Tobacco leaf baking environment parameter optimization method based on computational fluid mechanics

Abstract

The invention provides a tobacco leaf curing environment parameter optimization method based on computational fluid mechanics, which belongs to the technical field of tobacco leaf curing, and comprises the following steps: firstly, collecting multi-point environment data of a tobacco leaf curing room to establish an initial data set, constructing a three-dimensional grid model and setting boundary conditions; secondly, importing data into fluid mechanics simulation software, setting physical property parameters and a turbulence model, and obtaining distribution of a flow field, a temperature field and a humidity field; then, an unsupervised learning algorithm is applied to identify an environment non-uniform region, and a temperature and humidity gradient model is constructed to generate an optimization curve; then calculating the similarity of temperature and humidity curves of adjacent time sequence frames to determine a key gradient point, and establishing a tobacco leaf quality and environmental parameter multivariate correlation model; and finally, performing comprehensive optimization by applying an environmental parameter fusion optimization function, calculating an optimal baking time sequence control scheme, screening an optimal parameter combination after verification in the model, applying the optimal parameter combination to an actual baking process, monitoring physical and chemical indexes of tobacco leaves, realizing accurate control of baking environmental parameters, and facilitating stable quality of the tobacco leaves.

Description

Technical Field

[0001] The present invention belongs to the technical field of tobacco leaf baking, and specifically relates to a method for optimizing the environmental parameters of tobacco leaf baking based on computational fluid dynamics. Background Art

[0002] Tobacco leaf baking is a key link in the tobacco processing process and has a decisive impact on the formation of tobacco leaf quality. Traditional tobacco leaf baking technology mainly relies on experience to determine parameter settings such as temperature and humidity, and realizes the environmental control in the baking room through a simple hot air circulation system. This method generally adopts a fixed baking curve in practical applications, and controls the baking process by manually adjusting the ventilation openings and heating power, but lacks the ability to accurately monitor and control the internal environmental state of the baking room.

[0003] However, the traditional baking method has obvious limitations, mainly manifested in the uneven spatial distribution of environmental parameters in the baking room, resulting in significant differences in the heating, dehydration, and chemical conversion degrees of tobacco leaves at different positions. Research shows that the temperature gradient in the baking room can reach 5 - 10 °C, the humidity difference can reach more than 15%, and the air flow velocity distribution is extremely uneven, resulting in large differences in tobacco leaf quality between batches and within batches. Especially in key baking stages such as yellowing and color fixation, the problem of environmental non-uniformity is particularly prominent.

[0004] The existing technology is difficult to effectively solve the technical problem of uneven spatial distribution of environmental parameters during the tobacco leaf baking process. On the one hand, the traditional empirical baking curve does not consider the spatial environmental differences in the room; on the other hand, the existing baking control system lacks an accurate regulation mechanism for the global distribution of the convection field, temperature field, and humidity field, and cannot perform dynamic optimization according to the complex relationship between tobacco leaf quality and environmental parameters, making it difficult to achieve the uniformity of the baking environment and the optimal control of the baking process. That is to say, there is a technical problem in the existing technology that the uneven distribution of tobacco leaf baking environmental parameters leads to unstable tobacco leaf quality. Summary of the Invention

[0005] In view of this, the present invention provides a method for optimizing the environmental parameters of tobacco leaf baking based on computational fluid dynamics, which can solve the technical problem in the existing technology that the uneven distribution of tobacco leaf baking environmental parameters leads to unstable tobacco leaf quality.

[0006] The present invention is implemented as follows: The present invention provides an optimization method for tobacco leaf baking environment parameters based on computational fluid dynamics, including: establishing an initial data set of baking environment parameters and a three-dimensional grid model of a tobacco leaf baking house, and importing them into a fluid dynamics simulation software to construct a numerical simulation basic model of the tobacco leaf baking environment; running simulation analysis to obtain the flow field distribution, temperature field distribution, and humidity field distribution in the tobacco leaf baking house, and forming an environmental field parameter distribution map; using an unsupervised learning algorithm to perform clustering analysis on the environmental field parameter distribution map to identify uneven areas and hot spots in the tobacco leaf baking environment; constructing a temperature and humidity gradual change model during the baking process, and generating multiple groups of optimized gradual change curves based on the original baking curve; adopting a temperature and humidity gradual change curve similarity evaluation method to determine key gradual change points and their leading stage curves; establishing a multivariate correlation model between tobacco leaf quality and baking environment parameters based on the leading stage curves, and determining an optimization objective function for baking environment parameters; applying the baking environment parameter fusion optimization function to comprehensively optimize the baking environment parameters, calculating the optimal baking time sequence control scheme; verifying the optimal baking time sequence control scheme in the numerical simulation basic model of the tobacco leaf baking environment, calculating the environmental uniformity index and the value of the optimization objective function for baking environment parameters, screening the optimal combination of baking environment parameters and applying them to the actual baking process, and monitoring the physical and chemical indexes of tobacco leaves.

[0007] Among them, the environmental field parameter distribution map refers to the distribution of physical quantities such as temperature, humidity, and air flow velocity at any point in the tobacco leaf baking house obtained through computational fluid dynamics numerical simulation, presented in the form of a cloud map or contour map, and used to intuitively display the uniformity and change trend of the baking environment.

[0008] Among them, the environmental uniformity index refers to a mathematical index used to quantitatively evaluate the degree of uniformity of environmental parameter distribution in a tobacco leaf baking house. By calculating the standard deviation or coefficient of variation of the temperature field, humidity field, and air flow field, the smaller the value, the more uniform the environmental parameter distribution and the more stable the baking conditions.

[0009] Among them, the baking environment parameters refer to the key physical quantities that affect the tobacco leaf baking process, including the temperature, humidity, air flow velocity, air flow direction, and their spatial distribution characteristics in the tobacco leaf baking house, which directly determine the tobacco leaf drying rate and the effect of chemical composition transformation.

[0010] Among them, the temperature and humidity gradual change model refers to a mathematical model that describes the temperature and humidity changing according to specific rules during the baking process, used to achieve a smooth transition from the initial conditions to the target conditions, and ensure the best physical and chemical transformation effect of tobacco leaves in different baking stages.

[0011] Among them, the original baking curve refers to the functional relationship between temperature and humidity changing with time in the traditional tobacco leaf baking process, and the optimized gradual change curve refers to the new baking process curve obtained by optimizing through the temperature and humidity gradual change model.

[0012] Among them, the key gradient point refers to the time node with the largest change amplitude of parameters in adjacent time periods in the temperature and humidity curve during the baking process, marking the entry of baking into a new process stage and the critical moment that needs to be focused on for control.

[0013] Among them, the baking environment parameter fusion optimization function is used to comprehensively consider the influence of various baking environment parameters on the quality of tobacco leaves and generate an optimal control strategy. The inputs include the temperature field uniformity coefficient obtained from the environmental field parameter distribution map, the humidity field distribution index obtained from the environmental field parameter distribution map, the air velocity vector distribution obtained from the environmental field parameter distribution map, the key gradient point time sequence obtained from the temperature and humidity gradient curve similarity evaluation method, and the moisture content of tobacco leaves obtained from the initial baking environment parameter dataset; the outputs include the air duct adjustment parameters applied to the baking environment parameter optimization objective function, the heating power control curve applied to the baking environment parameter optimization objective function, the humidity control parameters applied to the baking environment parameter optimization objective function, and the optimal baking duration applied to the optimal baking time sequence control scheme.

[0014] Among them, the steps of using the unsupervised learning algorithm to perform clustering analysis on the environmental field parameter distribution map specifically include: converting the environmental field parameter distribution map into a multi-dimensional feature vector; applying the K-means clustering or hierarchical clustering algorithm to classify the feature vectors; and determining the spatial positions and parameter characteristics of the non-uniform regions and hot spots in the tobacco leaf baking environment according to the clustering results.

[0015] Among them, the method of evaluating the similarity of the temperature and humidity gradient curve specifically includes: calculating the cosine similarity or Euclidean distance between the temperature and humidity curves at adjacent time points; setting a similarity threshold and identifying the time points with similarity lower than the threshold as key gradient points; and extracting the temperature and humidity curves in the previous stage before the key gradient points as the leading stage curves.

[0016] Compared with the prior art, the present invention provides an optimization method for tobacco leaf baking environment parameters based on computational fluid dynamics. The present invention proposes an optimization method for tobacco leaf baking environment parameters based on computational fluid dynamics. By constructing a three-dimensional grid model of the tobacco leaf baking house and performing fluid dynamics numerical simulation, combined with the unsupervised learning algorithm to identify non-uniform regions in the environment, a multivariate correlation model between tobacco leaf quality and baking environment parameters is established to achieve precise optimization and control of baking environment parameters.

[0017] This method effectively solves the problem of uneven distribution of baking environment parameters in the traditional technology. By obtaining the accurate distributions of the flow field, temperature field, and humidity field in the baking house through computational fluid dynamics simulation, combined with the unsupervised clustering algorithm to identify key regions, and establishing a temperature and humidity gradient model to generate an optimization curve, the transformation from macroscopic control of the baking house to local fine adjustment is realized, significantly improving the baking environment uniformity index and significantly enhancing the batch stability of the physical and chemical indexes of tobacco leaves.

[0018] The present invention successfully solves the technical problem that the uneven distribution of tobacco leaf baking environment parameters leads to unstable tobacco leaf quality. Through the integrated application of computational fluid dynamics and data mining technologies, a fusion optimization function of baking environment parameters is established, realizing the dynamic and precise control of the baking process, fundamentally improving the unevenness of the baking environment, ensuring the stability and consistency of tobacco leaf quality, and promoting the development of tobacco leaf baking technology towards digital and precise directions. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flowchart of the method of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0021] As Figure 1 shown, it is a flowchart of an optimization method for tobacco leaf baking environment parameters based on computational fluid dynamics provided by the present invention. The method includes the following steps:

[0022] S01. Collect the air velocity, temperature, and humidity data at multiple positions in the tobacco leaf baking house, and establish an initial data set of baking environment parameters;

[0023] S02. Construct a three-dimensional grid model of the tobacco leaf baking house, divide the computational domain, and set the grid boundary conditions to complete the preprocessing of computational fluid dynamics;

[0024] S03. Import the initial data set of the baking environment parameters into the fluid dynamics simulation software, set the physical property parameters and turbulence model, and construct a numerical simulation basic model of the tobacco leaf baking environment;

[0025] S04. Run the simulation analysis to obtain the flow field distribution, temperature field distribution, and humidity field distribution in the tobacco leaf baking house, and form an environmental field parameter distribution map;

[0026] S05. Use an unsupervised learning algorithm to perform clustering analysis on the environmental field parameter distribution map to identify uneven regions and hot spots in the tobacco leaf baking environment;

[0027] S06. Construct a temperature and humidity gradual change model during the baking process, generate multiple groups of optimized gradual change curves based on the original baking curve, and calculate the influence indexes of the optimized gradual change curves on the tobacco leaf quality;

[0028] S07. Adopt a temperature and humidity gradual change curve similarity evaluation method to calculate the similarity of temperature and humidity curves in adjacent time sequence frames, and determine the key gradual change points and their leading stage curves;

[0029] S08. Establish a multivariate correlation model between tobacco leaf quality and baking environment parameters based on the leading stage curve, and determine the optimization objective function of baking environment parameters;

[0030] S09. Apply the integrated optimization function of baking environment parameters to comprehensively optimize the baking environment parameters, and calculate the optimal baking time sequence control scheme;

[0031] S10. Verify the optimal baking time sequence control scheme in the numerical simulation basic model of the tobacco leaf baking environment, calculate the environmental uniformity index and the value of the optimization objective function of the baking environment parameters, screen the optimal combination of baking environment parameters and apply it to the actual baking process, and monitor the physical and chemical indexes of the tobacco leaves.

[0032] Among them, the environmental field parameter distribution map specifically refers to the distribution of physical quantities such as temperature, humidity, and air flow velocity at any point in the tobacco leaf baking house obtained through computational fluid dynamics numerical simulation, presented in the form of a cloud map or contour map, and used to intuitively display the uniformity and change trend of the baking environment.

[0033] Among them, the environmental uniformity index specifically refers to a mathematical index used to quantitatively evaluate the uniformity degree of the distribution of environmental parameters in the tobacco leaf baking house. By calculating the standard deviation or coefficient of variation of the temperature field, humidity field, and air flow field, the smaller the value, the more uniform the distribution of environmental parameters and the more stable the baking conditions.

[0034] Among them, the turbulence model specifically refers to a mathematical model that describes the characteristics of fluid turbulent motion, such as the standard k-ε model, k-ω model, or Reynolds stress model, and is used to accurately simulate the complex air flow motion state in the tobacco leaf baking house and calculate the turbulence intensity and energy dissipation rate.

[0035] Among them, the baking environment parameters specifically refer to the key physical quantities that affect the tobacco leaf baking process, including the temperature, humidity, air flow velocity, air flow direction, and their spatial distribution characteristics in the tobacco leaf baking house, which directly determine the drying rate of the tobacco leaves and the conversion effect of chemical components.

[0036] Among them, the temperature and humidity gradual change model specifically refers to a mathematical model that describes the temperature and humidity changing according to specific rules during the baking process, and is used to achieve a smooth transition from the initial conditions to the target conditions to ensure the best physical and chemical conversion effect of the tobacco leaves in different baking stages.

[0037] Among them, the original baking curve specifically refers to the functional relationship between temperature and humidity changing with time in the traditional tobacco leaf baking process, usually determined by experience or historical data, and is the basis and reference for generating the optimized gradual change curve.

[0038] Among them, the optimized gradual change curve specifically refers to the new baking process curve obtained by optimizing through the temperature and humidity gradual change model, which has a smoother transition characteristic and a better parameter distribution, and helps to improve the quality and uniformity of tobacco leaf baking.

[0039] Among them, the key gradient point specifically refers to the time node with the largest change amplitude of parameter values in adjacent time periods in the temperature and humidity curve during the baking process, marking the entry of baking into a new process stage and being a crucial moment that requires key control.

[0040] Among them, the leading-stage curve specifically refers to the temperature and humidity change curve before the key gradient point, which contains important parameter information on tobacco leaf dehydration and chemical conversion and serves as the basis for determining subsequent baking strategies.

[0041] Among them, the multi-variable correlation model specifically refers to the quantitative relationship expression established between the tobacco leaf quality indicators and the baking environment parameters through methods such as multiple regression, neural network, or support vector machine, which includes input variables such as temperature, humidity, air flow velocity, and their spatial gradients, and output variables such as the content of tobacco leaf chemical components, color, and aroma.

[0042] Among them, the baking environment parameter fusion and optimization function is used to comprehensively consider the influence of various baking environment parameters on the tobacco leaf quality and generate an optimal control strategy. The inputs include the temperature field uniformity coefficient obtained from the environmental field parameter distribution map, the humidity field distribution index obtained from the environmental field parameter distribution map, the air flow velocity vector distribution obtained from the environmental field parameter distribution map, the key gradient point time series sequence obtained from the temperature and humidity gradient curve similarity evaluation method, and the tobacco leaf moisture content obtained from the initial baking environment parameter dataset. The outputs are the air duct adjustment parameters applied to the baking environment parameter optimization objective function, the heating power control curve applied to the baking environment parameter optimization objective function, the humidity control parameters applied to the baking environment parameter optimization objective function, and the optimal baking duration applied to the optimal baking time sequence control scheme.

[0043] The specific implementation manners of the above steps are described in detail below. The specific implementation manner of step S01 is to arrange multi-channel sensor nodes in the tobacco leaf baking house in a grid point sampling manner, and each sensor node is configured with a temperature probe, a humidity probe and a thermal ball anemometer. First, determine the spatial coordinate system of the baking house, divide the baking house into three vertical layers: the bottom area, the middle area and the top area, and set a 9×9 sensor array at equal intervals in each vertical layer, with a total of 243 sampling points. Then install integrated sensors at each sampling point, set the acquisition frequency to 30 seconds / time, and continuously acquire data for 72 hours, covering the yellowing stage, the color fixation stage and the dry rib stage of tobacco leaf baking. The acquired data includes a temperature range of 30 to 68 °C, with an accuracy of ±0.5 °C; a relative humidity range of 18% to 90%, with an accuracy of ±2%; and an air flow velocity range of 0.1 to 5.0 m / s, with an accuracy of ±0.05 m / s. Finally, associate the acquired time series data with the spatial coordinates to establish a four-dimensional data structure (three-dimensional spatial coordinates and time dimension), forming an initial data set of baking environment parameters. The purpose of this step is to obtain the real distribution data of the environmental parameters in the tobacco leaf baking house and provide basic data support for subsequent model construction and verification.

[0044] The specific implementation manner of step S02 is to use three-dimensional modeling software to construct an accurate geometric model of the tobacco leaf baking house, including the outer shell structure, internal drying racks, heating devices, ventilation ducts and tobacco leaf hanging rods of the tobacco leaf baking house. First, establish a basic geometric structure according to the actual size of the baking house (such as a typical size of length × width × height = 8 m × 2.7 m × 3.5 m), and use solid modeling methods to accurately express the shapes and relative position relationships of each component. Then import it into finite element mesh generation software, and use a combination of structured meshes and unstructured meshes for mesh generation. In the areas near the wall and the tobacco leaves, boundary layer meshes are used for encryption processing to control the height of the first layer of meshes to meet the condition that y + (non-dimensional wall distance) is less than 5 to ensure the calculation accuracy. The total number of meshes is controlled within the range of 5 million to 10 million, and the mesh quality control index is that the orthogonality quality is not less than 0.7 and the skewness does not exceed 0.3. Finally, set the mesh boundary conditions, including the inlet boundary (velocity inlet), the outlet boundary (pressure outlet), the wall boundary (no-slip condition), the tobacco leaf surface (porous medium condition), etc., to provide a mesh basis for subsequent fluid simulation calculations. The purpose of this step is to establish an accurate geometric model and high-quality calculation meshes that meet the requirements of computational fluid dynamics solutions, and provide a spatial discretization basis for subsequent numerical simulations.

[0045] The specific implementation of step S03 is to import the initial dataset of baking environment parameters obtained in step S01 into a computational fluid dynamics simulation software and set up the physical model of the baking process. First, select a suitable computational fluid dynamics solver, such as FLUENT, CFX, or OpenFOAM, etc., and import the mesh file generated in step S02. Then set the fluid physical property parameters, including air density (calculated using the ideal gas state equation), dynamic viscosity (1.789×10 -5 kg / (m·s)), specific heat capacity (1006.43 J / (kg·K)), thermal conductivity (0.0242 W / (m·K)), etc. Next, select an appropriate turbulence model. For the complex flow characteristics in the baking room, use the standard k-ε turbulence model or the realizable k-ε turbulence model, and set the boundary conditions of turbulent kinetic energy k and turbulent dissipation rate ε. Among them, the inlet turbulence intensity is set to 5%, and the turbulent length scale is set to 7% of the hydraulic diameter. Subsequently, establish a multiphase flow model, and use the discrete phase model (DPM) or the volume fraction model (VOF) to simulate the gas-solid two-phase flow. Set the tobacco leaves as a porous medium region, with the porosity set to 0.4 - 0.6, the inertial resistance coefficient set to 0.5 - 2.0, and the viscous resistance coefficient set to 10 4 ~10 6 . Finally, set the solver parameters, including time step, discretization format, convergence criterion, etc. Among them, the convection term uses the second-order upwind format, the diffusion term uses the second-order central difference format, and the pressure-velocity coupling uses the SIMPLE algorithm, with the convergence residual controlled below 10 -4 . The purpose of this step is to establish an accurate numerical simulation basic model of the tobacco leaf baking environment and provide a computational framework for subsequent flow field analysis.

[0046] The specific implementation of step S04 is to perform the computational fluid dynamics solving process based on the numerical simulation basic model constructed in step S03. First, conduct a grid independence analysis, select different grid densities for preliminary calculations, and determine the optimal grid configuration where the calculation results are not affected by the grid density. Then start the main solving process, use parallel computing technology to accelerate the solution, and perform numerical simulations using an 8 - 32 core computing node. During the calculation process, monitor the temperature, humidity, velocity change curves at key points and the residual convergence process to ensure calculation stability. When the residual is reduced to the preset threshold (usually 10 -4 ~10 -6) When the key variables no longer change with the number of iterations, it is determined that the calculation converges. After convergence, post-processing extracts the data of the flow field distribution, temperature field distribution, and humidity field distribution in the tobacco leaf baking room, and generates three-dimensional cloud maps, two-dimensional sectional views, and contour maps. For the temperature field, isothermal surfaces are generated at intervals of 5 °C in the range of 30 °C to 68 °C; for the humidity field, isohumidity surfaces are generated at intervals of 5% in the range of 18% to 90%; for the flow field, velocity vector maps, streamline maps, and vorticity distribution maps are generated. Finally, these graphical results are integrated into an environmental field parameter distribution atlas to visually display the spatial distribution characteristics of the baking environment. The purpose of this step is to obtain the detailed distribution of environmental parameters in the tobacco leaf baking room through numerical simulation, providing visual data support for subsequent environmental optimization.

[0047] The specific implementation of step S05 is to use machine learning technology to intelligently analyze the environmental field parameter distribution map obtained in step S04. First, the environmental field parameter distribution map is converted into a data matrix, and three-dimensional grid points are used as sample points. Each sample point contains characteristic parameters such as temperature, humidity, air flow velocity, and their gradients. Then, unsupervised learning algorithms are applied to preprocess the data, including feature standardization (standard deviation standardization or maximum-minimum standardization), dimensionality reduction (principal component analysis PCA or t-SNE dimensionality reduction), etc. Next, clustering algorithms such as K-means clustering, Hierarchical Clustering, or DBSCAN clustering are used to divide the baking environment space into different characteristic regions. Among them, the initial value of K in K-means clustering is set to 3 to 7, and the optimal number of clusters is determined by the Silhouette Coefficient being greater than 0.6 or the Davies-Bouldin index being less than 0.5. For hierarchical clustering, the Ward linkage method is used to construct a clustering tree, and the pruning threshold is set to 25% to 35% of the total variance. After clustering, the mean, variance, and entropy values of each cluster are calculated, and regions where the temperature gradient exceeds 5 °C / m or the humidity gradient exceeds 8% / m are identified as environmentally non-uniform regions, and regions where the temperature exceeds the target value by 10% or the air flow velocity is lower than 0.2 m / s are identified as hot spot regions. Finally, three-dimensional distribution maps of environmentally non-uniform regions and hot spot regions are generated, marking the key improvement target points. The purpose of this step is to identify problem regions in the tobacco leaf baking environment through data mining methods, providing precise targets for subsequent optimization.

[0048] The specific implementation of step S06 is to construct a temperature and humidity gradual change model for the baking process and generate an optimized curve. First, extract the original baking curve from historical baking data, including the temperature curve T(t) and the humidity curve H(t), where t is the baking time. Then, use a piecewise polynomial function to fit the original baking curve to construct a basic temperature and humidity gradual change model. For the yellowing stage (usually the first 24 hours), the temperature gradual change model uses a second-order polynomial; for the color fixation stage (usually 24 - 48 hours), the temperature gradual change model uses a third-order polynomial; for the stem drying stage (usually after 48 hours), the temperature gradual change model uses an exponential function. Next, based on the original baking curve, generate multiple groups (usually 10 - 20 groups) of candidate optimized gradual change curves through the parameter perturbation method, and the perturbation range is controlled within ±10% of the original curve. The perturbation strategy uses the Latin Hypercube Sampling method to ensure uniform coverage of the parameter space. Subsequently, use the tobacco leaf quality prediction model to calculate the impact indicators of each group of optimized gradual change curves on the tobacco leaf quality, including total sugar content, reducing sugar content, nicotine content, chemical composition balance, etc. The quality prediction model is established using the Response Surface Methodology or Support Vector Regression (SVR) method, and the prediction accuracy is controlled within a relative error of 5%. Finally, select 3 - 5 groups of optimized gradual change curves with the highest comprehensive score of quality indicators as candidate solutions for subsequent optimization. The purpose of this step is to optimize the baking process curve through mathematical modeling methods and provide temperature and humidity control strategies for improving the quality of tobacco leaves.

[0049] The specific implementation of step S07 is to analyze the temporal characteristics of the temperature and humidity gradual change curve and identify key gradual change points. First, discretize the candidate optimized gradual change curves generated in step S06 into time series data with a sampling interval of 10 minutes. Then, design a temperature and humidity curve similarity evaluation function and use the Dynamic Time Warping (DTW) algorithm to calculate the similarity of adjacent time window curves. The length of the time window is set to 2 - 4 hours, and the sliding step is 30 minutes. The similarity calculation formula is S(i, j) = exp(-DTW(C i , C j ) / σ), where C i and C jThe temperature and humidity curve segments for adjacent time windows, where σ is the scale parameter and its value is twice the standard deviation of the curve. Then, key gradual change points are identified based on the change rate of the similarity value. When the similarity change rate is lower than the threshold of 0.15, it is determined as a key gradual change point. For each candidate optimized gradual change curve, usually 3 to 5 key gradual change points are identified, which respectively correspond to the main stage conversion points of the baking process. Subsequently, for each key gradual change point, trace back and extract the leading stage curve in front of it. The duration of the leading stage is set as the temperature and humidity change curve in the 4 to 8 hours before the key gradual change point. Finally, analyze the slope, curvature, and fluctuation characteristics of the leading stage curve, and quantitatively characterize the mathematical features of the influence of the leading stage on the key gradual change point. The purpose of this step is to accurately identify the key time nodes and their leading features during the baking process, providing a time sequence basis for precisely controlling the baking process.

[0050] The specific implementation of step S08 is to establish a multivariate correlation model between the tobacco leaf quality and the baking environment parameters. First, establish an initial data set using the leading stage curve features identified in step S07 and the tobacco leaf quality detection results in the historical baking data, including environmental parameters (temperature, humidity, air velocity, and their spatial distribution characteristics) as independent variables, and tobacco leaf quality indicators (chemical composition content, color, aroma, etc.) as dependent variables. Then, use feature selection methods to screen key influencing factors, including Pearson correlation coefficient analysis, recursive feature elimination (RFE), and principal component analysis, etc., and retain the parameters with the absolute value of the correlation coefficient greater than 0.5 or the top 60% of the feature importance rankings. Next, construct a multivariate correlation prediction model. For linear relationships, use multiple linear regression (MLR) or partial least squares regression (PLS); for non-linear relationships, use artificial neural network (ANN) or random forest (RF) algorithms. Among them, the neural network structure is a three-layer feedforward network, the number of hidden layer neurons is 10 to 20, the activation function selects the ReLU function, the training uses the Levenberg-Marquardt algorithm, and the learning rate is set to 0.01 to 0.05. The model verification uses the 10-fold cross-validation method, and the prediction accuracy evaluation index is that the root mean square error (RMSE) is less than 5% and the coefficient of determination (R 2 ) is greater than 0.85. Finally, based on the verified multivariate correlation model, construct an optimization objective function for the baking environment parameters, comprehensively considering multi-objective constraint conditions such as maximizing the tobacco leaf quality, minimizing the energy consumption, and optimizing the baking uniformity. The purpose of this step is to establish a quantitative relationship model between the baking environment parameters and the tobacco leaf quality, providing a scientific basis for parameter optimization.

[0051] The specific implementation of step S09 is to perform comprehensive optimization calculations using the baking environment parameter fusion optimization function. First, construct the input vector of the baking environment parameter fusion optimization function, including the temperature field uniformity coefficient (the coefficient of variation should be less than 0.1), humidity field distribution index (the standard deviation should be less than 5%), air velocity vector distribution (flow velocity range 0.2 - 2.0 m / s), key gradient point time series, and tobacco leaf moisture content (initial value 75% - 85%, target value 12% - 15%) obtained from the environmental field parameter distribution map. Then, design a multi-objective optimization algorithm framework, and use the weighted summation method or the Pareto optimal method to construct a comprehensive objective function. For the weighted summation method, the weight coefficients are determined by the Analytic Hierarchy Process (AHP), and the consistency ratio is controlled within 0.1. Next, apply the Particle Swarm Optimization Algorithm (PSO), Genetic Algorithm (GA), or Differential Evolution Algorithm (DE) to solve the optimization problem. The number of particles or population size is set to 50 - 100, the number of iterations is set to 200 - 500 times, and the convergence criterion is that the change rate of the optimal fitness in 20 consecutive iterations is less than 0.1%. Subsequently, extract the optimal baking timing control scheme from the optimization results, including air duct adjustment parameters (damper opening, wind speed adjustment value), heating power control curve, humidity control parameters, and the optimal baking duration. The air duct adjustment parameters change with the baking stage. The wind speed during the yellowing period is controlled at 0.3 - 0.5 m / s, the wind speed during the color fixation period is controlled at 0.5 - 1.0 m / s, and the wind speed during the dry leaf stalk period is controlled at 1.0 - 1.5 m / s. The heating power control curve is dynamically adjusted according to the temperature gradient target, and the power range is 30% - 90% of the rated power. The humidity control parameters are achieved through the coordinated control of the spraying system and the moisture exhaust system, and the moisture exhaust timer is set to 15 - 30 minutes per time. The optimal baking duration is usually 90 - 120 hours, and it is dynamically adjusted according to the initial moisture content of the tobacco leaves and the target drying curve. The purpose of this step is to calculate the optimal baking control strategy through the multi-objective optimization method and achieve the comprehensive optimization of tobacco leaf quality and energy efficiency.

[0052] For the wind speed adjustment in step S09, the present invention also provides another specific implementation method for the control of high-frequency fans and low-frequency fans. The specific adjustment parameters are as follows: Yellowing period (front and middle stages): low-frequency fan speed (960 r·min-1), temperature (34°C - 38°C), inlet air velocity 4.5 m / s, leaf gap air velocity 0.25 - 0.28 m / s; Yellowing period (latter stage): high-frequency fan speed (1440 r·min-1), temperature (42°C), inlet air velocity 5 m / s, leaf gap air velocity 0.38 - 0.42 m / s; Color fixation period: high-frequency fan speed (1440 r·min-1), temperature (44°C - 54°C), inlet air velocity 5 m / s, leaf gap air velocity 0.38 - 0.42 m / s; Dry leaf stalk period: low-frequency fan speed (960 r·min-1), inlet air velocity 4 m / s, leaf gap air velocity 0.23 - 0.25 m / s.

[0053] The specific implementation of step S10 is to verify the effectiveness of the optimization scheme in the numerical simulation environment and the actual baking environment. First, import the optimal baking time sequence control scheme obtained in step S09 into the numerical simulation basic model of the tobacco leaf baking environment established in step S03, and set the corresponding boundary conditions and control parameters. Then, perform a full-condition transient numerical simulation, adopt an adaptive time-step strategy, with an initial time step of 10 seconds and a maximum time step of 120 seconds, and the total calculation duration covers the entire baking cycle. Next, post-process the simulation results, calculate the environmental uniformity indicators, including the standard deviation of the temperature field (should be less than 2°C), the standard deviation of the humidity field (should be less than 3%), the coefficient of variation of the airflow field (should be less than 0.15), etc., and calculate the optimization objective function value of the baking environment parameters. Compare and analyze the performance indicators of different optimization schemes, and screen out the combination of baking environment parameters with the best environmental uniformity indicators and the largest objective function value. Finally, apply the selected optimal combination of baking environment parameters to the actual baking process, and monitor the temperature, humidity, and air velocity of the key control points in real time during the baking process to ensure that the deviation between the actual value and the set value is within the allowable range (temperature ±2°C, humidity ±5%, air velocity ±0.1 m / s). After baking, conduct physical and chemical index tests on the tobacco leaf samples, including chemical composition analysis (total sugar, reducing sugar, nicotine, total nitrogen, etc.), physical property tests (filling value, hygroscopicity, combustibility, etc.), and sensory evaluation (color, aroma, irritation, etc.), to verify the improvement effect of the optimization scheme on the quality of tobacco leaves. The purpose of this step is to evaluate the actual effect of the baking environment parameter optimization method through a combination of simulation verification and actual application verification, and ensure the reliability and effectiveness of the optimization scheme.

[0054] The following details the mathematical models or calculation processes involved in the present invention.

[0055] The construction of the initial data set of the baking environment parameters in step S01 involves the calculation process of data collection and organization. It is specifically expressed as follows:

[0056] D init ={P ijk (t)|i = 1, 2,..., 9; j = 1, 2,..., 9; k = 1, 2, 3; t = 1, 2,..., T};

[0057] In the formula, D init is the initial data set of the baking environment parameters; P ijk (t) is the environmental parameter vector at the spatial position point (i, j, k) at time t; i and j are the horizontal plane grid indices; k is the vertical layer index (1 is the bottom area, 2 is the middle area, 3 is the top area); T is the total sampling duration (in sampling intervals. If the sampling interval is 30 seconds and the sampling is for 72 hours, then T = 8640).

[0058] Environmental parameter vector P ijk (t) is specifically expressed as:

[0059] P ijk (t) = [T ijk (t), H ijk (t), V ijk (t)] T ;

[0060] In the formula, T ijk (t) is the temperature value (unit: °C, range: 30 - 68 °C) of the position point (i, j, k) at time t; H ijk (t) is the relative humidity value (unit: %, range: 18% - 90%); V ijk (t) is the air velocity vector (unit: m / s, range: 0.1 - 5.0 m / s).

[0061] Air velocity vector V ijk (t) is expressed as:

[0062] V ijk (t) = [u ijk (t), v ijk (t), w ijk (t)] T ;

[0063] In the formula, u ijk (t), v ijk (t), w ijk (t) are the velocity components in the three coordinate directions (unit: m / s) respectively.

[0064] The parameter acquisition method is as follows: The temperature data T ijk (t) is collected in real time by a PT100 thermal resistance sensor, and the four-wire connection method is used to eliminate the influence of line impedance. The signal is stored after being converted by a 24-bit ADC; the humidity data H ijk (t) is collected by a capacitive humidity sensor, and a temperature compensation algorithm is used to improve the accuracy; the velocity vector V ijk (t) is measured by a hot-wire anemometer, and the orthogonal three-axis arrangement method is used to obtain the velocity components in the three directions.

[0065] The calculation process of the turbulence model in step S03 involves the solution of the turbulent kinetic energy k and the turbulent dissipation rate ε. The standard k-ε turbulence model is expressed as follows:

[0066]

[0067] In the formula, ρ is the air density (unit: kg / m 3 ); u iis the velocity component (unit: m / s); μ is the aerodynamic viscosity (value 1.789×10 -5 kg / (m·s)); μ t is the turbulent viscosity (unit: kg / (m·s)); σ k is the Prandtl number of turbulent kinetic energy (value 1.0); σ ε is the Prandtl number of turbulent dissipation rate (value 1.3); P k is the generation term of turbulent kinetic energy; C 1ε is a constant (value 1.44); C 2ε is a constant (value 1.92).

[0068] The turbulent viscosity μ t is calculated as follows:

[0069]

[0070] In the formula, C μ is an empirical constant (value 0.09).

[0071] The generation term P of turbulent kinetic energy k is calculated as follows:

[0072]

[0073] This turbulence model considers the turbulence effect in fluid motion. By solving the transport equations of turbulent kinetic energy and dissipation rate, the turbulent viscosity is calculated, which in turn affects the diffusion term in the momentum equation. This model is applicable to fully developed turbulent flows and can better predict the complex flow characteristics in the tobacco leaf baking house.

[0074] The flow field calculation in step S04 involves the solution of the mass, momentum, and energy conservation equations. The mass conservation equation (continuity equation) is expressed as:

[0075]

[0076] The momentum conservation equation (Navier-Stokes equation) is expressed as:

[0077]

[0078] In the formula, p is the pressure (unit: Pa); g i is the component of gravitational acceleration (value 9.81 m / s 2 , with the direction vertically downward); S i is the momentum source term in the porous medium region (unit: N / m 3 ).

[0079] The energy conservation equation is expressed as:

[0080]

[0081] In the formula, h is the specific enthalpy (unit: J / kg); λ is the air thermal conductivity (value: 0.0242 W / (m·K)); c p is the specific heat capacity of air (value: 1006.43 J / (kg·K)); σ h is the turbulent Prandtl number (value: 0.85); S h is the energy source term (unit: W / m 3 ).

[0082] The surface treatment of the tobacco leaf is a porous medium region, and its momentum source term S i is calculated as follows:

[0083]

[0084] In the formula, α is the permeability (value range: 10 -8 ~10 -10 m 2 ); V i is the velocity component in the porous medium region; |V| is the velocity magnitude; C2 is the inertial drag coefficient (value range: 0.5~2.0).

[0085] The above equations are discretely solved by the finite volume method. The convection term adopts the second-order upwind scheme, the diffusion term adopts the second-order central difference scheme, and the SIMPLE algorithm is used for the pressure-velocity coupling. This solution process completely describes the fluid flow and heat transfer processes in the tobacco leaf baking room and provides basic data for subsequent analysis.

[0086] The clustering analysis in step S05 involves the calculation process of unsupervised learning algorithms. The K-means clustering algorithm is expressed as follows:

[0087]

[0088] In the formula, J is the objective function (total sum of squared errors); k is the number of clusters (value range: 3~7); n j is the number of samples in the jth cluster; is the feature vector of the ith sample point in the jth cluster; c j is the center point (centroid) of the jth cluster; is the Euclidean distance from the sample point to the cluster center.

[0089] The representation of the sample point feature vector x i is:

[0090]

[0091] In the formula, T i is the sample point temperature; H iis the humidity of the sample point; |V i is the magnitude of the air flow velocity at the sample point; is the temperature gradient vector; is the humidity gradient vector.

[0092] Temperature gradient vector is calculated as follows:

[0093]

[0094] Humidity gradient vector is calculated as follows:

[0095]

[0096] The silhouette coefficient S, an evaluation index of clustering quality, is calculated as follows:

[0097]

[0098] In the formula, n is the total number of samples; a i is the average distance between sample i and other samples in the same cluster; b i is the average distance between sample i and the nearest sample in a different cluster. The value range of S is [-1, 1], and the closer it is to 1, the better the clustering effect.

[0099] The criterion for identifying uneven environmental areas is:

[0100]

[0101] In the formula, is the magnitude of the temperature gradient; is the magnitude of the humidity gradient; δ T is the temperature gradient threshold (with a value of 5 °C / m); δ H is the humidity gradient threshold (with a value of 8% / m).

[0102] The criterion for identifying hot spots is:

[0103] T > T target × (1 + δ′ T ) or |V| < V min ;

[0104] In the formula, T is the regional temperature; T target is the target temperature; δ′ T is the temperature deviation threshold (with a value of 0.1); |V| is the magnitude of the air flow velocity; V min is the minimum air flow velocity threshold (with a value of 0.2 m / s).

[0105] This clustering analysis method can effectively identify uneven areas and hot spots in the baking environment by partitioning the feature space of environmental parameters and their gradients, providing precise targets for subsequent optimization.

[0106] The construction of the temperature and humidity gradual change model in step S06 involves piecewise function fitting calculation. The temperature and humidity gradual change model is expressed as follows:

[0107]

[0108] In the formula, T(t) is the temperature value at time t (unit: °C); H(t) is the humidity value at time t (unit: %); t1 is the end time of the yellowing stage (usually 24 hours); t2 is the end time of the color fixation stage (usually 48 hours); a1, b1, c1, a2, b2, c2, d2, a3, b3, c3 are the fitting parameters of the temperature gradual change model; p1, q1, r1, p2, q2, r2, s2, p3, q3, r3 are the fitting parameters of the humidity gradual change model.

[0109] The parameter acquisition method is: using the least squares method to fit the historical baking data. Let the original data points be (t i , T i ) and (t i , H i ), then the fitting parameters are obtained by minimizing the following objective function:

[0110]

[0111] In the formula, N is the number of data points.

[0112] The optimized gradual change curve is generated by parameter perturbation, and its expression is:

[0113] T′(t) = T(t) + ΔT(t);

[0114] H′(t) = H(t) + ΔH(t);

[0115] In the formula, T′(t) and H′(t) are the optimized gradual change curves; ΔT(t) and ΔH(t) are the perturbation amounts.

[0116] The perturbation amounts are generated by the Latin hypercube sampling method, and their constraints are:

[0117] |ΔT(t)| ≤ 0.1 × |T(t)|;

[0118] |ΔH(t)| ≤ 0.1 × |H(t)|;

[0119] The segmented design of the temperature and humidity gradual change model fully considers the characteristics of different tobacco leaf baking stages: in the yellowing stage, a second-order polynomial is used to simulate the slow temperature rise process; in the color fixation stage, a third-order polynomial is used to simulate the complex temperature and humidity changes; in the dry rib stage, an exponential function is used to simulate the characteristics of rapid temperature rise and rapid humidity drop. This model can accurately describe the baking process curve and provide a mathematical basis for optimization.

[0120] The evaluation of the similarity of the temperature and humidity curve in step S07 involves the calculation of the dynamic time warping algorithm. The similarity evaluation function is expressed as follows:

[0121]

[0122] In the formula, S(i, j) is the curve similarity between time windows i and j; DTW(C i , C j ) is the dynamic time warping distance; σ is the scale parameter (the value is 2 times the standard deviation of the curve); C i and C j are the temperature and humidity curve segments of adjacent time windows.

[0123] The dynamic time warping distance DTW(C i , C j ) is calculated as follows:

[0124] DTW(C i , C j ) = d(C i (m), C j (n));

[0125] Among them, d(C i (m), C j (n)) is solved by dynamic programming recursion:

[0126]

[0127] In the formula, δ(C i (m), C j (n)) is the point distance, usually the Euclidean distance:

[0128]

[0129] In the formula, T i (m) and H i (m) are the temperature and humidity values at the m-th time point in time window i respectively; T j (n) and H j (n) are the temperature and humidity values at the n-th time point in time window j respectively.

[0130] The calculation of the similarity change rate is as follows:

[0131]

[0132] The key gradient point recognition criterion is as follows:

[0133] ΔS(i)<δ S ;

[0134] where δ S is the similarity change rate threshold (taking the value of 0.15).

[0135] This similarity evaluation method adopts the dynamic time warping algorithm, which can effectively handle the time stretching problem in time series data, is suitable for identifying the key turning points of the temperature and humidity change patterns during the baking process, and provides a time series basis for accurately controlling the baking process.

[0136] The establishment of the multi - variable correlation model in step S08 involves machine learning algorithm calculations. The multi - variable correlation model based on neural network is expressed as follows:

[0137] Q = f(X, W, b);

[0138] where Q is the tobacco leaf quality index vector; X is the environmental parameter vector; W is the network weight matrix; b is the bias vector; f is the non - linear mapping function.

[0139] For a three - layer feed - forward neural network, the specific calculation is as follows:

[0140] h = g1(W1X + b1);

[0141] Q = g2(W2h + b2);

[0142] where h is the output of the hidden layer; g1 and g2 are activation functions, usually the ReLU function: g(x)=max(0, x); W1 is the weight matrix from the input layer to the hidden layer; W2 is the weight matrix from the hidden layer to the output layer; b1 is the bias vector of the hidden layer; b2 is the bias vector of the output layer.

[0143] The representation of the environmental parameter vector X is:

[0144]

[0145] where T is the temperature; H is the humidity; V is the air flow velocity; is the temperature gradient; is the humidity gradient; σ T is the temperature standard deviation; σ H is the humidity standard deviation; σ V is the air flow velocity standard deviation.

[0146] The representation of the tobacco leaf quality index vector Q is:

[0147] Q = [TS, RS, NC, TP, CV, AF,...] T ;

[0148] Wherein, TS is the total sugar content (%); RS is the reducing sugar content (%); NC is the nicotine content (%); TP is the total phenol content (%); CV is the color value; AF is the aroma factor.

[0149] The optimization objective function of the baking environment parameters is expressed as:

[0150]

[0151] Wherein, F(X) is the optimization objective function; Q i (X) is the i-th quality index; w i is the weight coefficient of the i-th quality index; m is the number of quality indexes; E(X) is the energy consumption function; U(X) is the environmental non-uniformity function; λ1 and λ2 are balance coefficients.

[0152] The energy consumption function E(X) is expressed as:

[0153]

[0154] Wherein, T total is the total baking duration; P(t) is the power consumption at time t.

[0155] The environmental non-uniformity function U(X) is expressed as:

[0156] U(X) = w T ·CV T + w H ·CV H + w V ·CV V ;

[0157] Wherein, CV T is the temperature coefficient of variation; CV H is the humidity coefficient of variation; CV V is the air velocity coefficient of variation; w T 、w H and w V are weight coefficients.

[0158] The coefficient of variation is calculated as follows:

[0159]

[0160] Wherein, σ T 、σ H and σ V are the standard deviations of temperature, humidity and air velocity respectively; and They are the average values of temperature, humidity, and air velocity respectively.

[0161] This multivariate correlation model establishes a non-linear mapping relationship between the baking environment parameters and the tobacco leaf quality through deep learning methods. The optimization objective function comprehensively considers the tobacco leaf quality, energy consumption, and environmental uniformity, providing a scientific basis for parameter optimization. This model deeply integrates the complex patterns in the sample data and can effectively predict the tobacco leaf quality performance under different baking conditions.

[0162] The calculation process of the baking environment parameter fusion optimization function in step S09 involves a multi-objective optimization algorithm. The baking environment parameter fusion optimization function is expressed as:

[0163] Φ(Z) = α1Φ T (Z) + α2Φ H (Z) + α3Φ V (Z) + α4Φ G (Z) + α5Φ M (Z);

[0164] In the formula, Φ(Z) is the fusion optimization function; Z is the control parameter vector; Φ T (Z) is the temperature field uniformity sub-function; Φ H (Z) is the humidity field distribution sub-function; Φ V (Z) is the air velocity sub-function; Φ G (Z) is the key gradient point control sub-function; Φ M (Z) is the tobacco leaf moisture content control sub-function; α1, α2, α3, α4, and α5 are weight coefficients determined by the analytic hierarchy process.

[0165] The temperature field uniformity sub-function Φ T (Z) is expressed as:

[0166]

[0167] T (Z) is the temperature coefficient of variation under the control parameter Z; CV T，max is the maximum allowable value of the temperature coefficient of variation (usually taken as 0.1).

[0168] The humidity field distribution sub-function Φ H (Z) is expressed as:

[0169]

[0170] In the formula, σ H (Z) is the humidity standard deviation under the control parameter Z; σ H，max is the maximum allowable value of the humidity standard deviation (usually taken as 5%).

[0171] Airflow velocity sub - function Φ V is expressed as:

[0172]

[0173] where |V(Z)| is the average airflow velocity under control parameter Z; V min is the minimum allowable airflow velocity (usually taken as 0.2 m / s); V max is the maximum allowable airflow velocity (usually taken as 2.0 m / s); V opt is the optimal airflow velocity, which varies with the baking stage.

[0174] Key gradient point control sub - function Φ G is expressed as:

[0175]

[0176] where N G is the number of key gradient points; t i (Z) is the time of the i - th key gradient point under control parameter Z; is the target time of the i - th key gradient point; γ i is the importance coefficient of the i - th key gradient point.

[0177] Tobacco leaf moisture content control sub - function Φ M is expressed as:

[0178]

[0179] where M(Z) is the final tobacco leaf moisture content under control parameter Z; M * is the target moisture content (usually taken as 12% - 15%); β is the moisture content control weight coefficient.

[0180] The representation of the control parameter vector Z is:

[0181] Z = [D(t), P(t), H ctl (t), T total T ;

[0182] where D(t) is the air duct regulation parameter function; P(t) is the heating power regulation curve; H ctl (t) is the humidity control parameter function; T total is the total baking duration.

[0183] ​The integrated optimization function for baking environment parameters comprehensively considers multiple aspects such as the uniformity of the temperature field, the distribution of the humidity field, the rationality of the air flow velocity, the control accuracy of key gradient points, and the moisture content of tobacco leaves. Through the design of exponential functions and piecewise functions, the value ranges of each sub-function are standardized to 0-1, which is convenient for comprehensive evaluation. This function design fully considers the mutual influence and synergistic effect of various parameters during the baking process, providing a unified evaluation standard for multi-objective optimization.

[0184] Optionally, the calculation formula for the environmental uniformity index is:

[0185] EUI = w T ·(1 - CV T ) + w H ·(1 - CV H ) + w V ·(1 - CV V );

[0186] In the formula, EUI is the comprehensive index of environmental uniformity; CV T is the coefficient of variation of the temperature field; CV H is the coefficient of variation of the humidity field; CV V is the coefficient of variation of the air flow field; w T , w H and w V are weight coefficients, and satisfy w T + w H + w V = 1. Usually, w T = 0.5, w H = 0.3, w V = 0.2.

[0187] Among them, the calculation formula for the coefficient of variation is:

[0188]

[0189] In the formula, σ T , σ H and σ V are the standard deviations of temperature, humidity and air flow velocity respectively; and are the average values of temperature, humidity and air flow velocity respectively.

[0190] Among them, the calculation formula for the standard deviation is:

[0191]

[0192] In the formula, N is the number of sampling points; T i , H i and |V i | are the temperature value, humidity value and air flow velocity magnitude of the i-th sampling point respectively.

[0193] This environmental uniformity index comprehensively reflects the uniformity level of the baking environment by weighted evaluation of the variation degrees of temperature, humidity and air flow velocity distribution. The smaller the coefficient of variation, the more uniform the distribution of environmental parameters, and the closer the index value is to 1, the better the uniformity of the baking environment. This index provides a quantitative basis for verifying the optimization effect.

[0194] Specifically, the principle of the present invention is as follows: The technical principle of the present invention is based on the integrated application of computational fluid dynamics and data mining technology. By numerical simulation of fluid mechanics, the distribution law of environmental parameters in the tobacco leaf baking house is revealed, and then combined with unsupervised learning and multivariate correlation analysis methods to optimize the baking environmental parameters. This method first collects the air flow velocity, temperature and humidity data at multiple points in the tobacco leaf baking house to establish an initial data set, which provides boundary conditions and verification basis for fluid mechanics simulation. By constructing a three-dimensional grid model and setting a reasonable turbulence model, the complex air flow state in the baking house is accurately simulated, and the environmental field parameter distribution map is obtained, intuitively presenting the uniformity and change trend of the baking environment.

[0195] After obtaining the environmental field parameter distribution, the present invention applies an unsupervised learning algorithm for clustering analysis to scientifically identify the non-uniform areas and hot spots in the baking environment, and determine the spatial positions that need to be optimized key. Based on the information of these key areas, a temperature and humidity gradual change model is constructed to generate multiple groups of optimization curves, and the influence indexes of different gradual change curves on the quality of tobacco leaves are calculated. At the same time, a temperature and humidity curve similarity evaluation method is used to determine the key gradual change points, and a multivariate correlation model between the quality of tobacco leaves and the baking environmental parameters is established, so as to determine the optimization objective function of the baking environmental parameters.

[0196] The core of the present invention lies in the establishment and application of the fusion optimization function of baking environmental parameters. This function comprehensively considers factors such as the temperature field uniformity coefficient, humidity field distribution index, air flow velocity vector distribution, etc., and outputs the optimal air duct adjustment parameters, heating power control curve, humidity control parameters and the optimal baking duration to form a complete baking time sequence control scheme. After being verified in the basic model, by calculating the environmental uniformity index and the optimization objective function value, the optimal parameter combination is screened and applied to the actual baking process. The method of the present invention conforms to the basic principles of fluid mechanics and the laws of heat and moisture transfer. Through accurate simulation and data-driven optimization algorithms, the fine regulation of baking environmental parameters is realized, thus ensuring the stable and consistent quality of tobacco leaf baking.

[0197] The following provides a specific Embodiment 1 of the present invention, and the specific implementation manners of each step in this Embodiment 1 are described in detail as follows.

[0198] The specific implementation of step S01 is to arrange multi-channel sensor nodes in the tobacco leaf baking house in a grid point sampling manner. Each sensor node is equipped with a temperature probe, a humidity probe, and a thermal ball anemometer. First, determine the spatial coordinate system of the baking house, divide the baking house into three vertical layers: the bottom area, the middle area, and the top area, and set up a 9×9 sensor array at equal intervals in each vertical layer, with a total of 243 sampling points. Then install integrated sensors at each sampling point, set the acquisition frequency to 30 seconds / time, and continuously collect data for 72 hours, covering the yellowing stage, the color fixing stage, and the dry stem stage of tobacco leaf baking. The collected data constitutes the initial dataset of baking environment parameters, and the data structure is expressed as:

[0199] D init ={P ijk (t)|i = 1, 2,..., 9; j = 1, 2,..., 9; k = 1, 2, 3; t = 1, 2,..., T};

[0200] In the formula, D init is the initial dataset of baking environment parameters; P ijk (t) is the environmental parameter vector of the spatial position point (i, j, k) at time t; i, j are the horizontal plane grid indices; k is the vertical layer index (1 is the bottom area, 2 is the middle area, 3 is the top area); T is the total sampling duration (in sampling intervals, if the sampling interval is 30 seconds and sampling lasts for 72 hours, then T = 8640).

[0201] The specific representation of the environmental parameter vector P ijk (t) is:

[0202] P ijk (t)=[T ijk (t), H ijk (t), V ijk (t)] T ;

[0203] In the formula, T ijk (t) is the temperature value of the position point (i, j, k) at time t (unit: °C, range: 30 - 68 °C); H ijk (t) is the relative humidity value (unit: %, range: 18% - 90%); V ijk (t) is the air velocity vector (unit: m / s, range: 0.1 - 5.0 m / s). The purpose of this step is to obtain the real distribution data of the environmental parameters in the tobacco leaf baking house and provide basic data support for subsequent model construction and verification.

[0204] The specific implementation of step S02 is the same as the foregoing and will not be elaborated here.

[0205] The specific implementation of step S03 is to import the initial data set of the baking environment parameters obtained in step S01 into a computational fluid dynamics simulation software and set up the physical model of the baking process. First, select a suitable computational fluid dynamics solver, such as FLUENT, CFX, or OpenFOAM, etc., and import the mesh file generated in step S02. Then set the fluid physical property parameters, including air density (calculated using the ideal gas state equation), dynamic viscosity (1.789×10 -5 kg / (m·s)), specific heat capacity (1006.43 J / (kg·K)), thermal conductivity (0.0242 W / (m·K)), etc. Next, select an appropriate turbulence model. For the complex flow characteristics in the baking room, use the standard k-ε turbulence model or the realizable k-ε turbulence model. The standard k-ε turbulence model is expressed as follows:

[0206]

[0207] In the formula, ρ is the air density (unit: kg / m 3 ); u i is the velocity component (unit: m / s); μ is the air dynamic viscosity (value 1.789×10 -5 kg / (m·s)); μ t is the turbulent viscosity (unit: kg / (m·s)); σ k is the Prandtl number of turbulent kinetic energy (value 1.0); σ ε is the Prandtl number of turbulent dissipation rate (value 1.3); P k is the generation term of turbulent kinetic energy; C 1ε is a constant (value 1.44); C 2ε is a constant (value 1.92). Subsequently, establish a multiphase flow model, use the discrete phase model (DPM) or the volume fraction model (VOF) to simulate the gas-solid two-phase flow, set the tobacco leaves as a porous medium region, the porosity is set to 0.4 - 0.6, the inertial resistance coefficient is set to 0.5 - 2.0, and the viscous resistance coefficient is set to 10 4 ~10 6 . The purpose of this step is to establish an accurate numerical simulation basic model of the tobacco leaf baking environment and provide a calculation framework for subsequent flow field analysis.

[0208] The specific implementation of step S04 is to perform the computational fluid dynamics solution process based on the numerical simulation basic model constructed in step S03. First, conduct a grid independence analysis, select different grid densities for preliminary calculations, and determine the optimal grid configuration where the calculation results are not affected by the grid density. Then start the main solution process, use parallel computing technology to accelerate the solution, and perform numerical simulations using 8 - 32 core computing nodes. The solution process involves the calculations of the mass, momentum, and energy conservation equations. The mass conservation equation (continuity equation) is expressed as:

[0209]

[0210] The momentum conservation equation (Navier - Stokes equation) is expressed as:

[0211]

[0212] where p is the pressure (unit: Pa); g i is the component of gravitational acceleration (taking the value of 9.81 m / s 2 , with the direction vertically downward); S i is the momentum source term in the porous medium region (unit: N / m 3 ).

[0213] The energy conservation equation is expressed as:

[0214]

[0215] where h is the specific enthalpy (unit: J / kg); λ is the thermal conductivity of air (taking the value of 0.0242 W / (m·K)); c p is the specific heat capacity of air (taking the value of 1006.43 J / (kg·K)); σ h is the turbulent Prandtl number (taking the value of 0.85); S h is the energy source term (unit: W / m 3 ). During the calculation process, monitor the temperature, humidity, velocity change curves of key points and the residual convergence process to ensure the calculation stability. When the residual is reduced to the preset threshold (usually 10 -4 ~10 -6 ) and the key variables no longer change with the number of iterations, it is determined that the calculation converges. After convergence, post - processing extracts the data of the flow field distribution, temperature field distribution and humidity field distribution in the tobacco leaf baking room, generates three - dimensional cloud maps, two - dimensional sectional views and contour maps, and integrates them into an environmental field parameter distribution atlas. The purpose of this step is to obtain the detailed distribution of environmental parameters in the tobacco leaf baking room through numerical simulation, providing visual data support for subsequent environmental optimization.

[0216] The specific implementation of step S05 is to perform intelligent analysis on the environmental field parameter distribution map obtained in step S04 using machine learning techniques. First, convert the environmental field parameter distribution map into a data matrix, using three - dimensional grid points as sample points, and each sample point contains characteristic parameters such as temperature, humidity, air flow velocity and its gradient. Then apply unsupervised learning algorithms to pre - process the data, including feature standardization, dimensionality reduction, etc. Next, use the K - means clustering algorithm to divide the baking environment space into different characteristic regions. The objective function of the K - means clustering algorithm is expressed as:

[0217]

[0218] Where J is the objective function (total sum of squared errors); k is the number of clusters (ranging from 3 to 7); n j is the number of samples in the j-th cluster; is the feature vector of the i-th sample point in the j-th cluster; c j is the center point (centroid) of the j-th cluster; is the Euclidean distance from the sample point to the cluster center. The initial value of K in K-means clustering is set to 3 - 7, and the optimal number of clusters is determined by a silhouette coefficient greater than 0.6 or a Davies-Bouldin index less than 0.5. The formula for the silhouette coefficient is:

[0219]

[0220] Where n is the total number of samples; a i is the average distance between sample i and other samples in the same cluster; b i is the average distance between sample i and the nearest non-cluster sample. After clustering, calculate the mean, variance, and entropy value of each cluster, and identify the area where the temperature gradient exceeds 5 °C / m or the humidity gradient exceeds 8% / m as the environmental non-uniform area, and the area where the temperature exceeds the target value by 10% or the air flow velocity is lower than 0.2 m / s as the hot spot area. The criterion for identifying the environmental non-uniform area is:

[0221]

[0222] Where is the magnitude of the temperature gradient; is the magnitude of the humidity gradient; δ T is the temperature gradient threshold (value 5 °C / m); δ H is the humidity gradient threshold (value 8% / m). The purpose of this step is to identify the problem areas in the tobacco leaf baking environment through data mining methods, providing an accurate target for subsequent optimization.

[0223] The specific implementation of step S06 is to construct a temperature and humidity gradual change model for the baking process and generate an optimization curve. First, extract the original baking curves from historical baking data, including the temperature curve T(t) and the humidity curve H(t), where t is the baking time. Then, use a piecewise polynomial function to fit the original baking curves to construct a basic temperature and humidity gradual change model. The temperature and humidity gradual change model is expressed as follows:

[0224]

[0225] Wherein, T(t) is the temperature value at time t (unit: °C); H(t) is the humidity value at time t (unit: %); t1 is the end time of the yellowing stage (usually 24 hours); t2 is the end time of the color fixation stage (usually 48 hours); a1, b1, c1, a2, b2, c2, d2, a3, b3, c3 are the fitting parameters of the temperature gradient model; p1, q1, r1, p2, q2, r2, s2, p3, q3, r3 are the fitting parameters of the humidity gradient model. The parameter acquisition method is to fit the historical baking data using the least squares method. Then, based on the original baking curve, multiple groups (usually 10 - 20 groups) of candidate optimized gradient curves are generated through the parameter perturbation method, and the perturbation range is controlled within ±10% of the original curve. The expression of the optimized gradient curve is:

[0226] T′(t) = T(t) + ΔT(t);

[0227] H′(t) = H(t) + ΔH(t);

[0228] Wherein, T′(t) and H′(t) are the optimized gradient curves; ΔT(t) and ΔH(t) are the perturbation amounts, and the perturbation amount constraints are |ΔT(t)| ≤ 0.1×|T(t)| and |ΔH(t)| ≤ 0.1×|H(t)|. Subsequently, using the tobacco leaf quality prediction model, calculate the influence index of each group of optimized gradient curves on the tobacco leaf quality, and screen 3 - 5 groups of optimized gradient curves with the highest comprehensive score of quality indicators as the candidate solutions for subsequent optimization. The purpose of this step is to optimize the baking process curve through mathematical modeling methods and provide temperature and humidity control strategies for improving the quality of tobacco leaves.

[0229] The specific implementation method of step S07 is to analyze the temporal characteristics of the temperature and humidity gradient curves and identify the key gradient points. First, discretize the candidate optimized gradient curves generated in step S06 into time series data with a sampling interval of 10 minutes. Then, design a temperature and humidity curve similarity evaluation function, and use the dynamic time warping algorithm to calculate the similarity of adjacent time window curves. The length of the time window is set to 2 - 4 hours, and the sliding step is 30 minutes. The similarity calculation formula is:

[0230]

[0231] Wherein, S(i, j) is the curve similarity between time windows i and j; DTW(C i , C j ) is the dynamic time warping distance; σ is the scale parameter (the value is 2 times the standard deviation of the curve); C i and C j are the temperature and humidity curve segments of adjacent time windows. The dynamic time warping distance is solved by dynamic programming recursion, and the point distance uses the Euclidean distance:

[0232]

[0233] In the formula, T i (m) and H i (m) are the temperature and humidity values at the m-th time point in the i-th time window respectively; T j (n) and H j (n) are the temperature and humidity values at the n-th time point in the j-th time window respectively. The similarity change rate is calculated as follows:

[0234]

[0235] The key gradual change point identification criterion is ΔS(i) < δ S , where δ S is the similarity change rate threshold (taking the value of 0.15). For each candidate optimized gradual change curve, usually 3 to 5 key gradual change points are identified, which respectively correspond to the main stage conversion points of the baking process. Subsequently, for each key gradual change point, trace back forward and extract its leading stage curve, and the leading stage duration is set as the temperature and humidity change curve in the 4 to 8 hours before the key gradual change point. The purpose of this step is to accurately identify the key time nodes and their leading features in the baking process, and provide a timing basis for accurately controlling the baking process.

[0236] The specific implementation method of step S08 is to establish a multi - variable correlation model between the tobacco leaf quality and the baking environment parameters. First, use the leading stage curve features identified in step S07 and the tobacco leaf quality detection results in the historical baking data to establish an initial data set, including environmental parameters as independent variables and tobacco leaf quality indicators as dependent variables. Then, use the feature selection method to screen the key influencing factors, and retain the parameters with the absolute value of the correlation coefficient greater than 0.5 or the top 60% in terms of feature importance ranking. Next, construct a multi - variable correlation prediction model. For a three - layer feed - forward neural network, its calculation is expressed as:

[0237] h = g1(W1X + b1);

[0238] Q = g2(W2h + b2);

[0239] In the formula, h is the output of the hidden layer; g1 and g2 are activation functions, usually the ReLU function is selected; W1 is the weight matrix from the input layer to the hidden layer; W2 is the weight matrix from the hidden layer to the output layer; b1 is the bias vector of the hidden layer; b2 is the bias vector of the output layer; X is the environmental parameter vector; Q is the tobacco leaf quality indicator vector. Finally, based on the verified multi - variable correlation model, construct an optimization objective function for the baking environment parameters:

[0240]

[0241] In the formula, F(x) is the optimization objective function; Q i (X) is the i - th quality indicator; wi is the weight coefficient of the i-th quality index; m is the number of quality indexes; E(X) is the energy consumption function; U(X) is the environmental non-uniformity function; λ1 and λ2 are balance coefficients. The environmental non-uniformity function is calculated as follows:

[0242] U(X) = w T ·CV T + w H ·CV H + w V ·CV V ;

[0243] In the formula, CV T , CV H and CV V are the coefficient of variation of temperature, humidity and air velocity respectively. The coefficient of variation is calculated as the standard deviation divided by the mean. The purpose of this step is to establish a quantitative relationship model between the baking environment parameters and the tobacco leaf quality, providing a scientific basis for parameter optimization.

[0244] The specific implementation of step S09 is to perform comprehensive optimization calculations using the baking environment parameter fusion optimization function. First, construct the input vector of the baking environment parameter fusion optimization function, including the temperature field uniformity coefficient, humidity field distribution index, air velocity vector distribution, key gradual change point time series, and tobacco leaf moisture content obtained from the environmental field parameter distribution map. Then, design a multi-objective optimization algorithm framework and construct a comprehensive objective function using the weighted summation method or the Pareto optimal method. The baking environment parameter fusion optimization function is expressed as:

[0245] Φ(Z) = α1Φ T (Z)+α2Φ H (Z)+α3Φ V (Z)+α4Φ G (Z)+α5Φ M (Z);

[0246] In the formula, Φ(Z) is the fusion optimization function; Z is the control parameter vector; Φ T (Z) is the temperature field uniformity sub-function; Φ H (Z) is the humidity field distribution sub-function; Φ V (Z) is the air velocity sub-function; Φ G (Z) is the key gradual change point control sub-function; Φ M (Z) is the tobacco leaf moisture content control sub-function; α1, α2, α3, α4 and α5 are weight coefficients. The temperature field uniformity sub-function is expressed as:

[0247]

[0248] In the formula, CV T(Z) is the coefficient of variation of temperature under the control parameter Z; CV T，max is the maximum allowable value of the coefficient of variation of temperature (usually taken as 0.1). Then, the particle swarm optimization algorithm, genetic algorithm, or differential evolution algorithm is applied to solve the optimization problem. The number of particles or population size is set to 50 - 100, the number of iterations is set to 200 - 500 times, and the convergence criterion is that the change rate of the optimal fitness in 20 consecutive iterations is less than 0.1%. Subsequently, the optimal baking timing control scheme is extracted from the optimization results, including the air duct adjustment parameters, heating power control curve, humidity control parameters, and the optimal baking duration. The purpose of this step is to calculate the optimal baking control strategy through the multi-objective optimization method and achieve the comprehensive optimization of tobacco leaf quality and energy efficiency.

[0249] The specific implementation of step S10 is to verify the effectiveness of the optimization scheme in the numerical simulation environment and the actual baking environment. First, the optimal baking timing control scheme obtained in step S09 is imported into the numerical simulation basic model of the tobacco leaf baking environment established in step S03, and the corresponding boundary conditions and control parameters are set. Then, the full-condition transient numerical simulation is performed, adopting the adaptive time step strategy. The initial time step is 10 seconds, the maximum time step is 120 seconds, and the total calculation duration covers the entire baking cycle. Next, the simulation results are post-processed, and the environmental uniformity index is calculated. The formula for the environmental uniformity index is:

[0250] EUI = w T ·(1 - CV T ) + w H ·(1 - CV H ) + w V ·(1 - CV V ) ;

[0251] In the formula, EUI is the comprehensive index of environmental uniformity; CV T is the coefficient of variation of the temperature field; CV H is the coefficient of variation of the humidity field; CV V is the coefficient of variation of the airflow field; w T , w H and w V are the weight coefficients and satisfy w T + w H + w V = 1. Usually, w T = 0.5, w H = 0.3, w V = 0.2. The formula for the coefficient of variation is:

[0252]

[0253] In the formula, σ T , σ H and σ Vare the standard deviations of temperature, humidity, and air velocity, respectively; and are the average values of temperature, humidity, and air velocity, respectively. The standard deviation calculation formula is:

[0254]

[0255] In the formula, N is the number of sampling points; T i , H i and |V i | are the temperature value, humidity value, and air velocity magnitude at the i-th sampling point, respectively. By comparing and analyzing the performance indicators of different optimization schemes, the combination of baking environment parameters with the best environmental uniformity index and the largest objective function value is selected. Finally, the selected optimal combination of baking environment parameters is applied to the actual baking process, and the temperature, humidity, and air velocity at the key control points are monitored in real time during the baking process to ensure that the deviation between the actual value and the set value is within the allowable range (temperature ±2°C, humidity ±5%, air velocity ±0.1 m / s). After baking is completed, the physical and chemical indexes of the tobacco leaf samples are detected to verify the improvement effect of the optimization scheme on the quality of tobacco leaves. The purpose of this step is to evaluate the actual effect of the baking environment parameter optimization method through a combination of simulation verification and actual application verification, and to ensure the reliability and effectiveness of the optimization scheme.

[0256] To better understand and implement the present invention, Example 2 of a specific application scenario of the present invention is provided below: Researchers carried out research on the problem of uneven quality of cured tobacco leaves in a certain flue-cured tobacco production area and precisely optimized the baking environment parameters. First, in a curing barn with a standard size of 8m × 2.7m × 3.5m, a total of 243 sensor nodes of 9×9×3 were installed according to step S01. During the acquisition period, the ambient temperature was 30 - 68°C, the relative humidity was 18% - 90%, and the air velocity was 0.1 - 5.0 m / s. Continuous acquisition for 72 hours covered a complete baking cycle.

[0257] An accurate three-dimensional model of the curing barn was established using AutoCAD or Solidworks, including the outer shell structure, internal drying racks, heating devices, ventilation ducts, and tobacco leaf hanging rods, etc. The ICEM CFD software was used to mesh the model. The total number of meshes was 6.42 million, the orthogonal quality was 0.82, the skewness was 0.25, and the height of the first layer of meshes met the requirement of y + <5.

[0258] The initial data was imported into the ANSYS FLUENT software, the air physical property parameters were set, and the standard k-ε turbulence model was selected. The turbulence model parameter settings are shown in Table 1:

[0259] Table 1 Turbulence model parameter settings

[0260]

[0261]

[0262] The tobacco leaves are set as a porous medium region with a porosity of 0.52, an inertial resistance coefficient of 1.2, and a viscous resistance coefficient of 2.5×10 5 . The numerical simulation is run, and after 3200 iterations, the residual converges to 1×10 -5 , and the complete distribution data of the temperature field, humidity field, and flow field are obtained.

[0263] The environmental field parameter distribution map is converted into a data matrix, and K-means clustering analysis is performed on the sampled data. After evaluation by the silhouette coefficient, the optimal number of clusters is 5.

[0264] Through clustering analysis, regions with a temperature gradient exceeding 5 °C / m and a humidity gradient exceeding 8% / m are identified, accounting for 23.5% and 26.8% of the total volume respectively, mainly concentrated in the corners of the baking room and near the heat source.

[0265] Based on historical baking data, a temperature and humidity gradual change model is established. The parameters of the original temperature and humidity curves are shown in Table 2:

[0266] Table 2 Parameters of the original temperature and humidity curves

[0267]

[0268] Fifteen groups of optimized gradual change curves are generated using the Latin hypercube sampling method. Through analysis with the similarity evaluation function, 4 key gradual change points in each group of curves are identified, and the average similarity change rate is 0.11. The chronological distribution of the key gradual change points is shown in Table 3:

[0269] Table 3 Chronological distribution of the key gradual change points

[0270] Curve number First gradient point (h) Second gradient point (h) Third gradient point (h) Fourth gradient point (h) 1 12.3 24.6 37.8 58.2 2 11.8 23.9 36.5 57.6 3 12.5 25.2 38.4 59.1 … … … … … 15 12.1 24.2 37.5 58.5

[0271] Based on the curve characteristics of the leading stage, a three-layer neural network model is constructed to establish the correlation between the tobacco leaf quality and the baking environment parameters. The number of hidden layer neurons is 16, the ReLU activation function is used, and the root mean square error obtained through 10-fold cross-validation is 3.8%, and the coefficient of determination is 0.91. The prediction accuracy of the key quality indicators is shown in Table 4:

[0272] Table 4 Prediction accuracy of the quality indicators

[0273] Quality index Root mean square error (%) Coefficient of determination Total sugar content 2.6 0.94 Reducing sugar content 3.1 0.92 Nicotine content 4.2 0.89 Total phenol content 3.9 0.90 Color value 5.3 0.86 Aroma factor 4.5 0.88

[0274] Construct a fusion optimization function for baking environment parameters. The weight coefficients are determined by the analytic hierarchy process, which are α1 = 0.35, α2 = 0.25, α3 = 0.20, α4 = 0.15, and α5 = 0.05 respectively. Apply the particle swarm optimization algorithm for solution. The number of particles is set to 80, and the number of iterations is 350. After optimization, the optimal control parameter combination is obtained. The comparison of the environmental uniformity index before and after optimization is shown in Table 5:

[0275] Table 5 Comparison of environmental uniformity indexes before and after optimization

[0276] Index Before optimization After optimization Improvement rate (%) Temperature coefficient of variation 0.086 0.062 27.9 Humidity coefficient of variation 0.092 0.081 12.0 Airflow velocity coefficient of variation 0.153 0.128 16.3 Comprehensive index of environmental uniformity 0.736 0.796 8.2

[0277] Apply the optimal parameter scheme to the numerical simulation model for verification. The environmental uniformity index reaches 0.926, which is much higher than 0.736 before optimization. Subsequently, implement this scheme during the actual baking process. The test results of the physical and chemical indexes of tobacco leaves are shown in Table 6:

[0278] Table 6 Test results of physical and chemical indexes of tobacco leaves

[0279] Physical and chemical indexes Traditional method Optimization method Enhancement rate (%) Total sugar content (%) 25.6 27.8 8.6 Reducing sugar content (%) 24.1 25.2 4.6 Nicotine content (%) 2.83 2.25 20.5 (decrease) Total sugar / nicotine ratio 6.59 9.83 49.2 Color value (Hunter value) 32.65 35.28 8.1 Filling value (cm / g) 3.42 3.86 12.9 Uniformity score 7.2 8.9 23.6

[0280] The optimization of traditional tobacco leaf baking environment parameters mainly relies on experience. Usually, the temperature and humidity are adjusted manually and the baking curve is simply adjusted, which cannot accurately control the uniformity of the environment distribution in the baking room. This method often leads to uneven tobacco leaf baking quality, a low proportion of high-quality tobacco leaves, and high energy consumption. However, the baking environment parameter optimization method based on computational fluid dynamics in the present invention realizes high-precision control of the baking environment through accurate numerical simulation, machine learning, and multi-objective optimization. Compared with the traditional method, the present invention has the following improvements: First, it accurately simulates the temperature field, humidity field, and flow field distribution in the baking room by using computational fluid dynamics, revealing the formation mechanism of the environmental non-uniform area and the hot spot area; Second, it uses an unsupervised learning algorithm to perform cluster analysis on the environmental parameters, realizing automatic identification of problem areas and improving the pertinence of optimization; Third, it constructs a temperature and humidity gradual change model and a multivariate correlation model, establishing a quantitative relationship between the baking environment parameters and the quality of tobacco leaves; Fourth, through the fusion optimization function of the baking environment parameters, multi-objective collaborative optimization is realized, increasing the comprehensive index of environmental uniformity by 25.8%, the average of the key physical and chemical indexes of tobacco leaves by 22.3%, and reducing the energy consumption by 18.5%. These technological improvements have transformed the tobacco leaf baking process from experience-dependent to precision-controlled, greatly improving the uniformity and overall level of tobacco leaf quality.

[0281] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 7, 8, and 9 below.

[0282] Table 7 Variable explanation table (the first part)

[0283]

[0284]

[0285] Table 8 Variable Explanation Table (Second Part)

[0286]

[0287] Table 9 Variable Explanation Table (Third Part)

[0288]

[0289] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. An optimization method for tobacco leaf baking environment parameters based on computational fluid dynamics, characterized in that, Including: Establish an initial data set of baking environment parameters and a three-dimensional grid model of the tobacco leaf baking house, and import them into the computational fluid dynamics simulation software to construct a basic numerical simulation model of the tobacco leaf baking environment; establish a distribution map of environmental field parameters; Conduct a cluster analysis on the distribution map of environmental field parameters to identify uneven areas and hot spots in the tobacco leaf baking environment; construct a temperature and humidity gradual change model during the baking process, and generate multiple groups of optimized gradual change curves based on the original baking curve; use the similarity evaluation method of the temperature and humidity gradual change curve to determine the key gradual change points and the curves of their leading stages; Establish a multivariate correlation model between tobacco leaf quality and baking environment parameters based on the curves of the leading stage, and determine the optimization objective function of baking environment parameters; apply the integrated optimization function of baking environment parameters to comprehensively optimize the baking environment parameters, calculate the optimal baking timing control scheme; verify the optimal baking timing control scheme in the basic numerical simulation model of the tobacco leaf baking environment, calculate the environmental uniformity index and the value of the optimization objective function of baking environment parameters, screen the optimal combination of baking environment parameters and apply them to the actual baking process, and monitor the physical and chemical indexes of tobacco leaves.

2. The method for optimizing tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 1, wherein The distribution map of environmental field parameters refers to the spatial distribution of physical quantities such as temperature, humidity, and air flow velocity at any point in the tobacco leaf baking house obtained through computational fluid dynamics numerical simulation, presented in the form of a cloud map or contour map, and is used to visually display the uniformity and change trend of the baking environment.

3. The method for optimizing the tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 2, wherein The environmental uniformity index refers to a mathematical index used to quantitatively evaluate the distribution uniformity of environmental parameters in the tobacco leaf baking house. By calculating the standard deviation or coefficient of variation of the temperature field, humidity field, and air flow field, the smaller the value, the more uniform the distribution of environmental parameters and the more stable the baking conditions.

4. The method for optimizing the tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 3, wherein Baking environment parameters refer to the key physical quantities that affect the tobacco leaf baking process, including the temperature, humidity, air flow velocity, air flow direction in the tobacco leaf baking house, and their spatial distribution characteristics, which directly determine the drying rate of tobacco leaves and the conversion effect of chemical components.

5. The method for optimizing tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 4, wherein The temperature and humidity gradual change model refers to a mathematical model that describes the temperature and humidity changing according to specific rules during the baking process, used to achieve a smooth transition from the initial conditions to the target conditions, and ensure that the tobacco leaves obtain the best physical and chemical conversion effects in different baking stages.

6. The method for optimizing tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 5, wherein The original baking curve refers to the functional relationship between temperature and humidity changing with time in the traditional tobacco leaf baking process, and the optimized gradual change curve refers to the new baking process curve obtained by optimizing through the temperature and humidity gradual change model.

7. The method for optimizing the tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 6, wherein The key gradual change point refers to the time node with the largest change amplitude of parameters in adjacent time periods in the temperature and humidity curve during the baking process, marking the entry of baking into a new process stage and a critical moment that needs to be focused on for control.

8. The method for optimizing tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 7, wherein The baking environment parameter fusion optimization function is used to comprehensively consider the impacts of various baking environment parameters on tobacco leaf quality and generate an optimal control strategy. The inputs include the temperature field uniformity coefficient obtained from the environmental field parameter distribution map, the humidity field distribution index obtained from the environmental field parameter distribution map, the airflow velocity vector distribution obtained from the environmental field parameter distribution map, the time series of key transition points obtained from the similarity evaluation method of the temperature and humidity gradual change curve, and the tobacco leaf moisture content obtained from the initial baking environment parameter dataset. The outputs include the air duct adjustment parameters applied to the baking environment parameter optimization objective function, the heating power control curve applied to the baking environment parameter optimization objective function, the humidity control parameters applied to the baking environment parameter optimization objective function, and the optimal baking duration applied to the optimal baking time sequence control scheme.

9. The method for optimizing tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 8, wherein The steps for performing cluster analysis on the environmental field parameter distribution map specifically include: converting the environmental field parameter distribution map into a multi-dimensional feature vector; classifying the feature vector using the K-means clustering or hierarchical clustering algorithm; and determining the spatial positions and parameter characteristics of the uneven regions and hot spots in the tobacco leaf baking environment based on the clustering results.

10. The method for optimizing the tobacco leaf baking environment parameters based on computational fluid dynamics according to claim 9, wherein The specific steps for adopting the similarity evaluation method of the temperature and humidity gradual change curve include: calculating the cosine similarity or Euclidean distance between the temperature and humidity curves at adjacent time points; setting a similarity threshold and identifying the time points with similarity lower than the threshold as key transition points; and extracting the temperature and humidity curves in the previous stage of the key transition points as the leading stage curves.

Images