A method for determining the ventilation rate of a cigarette, electronic equipment, and storage medium.
By combining finite element simulation and graph neural network methods, the problems of low efficiency and insufficient accuracy in calculating the ventilation rate of cigarettes with perforated paper are solved, realizing fast and accurate prediction of ventilation rate and supporting the refined design of ventilation systems for low-tar cigarettes.
Patent Information
- Application Number
- CN202611312765.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-08-27
- Publication Date
- 2026-11-10
AI Technical Summary
In the calculation of ventilation rate of perforated cigarettes with paper filling, the traditional finite element simulation method has a long calculation cycle and low efficiency, while the pure data-driven prediction method lacks fluid dynamic mechanism constraints, resulting in distorted results and insufficient reliability, making it difficult to meet the requirements of accuracy and efficiency at the same time.
By combining finite element simulation and graph neural network, a model is built and simulated by obtaining the geometric, material and perforation parameters of the splicing paper. The ventilation rate is predicted by a pre-trained graph neural network, and the credibility of the results is judged by the ventilation rate difference and the Pearson coefficient of the air intake distribution matrix. Mechanism simulation is used as a fallback to ensure accuracy.
It achieves improved calculation efficiency while ensuring accuracy, can quickly assess the impact of perforation parameters on ventilation rate, and supports the optimized design of splicing paper ventilation systems.
Smart Images

Figure CN123069971A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of tobacco technology, and more specifically, to a method for determining the ventilation rate of tapped cigarettes, an electronic device, and a storage medium. Background Technology
[0002] In the calculation of ventilation rate of perforated cigarettes with tipped paper, the traditional full-domain finite element simulation method can ensure the accuracy and physical rationality of the calculation, but it has the disadvantages of long modeling cycle and long calculation time for a single case, making it difficult to support batch parameter traversal and rapid scheme iteration. On the other hand, the simple data-driven prediction method is fast in calculation, but it generally lacks fluid dynamic mechanism constraints, which can easily lead to prediction results that violate physical laws and distortion of flow field distribution. Moreover, it lacks an effective mechanism for mechanism verification and credibility judgment, resulting in insufficient reliability of the results. It cannot simultaneously meet the efficiency and accuracy requirements of ventilation rate calculation, and it is difficult to support the refined and rapid design of ventilation systems for low-tar cigarettes. Summary of the Invention
[0003] In view of this, the purpose of this application is to provide a method for determining the ventilation rate of a cigarette holder, an electronic device, and a storage medium, which can improve the problem.
[0004] To achieve the above technical objectives, the technical solution adopted in this application is as follows:
[0005] In a first aspect, embodiments of this application provide a method for determining the ventilation rate of a dispensed cigarette, the method comprising:
[0006] Obtain the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters of the tipping paper;
[0007] Based on the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters, a finite element model and a graphical structural model of the splicing paper are established.
[0008] The finite element model was used to perform finite element simulation to obtain the first air intake distribution matrix and the first ventilation rate of the splicing paper;
[0009] The graph structure model is input into a pre-trained graph neural network, and the graph neural network outputs the second air intake distribution matrix and the second ventilation rate of the splicing paper.
[0010] The final ventilation rate of the cigarette pack is determined based on the difference between the first ventilation rate and the second ventilation rate, and the Pearson coefficients of the first and second air intake distribution matrices.
[0011] Secondly, embodiments of this application also provide an electronic device, the electronic device including a processor and a memory coupled to each other, the memory storing a computer program, and when the computer program is executed by the processor, causing the electronic device to perform the method as described in the first aspect.
[0012] Thirdly, embodiments of this application also provide a computer-readable storage medium storing a computer program that, when run on a computer, causes the computer to perform the method described in the first aspect.
[0013] The invention employing the above technical solution has the following advantages:
[0014] In the technical solution provided in this application, the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters of the tipping paper are first obtained. A finite element simulation model and a graphical structure model are then simultaneously built. The ventilation rate is calculated using a combination of finite element mechanism simulation and graphical neural network intelligent prediction. The geometric parameters include the axial width of the glue-free permeable zone as a configurable variable, and the circumferential overlap width of the tipping paper and the cigarette paper are introduced to deduct the influence of the non-permeable overlapping area of the glue-sealed area on the effective permeable area. The glue-free permeable zone is set to ensure effective perforation of the tipping paper, and the perforated permeable zone must be located on the axis of the glue-free permeable zone. Within the range, the adhesive-free permeable zone is slightly wider than the perforated permeable zone. The forming paper corresponding to the adhesive-free permeable zone uses high-permeability forming paper to ensure smooth air penetration into the filter rod. The perforation parameters include structural indicators such as the number of perforation rows, number of perforations, and pore diameter, and also incorporate the axial position of the perforated zone as a configurable variable. This allows the coverage position of the perforated permeable zone within the adhesive-free permeable zone to be dynamically adjusted according to the parameters, accurately reflecting the influence of the perforation position on the path length and ventilation rate of the infiltrated air. The perforation parameters are used in the conversion of the equivalent permeability and resistance coefficient of the perforated zone in the finite element model, and also serve as input diagrams for local perforation structural features and global perforation working condition features. The structural model ensures that both models accurately reflect the influence of the number of perforation rows, number of perforations, perforation location, width of the glue-free zone, and circumferential overlap seal on the ventilation rate. For the suction resistance parameter, the suction resistance of the tobacco section is measured using cigarettes without perforations or without filters, avoiding interference from perforations in the filter section. The resistance characteristics of the tobacco section can also be characterized by tobacco density. The reliability of the prediction results is determined by dual indicators: the numerical deviation of the ventilation rate and the Pearson correlation coefficient of the spatial distribution of the air intake. When both indicators meet the requirements, the prediction results meet both the numerical accuracy requirements and closely match the actual flow patterns, allowing for direct neural network prediction. The result aims to improve efficiency. If any one of the criteria is not met, the finite element simulation result is used as the final value. Mechanism simulation is used as a safety net to ensure the physical rationality and accuracy of the result, avoiding the problems of physical distortion and insufficient reliability that are easily caused by pure data-driven prediction. Overall, it achieves a balance between measurement efficiency and calculation accuracy. In particular, it can quickly evaluate the impact of the number of perforation rows, the number of perforations, the axial position of the perforations, and the width adjustment of the glue-free area on the ventilation rate. Under the condition that other conditions remain unchanged, the ventilation rate prediction results under different schemes can be quickly obtained by simply adjusting the corresponding parameters, providing efficient measurement support for the optimized design of the splicing paper ventilation system. Attached Figure Description
[0015] This application can be further illustrated by the non-limiting embodiments given in the accompanying drawings. It should be understood that the following drawings only illustrate some embodiments of this application and should not be considered as limiting the scope. For those skilled in the art, other related drawings can be obtained from these drawings without any inventive effort.
[0016] Figure 1 A flowchart illustrating the method for determining the ventilation rate of a cigarette pack provided in this application embodiment.
[0017] Figure 2 This is a sub-flowchart of S130 provided in an embodiment of this application. Detailed Implementation
[0018] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that similar or identical parts are referred to by the same reference numerals in the drawings or description. Implementations not shown or described in the drawings are forms known to those skilled in the art. In the description of this application, terms such as "first" and "second" are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0019] Please refer to Figure 1 This application provides a method for determining the ventilation rate of a tipped-up cigarette, which can be applied to electronic devices and can be executed or implemented by the electronic devices. The electronic devices can be, but are not limited to, personal computers, smartphones, etc. The method for determining the ventilation rate of a tipped-up cigarette may include the following steps:
[0020] S110, obtain the geometric parameters, material air permeability parameters, suction resistance parameters and perforation parameters of the tipping paper;
[0021] S120: Based on the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters, establish the finite element model and graphical structural model of the splicing paper;
[0022] S130: Perform finite element simulation on the finite element model to obtain the first air intake distribution matrix and the first ventilation rate of the splicing paper;
[0023] S140: Input the graph structure model into a pre-trained graph neural network, and output the second air intake distribution matrix and the second ventilation rate of the splicing paper through the graph neural network;
[0024] S150: Determine the final ventilation rate of the cigarette pack based on the difference between the first ventilation rate and the second ventilation rate, and the Pearson coefficients of the first and second air intake distribution matrices.
[0025] In the above implementation, a pre-trained graph neural network is used to perform the main calculation work. A graph model is built according to the gridded air-permeable structure of the splicing paper, and the lateral diffusion characteristics of different air-permeable areas are embedded in advance. After inputting parameters, the ventilation rate and air intake distribution results can be obtained quickly. It is not necessary to model and run simulations from scratch every time, which greatly improves the efficiency of batch parameter traversal and scheme selection and solves the problems of long simulation cycle and slow iteration in traditional simulation. At the same time, it uses the finite element simulation results with the same parameters as the verification benchmark. The credibility of the results is judged by dual indicators of ventilation rate numerical deviation and air intake spatial distribution consistency. If the prediction result meets the standard, the rapid prediction value is used to improve efficiency. If it does not meet the standard, the finite element simulation result is used as a backup to ensure accuracy. This avoids the problem that pure data prediction is prone to violating physical laws and lacks reliability. It also takes into account the efficiency and accuracy of the calculation, and can support the refined and rapid design of low tar cigarette ventilation systems.
[0026] The following is a detailed explanation of each step in the method for determining the ventilation rate of packaged cigarettes:
[0027] In S110, the geometric parameters specifically include the total length of the cigarette, the cigarette diameter, the filter rod length, the tobacco section length, the tipping paper length and thickness, the high-permeability forming paper thickness, the cigarette paper thickness, the axial coordinate range of the perforated area, the adhesive area width, the 9mm clamping section size at the filter tip end, the coordinate dimensions of the overlapping section between the tipping paper and the cigarette paper, the axial width of the non-adhesive permeable area, the circumferential overlap width of the tipping paper, and the circumferential overlap width of the cigarette paper, among other spatial dimensional indicators. These parameters can be extracted from product design drawings and process specification documents, as well as... The actual tipping paper and cigarette samples can be measured using precision measuring equipment such as image measuring instruments and micrometers, following the testing methods corresponding to the tobacco industry. The boundary positions of each area are ultimately confirmed by combining the actual glue application, perforation, and overlap process positions. Among them, the axial width of the glue-free air permeable zone is a configurable variable along the axial direction of the cigarette. Its value directly determines the axial coverage and effective air permeability of the glue-free air permeable zone, which is a key geometric parameter affecting the ventilation of the tipping paper. The glue-free air permeable zone is specially designed to ensure effective perforation of the tipping paper. In the adhesive-free area, the perforated ventilation zone must be located within the axial range of the adhesive-free ventilation zone. The axial width of the adhesive-free ventilation zone is greater than the axial width of the perforated ventilation zone (i.e., the adhesive-free ventilation zone is slightly wider than the perforated ventilation zone) to ensure that the perforation position falls entirely within the adhesive-free area and to prevent the adhesive layer from clogging the ventilation holes. The filter rod forming paper corresponding to the adhesive-free ventilation zone uses high-permeability forming paper, allowing air that seeps in through the perforations and adhesive-free area to smoothly pass through the filter rod wrapping layer and enter the interior. The circumferential overlap width of the splice paper refers to the overlap of the first and last pieces of adhesive coating after the splice paper wraps around the filter rod. The width of the region, which is completely airtight due to the double-layer paper stacking and the adhesive layer barrier, needs to be deducted from the effective circumferential air-permeable arc length of the tipping paper; the circumferential overlap width of the cigarette paper refers to the width of the adhesive-coated area where the cigarette paper overlaps the tobacco shreds one lap. This area is also an airtight sealed section and needs to be deducted from the effective circumferential air-permeable arc length of the cigarette paper. The above parameters are not only the spatial reference for the geometric construction of the finite element model, the division of the air-permeable area, and the mesh generation, but also the core basis for the first mesh discretization, region category labeling, and position feature assignment in the graph structure model.
[0028] The material permeability parameters specifically include four core indicators: air permeability of the glue-free area of the tipping paper, air permeability of the perforated area of the tipping paper, air permeability of the high-permeability forming paper, and air permeability of the cigarette paper. These must be measured according to tobacco industry standards such as GB / T23227, using a permeability tester under standard temperature and humidity conditions. Samples from different areas are taken for actual measurement, and multiple samples are tested in parallel for each parameter group, with the average value taken to ensure data stability. After obtaining the original permeability, the composite equivalent permeability of the glue-free area is calculated based on the principle of superposition of series permeability resistance. Then, combined with the material thickness, it is converted into the normal viscous resistance coefficient of the porous medium based on Darcy's law of permeability. A transverse resistance coefficient is set at 10 times the normal value. These parameters are used both for assigning property values to the permeable thin layer in the finite element model and as components of the node features input into the graph structure model.
[0029] The draw resistance parameters specifically include two items: the total draw resistance of the filter rod and the total draw resistance of the tobacco section. These are obtained according to industry standards for cigarette draw resistance testing, using a comprehensive cigarette testing bench under standard draw flow rates. The filter rod draw resistance is obtained directly from the finished empty filter rod. The tobacco section draw resistance must be obtained after eliminating interference from the perforated tipping paper on the filter tip draw resistance. One of the following two implementation methods is used:
[0030] The first method is the non-perforated tipping paper replacement method. This method uses a cigarette sample of the same specifications as the target cigarette but with an unperforated tipping paper. The total draw resistance is tested, and then the draw resistance of the corresponding filter rod is subtracted to obtain the draw resistance of the tobacco section. Since the tipping paper is unperforated, the filter section has no additional air intake channel, and the difference between the total draw resistance and the filter rod draw resistance accurately reflects the flow resistance of the tobacco section itself. The second method is the filterless direct testing method. This method directly tests the draw resistance of a filterless cigarette containing only tobacco sections and cigarette paper. The result is the total draw resistance of the tobacco section. This method structurally isolates the influence of the filter rod and tipping paper. During testing, the average value of multiple parallel samples is taken to eliminate errors caused by individual differences. In addition, the tobacco section can also be characterized by tobacco density. Tobacco density can be calculated by instrumental testing to obtain the average density, or it can be measured manually—generally, several cigarettes are taken, and each… After the tobacco shreds of a cigarette are stripped and weighed, the average density is obtained by dividing the total weight of the tobacco shreds by the total volume of the tobacco segments. The measured tobacco density can be automatically collected by electronic equipment or manually entered into the system. Based on a preset density-resistance mapping relationship or calibration curve, the tobacco density can be converted into the viscous resistance coefficient and inertial resistance coefficient of the tobacco segments, which can be used as alternative or supplementary inputs for the suction resistance parameters. Based on the measured total suction resistance value or the tobacco density conversion result, combined with the length, cross-sectional area and design suction flow rate of the corresponding segment, the viscous resistance coefficient and inertial resistance coefficient of each segment can be obtained by fitting the Darcy-Forchheimer equation. This is used to assign the porous media properties of the filter rod segment and tobacco segment in the finite element model. At the same time, the suction resistance parameters are also used as a component of the global operating condition feature vector and participate in the global ventilation rate adjustment calculation of the graph neural network.
[0031] The perforation parameters specifically include perforation row number, number of perforations per row, total number of perforations, axial spacing of perforations in each row, single hole diameter, circumferential distribution range of the perforation area, and axial position of the perforation area. These parameters can be extracted from perforation process documents and laser perforation equipment parameter records, or the actual values can be obtained by scanning the perforation area of the feed paper with a microscopic imaging device and combining it with image recognition algorithms. During the statistical analysis, the average value of multiple samples is taken to eliminate perforation deviation of a single sample. The axial position of the perforation area refers to the starting or center coordinates of the perforation area on the filter tip axis, representing the arrangement position of the perforated air-permeable area along the filter rod axis. Its configurable range is outside the 9mm clamping and sealing area at the end of the filter tip and does not overlap with the feed paper or cigarette paper. The perforated area, located in the overlapping section of the filter rod, can be positioned at any axial location between the clamping section and the tipping paper / cigarette paper overlap section, or at any axial location on the side of the overlap section closest to the tobacco end. However, the tipping paper / cigarette paper overlap section itself will not have ventilation holes due to the double-layer paper stacking and adhesive sealing. It is important to note that the perforated ventilation area must be located within the axial range of the adhesive-free ventilation area. The adhesive-free ventilation area is designed to ensure effective perforation of the tipping paper—if the perforation location falls within the adhesive area, the adhesive layer will block the ventilation holes, causing the perforation to fail. Therefore, the axial width of the adhesive-free ventilation area is greater than the axial width of the perforated ventilation area (the adhesive-free ventilation area is slightly wider than the perforated ventilation area) to ensure that the perforation location is entirely within the adhesive-free area. The forming paper corresponding to the adhesive-free ventilation area uses… High-permeability forming paper ensures smooth air penetration; the axial position of the perforated area directly determines the length of the flow path of outside air after penetration to the filter rod outlet. The closer the perforation is to the end of the filter tip (outside the clamping section), the shorter the internal flow path of the infiltrated air and the lower the friction resistance, resulting in a higher ventilation rate. Conversely, the closer the perforation is to the tobacco end, the longer the flow path, the greater the friction resistance, and the lower the ventilation rate. Therefore, the axial position of the perforated area is a key positional parameter affecting the ventilation rate of the tipping paper. The axial layout of the glue-free permeable area matches the axial position of the perforated area. The glue-free permeable area extends a certain width to both sides of the perforated area, so that the perforated area is completely contained within the glue-free permeable area. The inclusion relationship and width difference between the two are determined by the perforation process and the glue coating of the tipping paper. The layout is determined by the perforation parameters, which directly determine the effective opening area, overall air permeability, and axial placement of the perforated air permeable zone. The more rows of perforations and the more holes per row, the larger the effective air permeable area per unit area of the perforated zone, and the stronger the overall air permeability of the corresponding area. This is a key structural parameter affecting the ventilation rate of the splicing paper. In finite element modeling, the perforation parameters are used to calculate the equivalent air permeability and resistance coefficient of the perforated zone and to determine the axial coverage position of the perforated air permeable zone within the glue-free air permeable zone. In the graph structure model, the perforation parameters are included as a feature vector of each first grid as a local perforated structural feature, and also participate in the global adjustment calculation of the ventilation rate as a global working condition feature in the form of the total number of perforation rows, the total number of perforations, and the axial position of the perforated zone.
[0032] Tipped cigarettes are a typical cigarette structure in the low-tar cigarette category that relies on filter ventilation to dilute and reduce tar in the smoke. They are also the core research object of this solution. Their core function is to allow outside air to seep into the filter rod through the air-permeable area of the tipped paper, diluting the mainstream smoke and reducing the release of tar and harmful components, while taking into account both the draw resistance and the taste experience.
[0033] Structurally, this type of cigarette is divided into two main parts along the axial direction: a tobacco section and a filter rod section. The filter rod section is wrapped with tipping paper, and the inner side of the tipping paper is wrapped with high-permeability forming paper corresponding to the filter rod filament bundle. The tobacco section is covered with cigarette paper. The tipping paper can be divided into several continuous areas with significantly different air permeability characteristics along the axial direction: a 9mm clamping and sealing area at the end of the filter tip, an overlapping and sealing area where the tipping paper and cigarette paper overlap, a non-adhesive air-permeable area without adhesive layer obstruction, and an adhesive sealing area with adhesive applied to the end. Among these, the clamping area, overlapping area, and adhesive area are basically non-permeable due to physical compaction or adhesive layer obstruction. The air permeability is achieved through the glue-free air permeable zone, which serves as the core channel for external air to infiltrate. The glue-free air permeable zone is specifically designed to ensure effective perforation of the splicing paper. The perforated air permeable zone must be located within the axial range of the glue-free air permeable zone. The axial width of the glue-free air permeable zone is greater than the axial width of the perforated air permeable zone (i.e., the glue-free air permeable zone is slightly wider than the perforated air permeable zone) to ensure that the perforation position falls entirely within the glue-free zone and to prevent the adhesive layer from clogging the ventilation holes. The filter rod forming paper corresponding to the glue-free air permeable zone is made of high-permeability forming paper, allowing the air that infiltrates through the perforation and glue-free zones to pass smoothly through the filter rod wrapping layer and enter the interior. The air permeability of the perforated air permeable zone is determined by perforation parameters such as the number of perforations, the number of perforations, and the perforation diameter. The axial arrangement of the perforated air permeable zone within the glue-free air permeable zone is determined by the axial position of the perforated zone. Its configurable range is the filter rod section outside the 9mm clamping and sealing area at the filter tip end and not overlapping with the tipping paper and cigarette paper overlap section. The tipping paper and cigarette paper overlap section itself does not have ventilation holes due to double-layer lamination and adhesive sealing. Different locations result in different flow path lengths for infiltrated air to reach the outlet, directly affecting the ventilation rate. The axial arrangement of the glue-free air permeable zone... The width of the configurable variable cover directly affects the effective air permeability area and paper ventilation volume in the glue-free area. In addition, after the tipping paper wraps the filter rod for one week, there is an overlap area in the circumferential direction. Similarly, after the cigarette paper wraps the tobacco section for one week, there is also an overlap area in the circumferential direction. The aforementioned circumferential overlap areas are all sealed with glue and are completely airtight. This results in the effective air permeability arc length of the tipping paper and cigarette paper in the circumferential direction being less than the corresponding circumference. The overlap sealing section needs to be deducted in the air permeability area calculation and finite element modeling to avoid overestimating the air intake volume of the glue-free area and paper ventilation.
[0034] From the perspective of air permeability mechanism, air infiltration does not only occur at the perforated location, but also diffuses laterally within the plane of the tipping paper in the glue-free permeable area. Since the perforated permeable area must be located within the axial range of the glue-free permeable area, and the glue-free permeable area is slightly wider than the perforated permeable area, the air infiltrated through the perforations can further diffuse laterally in the glue-free non-perforated area surrounding the perforated area. Furthermore, the tipping paper and the high-permeability forming paper form a series permeable structure, requiring air to pass through two layers of permeable material before entering the filter rod. The overall air permeability process exhibits a significant spatial coupling effect. The ventilation rate is influenced by multiple parameters, including material permeability, area layout, width of the glue-free area, circumferential overlap sealing range, perforation parameters (including the axial position of the perforated area), and cigarette draw resistance. Among these, the number of perforation rows and the number of perforations directly determine the effective open area of the perforated area, while the axial position of the perforated area directly determines the flow path length and friction resistance of the infiltrated air inside the filter rod. The closer the perforation location is to the filter tip end (clamping...), the more effective the air permeability. Outside of the holding section and the overlapping section, the shorter the path and the higher the ventilation rate, the closer to the tobacco end, the longer the path and the lower the ventilation rate. The axial width of the glue-free breathable area determines the range of the perforated area and the size of the glue-free area around the perforation that can diffuse laterally. The wider the glue-free breathable area, the larger the glue-free non-perforated area around the perforated area, and the more sufficient the lateral air diffusion. The circumferential overlap width between the tipping paper and the cigarette paper indirectly affects the actual air intake area of the glue-free area and the paper ventilation by compressing the effective breathable arc length. Under the condition that other conditions remain unchanged, the more perforated rows and the larger the number of perforations, the stronger the overall breathability of the perforated area and the higher the ventilation rate. The closer the perforation position is to the filter end, the wider the glue-free area, and the smaller the circumferential overlap width, the larger the effective breathable area and the shorter the flow path, and the higher the ventilation rate. Conversely, the lower the ventilation rate, the lower the effective breathable area and the shorter the flow path. This is also the core reason why this embodiment uses a combination of finite element mechanism simulation and graph neural network intelligent prediction to carry out accurate calculations.
[0035] In S120, the finite element model of the splicing paper is established based on the geometric parameters, material permeability parameters, suction resistance parameters, and perforation parameters as follows:
[0036] Based on the material's air permeability parameters and perforation parameters, the equivalent air permeability and normal viscous resistance coefficient of the tipping paper are calculated. The equivalent air permeability of the perforated air permeable zone is determined by converting the opening ratio calculated from the number of perforation rows, the number of perforations per row, and the diameter of a single hole. Based on the geometric parameters and suction resistance parameters, the viscous resistance coefficient and inertial resistance coefficient of each specified segment are calculated. The tipping paper is divided into multiple continuous specified segments along its axial direction. Based on the initial geometric model of the tipping paper, a finite element simulation is performed in finite element software. The fluid medium is air at normal temperature and pressure. Each pair of viscous resistance coefficients and inertial resistance coefficients is used as the amplitude of the porous medium region. The boundary conditions are the equivalent air permeability and normal viscous resistance coefficient, as well as the preset outlet volumetric flow rate and ambient pressure.
[0037] In this embodiment, the purpose of this step is to convert the measured permeability, suction resistance parameters, and perforation parameters into resistance coefficients that can be directly called by the porous media module in the finite element simulation, ensuring that all model parameters are based on measured data rather than empirical values.
[0038] The non-adhesive area of the tipping paper and the high-permeability forming paper below form a series permeable structure. Air must pass through the tipping paper and the forming paper successively before entering the filter rod. Therefore, the overall equivalent air permeability of the non-adhesive area is calculated based on the principle of superposition of series permeability resistance:
[0039]
[0040] The measured air permeability of the glue-free area of the tipping paper and the measured air permeability of the high-permeability forming paper at the corresponding position are used to calculate the composite equivalent air permeability of the glue-free area. The air permeability of the perforated area and the air permeability of the cigarette paper can be directly obtained from the measured values. The air permeability of the perforated area is determined by calculation based on the perforation parameters. The effective opening rate of the perforated area is calculated according to the number of perforation rows, the number of perforations per row, and the diameter of a single hole, and then converted to obtain the overall equivalent air permeability of the perforated area. At the same time, the coverage position of the perforated air permeable area on the tipping paper axis is determined by the axial position of the perforated area in the perforation parameters. That is, the specific range of the perforated air permeable area on the filter rod axis is locked according to the starting coordinate of the axial axis of the perforated area and the axial width of the perforated area (calculated from the number of perforation rows and the axial spacing of each row). This range must fall completely within the axial range of the glue-free air permeable area and avoid the overlap section between the tipping paper and the cigarette paper to ensure that the layout position of the perforated air permeable area in the finite element model is consistent with the actual process. The axial position of the glue-free air permeable area is determined by extending to both sides from the perforated area as the center, and its axial width is greater than the axial width of the perforated air permeable area, so that the perforated area is completely contained within the glue-free air permeable area. The air permeability of the cigarette paper can be directly obtained from the measured value. It should be noted that the effective air permeability area of both the glue-free air permeable area and the cigarette paper air permeable area must deduct the circumferential overlap sealing section: the effective air permeability arc length of the glue-free area of the tipping paper is equal to the circumference of the cigarette butt minus the circumferential overlap width of the tipping paper, and the effective air permeability arc length of the cigarette paper air permeable area is equal to the circumference of the cigarette butt minus the circumferential overlap width of the cigarette paper. The effective air permeability area calculated in this way is the true area involved in air intake, avoiding overestimation of the air intake volume of the glue-free area and the paper ventilation due to including the circumferential glued overlap section in the air permeability area.
[0041] To determine the air permeability of the high-permeability forming paper at the corresponding location, the composite equivalent air permeability of the glue-free area is calculated. The air permeability of the perforated area and the cigarette paper can be directly calculated using the measured values. The air permeability of the perforated area is determined by combining the perforation parameters. The effective open area ratio of the perforated area is calculated based on the number of perforation rows, the number of perforations per row, and the diameter of each hole, and then converted to obtain the overall equivalent air permeability of the perforated area. The air permeability of the cigarette paper can be directly calculated using the measured values.
[0042] For thin, breathable materials such as tipping paper layers, cigarette paper layers, and forming paper layers, based on Darcy's law of permeability, the air permeability is converted into the normal viscous resistance coefficient of the porous medium:
[0043]
[0044] in The aerodynamic viscosity at room temperature. For the corresponding material thickness, The air permeability of the corresponding area (the above composite equivalent air permeability is used for the glue-free area). At the same time, the in-plane lateral drag coefficient is set to be 10 times that of the normal drag, i.e. It restricts the unrestricted lateral diffusion of air within the paper surface, conforming to the actual fiber permeation characteristics.
[0045] The resistance coefficient fitting of the filter rod and tobacco section is based on the measured total draw resistance of the filter rod and the total draw resistance of the tobacco section. Combined with the length, cross-sectional area and standard draw flow rate of the corresponding section, the viscous resistance coefficient of each section is obtained by fitting using the Darcy-Forchheimer equation. With inertial drag coefficient This serves as the basis for assigning values to the internal porous media region.
[0046] Upon completion of this stage, the equivalent air permeability and a complete set of porous media resistance parameters for all regions are obtained. The air permeability and resistance coefficient of the perforated area are calculated from the perforation parameters and serve as the core input for subsequent model assignment. Specifically: the equivalent air permeability of the perforated air permeable area is obtained through an open area ratio conversion. First, the effective open area ratio of the perforated area is calculated based on the perforation parameters: the total open area of all ventilation holes in the perforated area is calculated from the number of perforation rows, the number of perforations per row, and the diameter of each hole, and then divided by the total area of the perforated area to obtain the open area ratio. Subsequently, based on the air permeability of the tipping paper substrate itself, the overall equivalent air permeability of the perforated area is calculated by combining the open area ratio and the air permeability characteristics of each hole. The higher the open area ratio, the greater the overall equivalent air permeability of the perforated area, and vice versa. The equivalent air permeability obtained in this way accurately reflects the overall air permeability of the perforated area under different numbers of perforation rows and perforations, ensuring that the air permeability properties of the perforated area in the finite element model are consistent with the actual perforation conditions.
[0047] In CFD finite element simulation software, a full three-dimensional computational domain comprising three core regions is constructed based on measured cigarette size parameters:
[0048] External environment domain: A cylindrical air domain extending outward from the cigarette as the center, 5 to 10 times the diameter of the cigarette, to simulate the atmospheric environment around the cigarette, ensuring sufficient development of air intake flow and avoiding boundary effects from interfering with the accuracy of air permeability calculation.
[0049] The internal domain of the cigarette is divided into two independent fluid domains along the axial direction: the tobacco section and the filter rod section, which correspond one-to-one with the internal structure of a real cigarette.
[0050] Delineation of the breathable thin layer and sealing area: Multiple continuous designated functional segments are divided along the tipping paper axis, with differentiated geometric processing. Thin-layer solid domains are established in the glue-free breathable area of the tipping paper (including its internal perforated breathable area) and at the corresponding positions on the cigarette paper. These domains are used to set the breathability properties of the porous medium. The perforated breathable area is a sub-region within the glue-free breathable area, and its breathability intensity is determined by perforation parameters such as the number of perforations and the number of perforations. The axial coverage position of the perforated breathable area is dynamically determined by the axial position of the perforated area in the perforation parameters and falls completely within the glue-free breathable area. Its configurable range is the filter rod segment area outside the 9mm clamping sealing area at the filter tip end and not overlapping with the tipping paper and cigarette paper overlap section. During modeling, the specific axis of the perforated breathable thin layer is locked based on the axial starting coordinate and axial width of the perforated area. In the axial direction, the axial coverage of the glue-free breathable zone is determined by the axial width of the glue-free breathable zone in the geometric parameters. Its axial width is greater than the axial width of the perforated breathable zone, and it expands to both sides with the perforated zone as the center. The tipping paper adhesive sealing zone, the 9mm clamping sealing zone at the end of the filter tip, the overlapping sealing zone of the tipping paper and the cigarette paper, as well as the circumferential overlapping area of the tipping paper and the circumferential overlapping area of the cigarette paper, do not have a breathable thin layer and are directly defined as solid walls to block the air infiltration channel. Among them, the overlapping sealing zone of the tipping paper and the cigarette paper will not have ventilation holes due to the double-layer paper stacking and the adhesive sealing itself. The circumferential overlapping area of the tipping paper is the adhesive section that overlaps at the beginning and end after the tipping paper wraps around the filter rod, and the circumferential overlapping area of the cigarette paper is the adhesive section that overlaps at the beginning and end after the cigarette paper wraps around the tobacco section. Both are distributed in a narrow strip along the circumference.
[0051] After completing the geometry construction, all computational domains and boundary surfaces are standardized and named, such as tobacco section domain, filter rod section domain, tipping paper breathable layer domain, perforated breathable layer domain, environmental pressure inlet, tobacco outlet, and adhesive sealing wall, to facilitate subsequent batch assignment and automatic result extraction.
[0052] The continuous geometric domain is discretized into computational units, and a hybrid structured and unstructured meshing strategy is adopted to balance computational accuracy and solution efficiency.
[0053] Main region mesh: The main regions, such as the external environment domain, tobacco shreds, and filter rods, are covered by a structured hexahedral mesh to ensure computational stability and solution efficiency.
[0054] Critical area densification: For the 50μm thick permeable thin layer area, a boundary layer densification method is adopted, dividing the area into 3 to 5 grid layers along the thickness direction to ensure the calculation accuracy of normal seepage flow; the grid in the perforated area can be further locally densified according to the perforation density to accurately capture the flow characteristics near the orifice; at the same time, a first grid height adapted to low Reynolds number flow is set in the near-wall area to match the flow characteristics of low-speed laminar flow of cigarettes.
[0055] Mesh quality verification: After mesh generation, check the mesh orthogonality quality, skewness, aspect ratio, and other indicators. The orthogonality quality should be greater than 0.3 and the skewness should be less than 0.7 to ensure that the mesh quality meets the requirements for computational convergence.
[0056] Assign physical properties to all computational domains, define inlet and outlet flow conditions, configure solution rules, and complete the full setup of the calculation case, corresponding to the user-specified medium, resistance amplitude, and boundary parameter configuration logic:
[0057] Fluid medium definition: Normal temperature and pressure air is selected as the flow medium, set as an incompressible Newtonian fluid, and assigned the corresponding air density and dynamic viscosity to replace real flue gas for flow simulation.
[0058] Porous media region assignment:
[0059] Tobacco section and filter rod section: These are set as isotropic porous media regions, and the viscous resistance coefficient and inertial resistance coefficient obtained from the previous fitting are assigned values respectively to simulate the flow resistance characteristics of the tobacco and filter rod bundles.
[0060] The tipping paper breathable thin layer and cigarette paper breathable thin layer are set as an anisotropic porous medium. The viscous resistance coefficient is calculated by assigning a normal value, and the viscous resistance coefficient in the two in-plane directions is assigned 10 times that of the normal. The inertial resistance coefficient is set to 0 (the influence of the inertial term can be ignored under low-speed seepage), thus restoring the seepage characteristics of breathable paper of "normal air permeability and lateral restriction". The normal resistance coefficient of the perforated breathable thin layer is calculated by perforation parameters. The more rows of perforations and the larger the number of perforations, the smaller the corresponding resistance coefficient and the stronger the air permeability.
[0061] Boundary condition settings:
[0062] Environmental boundary outside the domain: set as the pressure inlet boundary, with a gauge pressure of 0 Pa, corresponding to the real atmospheric environment.
[0063] Smoke outlet end face: set as volumetric flow rate outlet, constant flow rate 17.5mL / s, corresponding to the average flow rate of ISO standard suction conditions.
[0064] Non-air-permeable areas such as the adhesive area, clamping section, axial overlap section, circumferential overlap section of the tipping paper, circumferential overlap section of the cigarette paper, and tobacco ends are set as non-slip solid walls with a normal velocity of 0 to prevent air infiltration.
[0065] Interface between different areas: set as an internal interface to ensure continuous transmission of flow and pressure.
[0066] Solver configuration: A pressure-based steady-state solver is used; the pressure-velocity coupling uses the SIMPLE algorithm; momentum, pressure, and other terms are calculated using a second-order upwind scheme; the residual convergence criteria are set to continuity and velocity residuals ≤ 10. -6 The flow model selected is a laminar flow model, which is suitable for the laminar flow characteristics of low cigarette smoking velocity and small Reynolds number.
[0067] After completing the full configuration of the above four stages, a complete finite element simulation model of the cigarette with perforated paper is obtained. It can be directly submitted to the solver for iterative calculation. After convergence, the first air intake distribution matrix and the first ventilation rate result are extracted.
[0068] In this embodiment, the configured finite element model is submitted to the CFD solver for steady-state iterative calculation, until the continuity and velocity residuals decrease to 10. -6Below, after monitoring quantities such as outlet static pressure and total air intake volume fluctuate by less than 0.1% for 200 consecutive steps and meet the convergence criteria, the first ventilation rate and the first air intake volume distribution matrix are extracted through post-processing. The first ventilation rate is calculated based on the area flow integral. The normal air intake flow of the tipping paper permeable layer and the cigarette paper permeable layer are respectively integrated to obtain the total air intake of the corresponding area. The integration range is strictly limited to the effective permeable area after deducting the circumferential overlap sealing section. The integration interval of the perforated permeable area is dynamically adjusted with the axial position of the perforated area and falls completely within the glue-free permeable area and avoids the overlap section of the tipping paper and the cigarette paper. The integration interval of the glue-free permeable area covers the entire glue-free area including the perforated area. Then, it is calculated as a ratio with the total suction flow of 17.5 mL / s set at the outlet to obtain the filter ventilation rate and paper ventilation rate. The sum of the two is the total ventilation rate, which together constitutes the first ventilation rate as the mechanism benchmark. The first air intake volume distribution matrix is completely aligned with the graph structure model. Mesh resolution extraction involves generating corresponding sampling meshes on the ventilated surface of the splice paper according to the division rules of axial M segments and circumferential N segments. The circumferential mesh needs to mark the mesh position corresponding to the circumferential overlap sealing section of the splice paper and set its air intake to zero. The axial mesh needs to mark the mesh range corresponding to the perforated ventilated area within the glue-free ventilated area according to the axial position of the perforated area, and mark the mesh range corresponding to the non-perforated sub-area within the glue-free ventilated area. The normal air intake velocity of each effective ventilated mesh position is extracted one by one and converted into the air intake of a single mesh. The matrix is arranged in row and column order into a two-dimensional matrix of M rows and N columns, which is the first air intake distribution matrix that perfectly matches the dimension of the deep learning prediction result. The air intake distribution characteristics of the perforated area in the matrix directly reflect the influence of the number of perforation rows, perforation density, and perforation axial position on the local ventilation intensity. The air intake distribution characteristics of the glue-free non-perforated area reflect the combined effect of the axial width of the glue-free area, the inclusion relationship of the perforated area, and the circumferential effective ventilated arc length, which can be directly used for subsequent comparison and verification of the consistency of the flow field distribution.
[0069] In S120, the graph structure model is established as follows:
[0070] Based on the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters of the tipping paper, a feature vector of each first grid of the tipping paper is obtained, wherein the tipping paper is divided into multiple first grids along its axial and circumferential directions.
[0071] Among the geometric parameters, material permeability parameters, suction resistance parameters, and perforation parameters, engineering parameters that have an adjusting effect on ventilation rate are selected to form a global operating condition feature vector;
[0072] The graph structure model is established based on the first grid, wherein the nodes of the graph structure model are the first grids, the node features are the corresponding feature vectors and global working condition feature vectors, and the initial weights of the edges are obtained based on the specified regions of the splicing paper where the two connected first grids are located. The splicing paper is divided into multiple continuous specified regions along its axial direction.
[0073] After the cylindrical splicing paper is unfolded into a two-dimensional rectangular plane along the circumference, it is uniformly divided into multiple smallest air-permeable units of the same size along the axial and circumferential directions according to a preset resolution. These are the first grids, each grid corresponding to a unique axial and circumferential position coordinate. Then, based on the coordinate range of each functional area in the geometric parameters, the designated area of the splicing paper to which each first grid belongs is determined. The area air permeability and material thickness in the material air permeability parameters are then matched. The relative position coordinates of the grid itself, the air permeability of the high-permeability forming paper at the corresponding position, and the perforation structure features are superimposed. After all these are normalized, they are spliced into a feature vector of fixed length, so that each first grid fully carries its own area attributes, air permeability, spatial position, coupled material information, and perforation density features, forming the local feature basis of the graph model.
[0074] From all input parameters, engineering parameters that do not change with the spatial position of the tipping paper and have a global regulatory effect on the overall ventilation rate of the entire cigarette are selected to form a global operating condition feature vector. The core principle of the selection is that the parameter effect covers the entire cigarette and there are no local distribution differences. Specifically, this includes geometric parameters such as cigarette diameter, total cigarette length, filter rod length, tobacco segment length, axial width of the glue-free ventilation zone, circumferential overlap width of the tipping paper, and circumferential overlap width of the cigarette paper; suction resistance parameters such as total suction resistance of the filter rod and total suction resistance of the tobacco segment (or tobacco density conversion value); perforation parameters such as total number of perforation rows, total number of perforations, and axial position of the perforation area; and operating condition parameters such as the suction volume flow rate. Among these, the axial position of the perforation area determines the perforation. The axial layout of the ventilation zone within the glue-free ventilation zone requires modeling to ensure that the perforated area falls entirely within the glue-free ventilation zone and does not overlap with the tipping paper and cigarette paper overlap. The axial width of the glue-free ventilation zone is greater than the axial width of the perforated ventilation zone. All selected parameters are normalized by min-max to eliminate dimensional differences and then spliced into a one-dimensional vector in a fixed order. This vector is used to adjust the overall magnitude of the ventilation rate from a global perspective, allowing the model to adapt to the cigarette calculation needs of different specifications, different glue-free zone widths, different overlap sealing ranges, different perforation axial positions and perforation zone layouts, different perforation schemes, and different suction conditions. This compensates for the deficiency that relying solely on local mesh features cannot reflect the overall resistance and operating condition effects.
[0075] Using all first-order grids as nodes in the graph structure model, the local feature vector and global operating condition feature vector corresponding to each grid are concatenated to form the complete input feature of that node, allowing each node to simultaneously carry local air permeability attributes, perforation density features, and global operating condition information. The four-neighborhood rule is then used to define the edges between nodes; that is, a connection is established when two first-order grids share a common edge, corresponding to the physical channel for lateral air diffusion between adjacent units. Simultaneously, the initial weights of the edges are set based on multiple consecutive designated regions divided along the splicing paper axis: when two grids belonging to the same air permeability region are connected, the weight is set to a higher value, representing... With low lateral diffusion resistance and easier air exchange, the weights of the connected permeable and sealed areas are set to extremely low values, indicating that lateral diffusion is basically blocked. The initial weights of the edges of adjacent grids belonging to the perforated permeable areas need to be further corrected based on the local perforation density of the corresponding positions of the two grids. The higher the perforation density, the higher the weight, which more accurately reflects the lateral diffusion characteristics of densely perforated areas. Finally, the weighted nodes and connecting edges together form a complete graph structure model, which transforms the spatial permeable structure, perforation distribution law and diffusion physical logic of the splicing paper into a data structure that can be directly recognized and processed by the graph neural network.
[0076] The tipping paper is divided into four continuous designated areas along the axial direction, arranged sequentially from the filter tip to the tobacco end: a 9mm clamping and sealing area at the filter tip end, an overlap sealing area between the tipping paper and the cigarette paper, a non-adhesive breathable area, and an adhesive sealing area. The clamping and sealing area is airtight due to the compaction of the cigarette paper; the overlap sealing area is essentially airtight due to the double-layered paper and adhesive layer, and this area itself does not have ventilation holes; the adhesive sealing area is airtight because the adhesive blocks the air passages; and the non-adhesive breathable area allows outside air to permeate into the filter. The core ventilation channel of the rod; the glue-free ventilation zone is specially designed to ensure effective perforation of the splice paper. The perforation ventilation zone is a sub-region within the glue-free ventilation zone and must be located within the axial range of the glue-free ventilation zone. The axial width of the glue-free ventilation zone is greater than the axial width of the perforation ventilation zone (the glue-free ventilation zone is slightly wider than the perforation ventilation zone) to ensure that the perforation position falls completely within the glue-free area and to avoid the glue layer blocking the ventilation holes. The forming paper corresponding to the glue-free ventilation zone uses high-permeability forming paper to ensure smooth air penetration; perforation ventilation zone The air permeability is determined by perforation parameters such as the number of perforations, the number of perforations, and the perforation diameter. The axial arrangement of the perforated air permeable area within the glue-free air permeable area is determined by the axial position of the perforated area. It can be flexibly configured in the filter rod section area outside the 9mm clamping and sealing area at the end of the filter tip and does not overlap with the overlap section of the tipping paper and cigarette paper. Different positions will result in synchronous changes in the axial coordinate range of the corresponding first grid. The axial coverage of the glue-free air permeable area can be configured with variable width. The axial coordinate range of the four types of areas are all derived from geometric parameters and perforation parameters, which are the core basis for determining the attributes of each first grid area. In addition to the axial area division, the tipping paper and cigarette paper also have circumferential overlap sealing sections in the circumferential direction, respectively. These areas are distributed in a narrow strip along the circumference and are completely air-permeable with glue. They need to be marked separately in the determination of the area attributes of the first grid. The air permeability of the corresponding grid is set to 0, and the initial weight of the edge is set to an extremely low value to reflect the blocking effect of the circumferential overlap section on lateral diffusion and effective air permeable area.
[0077] The feature vector corresponding to each first grid is composed of six types of information concatenated together. All numerical terms have been normalized by min-max and uniformly mapped to the 0~1 interval to ensure the stability of model training and inference. Specifically, this includes: First, region category information, converting the axially specified region to which the grid belongs into a 4D unique thermal code (corresponding to clamping sealing area, overlapping sealing area, glue-free breathable area, and adhesive sealing area), and adding a 1D perforated area identifier (indicating whether the grid belongs to the perforated breathable area within the glue-free breathable area) and a 1D circumferential overlapping seal identifier, used to identify the breathability attribute category of the grid, whether it is a perforated area within the glue-free area, and whether it belongs to the circumferential overlapping seal section of the splice paper; Second, splice paper air permeability parameters, corresponding to the air permeability value of the specified region to which the grid belongs. The perforated area within the glue-free area takes the equivalent air permeability calculated based on the perforation parameter, the non-perforated area within the glue-free area takes the glue-free area air permeability, and the sealing area and circumferential overlapping section take 0. The data comes from the material air permeability parameters and perforation parameters; Third, material thickness parameters, i.e., the thickness value of the splice paper at the corresponding position, the data comes from geometric parameters; Fourth, spatial position information, including the axial relative coordinates and circumferential relative coordinates of the grid, used to identify the grid in the splice... The spatial position on the paper is calculated from the geometric parameters and the axial position of the perforated area. The axial coordinates need to reflect the start and end positions of the glue-free breathable area and the start and end positions of the perforated breathable area after dynamic adjustment of the axial position of the perforated area within the glue-free breathable area (and avoid the overlap section between the tipping paper and the cigarette paper). The circumferential coordinates need to reflect the start and end positions of the circumferential overlap section of the tipping paper. Fifth, the high-permeability forming paper coupling permeability, that is, the permeability of the high-permeability forming paper below the corresponding position of the grid. The data comes from the material permeability parameters, reflecting the series permeability coupling relationship between the tipping paper and the forming paper. The forming paper corresponding to the glue-free breathable area is made of high-permeability forming paper. Sixth, the perforation structure characteristics, including the number of perforation rows to which the grid belongs, the local opening rate, the proportion of the number of perforations in the row, and the axial relative position of the perforated area (that is, the ratio of the axial offset of the grid in the perforated breathable area to the total axial width of the perforated area). The data comes from the perforation parameters and is used to characterize the perforation density, permeability intensity difference of the grid position, and the layout position of the perforated area in the glue-free breathable area. The above information is concatenated in a fixed order to form a single, fixed-length first grid feature vector.
[0078] Based on this, in S130, the second intake volume distribution matrix and the second ventilation rate can be obtained through a graph neural network, such as... Figure 2 As shown, the specific steps include:
[0079] S131: Map the node features of each of the first grids to a high-dimensional space to obtain high-dimensional node features;
[0080] S132: Based on the high-dimensional node features of the adjacent first network, the learnable attention vector, and the corresponding initial weights, the contribution ratio of the neighboring nodes to the first grid is obtained. The contribution ratio is used to characterize the degree of influence of the air permeability of the adjacent nodes on the air intake of the first grid. The contribution ratio is positively correlated with the degree of influence.
[0081] S133: Aggregate the contribution ratio and node features corresponding to each of the first grids, and perform non-linear activation processing on the aggregated contribution ratio and node features to obtain the first node feature matrix of each of the first grids;
[0082] S134: Perform a convolution operation on the first node feature matrix to obtain the second node feature matrix;
[0083] S135: Input the second node feature matrix into the local branch, and the local branch obtains the second intake volume distribution matrix based on the second node feature matrix;
[0084] S136: Input the second node feature matrix into the global branch, and the global branch obtains the second ventilation rate based on the second node feature matrix.
[0085] In this embodiment, in S131, the initial node feature vector of each first grid is used as input. Through a set of learnable shared linear weight matrices, a unified linear mapping operation is performed on the features of all grids. The low-dimensional features that originally contain regional attributes, air permeability, material thickness, location information and global operating parameters are transformed into a higher-dimensional hidden feature space, so that the air permeability-related attributes of each grid can be more fully abstracted and expressed. All first grids share the same set of mapping rules to ensure that the features of different grids are in a unified metric space. Finally, the high-dimensional node features corresponding to each grid are output, providing a feature basis for the subsequent calculation of the influence of adjacent units.
[0086] In this embodiment, S132 includes the following specific steps:
[0087] Based on the high-dimensional node features, the learnable attention vector, and the corresponding initial weights, the attention scores of neighboring nodes to the first grid are calculated; all the attention scores are then min-max normalized to obtain the contribution ratio of neighboring nodes to the first grid.
[0088] For each pair of adjacent first grids, the high-dimensional node features of the two grids are first concatenated into a combined feature vector. The initial influence score is then calculated by combining the learnable attention vector. This initial influence score is then multiplied by the preset initial edge weights between the two grids for correction. The initial edge weights are determined by the type of splicing paper region to which the two grids belong. Adjacent grids belonging to the same ventilated region have higher weights, while grids adjacent to the ventilated and sealed regions have extremely low weights. Based on this, the influence scores of all neighbors of each first grid are subjected to softmax normalization so that the sum of the contribution percentages of all neighbors is 1. Finally, the contribution percentage of each neighbor grid to the current first grid is obtained. The higher the percentage, the stronger the influence of the ventilated state of the adjacent node on the lateral diffusion of the air intake of the current grid.
[0089] In this embodiment, in S133, for each first grid, the calculated contribution ratio of each neighbor is used as a weighting coefficient, and the high-dimensional node features of all its adjacent grids are weighted and summed to complete the aggregation of neighbor air permeability information, allowing the current grid to incorporate the air permeability influence of surrounding adjacent units; then, the aggregated feature vector is input into a nonlinear activation function for transformation, introducing nonlinear expressive power, so that the model can fit the complex lateral diffusion coupling law in the splicing paper plane; after all first grids have completed aggregation and activation, they are arranged in the order of node index into a two-dimensional matrix form, that is, the first node feature matrix is obtained. At this time, the features of each grid have incorporated its own attributes and the air permeability influence of directly adjacent units.
[0090] In this embodiment, in S134, the first node feature matrix is used as input to perform a new round of graph attention convolution operation. The features are linearly transformed through a new learnable weight matrix, the attention contribution ratio between adjacent grids is recalculated, and then weighted aggregation and nonlinear activation are completed. After this round of convolution operation, the feature perception range of each first grid is expanded from directly adjacent first-order neighbors to second-order neighbors, that is, neighbors' neighbors. It can capture the indirect seepage transmission effect of air across multiple grid cells. Finally, the second node feature matrix containing medium and long distance spatial coupling information is output, providing more comprehensive feature support for the accurate prediction of subsequent air intake distribution and ventilation rate.
[0091] In S135, the specific steps are as follows:
[0092] The local branch performs a Sigmoid mapping on the features of the second node to obtain the second intake volume distribution matrix.
[0093] Specifically: The local branch takes the second node feature matrix output by graph convolution as input. Each row of this matrix corresponds to a high-dimensional hidden feature of the first grid, which contains comprehensive information on the grid's own air permeability and global spatial coupling. Then, for each high-dimensional hidden feature of the first grid, a linear transformation is performed through a shared fully connected layer of weights and biases to reduce the high-dimensional feature to a single value. This value is then input into the Sigmoid activation function for mapping, constraining the output value to the range of 0 to 1, thus obtaining the normalized air intake volume corresponding to each grid. All grids complete this calculation in parallel, resulting in a one-dimensional air intake volume sequence with the same number of grids as the first grid. Finally, according to the row and column arrangement order of the first grid along the axial and circumferential directions, the one-dimensional air intake volume sequence is reshaped into an M-row N-column two-dimensional matrix, namely the second air intake volume distribution matrix. The position of each element in the matrix corresponds one-to-one with the spatial position of the corresponding first grid on the splicing paper. The value represents the relative air permeability intensity at that position. Its dimension is completely aligned with the first air intake volume distribution matrix obtained by finite element simulation, and it can be directly used for subsequent flow field distribution consistency verification.
[0094] In S136, the specific steps are as follows:
[0095] The global branch performs global average pooling on all the second node feature matrices to obtain a global feature vector; the global branch inputs the global feature vector into a fully connected layer, and the fully connected layer obtains the second ventilation rate by performing a Sigmoid mapping on the global feature vector.
[0096] Specifically:
[0097] The global branch takes the feature matrix of the second node as input. First, it performs a global average pooling operation on all the high-dimensional hidden features corresponding to the first grid, that is, it takes the arithmetic mean of the values of each feature dimension along the node dimension, and aggregates the two-dimensional feature matrix corresponding to V grids into a single one-dimensional global feature vector. This vector condenses the comprehensive information of the air permeability structure properties and lateral diffusion coupling of the entire tipping paper, and can characterize the overall air permeability level of the tipping paper. Then, the global feature vector is fed into the fully connected layer, and the global features are linearly transformed and nonlinearly mapped through learnable weight matrices and bias terms, gradually reducing the dimensionality to the target output dimension. Finally, the output value is constrained to the range of 0 to 1 through the Sigmoid activation function, and the normalized values corresponding to the total ventilation rate, filter ventilation rate and paper ventilation rate are obtained, that is, the second ventilation rate. Its output diameter is completely aligned with the first ventilation rate obtained by finite element simulation, and can be directly used for subsequent numerical deviation comparison and final ventilation rate determination.
[0098] Based on this, the graph neural network mentioned in this embodiment is obtained through supervised end-to-end training. All training data is generated in batches from high-precision finite element simulation models that have undergone actual measurement and calibration. The values of the following parameters are adjusted through parametric traversal: the geometric parameters of the tipping paper (including the axial width of the glue-free breathable area, the circumferential overlap width of the tipping paper, and the circumferential overlap width of the cigarette paper), material air permeability parameters, draw resistance parameters (including the converted value of tobacco density), and perforation parameters (including the axial position of the perforated area). The traversal range of the axial position of the perforated area is strictly limited to the filter rod section outside the 9mm clamping and sealing area at the end of the filter tip and does not overlap with the overlap section of the tipping paper and cigarette paper. Furthermore, the perforated breathable area must completely fall within the axial range of the glue-free breathable area, and the axial width of the glue-free breathable area is greater than the axial width of the perforated breathable area. This covers typical working conditions with different area sizes, glue-free area widths, overlap sealing ranges, perforation axial positions, and perforation area layouts, air permeability levels, and draw resistance levels, generating a large number of one-to-one corresponding graph structure model input samples and... The finite element simulation baseline labels include the first intake volume distribution matrix and the first ventilation rate under the corresponding working condition. These labels are divided into training and validation sets for model training and accuracy verification. The training adopts a dual-branch joint loss function, which is composed of a weighted sum of the ventilation rate numerical loss and the intake volume distribution loss. The ventilation rate branch uses the mean square error loss to constrain the numerical deviation between the predicted value and the baseline value. The intake volume distribution branch combines the mean square error loss and the negative Pearson correlation coefficient loss, while simultaneously constraining the numerical accuracy and morphological consistency of the distribution. This guides the model to learn the magnitude law of the ventilation rate and the spatial distribution physical characteristics of the intake volume. During the training process, graph structure samples are input into the network in batches. The forward propagation outputs the second intake volume distribution matrix and the second ventilation rate prediction results. After calculating the joint loss value, the learnable weights in the network are updated layer by layer through backpropagation and gradient descent algorithms. The training is iterated repeatedly until the validation set loss converges and no longer decreases. Finally, a pre-trained graph neural network model with strong generalization ability and satisfactory prediction accuracy is obtained.
[0099] This graph neural network employs an end-to-end architecture with two-layer graph attention convolutions and two-branch outputs. The input is a paper-based structural model with regional diffusion weights, using the first grid as nodes. Node features integrate local features such as regional attributes, air permeability, and location information of each grid, along with global operating condition features such as cigarette draw resistance and size. The first-layer graph attention convolution, through shared linear transformations, attention weight calculations, and weighted aggregation of neighborhood features, allows each grid to incorporate the air permeability effects of directly adjacent units, simulating direct lateral air diffusion. The second-layer graph attention convolution further executes the functions based on the features from the first layer. One round of convolution operation expands the sensing range to second-order neighbors, capturing the indirect seepage transmission effect across cells, and finally outputs a node feature matrix containing global spatial coupling information. Subsequently, parallel local and global branches complete the result calculation. The local branch performs a fully connected transformation and Sigmoid mapping on each grid feature and reshapes it into a two-dimensional matrix, outputting the second intake volume distribution matrix. The global branch performs global average pooling aggregation on all grid features and then maps it to Sigmoid through a fully connected layer to output the second ventilation rate, realizing the synchronous prediction of local flow field distribution and global ventilation index.
[0100] In S150, the specific process is as follows:
[0101] When the absolute value of the difference between the first ventilation rate and the second ventilation rate is greater than or equal to the difference threshold, and the Pearson coefficient is greater than or equal to the coefficient threshold, the final ventilation rate is the second ventilation rate. When the absolute value of the difference between the first ventilation rate and the second ventilation rate is less than the difference threshold, or when the Pearson coefficient is less than the coefficient threshold, the final ventilation rate is the first ventilation rate.
[0102] The reason for setting up this dual judgment rule is to simultaneously consider the efficiency and reliability of ventilation rate calculation, avoiding the limitations of a single indicator: the deviation of the ventilation rate value only reflects the degree of agreement of the overall magnitude, and cannot verify whether the prediction result conforms to the actual physical laws of air diffusion. It is possible that the value happens to be close, but the flow field distribution is completely distorted. On the other hand, the Pearson coefficient only reflects the consistency of the spatial distribution of the intake air volume, and cannot guarantee the accuracy of the overall ventilation rate value. Only by combining the two can the numerical accuracy and physical rationality of the prediction result be verified at the same time. When both indicators meet the standard, it means that the prediction result of the graph neural network is not only close to the finite element simulation benchmark, but also highly consistent with the spatial distribution law of the intake air volume and the actual seepage process. The prediction result is reliable. At this time, the second ventilation rate with faster calculation speed is used as the final result, which can give full play to the efficiency advantage of deep learning. If either indicator fails to meet the standard, either the numerical deviation exceeds the allowable range, or the flow field distribution is inconsistent and the prediction has physical distortion, and the reliability of the result is insufficient. At this time, the first ventilation rate obtained by finite element simulation is used as the final result, and the accuracy of the result is guaranteed by the mechanism simulation, thereby achieving a balance between calculation efficiency and accuracy.
[0103] This application provides an electronic device that may include a processing module and a memory. The memory stores a computer program, which, when executed by the processor, enables the electronic device to perform the corresponding steps in the above-described method for determining the ventilation rate of a dispensed cigarette.
[0104] In this embodiment, the processor can be an integrated circuit chip with signal processing capabilities. For example, the processor can be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application.
[0105] The memory can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc.
[0106] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the electronic device described above can be referred to the corresponding steps in the aforementioned method, and will not be elaborated further here.
[0107] This application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the method for determining the ventilation rate of the dispensed cigarette as described in the above embodiments.
[0108] Computer-readable storage media may be magnetic disks, optical disks, read-only memory, random access memory, flash memory, USB flash drives, hard disks, or solid-state drives, etc., and may also include combinations of the above types of memory. It is understood that computers, processors, microprocessor controllers, or programmable hardware include storage components capable of storing or receiving software or computer code, which, when accessed and executed by the computer, processor, or hardware, implement the methods shown in the above embodiments.
[0109] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps in the above-described method for determining the ventilation rate of dispensed cigarettes. The computer program product may exist in a computer-readable storage medium in forms including, but not limited to, source files, executable files, and installation package files.
[0110] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This software product can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, electronic device, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.
[0111] In the embodiments provided in this application, it should be understood that the disclosed methods can also be implemented in other ways. The method embodiments described above are merely illustrative. For example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, program segment, or part of code, which includes one or more executable instructions for implementing a specified logical function. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions. Furthermore, the functional modules in the various embodiments of this application can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0112] The above description is merely an embodiment of this application and is not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.
Claims
1. A method for determining the ventilation rate of a tapped cigarette, characterized in that, The method includes: Obtain the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters of the tipping paper; Based on the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters, a finite element model and a graphical structural model of the splicing paper are established. The finite element model was used to perform finite element simulation to obtain the first air intake distribution matrix and the first ventilation rate of the splicing paper; The graph structure model is input into a pre-trained graph neural network, and the graph neural network outputs the second air intake distribution matrix and the second ventilation rate of the splicing paper. The final ventilation rate of the cigarette pack is determined based on the difference between the first ventilation rate and the second ventilation rate, and the Pearson coefficients of the first and second air intake distribution matrices.
2. The method according to claim 1, characterized in that, The step of determining the final ventilation rate of the cigarette pack based on the difference between the first ventilation rate and the second ventilation rate, and the Pearson coefficients of the first and second air intake distribution matrices, includes: When the absolute value of the difference between the first ventilation rate and the second ventilation rate is greater than or equal to the difference threshold, and the Pearson coefficient is greater than or equal to the coefficient threshold, the final ventilation rate is the second ventilation rate. When the absolute value of the difference between the first ventilation rate and the second ventilation rate is less than the difference threshold, or when the Pearson coefficient is less than the coefficient threshold, the final ventilation rate is the first ventilation rate.
3. The method according to claim 1, characterized in that, Based on the aforementioned geometric parameters, material air permeability parameters, and suction resistance parameters, a graphical structural model of the splicing paper is established, including: Based on the geometric parameters, material air permeability parameters, suction resistance parameters, and perforation parameters of the tipping paper, a feature vector of each first grid of the tipping paper is obtained, wherein the tipping paper is divided into multiple first grids along its axial and circumferential directions. Among the geometric parameters, material permeability parameters, suction resistance parameters, and perforation parameters, engineering parameters that have an adjusting effect on ventilation rate are selected to form a global operating condition feature vector; The graph structure model is established based on the first grid, wherein the nodes of the graph structure model are the first grids, the node features are the corresponding feature vectors and global working condition feature vectors, and the initial weights of the edges are obtained based on the specified regions of the splicing paper where the two connected first grids are located. The splicing paper is divided into multiple continuous specified regions along its axial direction.
4. The method according to claim 1, characterized in that, Based on the aforementioned geometric parameters, material permeability parameters, suction resistance parameters, and perforation parameters, a finite element model of the splicing paper is established, including: Based on the material's air permeability parameters, perforation parameters, and the axial width of the adhesive-free air permeable zone in the geometric parameters, the equivalent air permeability and normal viscous resistance coefficient of each region of the splicing paper are calculated. The equivalent air permeability of the perforated air permeable zone is determined by converting the opening ratio calculated based on the number of perforation rows, the number of perforations per row, and the diameter of a single hole. The axial coverage position of the perforated air permeable zone is determined based on the axial position of the perforated zone in the perforation parameters and is located within the axial range of the adhesive-free air permeable zone. The axial width of the adhesive-free air permeable zone is greater than the axial width of the perforated air permeable zone. The effective air permeable arc length of the adhesive-free air permeable zone is determined by deducting the circumferential overlap sealing section from the circumferential overlap width of the splicing paper. Based on the geometric parameters and suction resistance parameters, the viscous resistance coefficient and inertial resistance coefficient of each specified segment are calculated. The tipping paper is divided into multiple continuous specified segments along its axial direction, and tipping paper overlap sealing segments are divided along the circumferential direction. The cigarette paper is divided into cigarette paper overlap sealing segments along the circumferential direction. Based on the initial geometric model of the tipping paper, a finite element simulation was performed in finite element software. The fluid medium was air at normal temperature and pressure. Each pair of viscous drag coefficients and inertial drag coefficients was used as the amplitude of the porous medium region. The boundary conditions were the equivalent air permeability and normal viscous drag coefficient, as well as the preset outlet volume flow rate and ambient pressure. The circumferential overlap area of the tipping paper and the circumferential overlap area of the cigarette paper were both set as impermeable solid walls.
5. The method according to claim 3, characterized in that, The graph neural network model includes local branches and global branches; The step of inputting the graph structure model into a pre-trained graph neural network, and outputting the second air intake distribution matrix and the second ventilation rate of the splicing paper through the graph neural network, includes: The node features of each of the first grids are mapped to a high-dimensional space to obtain high-dimensional node features; Based on the high-dimensional node features of the adjacent first network, the learnable attention vector, and the corresponding initial weights, the contribution ratio of the neighboring nodes to the first grid is obtained. The contribution ratio is used to characterize the degree of influence of the air permeability of the neighboring nodes on the air intake of the first grid. The contribution ratio is positively correlated with the degree of influence. Aggregate the contribution ratio and node features corresponding to each of the first grids, and perform non-linear activation processing on the aggregated contribution ratio and node features to obtain the first node feature matrix of each of the first grids. The first node feature matrix is convolved to obtain the second node feature matrix; The second node feature matrix is input into the local branch, and the local branch obtains the second intake volume distribution matrix based on the second node feature matrix. The second node feature matrix is input into the global branch, and the global branch obtains the second ventilation rate based on the second node feature matrix.
6. The method according to claim 5, characterized in that, Based on the high-dimensional node features of neighboring nodes in the first network, the learnable attention vector, and the corresponding initial weights, the contribution ratio of neighboring nodes to the first grid is obtained. This contribution ratio characterizes the degree of influence of the air permeability state of neighboring nodes on the air intake of the first grid, including: Based on the high-dimensional node features, the learnable attention vector, and the corresponding initial weights, the attention scores of neighboring nodes to the first grid are calculated. Min-max normalization is performed on all the attention scores to obtain the contribution ratio of adjacent nodes to the first grid.
7. The method according to claim 5, characterized in that, The step of inputting the second node feature matrix into the local branch, wherein the local branch obtains the second intake volume distribution matrix based on the second node feature matrix, includes: The local branch performs a Sigmoid mapping on the features of the second node to obtain the second intake volume distribution matrix.
8. The method according to claim 5, characterized in that, The second node feature matrix is input into all branches, and the global branch obtains the second ventilation rate based on the second node feature matrix, including: The global branch performs global average pooling on all the feature matrices of the second node to obtain a global feature vector; The global branch inputs the global feature vector into the fully connected layer, and the fully connected layer obtains the second ventilation rate by performing a Sigmoid mapping on the global feature vector.
9. An electronic device, characterized in that, The electronic device includes a processor and a memory coupled together, the memory storing a computer program that, when executed by the processor, causes the electronic device to perform the method as described in any one of claims 1 to 8.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when run on a computer, causes the computer to perform the method as described in any one of claims 1 to 8.
