturbances. For example, the pressure compensation estimation data can rely on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to achieve dynamic compensation.

[0071] Acquiring the cavity background pressure fluctuation data of the current vacuum testing system can be achieved by collecting the cavity pressure change sequence over time under conditions without samples or with standard samples. Furthermore, this operation can be implemented by performing a cavity sampling cycle before each test and inserting a background calibration cycle at fixed intervals during continuous production, thereby achieving the technical effect of establishing a benchmark model for environmental interference for subsequent compensation. Pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the decomposed test attenuation segment data can be performed by using a background fluctuation model to denoise and correct the measured attenuation signal, separating the pressure changes caused only by defects. In an exemplary embodiment, this operation can be achieved by using a sliding window adaptive filtering algorithm to subtract background drift in real time and using a state-space model to jointly estimate and compensate for multi-source interference, thereby achieving the technical effect of eliminating the interference of external factors such as temperature, vibration, and pump fluctuations on the test results.

[0072] Step S400: Construct a defect sensitivity response map based on the pressure compensation estimation data to obtain defect response characteristic map data. Quantify leakage risk based on the defect response characteristic map data to obtain leakage risk quantification data.

[0073] The defect sensitivity response map can be a mapping spectrum characterizing the sensitivity of different regions to actual leaks. It can be used to guide the detection system to set higher sensitivity thresholds in highly sensitive areas, improving the ability to detect minute defects. In a specific embodiment, the defect sensitivity response map may include, but is not limited to, statistical response maps based on historical defect samples, theoretical sensitivity maps based on physical simulation, and dynamic response maps based on real-time compensated signals. The defect response feature map data can be a feature response distribution map formed by mapping the compensated pressure signal to the spatial or temporal dimension. It can be used to visualize the location and intensity of potential leak points, providing input for risk quantification. For example, the defect response feature map data may include, but is not limited to, spatiotemporal joint response maps, regional sensitivity heat maps, and attenuation slope distribution maps. The leakage risk quantification data can be a numerical leakage risk index calculated based on the defect response feature map. It can be used to support graded judgments (such as qualified, minor defects, and serious leaks) to achieve refined quality control. Furthermore, the leakage risk quantification data may include, but is not limited to, probabilistic risk scores, physical equivalent pore size values, and cumulative deviation indices.

[0074] Constructing a defect-sensitive response map based on pressure compensation estimation data can involve mapping the compensated pressure signal along spatial or temporal dimensions to generate a distribution map reflecting leakage sensitivity. In an exemplary embodiment, this operation can be achieved by inverse mapping to associate time decay segments with physical regions to generate a heat map, and by using a convolutional neural network to extract local sensitive feature maps from the compensated signal. This achieves the technical effect of transforming an abstract pressure curve into interpretable spatial response features. Leakage risk quantification based on the defect response feature map data can be achieved by numerically converting features such as peak values, areas, or slopes in the response map to output a risk score or equivalent leakage index. Furthermore, this operation can be achieved by looking up a pre-set response-risk mapping table and fusing multi-dimensional features into a single risk index through a regression model, thereby achieving a leap from qualitative judgment to quantitative classification.

[0075] Step S500: Based on the leakage risk quantification data, a pass or fail judgment is output to obtain the online sealing test result, so as to realize high-precision dynamic sealing test of the condom production line.

[0076] The online sealing test result can be the final output of the condom sealing judgment conclusion and its associated data, which can be used for automatic sorting, quality traceability, and process feedback. In one specific embodiment, the online sealing test result can be generated by judging leakage risk quantification data according to preset threshold rules. The pass / fail judgment output based on the leakage risk quantification data can be achieved by comparing the quantified risk value with a preset threshold, outputting a binary or multi-level judgment result. For example, this operation can employ multi-threshold hierarchical judgment (pass / rework / scrap), combined with dynamic adjustment of the judgment boundary using confidence intervals to adapt to operational drift, thereby achieving the technical effect of generating structured test conclusions that can be used for automatic sorting and quality traceability.

[0077] Taking online testing on a high-speed condom production line as an example, the vacuum decay sealing test method on the condom production line in this embodiment can be implemented on an automated production line that processes 3,000 condoms per minute. The system first retrieves the micropore parameters of the current latex from the material batch database and obtains the three-dimensional structure of each product using a high-speed laser profilometer. Combining the high curvature characteristics and porosity distribution of the reservoir tip, the algorithm automatically marks the last 15 mm area as a high-risk zone. According to the accuracy requirements set by the standard, the system decomposes the vacuum decay process into three time periods: the first 200 milliseconds are used to compensate for reservoir tip deformation, the middle 300 milliseconds are used to monitor steady-state permeation, and the last 100 milliseconds are used to focus on the response to minor leaks. At the same time, the cavity base pressure is calibrated every 5 seconds to compensate for ambient temperature fluctuations in real time. Finally, the defect sensitivity response map shows that a certain product has an abnormal peak value in the reservoir tip area. The risk value is quantified to be 0.82, which exceeds the minor defect threshold of 0.7 but is lower than the scrap threshold of 0.9. The system marks it as a reworkable product and records the data for process optimization.

[0078] In one embodiment, sealing risk areas are divided based on the latex micropore porosity distribution and the structural parameters of the seminal vesicle, resulting in data on the risk areas to be inspected, including:

[0079] Based on the latex micropore porosity distribution data and seminal vesicle stress parameters, the defect-pressure response coupling relationship was analyzed to obtain the defect-pressure coupling matrix data.

[0080] Obtain statistical data on the distribution of defects in the condom production process, and calculate the frequency of risk occurrence based on the statistical data to obtain defect risk frequency data;

[0081] Based on the defect-pressure coupling matrix data and defect risk frequency data, a digital twin inspection domain is constructed to obtain the inspection risk area data.

[0082] The reservoir stress parameters can be a set of mechanical parameters describing the internal stress state of the condom reservoir under vacuum negative pressure. These parameters can be used to quantify the contribution of structural deformation to intracavity pressure changes, supporting defect-pressure response coupling modeling. In an exemplary embodiment, the reservoir stress parameters can be calculated using finite element simulation combined with product structural profile data, or derived from material constitutive models and measured deformation inversion. Furthermore, the reservoir stress parameters can include, but are not limited to, principal stress directions, equivalent stress values, and strain energy density. The defect-pressure response coupling relationship can be the pressure attenuation characteristic mode caused by a real leakage defect under specific material micropore distribution and structural stress conditions. This can be used to establish a mapping mechanism between the presence or absence of a defect and pressure signal changes, distinguishing between real and pseudo-leakage. For example, the defect-pressure response coupling relationship relies on the latex micropore porosity distribution data and reservoir stress parameters for joint analysis, forming a computable coupling model. The defect-pressure coupling matrix data can be a data entity that structurally expresses the defect-pressure response coupling relationship in matrix form. It can be used to provide a multi-dimensional feature space, supporting the quantitative characterization of leakage sensitivity in different areas of the digital twin inspection domain. The formula for constructing the defect-pressure coupling matrix is ​​used to establish the relationship between material microstructures and pressure response, and is the basis for risk area delineation. The formula is as follows:

[0083]

[0084] In the formula, The defect-pressure coupled response value represents the specific defect type at region i. The higher the value, the greater the risk, indicating sensitivity to changes in stress. The micropore porosity of latex describes the microscopic air permeability of the material at position i, and is used to quantify the basic pressure loss caused by the permeation of polymer micropores. The elastic modulus of latex reflects the stiffness of the material at position i and is used to correct the resistance to deformation under different thicknesses or curing degrees. These are the stress parameters of the seminal vesicle, corresponding to the structural parameters of the seminal vesicle, describing the degree of stress concentration in a specific area (such as the tip of the seminal vesicle) under vacuum negative pressure. is the volumetric deformation coefficient, a geometric parameter used to convert stress into volumetric change; For defect risk frequency, it is a probability value based on production process statistics, used to weight the types of defects that are historically prone to problems; , , The weighting coefficients adjust the contribution ratios of micropore permeability, structural deformation, and historical statistical data to the overall risk, respectively. This formula, through coupled analysis, combines physical properties (porosity, stress) with statistical data (defect frequency) to generate the initial matrix required for the digital twin model, used to identify high-risk areas requiring focused monitoring. Based on latex micropore porosity distribution data and reservoir stress parameters, the defect-pressure response coupling relationship is analyzed, which can combine material permeability characteristics with structural mechanical response to simulate or derive the differences in pressure decay curves under different defect states. Furthermore, this operation can be achieved through multiphysics coupled simulation to generate a response sample library and training a regression mapping model based on measured leakage samples and compensated signals, thereby establishing a response criterion that can distinguish between actual leakage and material / structural disturbances.

[0085] The statistical data on the distribution of defects in the condom production process can be the frequency and type distribution data of various sealing defects recorded in the historical production process, showing their spatial location on the product. This data can be used to reflect the statistical patterns of weak points in the actual process, providing empirical basis for risk frequency calculation. In this embodiment, the statistical data on the distribution of defects in the condom production process can be aggregated from a quality traceability system, automatic visual inspection records, or a manual sampling database. Defect risk frequency data can be a probabilistic index of leakage defects occurring in each area, calculated based on the statistical data on defect distribution. This data can be used to integrate process experience information and enhance the practical relevance of risk area classification. For example, defect risk frequency data can include, but is not limited to, Poisson event frequency, Bayesian posterior probability, and sliding window local density. Obtaining the statistical data on the distribution of defects in the condom production process can be achieved by extracting historical defect spatial location and type records from the production quality management system. Furthermore, this operation can be implemented by pulling defect logs from the MES in real time via API and periodically synchronizing aggregated reports from the offline quality inspection database, thereby introducing actual process experience and improving the practicality of risk area classification. Calculating the frequency of risk occurrence based on defect distribution statistics can involve normalizing or probabilistically modeling the number of defects occurring in each region and outputting a frequency index. In one specific embodiment, this operation can be achieved by using kernel density estimation to generate a continuous risk field and using time-weighted moving averages to highlight recent defect trends, thereby transforming empirical data into quantifiable risk weights.

[0086] A digital twin inspection domain can be a virtual inspection area mapping model that integrates a physical mechanism model and historical defect statistics. It can be used to dynamically generate high-priority inspection areas, guiding vacuum attenuation testing to focus on the most likely and most difficult-to-identify failure points. In this embodiment, the digital twin inspection domain is constructed by jointly driving defect-pressure coupling matrix data and defect risk frequency data, outputting inspection risk area data. Constructing the digital twin inspection domain based on defect-pressure coupling matrix data and defect risk frequency data can involve weighted fusion of physical response sensitivity and historical defect probabilities to generate a virtual inspection priority map. Furthermore, this operation can be achieved by superimposing the coupling matrix and frequency map using tensor fusion to generate a comprehensive risk domain, and then using a graph neural network to model the inter-regional correlations before outputting the inspection domain. This enables intelligent inspection area focusing driven by multi-source information.

[0087] Taking intelligent focusing detection based on digital twins as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be as follows: In a high-end condom production line, the system first calls the porosity distribution map of the current batch of latex and combines it with the equivalent stress cloud map obtained from the finite element simulation of the reservoir to analyze that when a 0.5-micron pinhole exists, the pressure attenuation rate in the reservoir area will be 1.8 times faster than that in the body. At the same time, 237 leakage defects are extracted from the quality inspection database of the past three months, and it is found that 82% are concentrated in the range of 8 to 12 mm from the top. Based on this, the system calculates the defect risk frequency of this range as 0.63. Then, the regional sensitivity weight in the defect-pressure coupling matrix is ​​fused with the frequency value of 0.63 to construct a digital twin inspection domain and mark this ring as an ultra-high priority detection area. In the subsequent vacuum test, the attenuation segment corresponding to this area is assigned a longer observation window and a lower judgment threshold, successfully capturing a 0.3-micron pinhole that was missed by the traditional method.

[0088] In one embodiment, a digital twin inspection domain is constructed based on defect-pressure coupling matrix data and defect risk frequency data to obtain inspection risk area data, including:

[0089] Based on the defect-pressure coupling matrix data and defect risk frequency data, virtual vacuum test field coverage features and pressure attenuation transmission features are extracted, resulting in virtual test field coverage feature data and pressure attenuation transmission feature data, respectively.

[0090] The virtual vacuum test field coverage feature data can be extracted from the defect-pressure coupling matrix and defect risk frequency, representing the effective coverage range of the digital twin model on the physical test space. This data can be used to define the completeness of the twin model's spatial mapping to the real test cavity, avoiding detection blind spots. In an exemplary embodiment, the virtual vacuum test field coverage feature data can be spatially resolved by extracting the main coverage direction features through principal component analysis in conjunction with the context. Further, the virtual vacuum test field coverage feature data can include, but is not limited to, geometric coverage, response sensitivity coverage domain, and risk weight coverage density. The pressure attenuation and transmission feature data can be feature data describing the dynamic propagation and attenuation of the pressure signal within the condom cavity over time. This data can be used to characterize the temporal correlation of the leakage signal between different regions, supporting spatiotemporal coupling modeling. For example, the pressure attenuation and transmission feature data can be dynamically modeled by encoding the attenuation temporal pattern using LSTM in conjunction with the context. In a specific embodiment, the pressure attenuation and transmission feature data can include feature vectors corresponding to deformation-dominated attenuation segments, steady-state penetration segments, and abrupt leakage segments. Based on defect-pressure coupling matrix data and defect risk frequency data, virtual vacuum test field coverage feature extraction and pressure attenuation transfer feature extraction can be performed. This allows for the analysis of implicit patterns in the coupling matrix and frequency data from two dimensions: spatial coverage integrity and temporal attenuation dynamics. Furthermore, this operation can be achieved by employing a graph convolutional network to simultaneously learn spatial coverage and temporal transfer features on a gridded condom surface, thereby constructing a digital twin with spatiotemporal resolution capabilities.

[0091] Based on the virtual test field coverage feature data and pressure attenuation transmission feature data, twin defect response maps are generated from the defect-pressure coupling matrix data to obtain twin defect response map data.

[0092] The twin defect response map data can be a spatial distribution map reflecting the defect response predicted by the digital twin model, generated based on the virtual test field coverage features and pressure attenuation transfer features. It can be used as a benchmark for virtual-to-real comparison, identifying systematic deviations between simulation and actual measurements. In this embodiment, the twin defect response map data can be reconstructed from defect-pressure coupling matrix data after feature fusion, serving as a prerequisite input for virtual-to-real calibration. Generating the twin defect response map from the defect-pressure coupling matrix data based on the virtual test field coverage feature data and pressure attenuation transfer feature data can involve re-injecting the extracted spatiotemporal features into the original coupling matrix to reconstruct a response distribution closer to real test conditions. Furthermore, this operation can be achieved by weighted fusion of coverage and transfer features using an attention mechanism to reconstruct the response map, thereby generating a high-fidelity twin response benchmark suitable for virtual-to-real comparison.

[0093] The virtual-real mismatch region detection was performed on the twin defect response map data to obtain virtual-real calibration map data. The virtual-real mismatch region detection included environmental temperature fluctuation mismatch detection, cavity leakage background anomaly detection, and latex elastic creep deviation detection.

[0094] The virtual-real mismatch region can be an area or period where there is a significant deviation between the twin defect response map and the actual test response. It can be used to identify the areas affected by non-ideal factors that need calibration, thereby improving model fidelity. In a specific embodiment, the virtual-real mismatch region may include an environmental temperature fluctuation mismatch region, a cavity leakage background anomaly region, and a latex elastic creep deviation region. Environmental temperature fluctuation mismatch detection can be a process of identifying the difference between the simulation and measured responses caused by changes in material permeability or sensor drift due to changes in environmental temperature. It can be used to separate the interference components of temperature disturbances on the pressure attenuation signal. In this embodiment, environmental temperature fluctuation mismatch detection can be combined with context by comparing the residuals of the twin model prediction value and the measured value in the steady-state segment, and attributing the cause by combining cavity temperature sensor data. Cavity leakage background anomaly detection can determine whether there are abnormal background noises caused by minor leaks or seal degradation in the vacuum testing system itself. It can be used to prevent equipment condition deterioration from being misjudged as product defects. For example, cavity leakage background anomaly detection can be combined with context by comparing the deviation of the current background decay rate from the historical baseline in no-load or standard part testing. Latex elastic creep deviation detection can identify additional pressure changes caused by time-dependent deformation (creep) of latex materials under continuous negative pressure, and can be used to compensate for the interference of time-varying mechanical behavior of materials on short-term test signals. In an exemplary embodiment, latex elastic creep deviation detection can predict long-term deformation trends based on the material creep model in conjunction with context, and perform residual analysis with the measured initial segment attenuation slope. Real-virtual mismatch zone detection of twin defect response map data can be performed by performing residual analysis between the twin response map and the measured compensated pressure signal to identify systematic deviation areas. Furthermore, this operation can be achieved by using multi-channel residual thermograms to correspond to three types of mismatch sources: temperature, background, and creep, thereby locating and classifying model-reality inconsistencies caused by non-defect factors. Real-virtual calibration map data can be a space-time calibration weight map generated after real-virtual mismatch zone detection, used to correct twin model biases, and can be used to dynamically adjust the digital twin response to make it closer to real test conditions. In this embodiment, the virtual and real calibration map data can be fused together with three types of mismatch detection results, and the output serves as the calibration basis for subsequent time-series constraint injection.

[0095] The production line cycle time constraint is injected into the virtual and real calibration diagram data to obtain the time constraint diagram data;

[0096] The production line cycle time constraint can be the time window and synchronous triggering logic for each product's stay at the test station in high-speed continuous production. This can be used to ensure that the response calculation of the digital twin model is strictly aligned with the production line rhythm. For example, the production line cycle time constraint can obtain cycle signals and position synchronization pulses from a PLC or motion control system. The time constraint graph data can be a spatiotemporal joint constraint graph mapping the production line cycle constraints onto spatial risk areas. This can be used to limit the detection of high-risk areas to be completed within the effective cycle window, ensuring real-time performance. In a specific embodiment, the time constraint graph data can include a time window mask map, a synchronous trigger marker map, and a cycle fault tolerance boundary map. Injecting production line cycle time constraints into the virtual-real calibration graph data can be achieved by embedding the time window information of the production line cycle into the spatial dimension of the calibration graph, forming a spatiotemporal joint constraint. Furthermore, this operation can be implemented by applying a sliding window mask to the calibration graph on the time axis, thereby ensuring that subsequent region extraction meets the real-time requirements of high-speed production.

[0097] High-risk connected subdomains are extracted from the time-series constraint graph data to obtain the data of the risk areas to be inspected.

[0098] High-risk connected subdomains can be a set of spatiotemporally continuous connected regions with risk values ​​exceeding a threshold, extracted from the calibrated risk region under the guidance of a temporal constraint graph. These subdomains can be used to output the dynamic detection region ultimately used for vacuum attenuation testing. In an exemplary embodiment, high-risk connected subdomains may include isolated high-risk areas, axially extending band-shaped areas, and circumferentially closed ring-shaped areas. Extracting high-risk connected subdomains based on temporal constraint graph data can be achieved by performing connected component analysis on the calibrated risk region, extracting significantly high-risk clusters, provided that the beat time window is met. Furthermore, this operation can be implemented using a spatiotemporal connected component labeling algorithm (3Dfloodfill) to extract subdomains, thereby outputting the dynamic focusing region ultimately used for differential detection.

[0099] Taking four-dimensional driven intelligent risk positioning as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be as follows: On a high-speed production line with 4,000 condoms per minute, the system extracts the high pressure attenuation transmission sensitivity of the reservoir area based on the defect-pressure coupling matrix. At the same time, the virtual test field coverage features show that there is a low coverage blind zone in the middle of the tube. The generated twin defect response map predicts that a certain product should have an attenuation rate of 0.15 kPa / s at the top 10 mm, but the actual measurement is only 0.09 kPa / s. The virtual-real mismatch detection module identifies that the deviation is mainly due to the 3°C increase in workshop temperature that day, which caused the latex to permeate. The gas content was enhanced and marked as a mismatch zone due to ambient temperature fluctuations; the cavity background test confirmed that there was no leakage abnormality in the equipment, and the creep test found that the batch of latex showed slight relaxation after 200 milliseconds; the system generated a virtual-real calibration map based on this and applied a compensation of +0.06 kPa / s to the seminal vesicle area; then the production line cycle time constraint (the single product test window is 350 milliseconds) was injected to form a time series constraint map; finally, the high-risk connected subdomain extraction module output a connected subdomain located in the axial range of 8-12 mm and lasting for 250 milliseconds as the key point to be tested, guiding the subsequent vacuum decay test to use the ultra-high sensitivity threshold within this time window.

[0100] In one embodiment, the test accuracy requirement data and the data of the risk area to be inspected are decomposed into attenuation segment data by a vacuum test unit to obtain test attenuation segment decomposition data, including:

[0101] The topology of the vacuum attenuation test section is divided based on the test accuracy requirement data and the data of the risk area to be inspected, and the test section topology division data is obtained.

[0102] The vacuum attenuation test segment topology partitioning can be a method of dividing the vacuum attenuation test process into logically related structured test segments based on physical regions or functional characteristics. This can provide a structural basis for subsequent differentiated modeling based on regional characteristics, breaking the limitations of traditional holistic pressure monitoring. In an exemplary embodiment, the vacuum attenuation test segment topology partitioning can combine accuracy constraints and risk distribution to divide the test space or time domain into multiple topological units with independent physical semantics. Furthermore, the vacuum attenuation test segment topology partitioning can use graph segmentation algorithms to divide the condom surface into connected risk regions and divide the attenuation time axis into deformation segments, permeation segments, and leakage response segments according to preset rules, thereby realizing the transformation from homogeneous holistic detection to structured segmented detection. The test segment topology partitioning data can be structured data output after vacuum attenuation test segment topology partitioning, containing the spatial location, boundaries, and attribute labels of each test segment, which can be used as input for gas permeation characteristic extraction and key defect node extraction. For example, the test segment topology partitioning data can be coordinated with test accuracy requirement data and data of the risk regions to be inspected to jointly determine the granularity and semantics of the topological units. In one specific embodiment, the topology division of the vacuum attenuation test section may include a region topology based on geometric contours, a functional topology based on risk level, a permeation topology based on material properties, etc.

[0103] Based on the test section topology division data, gas permeation characteristics and key defect nodes are extracted, resulting in gas permeation characteristic data and key defect node data, respectively.

[0104] The gas permeation characteristic data can be a set of quantitative parameters describing the gas permeation behavior of the latex material within each test section due to its microporous structure. This data can be used to characterize the source of non-defective, slowly varying pressure signals and to construct a background noise model. In one exemplary embodiment, the gas permeation characteristic data can be calculated based on the microporous parameters of the latex polymer material and the geometric characteristics of the test section. Further, the gas permeation characteristic data can include local permeability, cumulative permeation over time, and permeation rate gradient. The critical defect node data can be the coordinates and attribute information of structural weak points or stress concentration points identified in the topologically divided test section. This data can be used to identify the potential locations where actual leaks are most likely to occur and guide the setting of a high-sensitivity detection window. In one specific embodiment, the critical defect node data can be extracted through joint analysis of product structural contour data and a mechanical simulation model. For example, the critical defect node data can include edge sealing seam nodes, reservoir curvature transition nodes, and wall thickness abrupt change nodes.

[0105] Gas permeation characteristic extraction based on test segment topology data can be achieved by calculating local gas permeation behavior for each test segment, combining its material micropore parameters and geometric features. Furthermore, this operation can be implemented by calculating the steady-state permeation flux of each segment using Fick's diffusion law and fitting a segment-level permeation rate model using historical batch data, thus explicitly modeling the slowly varying pressure signal caused by the material's inherent permeability. Key defect node extraction based on test segment topology data can be achieved by marking locations of geometric abrupt changes, stress concentrations, or process weaknesses in the topology results as key nodes. Further, this operation can be achieved by automatically identifying nodes in the reservoir transition zone based on curvature change thresholds and extracting nodes in high-strain regions through finite element stress simulation, thereby locating high-risk points most likely to cause actual leaks and improving the targeting of detection.

[0106] Based on the gas permeation characteristic data and key defect node data, the pressure decay time-series dependency processing of the test section topology partitioning data is performed to obtain time-series dependency data.

[0107] The time-dependent data can be characteristic data describing the interrelationship and dynamic evolution of the pressure decay response of each test segment in the time dimension. It can be used to capture the temporal separation characteristics of the slowly varying signals caused by material infiltration and structural deformation and the abrupt signals of actual leakage. In a specific embodiment, the time-dependent data can be fused with gas infiltration characteristic data and key defect node data to establish a pressure response model with spatiotemporal resolution. Applying pressure decay time-dependent processing to the test segment topology partitioning data based on gas infiltration characteristic data and key defect node data can fuse the slowly varying characteristics of infiltration and the abrupt characteristics of defects, modeling the dynamic coupling relationship of pressure decay in each test segment over time. Furthermore, this operation can be achieved by constructing a state-space model to describe the superimposed decay process of infiltration and leakage, and using sliding window correlation analysis to extract the inter-segment time-dependent patterns, thereby separating the instantaneous response of actual leakage from the delayed response caused by materials / structure.

[0108] A dynamic decay response map is constructed based on time-dependent data to obtain the decomposed data of the test decay segment.

[0109] The dynamic decay response map, constructed based on time-dependent data, is a dynamic mapping reflecting the pressure decay sensitivity of different test segments over time. It can be used to achieve spatiotemporal decoupling between real leak signals and pseudo-leakage backgrounds, supporting refined defect identification. In an exemplary embodiment, the dynamic decay response map may include a time-region joint response map, a penetration-leakage separate response map, and a multi-scale decay feature map. Constructing the dynamic decay response map based on time-dependent data involves mapping time-dependent features into a visualized or computable response spectrum, characterizing the leakage sensitivity of each region at different times. Furthermore, this operation can display the time-region two-dimensional response intensity in the form of a heatmap and generate multi-channel time-series feature maps for input into machine learning models, thereby generating a detection benchmark with spatiotemporal resolution capabilities to support subsequent accurate compensation and judgment.

[0110] Taking the online detection of high-sensitivity condoms as an example, the vacuum decay sealing test method on the condom production line in this embodiment can be as follows: The system receives the test accuracy requirements of a batch of condoms (needs to detect pinholes ≥5 micrometers) and the data of the risk areas to be inspected (10 mm at the end of the reservoir tip is considered high risk). First, the vacuum decay test section topology is divided, dividing the product into three topological units: the body section, the transition section, and the reservoir tip section. Then, the gas permeation characteristic data of each section (e.g., the reservoir tip section has high permeability due to its thin wall) and the key defect node data (e.g., the curvature change point in the transition section is marked as a node) are extracted. Next, the pressure decay time-series dependency processing is performed, and it is found that the reservoir tip section has a slow-changing permeability in the first 300 milliseconds, while the transition section shows an abnormal steep drop after 400 milliseconds. Finally, a dynamic decay response diagram is constructed, which clearly shows that the transition section has a high-sensitivity response in the later stage. Based on this, the system determines that the product has micro-pinholes rather than material interference, outputs a non-conforming conclusion, and records the node position for process traceability.

[0111] In one embodiment, the cavity background pressure fluctuation data of the current vacuum testing system is acquired, and pressure deviation compensation estimation is performed based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data, including:

[0112] Based on the cavity background pressure fluctuation data and test attenuation segment decomposition data of the current vacuum testing system, multi-cycle pressure attenuation superposition calculation is performed to obtain attenuation superposition data.

[0113] The pressure deviation kernel distribution is calculated based on the attenuation superposition data to obtain the local pressure deviation intensity data;

[0114] Based on the current background pressure data of the vacuum test chamber, the ultimate vacuum degree of the chamber and the response error of the pressure sensor are calculated, and the ultimate vacuum data and sensor error data are obtained respectively.

[0115] Based on the ultimate vacuum data and sensor error data, the dynamic pressure compensation amount is extrapolated from the local pressure deviation intensity data to obtain the pressure compensation estimate data.

[0116] The multi-cycle pressure attenuation superposition calculation can be a signal processing procedure that aligns and superimposes the pressure attenuation signals of corresponding attenuation segments in multiple test cycles. This can enhance the repetitive pressure change characteristics caused by actual leaks and suppress random environmental noise. In an exemplary embodiment, the multi-cycle pressure attenuation superposition calculation can align and average the pressure signals of the same attenuation segment in multiple consecutive test cycles. Furthermore, the multi-cycle pressure attenuation superposition calculation can be achieved by using moving average superposition or time-aligned superposition based on cross-correlation, thereby improving the signal-to-noise ratio of the systematic signal and reducing the impact of random disturbances. The superimposed attenuation data can be a synthetic pressure attenuation signal generated after the multi-cycle pressure attenuation superposition calculation, which can be used as input for pressure deviation kernel distribution calculation, improving the stability of systematic deviation identification.

[0117] Pressure deviation kernel distribution calculation can be a process of statistically modeling attenuation superposition data in the time-amplitude domain to extract systematic deviation patterns. It can be used to identify non-random pressure change characteristics caused by material permeability or structural deformation. In a specific embodiment, pressure deviation kernel distribution calculation can statistically model the superimposed pressure attenuation curve to extract its deviation distribution characteristics in the time-amplitude space. For example, pressure deviation kernel distribution calculation can be achieved by fitting the deviation distribution using a Gaussian mixture model and extracting local time-frequency features through wavelet transform, thereby separating non-random pseudo-leakage signal patterns caused by materials or structures. Local pressure deviation intensity data can be quantitative data characterizing the amplitude and persistence of systematic pressure deviations within different attenuation periods. It can be used to reflect the degree of interference from non-defect factors in each test segment and for subsequent compensation calculation. Furthermore, local pressure deviation intensity data can include, but is not limited to, material permeation-dominated deviations, structural deformation-dominated deviations, and mixed-coupling deviations.

[0118] The current vacuum test chamber background pressure data can be the raw pressure time-series data collected by the vacuum test chamber in a state without the tested object, which can be used as the basic input for calculating the chamber's ultimate vacuum and analyzing sensor errors. In an exemplary embodiment, the current vacuum test chamber background pressure data can be from the same source as the chamber's background pressure fluctuation data, but with an emphasis on static baseline values ​​rather than dynamic fluctuation characteristics. The chamber's ultimate vacuum calculation can be a process of evaluating the lowest stable pressure level that the vacuum system can achieve based on the background pressure data, which can be used to determine the physical capability boundary of the system and constrain the reasonable range of the compensation amount. In a specific embodiment, the chamber's ultimate vacuum calculation can analyze the convergence value and lower limit of fluctuation of the background pressure in the steady state stage to estimate the lowest pressure that the system can reach. Furthermore, the chamber's ultimate vacuum calculation can be achieved by taking the lowest percentile value of the long-term steady-state pressure as the limit estimate and combining it with the pump performance curve to back-calculate the theoretical limit, thereby setting a physical feasibility boundary for the compensation amount.

[0119] Ultimate vacuum data can be the theoretically lowest pressure value achievable by the cavity under current operating conditions, and can be used to prevent compensation algorithms from generating false corrections in physically inaccessible areas. Pressure sensor response error calculation can be a process of evaluating the measurement deviation of the pressure sensor under current temperature, frequency, and range, and can be used to quantify uncertainties introduced by the measurement link for accuracy correction. In an exemplary embodiment, pressure sensor response error calculation can utilize the known stationary characteristics of the background pressure, compare the sensor output with the theoretical value, and inversely deduce its measurement error. For example, pressure sensor response error calculation can be achieved by estimating drift error through zero-pressure calibration residuals and testing the dynamic response bandwidth using high-frequency pressure perturbations, thereby quantifying the measurement uncertainty of the sensor under current operating conditions. Sensor error data can be a quantitative indicator describing the deviation between the pressure sensor output and the true pressure, and can be used to correct reading distortions caused by sensor nonlinearity, hysteresis, or temperature drift. Furthermore, sensor error data can include, but is not limited to, zero-point drift error, gain error, and dynamic response hysteresis error.

[0120] Dynamic pressure compensation calculation can be a process of combining ultimate vacuum data, sensor error data, and local pressure deviation intensity data to calculate a real-time compensation value adapted to the current attenuation range. This can be used to generate a pressure correction amount that matches the operating conditions, achieving accurate noise reduction. In one specific embodiment, dynamic pressure compensation calculation can correct the local deviation intensity under physical limits and measurement error constraints to generate the final compensation value. Furthermore, dynamic pressure compensation calculation can be achieved by using weighted constraint optimization to solve for the compensation amount, and by using a lookup table method combined with real-time error parameter interpolation to obtain the compensation value, thus ensuring that the compensation amount is both accurate and conforms to the actual physical and measurement conditions of the system. For example, the dynamic pressure deviation compensation estimation formula can eliminate the interference of environmental fluctuations and system background noise on the test results. The formula expression is as follows:

[0121]

[0122] In the formula, The pressure compensation estimate is the amount of correction that needs to be applied to the test system at time t. This is the starting time point for testing or sampling; This is a temporary variable used during the integration process, representing the value from the start time. Any point in time between the current time t and the present time t; This data represents the pressure fluctuations within the cavity. It shows the real-time pressure changes within the vacuum cavity over time. The baseline pressure is the average pressure value of the system under ideal steady-state conditions. It is a kernel distribution function, a window function used to calculate the intensity of local pressure deviation, highlighting short-term fluctuations and suppressing long-term drift; Sensor response error is a fixed or systematic deviation caused by the accuracy and hysteresis characteristics of the pressure sensor. This is the dynamic vacuum source fluctuation term, representing the pressure pulsation component generated by the vacuum pump during its operating cycle. This formula dynamically calculates the pressure deviation caused by non-product factors in the current testing environment through superposition calculation and kernel distribution analysis of multi-cycle background fluctuations. This deviation is then eliminated in subsequent testing, ensuring the purity of the detection signal.

[0123] Taking high-stability online sealing detection and compensation as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be performed on a production line that runs continuously for 8 hours. Every 10 tests, the system performs a multi-cycle superposition calculation to align and superimpose the pressure signals of the corresponding attenuation segment of the reservoir. Then, the deviation distribution of this segment is extracted by Gaussian kernel density estimation, and the local pressure deviation intensity is obtained as 0.12 kPa / s. At the same time, the system uses the no-load background data to calculate that the current cavity ultimate vacuum is 0.8 Pa, and the sensor has a zero-point drift of +0.05 kPa. The dynamic compensation amount extrapolation module reduces the original deviation intensity by 8% accordingly and generates the final pressure compensation estimation data, which is used to correct the measured attenuation curve of the next cycle to ensure that the real leakage of 0.03 kPa / s caused by micron-level pinholes is not masked.

[0124] In one embodiment, a defect-sensitive response map is constructed based on pressure compensation estimation data to obtain defect response feature map data, including:

[0125] Based on the current background pressure data of the vacuum test chamber and the decomposed data of the test attenuation segment, the evening response sub-graph of the test attenuation day is constructed to obtain the response sub-graph data;

[0126] The current vacuum test chamber background pressure data can be the time-series data of the baseline pressure fluctuations caused by environmental and equipment factors in the vacuum test chamber during the current test cycle, under conditions of no effective leakage signal interference. This data can be used as a reference baseline for constructing the test attenuation segment response sub-graph to separate the actual response from the system's background noise. In an exemplary embodiment, the current vacuum test chamber background pressure data can be acquired in real time by a high-precision pressure sensor during the test preparation phase or under no-load conditions. The test attenuation segment response sub-graph can be a local mapping of the pressure response corresponding to a specific test attenuation segment in spatial or logical dimensions. It can be used to achieve independent modeling of attenuation behavior in different physical regions or time stages, supporting differentiated processing of local sensitivity. For example, the test attenuation segment response sub-graph can include, but is not limited to, time segment response sub-graphs, structural region response sub-graphs, and physical mechanism response sub-graphs. The response sub-graph data can be a set of structured local response information output after the test attenuation segment response sub-graph is constructed, and can be used as the input basis for dynamic weighting of defect sensitivity.

[0127] Based on the current background pressure data of the vacuum test chamber and the decomposed data of the test attenuation segment, a test attenuation segment response sub-graph is constructed. This can be achieved by dividing the total pressure response into several sub-segments according to the test attenuation segment decomposed data, and subtracting system noise from the background pressure data to generate a local response graph corresponding to each sub-segment. Furthermore, this operation can be achieved by mapping the pressure slope within each attenuation segment's time window to the response intensity of the corresponding structural region, and by using a background pressure fluctuation model to extract residuals from the sub-segment signals to generate a clean response sub-graph. This allows for refined modeling of the local pressure attenuation behavior, laying the foundation for subsequent differentiated weight assignment.

[0128] Based on the pressure compensation estimation data, dynamic weights for defect sensitivity are assigned to the response subgraph data to obtain dynamic weighted response graph data.

[0129] The dynamic weighting of defect sensitivity can be a process of assigning variable weights to each response submap based on pressure compensation estimation data, reflecting its sensitivity to the actual defect. This allows the system to dynamically adjust the detection sensitivity of different regions based on factors such as material permeability and structural deformation compensation results. In one specific embodiment, the dynamic weighting of defect sensitivity can rely on the quantitative information of non-defect pressure components in each region provided by pressure compensation estimation data, thereby adjusting the weights accordingly. The dynamic weighted response map data can be intermediate spectrum data containing weighted response intensities generated after the dynamic weighting of defect sensitivity. This can be used to highlight the response of the actual leakage area and suppress spurious responses caused by material permeability or deformation. Furthermore, the dynamic weighted response map data can include, but is not limited to, weighted maps based on permeability compensation, weighted maps based on deformation models, and weighted maps based on environmental disturbance correction.

[0130] Dynamically assigning defect sensitivity weights to the response submap data based on pressure compensation estimation data can be achieved by inversely calculating the sensitivity weight of each region to the actual leak based on the magnitude of the non-defect pressure components in the pressure compensation estimation data, and then applying this weight to the response submap. For example, this operation can be implemented by normalizing the weights using the inverse of the compensation residual as a weighting factor and mapping the compensation estimation features to nonlinear sensitivity weights using a machine learning model. This adaptively enhances the response contribution of highly sensitive regions and suppresses the interference effects of low signal-to-noise ratio regions.

[0131] Acyclic optimization of abnormal responses is performed based on dynamic weighted response graph data to obtain defect response feature graph data.

[0132] The acyclic optimization of anomaly response can be a post-processing operation that removes or smooths isolated anomalies in the dynamic weighted response graph that do not conform to physical continuity or logical consistency. It can be used to eliminate false anomalies caused by sensor transient noise and non-steady-state fluctuations, ensuring the spatial coherence and causal rationality of the defect response. In this embodiment, acyclic optimization of anomaly response can include, but is not limited to, topological connectivity constraint optimization, temporal consistency filtering, and energy minimization smoothing. The defect response feature map data can be the final defect response map output after acyclic optimization of anomaly response, possessing spatial resolution and logical closed-loop characteristics, and can be used to provide high-fidelity, low-noise input for leakage risk quantification. The defect-sensitive response map can be a comprehensive map integrating multiple weighted responses and logically optimized, characterizing the sensitivity of each region of the condom to actual leakage, and can be used to support the shift in detection paradigm from overall average judgment to local intelligent perception. For example, the defect-sensitive response map can include, but is not limited to, segment-level weighted fusion maps, regional sensitivity correction maps, and logical closed-loop verification maps.

[0133] Acyclic optimization of abnormal responses based on dynamic weighted response graph data can identify isolated, abrupt, or physically inconsistent response points in the dynamic weighted response graph and remove or smooth them through constraints. Furthermore, this operation can be achieved by applying connected component analysis in graph theory to remove isolated points and using a Markov random field model to optimize the global consistency of the response graph, thereby improving the spatial consistency and logical reliability of the defective response graph and avoiding misjudgments caused by false anomalies.

[0134] Taking the optimization scenario of micron-level pinhole detection as an example, the vacuum decay sealing test method on the condom production line in this embodiment can be as follows: When testing a condom, the system decomposes its vacuum decay process into three decay segments: the reservoir section, the front section of the tube, and the rear section of the tube. First, based on the current cavity background pressure data, response sub-graphs of the three segments are constructed respectively. It is found that the initial response of the reservoir section is strong, but after pressure compensation estimation, it is shown that it mainly originates from elastic deformation. Based on this, the system assigns a low dynamic weight to the response sub-graph of this segment, while assigning a high weight to the rear section of the tube (which still has continuous decay after compensation). Subsequently, an isolated high response point is found in the dynamic weight response graph, but there is no connected response area around it. After abnormal response loop optimization, it is determined to be sensor transient interference and is removed. The final defect response feature map only retains one continuous abnormality in the middle of the tube. Based on this, the risk quantification module determines it to be a micron-level pinhole leak and triggers a non-conformance mark.

[0135] In one embodiment, leakage risk quantification is performed based on defect response feature map data to obtain leakage risk quantification data, including:

[0136] Pressure attenuation feature mapping is performed based on defect response feature map data to obtain attenuation feature data;

[0137] Abnormal leakage tuples are extracted based on attenuation characteristic data to obtain leakage tuple data;

[0138] Based on the leakage tuple data, micropore permeation stress deformation coupling quantification is performed to obtain leakage risk quantification data.

[0139] The pressure attenuation feature mapping can be a mathematical or algorithmic process of converting defect response feature map data into structured attenuation feature parameters. It can be used to extract key attenuation patterns of the pressure signal in the time, space, or frequency domain, forming feature vectors that can be used for anomaly identification. In this embodiment, the pressure attenuation feature mapping can parameterize the spatiotemporal response patterns in the defect response feature map to generate structured feature vectors. For example, the pressure attenuation feature mapping can include time-domain slope features, frequency-domain energy distribution features, spatial gradient features, etc. Furthermore, this mapping can be implemented by extracting multi-scale attenuation features through wavelet transform and calculating local attenuation slope and curvature through a sliding window, thereby achieving a refined characterization of the dynamic characteristics in the original pressure signal and providing high-dimensional input for subsequent anomaly identification. The attenuation feature data can be a set of structured feature parameters obtained after pressure attenuation feature mapping, which can be used as input for extracting abnormal leakage tuples to characterize the multidimensional dynamic characteristics during the pressure attenuation process. In an exemplary embodiment, the attenuation feature data can include instantaneous attenuation rate, local curvature change, regional response asymmetry, etc.

[0140] Anomaly leakage tuple extraction can be a process of identifying and combining feature subsets that conform to the physical patterns of actual leakage from attenuation feature data. It can be used to filter out smooth background responses caused by the uniform permeability of materials and focus on nonlinear, localized anomalous signals. In a specific embodiment, anomaly leakage tuple extraction can identify feature combinations in attenuation features that conform to the physical laws of actual leakage (such as rapid local attenuation, asymmetric response). For example, anomaly leakage tuple extraction can include rule-based tuple screening, clustering-based anomaly pattern clustering, and classifier-based leakage feature identification. Furthermore, this extraction can be achieved by using a preset leakage pattern template for matching and screening, and training an unsupervised anomaly detection model to automatically cluster leakage-related features, thereby effectively eliminating smooth background signals caused by the uniform permeability of materials and improving the specificity of identifying minute real defects. Leakage tuple data can be a multidimensional feature combination characterizing potential real leaks, containing temporal, spatial, and morphological correlation information, and can be used as a direct input for the coupling quantification of micropore permeation and stress deformation, supporting the physical tracing of leakage mechanisms. In this embodiment, the leakage tuple data may include peak-location coupling pairs, attenuation slope-region curvature correlation pairs, response delay-wall thickness mapping triplet pairs, etc.

[0141] Micropore permeation stress-deformation coupling quantification can be a process of jointly modeling the micropore permeation behavior of latex materials and the elastic deformation of the reservoir structure under vacuum negative pressure, and outputting a comprehensive risk value. This can be used to distinguish between actual pinhole leaks and systemic pressure changes caused by materials / structures, achieving risk scoring based on physical mechanisms. In a specific embodiment, micropore permeation stress-deformation coupling quantification can jointly calculate the leakage tuple with known latex micropore permeation models and reservoir stress-deformation response models, outputting a fused risk value. For example, micropore permeation stress-deformation coupling quantification can include product-type coupling models, superimposed compensation-type coupling models, state-space coupling estimation models, etc. Furthermore, this quantification can be achieved by constructing a coupling function to weightedly fuse permeation rate and deformation energy density, using a Bayesian network to fuse multi-source features, and outputting a posterior leakage probability, thereby realizing risk scoring based on multiple physical mechanisms, supporting the distinction between minor permeation and penetrating leaks, and promoting the evolution of quality assessment from binary to continuous evaluation. For example, the leakage risk micropore-deformation coupling quantification formula can be used to finally determine whether there is a real leak and distinguish between minor leaks and material elastic deformation. The formula expression is as follows:

[0142]

[0143] In the formula, It is a comprehensive indicator used to determine whether a product is qualified, and is a quantitative value for leakage risk. The dynamic decay deviation rate is the actual monitored pressure decay rate, which includes the dual effects of actual leakage and volume expansion; t is the time variable. The deformation rate of the seminal vesicle is the rate of volume change caused by stress deformation of the seminal vesicle under negative pressure. The permeability-deformation coupling coefficient is a parameter used to convert the volumetric deformation rate into the equivalent pressure change rate, reflecting the coupling relationship between the elasticity and permeability of latex. This is a highly sensitive defect risk threshold, an optimized judgment boundary used to distinguish between qualified and unqualified products. This formula is the core of the detection process; it achieves precise analysis of microporous leakage characteristics by calculating the residual between the actual pressure decay and the pseudo-decay caused by structural deformation. Only when the pressure change rate exceeds the range explained by structural deformation, i.e., the residual exceeds the threshold, is it judged as a real leakage risk, thus significantly reducing the false positive rate.

[0144] Taking the identification of micron-level pinhole defects as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be as follows: After the system receives the defect response feature map, it first maps it into attenuation feature data containing the initial attenuation slope, local peak intensity and response asymmetry through time-frequency analysis; then, the abnormal leakage element extraction module identifies a set of features: a sudden pressure drop of 15Pa within 0.8 seconds accompanied by a significant spatial gradient occurs in the reservoir area, and this combination is marked as a leakage element; next, the micropore permeation stress deformation coupling quantification module calls the permeability coefficient (0.02cm³·mm) and reservoir wall thickness (0.06mm) of the current batch of latex to calculate that under the same vacuum conditions, the material permeability will only cause a slow attenuation of 5Pa, while the measured 15Pa is far beyond expectations. Combined with deformation simulation, it shows that the stress concentration amplification effect in this area is 1.3 times, and finally outputs a leakage risk quantification value of 0.88, which is determined to be a penetrating leakage rather than material permeation.

[0145] In one embodiment, micropore permeation stress-deformation coupling quantification is performed based on leakage tuple data to obtain leakage risk quantification data, including:

[0146] Based on the leakage tuple data, the unsteady pressure attenuation deviation and the deformation amplitude deviation of the seminal vesicle were calculated to obtain dynamic attenuation deviation data and deformation amplitude deviation data, respectively.

[0147] The unsteady-state pressure decay deviation can be a quantitative indicator describing the degree of deviation between the actual pressure decay curve and the predicted curve of the ideal material permeability model in the unsteady-state stage. It can be used to characterize the nonlinear, rapid pressure changes caused by real micro-leakage or local anomalies, distinguishing it from the slow-changing response of uniform material permeability. In an exemplary embodiment, the unsteady-state pressure decay deviation may include, but is not limited to, initial slope deviation, instantaneous acceleration deviation, and local curvature abrupt change. The seminal vesicle deformation amplitude deviation can be the difference between the actual deformation of the seminal vesicle under vacuum negative pressure and its theoretical elastic deformation expectation value. It can be used to reflect unexpected deformation caused by abnormal local stiffness or micro-defects in the structure, serving as one of the important sources of false leakage signals. For example, the seminal vesicle deformation amplitude deviation may include, but is not limited to, radial expansion deviation, axial contraction deviation, and local wrinkle deformation energy deviation. The dynamic decay deviation data can be a time-series or regionalized numerical set of the calculated unsteady-state pressure decay deviation, which can be used as input for the analysis of micropore leakage characteristics to identify abnormal decay patterns. Furthermore, the dynamic attenuation deviation data, together with the deformation amplitude deviation data, can constitute a dual-channel physical basis for leakage characteristic analysis. The deformation amplitude deviation data can be a structured expression of the seminal vesicle deformation amplitude deviation in the spatial or temporal dimensions, which can be used to provide quantitative evidence of structural response anomalies and help distinguish between actual leakage and deformation interference. In a specific embodiment, the deformation amplitude deviation data can be jointly input with the dynamic attenuation deviation data into the micropore leakage characteristic analysis module.

[0148] Based on the leakage element data, the unsteady-state pressure attenuation deviation and the deformation amplitude deviation of the reservoir are calculated, yielding dynamic attenuation deviation data and deformation amplitude deviation data. This can be achieved by comparing the pressure and structural response data in the leakage element with the theoretical permeation model and the elastic deformation model, respectively, to calculate the deviation. Furthermore, this operation can be implemented by generating a theoretical deformation curve using finite element simulation and calculating the residual, deriving the ideal attenuation curve using the gas diffusion equation, and obtaining the time-varying deviation. This allows for the physical tracing of abnormal signals in the unsteady-state stage and the separation of material and structural interference components.

[0149] Micropore leakage characteristics were analyzed based on dynamic attenuation deviation data and deformation amplitude deviation data to obtain micropore leakage analysis data;

[0150] The micropore leakage feature analysis can be an analytical process that extracts the unique signal patterns of real microleakage from dynamic attenuation and deformation deviation data. It can be used to separate high-frequency, localized real leakage features and suppress background noise caused by overall material permeability and structural deformation. In an exemplary embodiment, the micropore leakage feature analysis may include, but is not limited to, separation of time-frequency domain anomalies, joint detection of spatial gradient and time derivative, and verification of multi-source deviation consistency. The micropore leakage analysis data can be a structured feature set representing the existence and intensity of real microleakage, output after feature analysis. This can be used as direct input for micropore permeation-deformation response coupling modeling, supporting physical mechanism-level decoupling. For example, the micropore leakage analysis data may include, but is not limited to, leakage initiation time, local attenuation dominant frequency, and deformation-pressure phase difference.

[0151] Micropore leakage characteristics are analyzed based on dynamic attenuation deviation data and deformation amplitude deviation data to obtain micropore leakage analysis data. This can be achieved by fusing the two types of deviation data and extracting feature combinations caused only by real micro-leakage using pattern recognition algorithms. Furthermore, this operation can be achieved by applying wavelet packet decomposition to extract high-frequency attenuation abrupt changes and verifying their spatial consistency with deformation deviation, and constructing a dual-input convolutional network to jointly analyze the pressure and deformation deviation maps. This effectively filters out systematic background caused by overall material permeability or uniform deformation, focusing on localized anomalies.

[0152] Based on the micropore leakage analysis data, a coupled model of micropore permeation-deformation response was performed to obtain the coupled response matrix data.

[0153] Among these methods, micropore permeation-deformation response coupling modeling can establish a dynamic mathematical model of the combined influence of latex micropore permeation behavior and seminal vesicle structural deformation on pressure attenuation signals. This model can be used to quantify the superposition weights of two interference sources in different regions under specific stress states, providing a physical basis for threshold control. In a specific embodiment, micropore permeation-deformation response coupling modeling can include, but is not limited to, state-space coupling models, tensor product response surface models, and graph neural network coupling mappings. The coupling response matrix data can be a multidimensional weight matrix output from micropore permeation-deformation response coupling modeling, representing the contribution ratio of permeation and deformation in each region to the total signal. This can be used to guide differentiated compensation of high-sensitivity defect risk thresholds and achieve region-adaptive judgment. Furthermore, the coupling response matrix data can include, but is not limited to, region-time coupling weight matrices, permeability coefficient-Young's modulus cross-response matrices, and deformation energy-attenuation rate correlation matrices.

[0154] Based on micropore leakage analysis data, a coupled model of micropore permeation-deformation response is performed to obtain coupled response matrix data. This can be achieved by jointly fitting the micropore leakage analysis data with known material and structural parameters to construct the coupled response relationship between permeation and deformation to pressure signals. For example, this operation can be implemented by constructing a three-dimensional coupling matrix of region-parameter-response using tensor decomposition and generating posterior response weights by fusing prior material models and measured analysis data using Bayesian inference. This allows for the establishment of a regionalized physical disturbance model, quantifying the contribution weights of each factor to the overall signal.

[0155] Based on the coupled response matrix data, a high-sensitivity defect risk threshold compensation and control is performed to obtain quantitative data on leakage risk.

[0156] The high-sensitivity defect risk threshold compensation adjustment can be an operational mechanism that dynamically adjusts the risk judgment threshold of each test area or time period based on the coupling response matrix. This can be used to moderately relax the threshold in high-interference areas to avoid false positives, and tighten the threshold in low-interference areas to improve the detection capability of micro-defects. In an exemplary embodiment, the high-sensitivity defect risk threshold compensation adjustment can include, but is not limited to, proportional scaling compensation, nonlinear mapping compensation, and confidence interval adaptive compensation. High-sensitivity defect risk threshold compensation adjustment based on coupling response matrix data yields leakage risk quantification data. This can be achieved by dynamically adjusting the risk judgment threshold of each area in the coupling response matrix based on the interference intensity of that area, and calculating the final risk value accordingly. Furthermore, this operation can be achieved by proportionally amplifying the original threshold in high-coupling weight areas and constructing a nonlinear mapping function based on the response matrix to convert the original risk value into a compensated risk value. This enables spatial adaptive optimization of detection sensitivity, improving the detection capability of small real defects while ensuring a low false positive rate.

[0157] Taking the identification of micro-pinholes in high-deformation areas as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be as follows: A condom shows a suspected leakage signal in the reservoir area. The system first calculates its unsteady-state pressure attenuation deviation as +12 Pa / s (exceeding the predicted value of 8 Pa / s from the material permeation model), while the reservoir deformation amplitude deviation reaches +0.15 mm (theoretical value is 0.3 mm, actual measurement is 0.45 mm). The micropore leakage feature analysis module identifies that the two are highly synchronized in time and space and exhibit high-frequency abrupt change characteristics, which is determined to be a potential real leak. Subsequently, the micropore permeation-deformation response coupling modeling calls the latex porosity of 0.08 and wall thickness of 0.055 mm in this area to generate a coupling response matrix showing that the amplification factor of the pressure signal due to deformation in this area is 1.6. Based on this, the high-sensitivity defect risk threshold compensation adjustment raised the original threshold from 0.75 to 0.90. The calculated actual risk value was 0.92, which still exceeded the compensated threshold. Ultimately, it was determined to be a penetrating micropore leakage, thus avoiding missed detection due to high deformation.

[0158] In one embodiment, a pass / fail judgment is output based on leakage risk quantification data to obtain online sealing test results, thereby achieving high-precision dynamic sealing test for condom production lines, including:

[0159] High-risk defect clusters are identified based on leakage risk quantification data to obtain core defect cluster data.

[0160] High-risk defect clusters can be sets of defect points with spatial proximity and consistent risk intensity identified through clustering in leakage risk quantification data. These clusters can be used to integrate discrete risk signals into physically meaningful potential leakage areas, improving the robustness and interpretability of the judgment. In an exemplary embodiment, high-risk defect clusters can include, but are not limited to, point clusters, line clusters, and area clusters. Identifying high-risk defect clusters based on leakage risk quantification data can involve clustering spatial or temporal risk points in the data to extract significant clustering regions. Furthermore, high-risk defect cluster identification based on leakage risk quantification data can be achieved by using the DBSCAN density clustering algorithm to identify spatially adjacent high-risk points and by using a sliding window statistical method to detect continuous abnormal segments in the time decay sequence. This can filter out isolated noise points and focus on defect groups with actual leakage significance.

[0161] Based on the defective core cluster data, the vacuum test accuracy level judgment threshold is optimized to obtain the judgment threshold optimization data;

[0162] The defect core cluster data can be structured data output after identifying high-risk defect clusters, including cluster center coordinates, risk density, coverage, and confidence level. This data can be used as input for optimizing the judgment threshold, supporting the generation of dynamic threshold strategies. The vacuum test accuracy-level judgment threshold can be a dynamic risk judgment boundary value used to distinguish between qualified and unqualified products. It can replace fixed thresholds to achieve adaptive responses to different defect modes. In one specific embodiment, the vacuum test accuracy-level judgment threshold can work in conjunction with the defect core cluster data, adjusting the threshold level in real time based on the cluster's strength and distribution characteristics. The judgment threshold optimization data can be an updated set of judgment boundary parameters generated by a threshold optimization algorithm. This can guide the generation of online sealing test results, improving the detection capability of small concentrated leaks and suppressing false judgments of uniform permeability. Optimizing the vacuum test accuracy-level judgment threshold based on the defect core cluster data can be achieved by dynamically calculating a judgment threshold adapted to the current operating conditions based on the defect core cluster's strength, density, and historical false judgment rate. For example, optimizing the vacuum test accuracy threshold based on defect core cluster data can be achieved by constructing a cluster feature-threshold mapping function to implement lookup table optimization and by using an online learning model to update the threshold parameters in real time based on new cluster data. This allows the judgment criteria to be adaptively adjusted according to the actual defect characteristics, thereby improving detection sensitivity and specificity.

[0163] Based on the judgment threshold, the optimized data generates online sealing test results to drive the production line sorting mechanism to dynamically reject unqualified products, thereby achieving high-precision online dynamic sealing test of the condom production line.

[0164] The production line sorting mechanism can be an automated mechanical device that performs physical rejection or diversion actions based on the detection results. It can be used to dynamically and accurately reject substandard condoms, ensuring consistent sealing performance of outgoing products. In one exemplary embodiment, the production line sorting mechanism can consist of a pneumatic pusher, a servo fork, or a vision-guided robotic arm, which performs sorting after receiving a trigger signal from the detection system. Substandard products can be condoms that, according to an optimized judgment threshold, pose a leakage risk exceeding the permissible range. These can be rejected by the sorting mechanism to prevent them from flowing into subsequent packaging stages. Generating online sealing performance detection results based on the optimized judgment threshold data can involve comparing the quantified leakage risk data with the optimized judgment threshold and outputting a structured judgment result containing risk level and location information. Furthermore, generating online sealing performance detection results based on the optimized judgment threshold data can be achieved by outputting multi-level labels (qualified / observed / rejected) with cluster coordinates and generating probabilistic judgment results with confidence levels. This allows for the generation of refined detection conclusions with process feedback value, going beyond simple binary judgment. The dynamic rejection of non-conforming products by driving the sorting mechanism on the production line can be achieved by transmitting rejection commands from online sealing test results to the sorting mechanism control unit in real time, triggering corresponding actions. For example, the dynamic rejection of non-conforming products by driving the sorting mechanism can be accomplished by the PLC sending pulse signals to control the pneumatic rejection valve and sending timing-synchronized rejection commands to the servo sorting arm via an industrial bus, thereby achieving precise physical isolation of non-conforming products in high-speed production lines.

[0165] Taking adaptive threshold-driven high-speed sorting as an example, the vacuum attenuation sealing test method on the condom production line in this embodiment can be as follows: during continuous production, the system performs real-time clustering of the leakage risk quantification data of a batch of condoms, identifies a high-density risk cluster in the reservoir area, with a risk mean of 0.78 and continuous spatial coverage; based on the core cluster data of this defect, the algorithm dynamically adjusts the original fixed threshold of 0.75 to 0.80 to avoid false triggering due to material background fluctuations, and at the same time, a sub-region sensitive threshold of 0.72 is used for local peaks within the cluster; the final test result shows that the product is unqualified, and the cluster center is marked as being located 12 mm from the top; the sorting mechanism receives a timestamped rejection instruction 50 milliseconds before the product arrives at the rejection station, accurately blows it into the waste channel, while the low-amplitude global signal of adjacent unclustered products is not mistakenly rejected because it is below the dynamic threshold.

[0166] In addition, refer to Figure 2 To achieve the above objectives, the present invention also provides a vacuum attenuation sealing performance testing system for a condom production line, the system comprising:

[0167] The risk classification module 10 is used to obtain the micropore parameters of the latex polymer material of the condom to be tested and the product structure contour data, and to classify the sealing risk area according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip, so as to obtain the risk area data to be tested.

[0168] The attenuation decomposition module 20 is used to acquire the accuracy requirement data of the production line sealing test; and to decompose the test accuracy requirement data and the data of the risk area to be inspected into attenuation segment decomposition in the vacuum test unit to obtain test attenuation segment decomposition data.

[0169] The deviation compensation module 30 is used to acquire the cavity background pressure fluctuation data of the current vacuum test system, and to perform pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data.

[0170] The quantification and judgment module 40 is used to construct a defect-sensitive response map based on the pressure compensation estimation data to obtain defect response feature map data; and to quantify leakage risk based on the defect response feature map data to obtain leakage risk quantification data.

[0171] The result output module 50 is used to determine whether a product is qualified or unqualified based on the leakage risk quantification data, and to obtain the online sealing test results, so as to realize high-precision dynamic sealing test of the condom production line.

[0172] Other embodiments or specific implementations of the vacuum attenuation sealing performance testing system on the condom production line described in this invention can be referred to the above-described method embodiments, and will not be repeated here.

[0173] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for testing vacuum attenuation sealing performance on a condom production line, characterized in that, The method includes: The micropore parameters and product structure contour data of the latex polymer material of the condom to be tested are obtained, and the sealing risk area is divided according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip to obtain the data of the risk area to be tested. Obtain the accuracy requirement data for the production line sealing test; decompose the test accuracy requirement data and the data of the risk area to be inspected into the attenuation section of the vacuum test unit to obtain the test attenuation section decomposition data; Acquire the cavity background pressure fluctuation data of the current vacuum test system, and estimate the pressure deviation compensation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain the pressure compensation estimation data; Based on the pressure compensation estimation data, a defect sensitivity response map is constructed to obtain defect response characteristic map data; based on the defect response characteristic map data, leakage risk is quantified to obtain leakage risk quantification data. Based on the quantitative data of leakage risk, a pass or fail judgment is output to obtain the online sealing test result, so as to realize high-precision dynamic sealing test of condom production line; The step of dividing the sealing risk area based on the latex micropore porosity distribution and the seminal vesicle structural parameters to obtain the risk area data to be inspected includes: Based on the latex micropore porosity distribution data and seminal vesicle stress parameters, the defect-pressure response coupling relationship was analyzed to obtain the defect-pressure coupling matrix data. Obtain statistical data on the distribution of defects in the condom production process, and calculate the frequency of risk occurrence based on the statistical data to obtain defect risk frequency data; Based on the defect-pressure coupling matrix data and defect risk frequency data, a digital twin inspection domain is constructed to obtain the inspection risk area data.

2. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The process of constructing a digital twin inspection domain based on defect-pressure coupling matrix data and defect risk frequency data yields inspection risk area data, including: Based on the defect-pressure coupling matrix data and defect risk frequency data, virtual vacuum test field coverage features and pressure attenuation transmission features are extracted, resulting in virtual test field coverage feature data and pressure attenuation transmission feature data, respectively. Based on the virtual test field coverage feature data and pressure attenuation transmission feature data, twin defect response maps are generated from the defect-pressure coupling matrix data to obtain twin defect response map data. The virtual-real mismatch region detection was performed on the twin defect response map data to obtain virtual-real calibration map data. The virtual-real mismatch region detection included environmental temperature fluctuation mismatch detection, cavity leakage background anomaly detection, and latex elastic creep deviation detection. The production line cycle time constraint is injected into the virtual and real calibration diagram data to obtain the time constraint diagram data; High-risk connected subdomains are extracted from the time-series constraint graph data to obtain the data of the risk areas to be inspected.

3. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The process of decomposing the test accuracy requirement data and the data of the risk area to be inspected into attenuation segment using a vacuum test unit yields test attenuation segment decomposition data, including: The topology of the vacuum attenuation test section is divided based on the test accuracy requirement data and the data of the risk area to be inspected, and the test section topology division data is obtained. Based on the test section topology division data, gas permeation characteristics and key defect nodes are extracted, and gas permeation characteristic data and key defect node data are obtained respectively. Based on the gas permeation characteristic data and key defect node data, the pressure decay time-series dependency processing of the test section topology partitioning data is performed to obtain time-series dependency data. A dynamic decay response map is constructed based on time-dependent data to obtain the decomposed data of the test decay segment.

4. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The process involves acquiring the cavity floor pressure fluctuation data of the current vacuum testing system, and performing pressure deviation compensation estimation based on the cavity floor pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data, including: Based on the cavity background pressure fluctuation data and test attenuation segment decomposition data of the current vacuum testing system, multi-cycle pressure attenuation superposition calculation is performed to obtain attenuation superposition data. The pressure deviation kernel distribution is calculated based on the attenuation superposition data to obtain the local pressure deviation intensity data; Based on the current background pressure data of the vacuum test chamber, the ultimate vacuum degree of the chamber and the response error of the pressure sensor are calculated, and the ultimate vacuum data and sensor error data are obtained respectively. Based on the ultimate vacuum data and sensor error data, the dynamic pressure compensation amount is extrapolated from the local pressure deviation intensity data to obtain the pressure compensation estimate data.

5. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The step of constructing a defect-sensitive response map based on pressure compensation estimation data to obtain defect response feature map data includes: Based on the current background pressure data of the vacuum test chamber and the decomposed data of the test attenuation section, the response subgraph of the test attenuation section is constructed to obtain the response subgraph data. Based on the pressure compensation estimation data, dynamic weights for defect sensitivity are assigned to the response subgraph data to obtain dynamic weighted response graph data. Acyclic optimization of abnormal responses is performed based on dynamic weighted response graph data to obtain defect response feature graph data.

6. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The step of quantifying leakage risk based on defect response feature map data to obtain leakage risk quantification data includes: Pressure attenuation feature mapping is performed based on defect response feature map data to obtain attenuation feature data; Abnormal leakage tuples are extracted based on attenuation characteristic data to obtain leakage tuple data; Based on the leakage tuple data, micropore permeation stress deformation coupling quantification is performed to obtain leakage risk quantification data.

7. The vacuum attenuation sealing test method on a condom production line as described in claim 6, characterized in that, The step of performing micropore permeation stress-deformation coupling quantification based on leakage tuple data to obtain leakage risk quantification data includes: Based on the leakage tuple data, the unsteady pressure attenuation deviation and the deformation amplitude deviation of the seminal vesicle were calculated to obtain dynamic attenuation deviation data and deformation amplitude deviation data, respectively. Micropore leakage characteristics were analyzed based on dynamic attenuation deviation data and deformation amplitude deviation data to obtain micropore leakage analysis data; Based on the micropore leakage analysis data, a coupled model of micropore permeation-deformation response was performed to obtain the coupled response matrix data. Based on the coupled response matrix data, a high-sensitivity defect risk threshold compensation and control is performed to obtain quantitative data on leakage risk.

8. The vacuum attenuation sealing test method on a condom production line as described in claim 1, characterized in that, The process of determining pass or fail based on leakage risk quantification data to obtain online sealing test results enables high-precision dynamic sealing testing of the condom production line, including: High-risk defect clusters are identified based on leakage risk quantification data to obtain core defect cluster data. Based on the defective core cluster data, the vacuum test accuracy level judgment threshold is optimized to obtain the judgment threshold optimization data; Based on the judgment threshold, the optimized data generates online sealing test results to drive the production line sorting mechanism to dynamically reject unqualified products, thereby achieving high-precision online dynamic sealing test of the condom production line.

9. A vacuum attenuation sealing performance testing system for a condom production line, characterized in that, The system includes: The risk classification module is used to obtain the micropore parameters of the latex polymer material of the condom to be tested and the product structure contour data, and to classify the sealing risk area according to the micropore porosity distribution of latex and the structural parameters of the reservoir tip, so as to obtain the data of the risk area to be tested. The attenuation decomposition module is used to acquire the accuracy requirement data for the production line sealing test; it decomposes the test accuracy requirement data and the data of the risk area to be inspected into attenuation segments in the vacuum test unit to obtain test attenuation segment decomposition data. The deviation compensation module is used to acquire the cavity background pressure fluctuation data of the current vacuum test system, and to perform pressure deviation compensation estimation based on the cavity background pressure fluctuation data and the test attenuation segment decomposition data to obtain pressure compensation estimation data. The quantification and judgment module is used to construct a defect sensitivity response map based on the pressure compensation estimation data to obtain defect response feature map data; and to quantify leakage risk based on the defect response feature map data to obtain leakage risk quantification data. The result output module is used to determine whether a product is qualified or unqualified based on the leakage risk quantification data, and to obtain the online sealing test results, so as to realize high-precision dynamic sealing test of the condom production line. The step of dividing the sealing risk area based on the latex micropore porosity distribution and the seminal vesicle structural parameters to obtain the risk area data to be inspected includes: Based on the latex micropore porosity distribution data and seminal vesicle stress parameters, the defect-pressure response coupling relationship was analyzed to obtain the defect-pressure coupling matrix data. Obtain statistical data on the distribution of defects in the condom production process, and calculate the frequency of risk occurrence based on the statistical data to obtain defect risk frequency data; Based on the defect-pressure coupling matrix data and defect risk frequency data, a digital twin inspection domain is constructed to obtain the inspection risk area data.

Images