An evaluation method for predicting cigarette softness based on superposition of Gaussian flow functions

Abstract

This invention provides an evaluation method for predicting the ease of smoking based on Gaussian flow function superposition. It relates to tobacco product processing, first collecting historical cigarette data, then cleaning, clustering, and standardizing it before using a Bayesian optimized XGBoost algorithm to train a puff flow prediction model. Inputting the parameters of the cigarette to be tested, the method outputs a puff flow time-series curve. A dual Gaussian flow function is used to fit and extract peak flow and time distribution width characteristic parameters. Combining consumer profiles with weighted calibration, an ease-of-sense factor is constructed. Then, an adaptive Lasso regression model is used to establish a quantitative correlation between the ease-of-sense factor and expert evaluation scores. Finally, the method outputs a predicted ease-of-sense score and achieves closed-loop optimization of cigarette structural parameters. The beneficial effects of this invention are: accurately reproducing the dynamic airflow and physiological compensation behavior of human puffing using dual Gaussian functions, eliminating subjective interference from manual evaluation, and achieving pre-prediction; the model has high prediction accuracy, strong generalization ability, and a wide range of applications, shortening product development cycles and reducing testing costs.

Description

Technical Field

[0001] This invention relates to the field of tobacco product processing technology, and specifically to an evaluation method for predicting the comfort of cigarettes based on the superposition of Gaussian flow functions. Background Technology

[0002] The sensory experience of consumers during the cigarette burning process exhibits significant dynamic evolution characteristics, with sensory sensations changing in real time as a single cigarette continues to burn and be smoked. The instantaneous flow rate evolution directly determines the sensory comfort of the smoke entering the mouth. The dynamic characteristics of the flow rate throughout the smoking process are a core factor affecting the overall smoking experience for consumers, and also an indispensable core technical evaluation indicator for cigarette quality assessment, product structure, and auxiliary material optimization design. Currently, the tobacco industry generally faces common technical challenges in the development of low-tar and slim cigarette products, such as insufficient smoke smoothness and large fluctuations in dynamic draw resistance. Furthermore, the existing cigarette development model relies heavily on the experience and judgment of R&D personnel, and a standardized, digital R&D evaluation system driven by quantitative mathematical models has not yet been established.

[0003] Currently, the industry has conducted multi-dimensional research on the smoking comfort of cigarettes: some studies use computational fluid dynamics simulation to analyze the airflow field distribution inside grooved filters and compare the differences in airflow velocity fields between conventional cigarettes and cigarettes with grooved filters; some studies simultaneously conduct physicochemical index testing of smoke and professional sensory evaluation, relying on non-constant flow smoking tests by evaluators to complete the scoring of smoking ease and explore the coupling relationship between the three sensory indicators of smoking ease, satisfaction, and smoothness; other studies construct a simple correspondence between the ratio of the total amount of smoke in the first three puffs of a cigarette and dynamic draw resistance and smoking ease; still other studies use airflow detection devices to collect smoke flow and dynamic draw resistance data, and achieve a rough quantification of smoking ease by analyzing the amplitude and fluctuation stability of draw resistance.

[0004] However, existing technical solutions still have significant drawbacks: First, existing evaluation methods mostly select isolated static physical parameters such as static suction resistance and total amount of smoke per puff to construct evaluation standards, which cannot fully characterize the dynamic evolution of airflow during actual vaping by consumers. It is difficult to establish the intrinsic relationship between airflow temporal changes and physiological compensation behaviors such as oral cavity negative pressure compensation and breathing inertia tailing, resulting in a significant lack of evaluation representation dimensions. Second, existing methods for quantifying the feeling of ease rely heavily on manual scoring by professional vaping evaluators. The evaluation results are easily affected by multiple factors such as the evaluator's physical condition, personal vaping habits, and temperature and humidity of the test environment. The data repeatability and objectivity are poor, and it is impossible to construct a stable, reproducible, and fully digitalized quantitative evaluation system.

[0005] In summary, the industry urgently needs to build a quantitative representation model of smoking ease that can integrate the dynamic characteristics of smoking flow time sequence, human smoking physiological compensation behavior, and standardized sensory evaluation data, so as to achieve rapid, objective, and high-precision digital prediction of smoking ease in cigarettes. This technical requirement has become a key technical bottleneck restricting the refined structural design of cigarettes and the iterative upgrading of product quality. Summary of the Invention

[0006] To address the shortcomings of existing cigarette product development, such as the difficulty in quantifying and characterizing the ease of smoking, reliance on manual evaluation, and unclear correlation between physical indicators and sensory experience, this invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. This method aims to simulate the human sensory preference behavior for dynamic changes in smoke flow rate, thereby studying the sensory evaluation index of cigarette smoking ease. This achieves objectivity and standardization in the evaluation of ease of smoking, reducing the impact of subjective fluctuations on the stability of results in traditional manual evaluation processes.

[0007] This invention provides a method for evaluating the ease of smoking based on the superposition of Gaussian flow functions, including: Step S1: Obtain historical cigarette data; Step S2: Select feature indicators based on historical cigarette data, and use extreme gradient boosting ensemble learning algorithm combined with Bayesian optimization to train the smoke flow prediction model; Step S3: Input the design parameters and real-time parameters of the cigarette to be predicted into the suction flow prediction model to obtain the time series curve of the suction flow of the cigarette to be predicted. Step S4: Use a double Gaussian flow function to perform nonlinear fitting on the pumping flow time series curve to obtain the characteristic parameters of the double Gaussian flow function; Step S5: Construct a relaxation factor based on the characteristic parameters of the double Gaussian flow function; Step S6: Combine consumer profiles and expert evaluation of ease of inhalation scores to establish a pre-defined correlation model between ease of inhalation factors and ease of inhalation scores; Step S7: Optimize and adjust the cigarette structural parameters based on the cigarette comfort prediction results output by the preset association model.

[0008] Furthermore, the method for establishing the suction flow prediction model in step S2 is as follows: The selected feature metrics are used as the training set input to the initial XGBoost model; Model error is monitored using leave-one-out cross-validation. With the goal of achieving the best model prediction accuracy, Bayesian optimization iterations are performed on the learning rate, subsampling rate, L1 regularization coefficient, and L2 regularization coefficient. When the model prediction error converges and there is no obvious overfitting, the optimized pumping flow prediction model is output.

[0009] Furthermore, in step S3, the design parameters of the cigarette to be tested include structural parameters and auxiliary material parameters, and the real-time parameters are the smoking parameters.

[0010] Furthermore, the expression for the double Gaussian flow function in step S4 is: ; Where Q(x) represents the pumping flow rate at time x; A1 and A2 represent the peak values ​​of the two Gaussian flow functions; μ1 and μ2 represent the locations of the peak values ​​of the two Gaussian flow functions; and σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function.

[0011] Further, step S5 specifically includes: The weighting parameters n1 and n2 in the ease factor are determined based on the consumer group profile. The ease factor is obtained using the characteristic parameters of the double Gaussian flow function; The expression for the ease factor is:

[0012] Where σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function; n1 is the flow peak feature weight parameter; and n2 is the time distribution feature weight parameter.

[0013] Further, step S6 specifically includes: Collect data on the actual smoking comfort scores from smoking experts; Using the feeling of ease factor as the explanatory variable and the feeling of ease during inhalation as the target variable, a regression model of the feeling of ease during inhalation is established. The expression for the regression model of the suction comfort score is:

[0014] In the formula, y is the explained variable, i.e., the score of ease of sucking; β0 is the constant term; β n ε is the regression coefficient; x is the explanatory variable, i.e., the ease factor; ε refers to the residual term independent of x. Based on the regression model, a correlation analysis model for the feeling of relief from sucking is established. The expression of the correlation analysis model is as follows:

[0015]

[0016] In the formula, λ1 refers to the L1 penalty norm; λ1 is a hyperparameter that adjusts with the model. The adaptive penalty weights are based on the importance of the explanatory variables; η1 and γ1 are the adaptive penalty parameters, which are the frequency of occurrence of the corresponding features or the initial importance index. To optimize the obtained regression coefficient estimates; Adaptive Lasso regression was used to select features and optimize stability of the model, and a pre-defined correlation model between the ease factor and the pulsation ease score was established.

[0017] Furthermore, the method for establishing the pre-defined correlation model between the ease factor and the snoring ease score is as follows: S601. Determine the initial hyperparameter λ1 and the corresponding adaptive penalty term based on the correlation analysis model. S602. Based on the initial hyperparameter λ1 and the adaptive penalty term, the coefficient estimates of the snoring ease sample data are obtained; S603. Calculate the stability index value St of the correlation analysis model based on the coefficient estimates; S604. Determine whether the stability index value St of the correlation analysis model is greater than the preset stability threshold. If not, proceed to step S05; if yes, proceed to step SF606. S605. Iteratively adjust the hyperparameter λ1 and the adaptive penalty term, and then return to step S602. S606. Retain the coefficient estimates and corresponding sample data under the current iteration; S607. Determine whether the currently retained sample data meets the preset model output conditions. If not, return to step S605; if yes, proceed to step S608. S608. Based on the coefficient estimates and sample data retained in the final iteration, a preset correlation model between the ease factor and the snoring ease score is constructed.

[0018] Furthermore, step S1 also includes: Historical data is preprocessed, clustered, and standardized to obtain standardized sample data.

[0019] Furthermore, in step S1, the historical data includes cigarette physical index data, auxiliary material structural parameter data, smoking condition parameter data, and sensory evaluation data.

[0020] Furthermore, in step S2, the feature indicators are filtered based on the correlation between changes in suction flow rate and the ease score.

[0021] The present invention has the following beneficial effects: 1. This invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. Addressing the shortcomings of existing technologies that rely on manual evaluation, have poor repeatability, and suffer from significant subjective fluctuations, this invention, for the first time, employs a double Gaussian superposition function to mathematically model the time-series curve of the smoking flow rate. This transforms the abstract and difficult-to-quantify feeling of smoking into fitting feature parameters and a feeling factor with unified calculation standards, repeatable experiments, and horizontal comparability, thus achieving a digital and quantitative expression of sensory experience. The entire process eliminates the need for on-site scoring by evaluators, and the evaluation results are unaffected by human factors such as the evaluator's physical condition, smoking habits, and experimental environment. This eliminates subjective interference from manual evaluation, achieving a digital, objective, and quantitative representation of the feeling of smoking ease, and significantly improving the stability and objectivity of the evaluation system.

[0022] 2. This invention provides an evaluation method for predicting the comfort of cigarettes based on the superposition of Gaussian flow functions. It uses dual Gaussian functions to simultaneously and accurately fit the main peak of the inhalation flow and the secondary distortion peak formed by oral negative pressure compensation. The two types of peaks correspond to real physiological inhalation behaviors such as active inhalation airflow, oral negative pressure compensation, and airway airflow inertial tailing. The fitting parameters and the constructed comfort factor can intuitively quantify the intrinsic relationship between the amplitude of smoke flow, the amplitude of airflow fluctuation, the continuous stability of airflow, and the comfort of inhalation. The physical meaning of each parameter is clear, which is convenient for researchers to deconstruct and analyze the mechanism of the effect of cigarette structure on the inhalation experience. This solves the problem that existing static single indicators cannot reflect the coupling relationship between the dynamic changes of inhalation airflow and human physiological compensation behavior.

[0023] 3. This invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. It uses the XGBoost algorithm with Bayesian optimization to construct a puff flow prediction model, accurately capturing the complex nonlinear coupling relationship between cigarette physics, auxiliary material parameters, and puff flow. At the same time, it introduces an adaptive Lasso regression algorithm to optimize the ease-of-smoothness association model, automatically selecting effective features and suppressing redundant variable interference, ensuring the stability and statistical significance of the association model fitting. It can train dedicated prediction sub-models adapted to regular, medium, slim, and low-tar cigarettes, and is compatible with various filter rod material structures such as grooved filter rods and ordinary filter rods, with a wide range of product compatibility. This solves the problems of weak nonlinear fitting ability and limited applicable scenarios of traditional evaluation models.

[0024] 4. This invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. Before product prototyping, only the cigarette structure and auxiliary material design parameters need to be input to output a predicted ease score. It establishes a closed-loop forward design process of "parameter input—flow prediction—sensory scoring—structural iteration," enabling rapid iterative optimization of cigarettes with low draw resistance and high draw smoothness. This significantly reduces the workload of multiple rounds of sample production and manual vaping tests, effectively shortening the product development cycle and reducing raw material testing costs. It specifically addresses common industry challenges such as large fluctuations in draw resistance, insufficient draw smoothness, and weak ease of smoking in slim and low-tar cigarettes. Compared to existing evaluation methods that only perform post-production testing of finished cigarettes and cannot predict the vaping experience during the design phase, this invention supports forward-looking cigarette development, shortens the development cycle, and solves common industry technical pain points.

[0025] 5. The present invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. It has good compatibility and scalability, and can be integrated with physicochemical and sensory evaluation modules such as static draw resistance, ventilation rate, smoke release, irritation, and aroma indicators to construct a multi-dimensional integrated digital evaluation platform for cigarette sensory evaluation. At the same time, it can connect to consumer smoking behavior databases and population profile data to establish a weighted parameter library for different population groups. This provides quantitative and actionable data decision support for precise positioning of cigarette sub-markets, differentiated and personalized product development, and iterative upgrading of existing product quality, thus helping to build a digital R&D system for the tobacco industry. Attached Figure Description

[0026] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0027] Figure 1 This is a flowchart of an evaluation method for predicting the comfort of smoking based on the superposition of Gaussian flow functions according to the present invention; Figure 2 It is a heatmap showing the correlation coefficient between suction behavior data and scores; Figure 3 This is a graph showing the fitting effect of superimposed double Gaussian flow functions; Figure 4 This is a graph showing the fitting results of the inhalation flow rate when smoking a certain brand of cigarette. Figure 5 This is a heatmap of the correlation coefficient between the ease factor and the score, where A2S4 represents the ease factor when n1=2 and 2n2=4. Detailed Implementation

[0028] The detailed features and advantages of this application are described below in the specific embodiments. The content of this description is sufficient to enable any person skilled in the art to understand the technical content of this application and implement it accordingly. Based on the specification, claims and drawings disclosed in this specification, a person skilled in the art can easily understand the related objectives and advantages of this application.

[0029] The invention will now be described with reference to the accompanying drawings, in which similar reference numerals denote similar elements. While specific structures and arrangements are discussed, it should be understood that this is done merely for illustrative purposes. Those skilled in the art will recognize that other structures and arrangements can be used without departing from the spirit and scope of the invention. It will be apparent to those skilled in the art that the invention can also be used in a variety of other applications.

[0030] In this specification and claims, several terms will be used, and unless otherwise indicated, these terms will be defined to have the following meanings: The singular forms “a” and “” include their corresponding plural forms. “At least one” means one or more, and “more” means two or more. “At least one of the following” or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b, or c can be expressed as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0031] All figures used to represent component amounts, properties (e.g., molecular weight), reaction conditions, etc., should be considered to be modified in all cases by the terms "within the unavoidable margin of error" or "about". Therefore, the numerical values ​​set forth herein are approximate and may vary depending on the desired properties sought to be obtained by the present invention. The principles of equivalents, which are applied to a minimum and not intended to limit the scope of the claims, should be applied, for example, each value should be interpreted at least according to the reported significant digits and by applying conventional rounding techniques.

[0032] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0033] In the description of this embodiment, it should be noted that the terms "upper", "lower", "inner", "bottom", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, or the orientation or positional relationship in which the product is usually placed during use. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0034] Unless otherwise indicated, the following abbreviations have the following meanings, and any other abbreviations used herein but not defined have their generally accepted standard meanings: All other terms used herein for special definition are intended to have the general meaning understood by one of ordinary skill in the art, in particular meaning that one of ordinary skill in the art can directly and without doubt determine how the technical solutions of the invention can be implemented after reading the claims, specification and drawings.

[0035] Even if there are incomplete descriptions, omissions, or ambiguities in the grammar, words, punctuation, graphics, symbols, etc. of the claims, specification, and drawings of this patent, a person skilled in the art can still arrive at the only correct understanding by reading the claims, specification, and drawings as a whole without extensive reasoning or experimentation, and effectively exclude various incorrect interpretations that are not aimed at achieving the purpose of this patent.

[0036] Those skilled in the art would first choose to read the claims, specification, and drawings of this patent to reasonably interpret the terms; secondly, they would choose to refer to the relevant definitions in other documents published by the applicant before the filing date to reasonably interpret the terms; thirdly, they would choose the references cited in this patent to reasonably interpret the terms; and finally, they would choose to combine the technical dictionaries, technical manuals, reference books, textbooks, national or industry technical standards, etc., commonly used by those skilled in the art to reasonably interpret the terms.

[0037] To make the objectives, technical solutions, and advantages of this application clearer, the following will describe in detail its implementation methods in conjunction with the specific structure and working process of the device of the present invention.

[0038] Please refer to Figure 1 This invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions. By establishing the correlation between cigarette structural parameters and dynamic characteristics of inhalation flow, the invention uses double Gaussian flow functions to mathematically represent human smoking behavior, and further constructs a mapping relationship between ease factor and sensory evaluation score, thereby realizing the digitization, quantification and rapid prediction of cigarette ease.

[0039] This invention provides an evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions, which mainly includes the following steps S1 to S7.

[0040] Step S1: Obtain historical data on cigarettes.

[0041] Historical data generally includes physical index data of cigarettes, structural parameter data of auxiliary materials, smoking condition parameter data, and sensory evaluation data.

[0042] The physical properties of the cigarette include, but are not limited to, cigarette weight, circumference, length, hardness, static draw resistance, and dynamic draw resistance.

[0043] The structural parameters of auxiliary materials include, but are not limited to, the air permeability of cigarette paper, the air permeability of tipping paper, the suction resistance of filter rods, and the structural parameters of filter rod grooves.

[0044] Suction operating parameters include, but are not limited to, suction capacity, suction frequency, single-port duration, peak flow rate, and average flow rate.

[0045] The sensory evaluation data are the sensory evaluation results obtained by professional sensory evaluation experts in rating the sense of relaxation.

[0046] The suction operating parameters are collected in real time by a suction behavior recording device, and the sensory evaluation data is obtained by multiple professional suction evaluators who score according to a unified suction standard.

[0047] To improve data stability, the same ambient temperature and humidity conditions are preferred during the sampling process, and the cigarette equilibration time is standardized to reduce the impact of the external environment on the smoking flow curve.

[0048] In some embodiments, due to the potential presence of invalid data, abnormal fluctuations, and specification differences between different brands of cigarettes in the original historical data, data preprocessing is necessary. This can be achieved by preprocessing, clustering, and standardizing the historical data to obtain standardized sample data.

[0049] Based on the preset cigarette detection technical standards, missing, abnormal, and invalid data in the historical data are removed. Use the 3σ principle to remove gross error data; The re-clustering data was re-clustered to eliminate the distribution differences between different brands and specifications of cigarettes; The clustered data is normalized to eliminate the influence of different dimensions between indicators.

[0050] The normalization formula is as follows:

[0051] In the formula, x refers to the data before normalization, i.e., the clustered data; This refers to data obtained after normalization, i.e., sample data; x min It refers to the minimum value of each indicator data in the clustered data, x max It is the maximum value of each indicator data in the value clustering data.

[0052] After standardization, the obtained historical data will be used to build a training database for subsequent use in establishing a pump flow prediction model.

[0053] Step S2: Select feature indicators based on historical cigarette data, and use extreme gradient boosting ensemble learning algorithm combined with Bayesian optimization to train the smoke flow prediction model.

[0054] The characteristic indicators are selected based on the correlation between changes in puff flow rate and ease score. By selecting key characteristic indicators that are highly correlated with changes in puff flow rate and ease score, the correlation requirements can be preset according to the actual needs of cigarette design.

[0055] The selected feature indicators are used to build a model training set, and the extreme gradient boosting ensemble learning (XGBoost) algorithm is used to build a pumping flow prediction model.

[0056] The XGBoost model is composed of multiple decision trees that iteratively build up, possessing strong nonlinear feature learning capabilities and effectively describing the complex coupling relationship between cigarette structural parameters and inhalation flow rate.

[0057] The method for establishing the suction flow prediction model is as follows: The feature metrics selected from the model training set are used as the input to the initial XGBoost model. Model error is monitored using leave-one-out cross-validation. With the goal of achieving the best model prediction accuracy, Bayesian optimization iterations are performed on the learning rate, subsampling rate, L1 regularization coefficient, and L2 regularization coefficient. When the model prediction error converges and there is no obvious overfitting, the optimized pumping flow prediction model is output.

[0058] Step S3: Input the design parameters and real-time parameters of the cigarette to be predicted into the suction flow prediction model to obtain the time series curve of the suction flow of the cigarette to be predicted.

[0059] The design parameters of the cigarette to be tested include structural parameters and auxiliary material parameters, while the real-time parameters are the suction parameters.

[0060] In the actual prediction process, the structural parameters, auxiliary material parameters and suction parameters of the cigarette to be tested are input into the suction flow prediction model, and the suction flow time series curve Q(x) is output.

[0061] The structural parameters of the cigarette to be tested include: cigarette length, circumference, filter rod structural parameters, and cigarette paper air permeability; the suction parameters include dynamic draw resistance, etc.

[0062] In this embodiment, the suction flow prediction model is applicable not only to conventional cigarettes, but also to different specifications of cigarettes such as slim cigarettes, medium cigarettes, and low-tar cigarettes.

[0063] Step S4: Use a double Gaussian flow function to perform nonlinear fitting on the pumping flow time series curve to obtain the characteristic parameters of the double Gaussian flow function.

[0064] After obtaining the time-series pumping flow rate curve, a double Gaussian flow function is used to perform nonlinear fitting on the pumping curve.

[0065] The expression for the fitting formula of the double Gaussian flow function is as follows: ; Where Q(x) represents the suction flow rate at time x; A1 and A2 represent the peak values ​​of the two Gaussian flow functions; μ1 and μ2 represent the locations of the peak values ​​of the two Gaussian flow functions; σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function. A1 and A2 reflect the flow build-up capability and suction response intensity; μ1 and μ2 reflect the distribution location of the suction flow peaks in the time dimension; σ1 and σ2 reflect the smoothness of the suction flow rate change and the stability of airflow maintenance.

[0066] It should be noted that, compared with the single Gaussian model, the invention uses a double Gaussian function, which can more accurately describe the dynamic relationship between the main peak flow and the secondary compensation flow during human aspiration, thus more closely reflecting the actual aspiration behavior.

[0067] Step S5: Construct a relaxation factor based on the characteristic parameters of the double Gaussian flow function.

[0068] After obtaining the characteristic parameters of the dual Gaussian flow function, a comfort factor λ is further constructed to quantify the correlation between dynamic changes in airflow and breathing comfort during human inhalation. Specifically: The weighting parameters n1 and n2 in the ease factor are determined based on the consumer group profile. The ease factor is obtained using the characteristic parameters of the double Gaussian flow function; The expression for the ease factor is:

[0069] Where σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function; n1 is the flow peak feature weight parameter; and n2 is the time distribution feature weight parameter.

[0070] The ease factor essentially reflects the comprehensive relationship between the body's ability to build up airflow and maintain stable flow during the inhalation process, including: The peak value reflects the airflow establishment capability, while the time width reflects the physiological requirements for maintaining airflow stability. In practical applications, the values ​​of n1 and n2 can be optimized and adjusted according to the target consumer group.

[0071] If targeting a specific consumer group, it is necessary to sample the suction behavior of typical consumers in that group, fit the actual suction flow scatter points into a Gaussian flow curve, output Gaussian feature parameters, calculate the ease factor by traversing different combinations of n1 and n2, and select the set of n1 and n2 with the highest correlation as the dedicated weight parameters for that consumer group, using the ease sensory score as a reference.

[0072] If the target audience is a broad consumer group, a large-sample statistical method is used to determine the optimal weight combination with wide applicability.

[0073] Step S6: Combine consumer profiles and expert evaluation of ease of inhalation scores to establish a pre-defined correlation model between ease of inhalation factors and ease of inhalation scores.

[0074] Specifically, data on the actual ease of smoking experienced by evaluation experts will be collected; Using the feeling of ease factor as the explanatory variable and the feeling of ease during inhalation as the target variable, a regression model of the feeling of ease during inhalation is established. The expression for the regression model of the suction comfort score is:

[0075] In the formula, y is the explained variable, i.e., the score of ease of sucking; β0 is the constant term; β n ε is the regression coefficient; x is the explanatory variable, i.e., the ease factor; ε refers to the residual term independent of x.

[0076] The association model is constructed using adaptive Lasso regression, which introduces L1 regularization and adaptive penalty weights on the basis of least squares fitting, automatically filters key features and removes redundant variables, thereby improving the model's stability and interpretability.

[0077] The initial hyperparameter λ1 and the corresponding adaptive penalty term are determined based on the correlation analysis model.

[0078] The coefficient estimates are obtained based on the initial hyperparameters and the penalty term.

[0079] Based on the regression model, a correlation analysis model for the feeling of relief from sucking is established. The expression of the correlation analysis model is as follows:

[0080]

[0081] In the formula, λ1 refers to the L1 penalty norm, which is the sum of the absolute values ​​of the weights of each variable; λ1 is a hyperparameter that is adjusted with the model and is used to balance the fitting error and the penalty intensity. The adaptive penalty weights are based on the importance of the explanatory variables; η1 and γ1 are the adaptive penalty parameters, which are the frequency of occurrence of the corresponding features or the initial importance index. The estimated regression coefficients are obtained for optimization.

[0082] Adaptive Lasso regression was used to select features and optimize stability of the model, and a pre-defined correlation model between the ease factor and the pulsation ease score was established.

[0083] The method for establishing the pre-defined correlation model between the ease factor and the snoring ease score is as follows: S601. Determine the initial hyperparameter λ1 and the corresponding adaptive penalty term based on the correlation analysis model. S602. Based on the initial hyperparameter λ1 and the adaptive penalty term, the coefficient estimates of the snoring ease sample data are obtained; S603. Calculate the stability index value St of the correlation analysis model based on the coefficient estimates; S604. Determine whether the stability index value St of the correlation analysis model is greater than the preset stability threshold. If not, proceed to step S05; if yes, proceed to step SF606. S605. Iteratively adjust the hyperparameter λ1 and the adaptive penalty term, and then return to step S602. S606. Retain the coefficient estimates and corresponding sample data under the current iteration; S607. Determine whether the currently retained sample data meets the preset model output conditions. If not, return to step S605; if yes, proceed to step S608. S608. Based on the coefficient estimates and sample data retained in the final iteration, a preset correlation model between the ease factor and the snoring ease score is constructed.

[0084] During the establishment of the preset association model: the model stability index is calculated based on the coefficient estimates; if the stability index is lower than the preset threshold, the hyperparameters and penalty terms are iteratively adjusted; the iteration is repeated until the stability requirements are met, and finally the preset association model of the relaxation factor and the snoring relaxation score is constructed.

[0085] To ensure the reliability of the model, during the modeling process, multiple professional smokers uniformly smoked the sample cigarettes one puff at a time according to their usual smoking methods, and used a smoking behavior recording device to simultaneously collect the smoking flow curve and the ease score data.

[0086] After iterative training on a large amount of sample data, the system establishes a stable pre-set association model.

[0087] Step S7: Optimize and adjust the cigarette structural parameters based on the cigarette comfort prediction results output by the preset association model.

[0088] Based on the preset association model, the system outputs a prediction result of the cigarette's comfort level and determines whether the prediction result meets the design expectations. If not, it returns to adjust the cigarette design parameters; if so, it determines the cigarette structure parameters.

[0089] In practical applications, the ease rating score of the tested cigarette is output based on the ease rating prediction model to evaluate the cigarette: If the predicted score does not reach the preset design target, the cigarette structure parameters, filter rod structure parameters and auxiliary material design parameters will be further adjusted, and the suction flow prediction model will be re-entered for iterative prediction. If the predicted score meets the preset design requirements, the final cigarette design scheme will be output.

[0090] Through the above feedback optimization process, a closed-loop design of cigarette products is achieved, encompassing "structural design - flow prediction - ease of use evaluation - parameter optimization," thereby improving cigarette R&D efficiency and shortening the product development cycle.

[0091] Example 1: Digital Prediction and Optimization of Cigarette Ease of Use

[0092] (1) Database setup

[0093] Ten samples of slim cigarettes with different filter groove structures and cigarette paper permeability were collected and balanced in a uniform environment of 22℃ and 60% relative humidity for 48 hours. Static / dynamic draw resistance, filter rod draw resistance, cigarette paper permeability and other physical auxiliary parameters were collected simultaneously. A single puff flow time series data was collected using a puff behavior recorder. Nine national-level smoke testers completed the ease of use scoring according to a unified standard to build the original database.

[0094] Please refer to Figure 2 The study correlated the testers' sucking behavior data with their scores and calculated the correlation coefficient to identify features that had a significant impact on the ease rating process.

[0095] Table 1. Correlation coefficients between the first suction flow rate and the scores.

[0096]

[0097] (2) Data preprocessing

[0098] Eight groups of missing or invalid samples were removed; two groups of outlier samples with abnormal flow were removed using the 3σ principle; the filter was clustered according to single-groove / multi-groove design; extreme value normalization was used to eliminate dimensions, and 14 features strongly correlated with suction flow were selected to form a training set of 64 groups. Leave-one-out cross-validation was used for training.

[0099] (3) Training of XGBoost traffic model

[0100] Please refer to Figure 3 and Figure 4 Bayesian optimization range: learning rate 0.02~0.15, subsampling rate 0.7~1.0, feature sampling rate 0.6~1.0, L1 and L2 regularization coefficients 0~5; number of trees 100~500, tree depth 3~12, minimum child node weight 1~8, split threshold 0.01~2.0; convergence after 10 iterations, and R0 of the flow curve fitting on the validation set under the optimal hyperparameter combination. 2 >0.9, indicating no significant overfitting.

[0101] (4) Double Gaussian fitting and calculation of ease factor

[0102] Select a slim cigarette model to be tested, input its parameters, and output the time-series curve of its suction flow rate; obtain the result through double Gaussian fitting.

[0103] A1=14.36, A2=21.59, σ1=0.247, σ2=0.392;

[0104] Please refer to Figure 5 For young consumers, the sampling and labeling weights were determined. When n1=3 and 2n2=2, the correlation coefficient between the ease factor and the corresponding cigarette score was calculated to be 0.85, which is the highest among different weight combinations.

[0105] (5) Sensory score association prediction

[0106] Input n1=3 and 2n2=2 into the adaptive Lasso association model, and output the prediction score of the cigarette ease feeling of a part of the external independent dataset; if the difference between the product design target score and the predicted score is ≥0.5 points, it is judged as not meeting the standard.

[0107] (6) Parameter Iterative Optimization

[0108] Adjusting cigarette parameters such as filter rod pressure drop, cigarette paper air permeability, and ventilation rate, and re-entering them into the model for calculation, the correlation coefficient between the ease factor and the score after optimization increased to 0.88, and the difference between the product design target score and the predicted score decreased. Further iterations were conducted until the design threshold was met, and the set of auxiliary materials and structural parameters were locked to complete the finalization.

[0109] Example 2: Construction of a general evaluation model for broad-spectrum cigarettes

[0110] Eight additional medium-length regular cigarettes were selected to cover different tar contents and filter types; the consumer sampling sample was expanded, and general weights were obtained by large-sample statistics; after the model was trained, 20 blind samples were randomly selected from the dataset for prediction. The average absolute error between the prediction score and the expert manual evaluation score was <0.3 points, and the prediction consistency was far better than the traditional static draw resistance evaluation method.

[0111] Comparative Example 1

[0112] The similarities between Comparative Example 1 and Example 2 will not be repeated here. The difference is that Comparative Example 1 uses the static draw resistance of cigarettes as the evaluation basis; Example 2 focuses on the dynamic process of cigarette smoking, collects time-series data such as dynamic draw resistance and draw flow rate during the smoking process, and establishes a comfort evaluation model based on dynamic characteristics.

[0113] Table 2 Comparison of technical solutions and evaluation results between Example 2 and Comparative Example 1

[0114] The results show that, compared with Comparative Example 1, which only evaluated ease of smoking based on static draw resistance, Example 2 considers dynamic draw resistance and consumer smoking behavior during cigarette smoking, making the evaluation process closer to the actual smoking state of consumers. After the model is established, iterative updates can be made by introducing new sample data to further expand the applicable cigarette categories and evaluation scope, demonstrating good scalability. Compared with directly determining the evaluation results based on static indicators, Example 2 establishes an evaluation model based on data-driven methods, making the evaluation process more standardized and reducing the influence of human experience judgment. Example 2 uses 18 types of cigarettes for verification, covering a wider range of cigarette categories than Comparative Example 1, and has a broader applicability while maintaining good consistency with human smoking evaluation results.

[0115] Comparative Example 2

[0116] The similarities between Comparative Example 2 and Example 2 will not be repeated here. The difference is that the characteristic parameters characterizing the ease of smoking cigarettes are different. Comparative Example 2 uses the dynamic draw resistance value and smoke volume of cigarettes to characterize the ease of smoking cigarettes. Example 2, on the basis of parameters such as dynamic draw resistance, further introduces data reflecting the characteristics of consumer smoking behavior as evaluation parameters.

[0117] Table 3 Comparison of technical solutions and evaluation results between Example 2 and Comparative Example 2

[0118] The results show that Example 2 supports iterative optimization of the model based on new data and can adjust the model's applicability by incorporating different consumer profiles. Compared to Comparative Example 2, Example 2 uses a wider variety of cigarette types for model training and validation, covering various cigarette samples with different tar content and filter types. During the evaluation process, it comprehensively utilizes multi-dimensional data such as dynamic draw resistance, draw flow rate, and characteristics reflecting consumer smoking behavior for modeling. Validation results show that Example 2 and Comparative Example 2 are similar in terms of consistency with human evaluation, but Example 2 has better applicability and model scalability due to its wider coverage of validation samples and richer input features.

[0119] The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed.

[0120] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0121] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0122] In this specification, references to "an embodiment" or "a specific implementation" mean that a particular feature, structure, or characteristic described in connection with that embodiment / specific implementation is included in at least one embodiment / specific implementation of the invention. Therefore, the phrase "in one embodiment / specific implementation" appearing in various places in this specification does not necessarily refer to the same embodiment / setting, but rather to potentially different embodiments. Furthermore, specific features, structures, or characteristics may be combined in one or more embodiments / settings in any suitable manner, as will be apparent to those skilled in the art from this disclosure.

[0123] Similarly, it should be understood that in the above description of exemplary embodiments / specific implementations of the invention, various features of the invention are sometimes combined in a single embodiment / specific implementation or its figures and description, with the aim of simplifying the disclosure and aiding in the understanding of one or more of the various aspects of the invention. However, the method of description of the invention should not be construed as reflecting an intention that, except for expressly stated instructions to the contrary or obvious technical contradictions or exclusions, the claimed features of the invention are more than those expressly stated in each claim.

[0124] Conversely, the inventive aspect reflected in the claims lies in not all the features of a single foregoing disclosed embodiment / specification. Therefore, the claims following the detailed description are expressly incorporated herein, each claim existing independently as a separate embodiment / specification of the invention.

[0125] Furthermore, while some embodiments / specific implementations described herein include, but are not limited to, other features included in other embodiments / specific implementations, combinations of features from different embodiments / specific implementations are intended to be within the scope of the invention and form different embodiments / specific implementations, as will be understood by those skilled in the art. For example, in the following claims, embodiments / specific implementations of any claim can be used in any combination.

[0126] The terms and expressions used in this specification are for illustrative purposes and not for limitation. In using these terms and expressions, it is not intended to exclude any equivalents of the features or portions thereof shown and described, but rather to recognize that various modifications may be possible within the scope of the invention.

[0127] Therefore, it should be understood that although the invention has been specifically disclosed through preferred embodiments, exemplary embodiments and optional features, those skilled in the art may take variations or modifications of the concepts disclosed herein, and such variations and modifications are therefore considered to be within the scope of the invention as defined by the appended claims.

[0128] The specific embodiments given in this specification are examples of useful implementations of the present invention. It will be apparent to those skilled in the art that the present invention can be implemented using many variations of the devices, device components, and method steps disclosed in this specification.

[0129] The foregoing description of specific embodiments fully discloses the general features of the present invention, enabling others to easily modify and / or adapt such specific embodiments for various applications by applying knowledge within the scope of the art, without conducting excessive experimentation and without departing from the general concept of the present invention.

[0130] Therefore, based on the teachings and guidance provided herein, it is intended that such modifications and alterations be included within the meaning and scope of equivalents of the disclosed embodiments. It should be understood that the wording or terminology used herein is for descriptive purposes and is not intended to be limiting; thus, the wording or terminology in this specification will be interpreted by those skilled in the art based on the foregoing teachings and guidance.

[0131] Furthermore, the scope of this invention should not be limited to any of the exemplary embodiments described above, but only to the appended claims and their equivalents. The terminology and expressions used herein are for descriptive purposes only, and the invention should not be limited to these terms and expressions. The use of these terms and expressions does not imply the exclusion of any illustrative and descriptive equivalents (or parts thereof), and it should be recognized that various modifications that may exist should also be included within the scope of the claims. Other modifications, variations, and substitutions may also exist. Accordingly, the claims should be considered to cover all such equivalents.

[0132] Similarly, it should be noted that although the present invention has been described with reference to the specific embodiments described above, those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention, and various equivalent changes or substitutions can be made without departing from the spirit of the present invention. Therefore, any changes or modifications to the above embodiments within the scope of the essential spirit of the present invention will fall within the scope of the claims of the present invention.

Claims

1. A method for evaluating the ease of smoking based on the superposition of Gaussian flow functions, characterized in that, include: Step S1: Obtain historical cigarette data; Step S2: Select feature indicators based on historical cigarette data, and use extreme gradient boosting ensemble learning algorithm combined with Bayesian optimization to train the smoke flow prediction model. Step S3: Input the design parameters and real-time parameters of the cigarette to be predicted into the suction flow prediction model to obtain the time series curve of the suction flow of the cigarette to be predicted. Step S4: Use a double Gaussian flow function to perform nonlinear fitting on the pumping flow time series curve to obtain the characteristic parameters of the double Gaussian flow function; Step S5: Construct a relaxation factor based on the characteristic parameters of the double Gaussian flow function; Step S6: Combine consumer profiles and expert evaluation of ease of inhalation scores to establish a pre-defined correlation model between ease of inhalation factors and ease of inhalation scores; Step S7: Optimize and adjust the cigarette structural parameters based on the cigarette comfort prediction results output by the preset association model.

2. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: The method for establishing the suction flow prediction model in step S2 is as follows: The selected feature metrics are used as the training set input to the initial XGBoost model; Model error is monitored using leave-one-out cross-validation. With the goal of achieving the best model prediction accuracy, Bayesian optimization iterations are performed on the learning rate, subsampling rate, L1 regularization coefficient, and L2 regularization coefficient. When the model prediction error converges and there is no obvious overfitting, the optimized pumping flow prediction model is output.

3. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: In step S3, the design parameters of the cigarette to be tested include structural parameters and auxiliary material parameters, while the real-time parameters are the suction parameters.

4. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: The expression for the double Gaussian flow function in step S4 is: ； Where Q(x) represents the pumping flow rate at time x; A1 and A2 represent the peak values ​​of the two Gaussian flow functions; μ1 and μ2 represent the locations of the peak values ​​of the two Gaussian flow functions; and σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function.

5. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to any one of claims 4, characterized in that: Step S5 specifically involves: The weighting parameters n1 and n2 in the ease factor are determined based on the consumer group profile. The ease factor is obtained using the characteristic parameters of the double Gaussian flow function; The expression for the ease factor is: Where σ1 and σ2 represent the time distribution width parameters of the Gaussian flow distribution function; n1 is the flow peak feature weight parameter; and n2 is the time distribution feature weight parameter.

6. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: Step S6 specifically involves: Collect data on the actual smoking comfort scores from smoking experts; Using the feeling of ease factor as the explanatory variable and the feeling of ease during inhalation as the target variable, a regression model of the feeling of ease during inhalation is established. The expression for the regression model of the suction comfort score is: In the formula, y is the score for ease of breathing; β0 is a constant term; β n It is the regression coefficient; x is the explanatory variable, namely the ease factor; ε refers to the residual term independent of x. Based on the regression model, a correlation analysis model for the feeling of relief from sucking is established. The expression of the correlation analysis model is as follows: In the formula, λ1 refers to the L1 penalty norm; λ1 is a hyperparameter that adjusts with the model. Adaptive penalty weights based on the importance of explanatory variables; η1 and γ1 are adaptive penalty parameters that represent the frequency of occurrence or initial importance of the corresponding features. To optimize the obtained regression coefficient estimates; Adaptive Lasso regression was used to select features and optimize stability of the model, and a pre-defined correlation model between the ease factor and the pulsation ease score was established.

7. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 6, characterized in that: The method for establishing the pre-defined correlation model between the ease factor and the snoring ease score is as follows: S601. Determine the initial hyperparameter λ1 and the corresponding adaptive penalty term based on the correlation analysis model. S602. Based on the initial hyperparameter λ1 and the adaptive penalty term, the coefficient estimates of the snoring ease sample data are obtained. S603. Calculate the stability index value St of the correlation analysis model based on the coefficient estimates; S604. Determine whether the stability index value St of the correlation analysis model is greater than the preset stability threshold. If not, proceed to step S05; if yes, proceed to step SF606. S605. Iteratively adjust the hyperparameter λ1 and the adaptive penalty term, and then return to step S602. S606. Retain the coefficient estimates and corresponding sample data under the current iteration; S607. Determine whether the currently retained sample data meets the preset model output conditions. If not, return to step S605; if yes, proceed to step S608. S608. Based on the coefficient estimates and sample data retained in the final iteration, a preset correlation model between the ease factor and the snoring ease score is constructed.

8. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that, Step S1 further includes: Historical data is preprocessed, clustered, and standardized to obtain standardized sample data.

9. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: In step S1, the historical data includes cigarette physical index data, auxiliary material structural parameter data, smoking condition parameter data, and sensory evaluation data.

10. The evaluation method for predicting the ease of smoking based on the superposition of Gaussian flow functions according to claim 1, characterized in that: In step S2, the feature indicators are filtered based on the correlation between changes in suction flow rate and the ease score.

Images