Airing method of aromatic tobacco for heating cigarettes

Abstract

The invention discloses an air-curing method of aromatic tobacco for heating cigarettes, and belongs to the technical field of heating cigarette raw material processing. According to the method, raw materials are graded in a two-dimensional mode through maturity and tobacco leaf parts, a multi-dimensional base line is established, differential withering, yellowing and drying treatment are carried out, multi-mode monitoring, local and cloud intelligent regulation and control and traceability packaging are combined, and the whole-process refined technology is formed. The aroma conversion efficiency and quality uniformity of the aromatic tobaccos are effectively improved, full-life-cycle traceability of the raw materials is achieved, the requirement for heating the cigarette raw materials is met, and intelligentization and practicability are achieved.

Description

Technical Field

[0001] This invention belongs to the field of heated cigarette raw material processing technology, specifically relating to a method for drying flavored tobacco for heated cigarettes. Background Technology

[0002] Aromatic tobacco is a crucial raw material that imparts the distinctive aroma to cigarettes, and its processing directly determines the aroma quality and industrial usability of the tobacco leaves. Traditional aromatic tobacco processing is extremely complex, requiring the construction of specialized processing sheds where tobacco leaves are strung together and placed on racks one by one. The process involves precise control of temperature, humidity, and light during stages such as wilting, yellowing, and color fixing, taking 15 to 25 days. Subsequent processes include rehydration and stacking. A core premise of this traditional process is the need to maintain the integrity of the tobacco leaves to the greatest extent possible to meet the demands of subsequent manual or machine rolling.

[0003] However, heated cigarettes, as a new type of tobacco product, are made from raw materials in the form of fragments or powder, with no requirement for the integrity of the tobacco leaves. Therefore, the numerous complex processes designed in traditional manufacturing to protect and process intact leaves have become unnecessary costs and technical burdens for heated cigarette raw materials. This constitutes a prominent technical contradiction: the complex traditional processes designed to meet the "shape preservation" requirement are severely incompatible with the fundamental characteristic of heated cigarettes that "do not require shape preservation." Continuing to use traditional processes would lead to low production efficiency, high energy and labor costs, and poor process adaptability.

[0004] If a crude natural sun-drying method is adopted for the sake of simplicity, the tobacco leaves will turn yellow unevenly and become dull in color due to the uncontrollable temperature and humidity. The precursors of aroma substances will not be fully converted or will be lost in large quantities due to volatilization. In the end, the product quality will be poor and will not be able to meet the requirements of high-end heated cigarettes for the aroma of raw materials.

[0005] Furthermore, existing technologies lack real-time, non-destructive, and multi-dimensional monitoring methods for the internal chemical transformation and physical state of tobacco leaves, making it impossible to establish a dynamic and accurate correlation model between process parameters and final quality. This results in the production process relying on experience, large batch-to-batch quality fluctuations, and difficulty in achieving standardized and intelligent production.

[0006] Therefore, there is an urgent need in this field for an innovative drying method that can directly address the above-mentioned technical contradictions and, based on the fundamental characteristic that heated cigarette raw materials "do not need to maintain their shape," can significantly simplify the process and reduce costs, while also ensuring or even improving the characteristic aroma quality of aromatic tobacco through multimodal sensing and intelligent feedback control. Summary of the Invention

[0007] To address the aforementioned problems in the prior art, the present invention provides a method for drying aromatic tobacco for heated cigarettes.

[0008] The objective of this invention can be achieved through the following technical solutions: A method for drying flavored tobacco for heated cigarettes, comprising: Harvest aromatic tobacco at the appropriate maturity stage and grade it according to two dimensions. Remove invalid raw materials, select representative samples to collect multimodal data and record initial parameters, and establish a multidimensional initial baseline. Graded tobacco leaves are laid flat on a breathable rack according to their different thicknesses, placed in the withering area, and environmental parameters are dynamically adjusted according to maturity. Data is collected in real time through a distributed sensing unit. After wilting, the tobacco leaves are transferred to a yellowing room, where differentiated temperature and humidity parameters and treatment cycles are set according to grade, a transition section is added, and the tobacco leaves are confirmed to meet the standards through a multi-dimensional monitoring system. Cut the qualified yellowed tobacco leaves into pieces of a predetermined thickness, with the cutting direction parallel to the main vein of the tobacco leaf. Store each piece separately and label it according to grade. The graded fragments were spread out according to different thicknesses, dried by natural light and adapted to a backup drying mode for cloudy and rainy days, and the drying endpoint was determined by multiple indicators. Collect data from the entire process, use a fusion algorithm to build a four-dimensional prediction model, and form an optimal process parameter database; The model is deployed in a local and cloud architecture, and the initial parameters are automatically called when the hierarchical label is input. At key nodes, the process parameters are dynamically adjusted based on real-time data. After the dried and qualified graded fragments are cooled, they are graded, sealed, and packaged, embedded with traceability chips, and stored in a constant temperature and humidity warehouse.

[0009] In a further embodiment of the present invention, the step of "harvesting aromatic tobacco at the appropriate maturity period and grading it according to two dimensions, eliminating invalid raw materials, selecting representative samples to collect multimodal data and recording initial parameters, and establishing a multidimensional initial baseline" specifically includes: First, harvested aromatic tobacco at the appropriate maturity stage is manually screened and graded according to both maturity and leaf part. Impurities, moldy and diseased leaves are manually removed. From each grade, several representative samples covering typical characteristics are randomly selected. Appearance image data is collected sequentially using an RGB camera and a multispectral camera, and internal characteristic data is collected using a hyperspectral sensor. At the same time, the total weight of the sample is recorded using an IoT weight sensor, and the ambient temperature and humidity at the time of harvest are recorded using an environmental sensor. The collected data are associated with the grading labels and integrated to form a multi-dimensional initial baseline, which is then stored in the data management system.

[0010] In a further embodiment of the present invention, the step of "laying graded tobacco leaves on a breathable support frame according to their differentiated thickness, placing them in the wilting area and dynamically adjusting environmental parameters according to their maturity, and collecting data in real time through a distributed sensing unit" specifically includes: According to the grading results, the tobacco leaves are placed on breathable racks and spread evenly along the length of the racks, with different spreading thicknesses designed according to maturity. The racks are then moved into the withering area equipped with an IoT temperature and humidity control module. Basic environmental parameters are first set, and then the heating rate is gradually adjusted according to the maturity gradient. Several sets of distributed sensing units are evenly arranged in each area of ​​the withering area. Each set synchronously collects the color change, morphological shrinkage rate, and weight reduction ratio of the tobacco leaves at fixed time intervals. The collected data is uploaded to the local edge computing module in real time through the IoT gateway.

[0011] In a further embodiment of the present invention, the step of "transferring the withered tobacco leaves to a yellowing room, setting differentiated temperature and humidity parameters and treatment cycles according to grade, adding a transition section, and confirming that the tobacco leaves meet the standards through a multi-dimensional monitoring system" specifically includes: After withering, the tobacco leaves are transferred to a temperature and humidity-controlled yellowing room according to grade and batch. The preset differentiated temperature and humidity parameters and processing cycle are entered into the control system. During the yellowing process, the basic parameters are maintained first. In the middle of the yellowing process, the color fixation transition section is connected and the corresponding temperature and wet-bulb temperature are switched. Data is continuously collected through a multi-dimensional monitoring system. The spectral sensor analyzes the internal chemical composition in real time, the imaging device captures color changes and quantifies uniformity, and the weight sensor records water loss. The data is compared with the preset standards in real time. Once the tobacco leaves are confirmed to meet the requirements, the yellowing process ends.

[0012] In a further embodiment of the present invention, the step of "cutting the qualified yellowed tobacco leaves into fragments of a predetermined thickness, with the slicing direction parallel to the main vein of the tobacco leaf, and storing and labeling them separately according to grade" specifically includes: Yellowed and qualified tobacco leaves are fed into the tobacco cutting equipment according to their grade. First, the position of the tobacco leaves is fixed by the equipment clamps to ensure that the direction of the main vein of the tobacco leaf is parallel to the movement direction of the slicing blade. The tobacco cutting equipment is started and the leaves are sliced ​​according to the preset thickness. After slicing, large lumps are removed by sieve. The processed fragments are placed into special sealed containers, and labels with clear markings of the location and maturity are affixed to the outer wall of the containers to avoid mixing raw materials of different grades.

[0013] In a further embodiment of the present invention, the step of "spreading graded fragments according to differentiated thicknesses, drying them under natural light and adapting to a backup drying mode for rainy days, and determining the drying endpoint through multiple indicators" specifically includes: Graded fragments are evenly spread on a drying platform with supplemental lighting modules according to their varying thicknesses to ensure no accumulation. Ambient light intensity is monitored in real time by a light sensor. When the intensity is lower than the target value, the control system automatically activates the supplemental lighting module. In case of cloudy or rainy weather or when the light intensity is consistently below the target, the system switches to hot air drying mode and operates at a preset temperature and wind speed. At fixed intervals, moisture content is monitored simultaneously by a weight sensor, and texture roughness is analyzed by an image processing device. The drying process stops when both indicators reach preset thresholds.

[0014] In a further embodiment of the present invention, the step of "collecting full-process data, constructing a four-dimensional prediction model using a fusion algorithm, and forming an optimal process parameter database" specifically includes: The system collects hierarchical labels, multimodal monitoring data, quantitative scoring data, and sensory evaluation results synchronously at each step node. The data is then classified, organized, and cleaned. First, values ​​are assigned to each quantitative indicator, and a comprehensive quantitative score is calculated according to preset weighting coefficients. A combination algorithm of random forest and gradient boosting tree is used to normalize the cleaned data, screen key feature variables, build a basic prediction model, and output preliminary results. The model residuals are corrected through gradient boosting tree, and after multiple rounds of verification, a four-dimensional prediction model is formed. Based on the model, the combination of process parameters is derived and a database is established.

[0015] In a further embodiment of the present invention, the phrase "deploying the model in a local and cloud architecture, automatically calling initial parameters by inputting hierarchical labels, and dynamically adjusting process parameters based on real-time data at key nodes" specifically includes: The trained prediction model is deployed to both the local edge computing module and the cloud platform. The local module is responsible for real-time data processing, while the cloud platform is responsible for model storage and upgrades. During subsequent production, the raw material grading label is input, and the system automatically retrieves the corresponding initial process parameters from the parameter database and sends them to each control device. At critical nodes such as wilting, yellowing, and drying, the local module inputs the real-time collected multimodal data into the model to predict the current quality level. If the preset standard is not met, the system automatically sends adjustment instructions to the control devices to increase temperature, humidity, or extend the processing time, or issues manual intervention prompts to the operators.

[0016] In a further embodiment of the present invention, the step of "cooling the dried and qualified graded fragments, then grading and sealing them, embedding a traceability chip, and storing them in a constant temperature and humidity warehouse" specifically includes: The dried and graded fragments are first allowed to cool naturally to a set temperature at room temperature. They are then divided into portions according to fixed weight and placed into moisture-proof sealed bags with built-in desiccant. Before sealing, an RFID chip is embedded in a designated position inside the bag. Information such as batch number, grading label, full-process parameters, and quality prediction level are recorded through a reading and writing device. After sealing, the packaging bags are sorted and stacked according to the grading label in a temperature- and humidity-controlled warehouse with reserved ventilation channels. The temperature and humidity of the warehouse are continuously regulated within a preset range to achieve full life-cycle traceability of raw materials.

[0017] In a further embodiment of the present invention, the monitoring data specifically includes: distributed sensing units in each step synchronously collect corresponding monitoring data, which is then aggregated through an IoT gateway and synchronously sent to the local edge computing module and the cloud platform using encrypted transmission; the local module processes the data in real time and generates monitoring reports, while the cloud platform stores the data by batch and hierarchical tags and associates it with timestamps; the integrity is verified in real time during data transmission, and remote login to the cloud platform is supported to view real-time data and historical backtracking records, providing complete data support for machine learning models.

[0018] The present invention has at least the following beneficial effects: 1. Dual-dimensional grading and differentiated process parameter adaptation enable raw materials of different maturity and parts to obtain optimal drying conditions, greatly improving the uniformity of aromatic tobacco quality.

[0019] 2. By integrating the prediction model constructed by the fusion algorithm with the process parameter database, the system can automatically call up initial parameters and dynamically adjust key nodes, thereby reducing manual intervention and improving production efficiency.

[0020] 3. RFID traceability chips and full-process data storage enable the tracking of information from raw material harvesting to storage, facilitating quality control and problem tracing. Attached Figure Description

[0021] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart of a method provided in one embodiment of the present invention; Figure 2 This is a flowchart of the harvesting grading and initial baseline establishment process provided in one embodiment of the present invention; Figure 3 This is a flowchart of the wilting process provided in one embodiment of the present invention. Detailed Implementation

[0023] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided.

[0024] In one embodiment of the present invention, a method for drying flavored tobacco for cigarettes is provided, comprising: Harvest aromatic tobacco at the appropriate maturity stage and grade it according to two dimensions. Remove invalid raw materials, select representative samples to collect multimodal data and record initial parameters, and establish a multidimensional initial baseline. Graded tobacco leaves are laid flat on a breathable rack according to their different thicknesses, placed in the withering area, and environmental parameters are dynamically adjusted according to maturity. Data is collected in real time through a distributed sensing unit. After wilting, the tobacco leaves are transferred to a yellowing room, where differentiated temperature and humidity parameters and treatment cycles are set according to grade, a transition section is added, and the tobacco leaves are confirmed to meet the standards through a multi-dimensional monitoring system. Cut the qualified yellowed tobacco leaves into pieces of a predetermined thickness, with the cutting direction parallel to the main vein of the tobacco leaf. Store each piece separately and label it according to grade. The graded fragments were spread out according to different thicknesses, dried by natural light and adapted to a backup drying mode for cloudy and rainy days, and the drying endpoint was determined by multiple indicators. Collect data from the entire process, use a fusion algorithm to build a four-dimensional prediction model, and form an optimal process parameter database; The model is deployed in a local and cloud architecture, and the initial parameters are automatically called when the hierarchical label is input. At key nodes, the process parameters are dynamically adjusted based on real-time data. After the dried and qualified graded fragments are cooled, they are graded, sealed, and packaged, embedded with traceability chips, and stored in a constant temperature and humidity warehouse.

[0025] The core of the raw material harvesting and grading stage lies in establishing initial identification and quality baselines for the raw materials. First, considering the specific requirements of heated tobacco for the optimal maturity period of aromatic tobacco, raw materials at this stage are selected to avoid insufficient accumulation or excessive consumption of aroma substances due to improper harvesting timing. The design principle of the two-dimensional grading is based on the core influencing factors of raw material quality: maturity directly relates to the content and types of aroma substances, while the part of the tobacco leaf determines the physical structure and component distribution of the raw material. Combining these two aspects allows for a refined classification of raw material quality. The process of eliminating ineffective raw materials must balance the intuitiveness and accuracy of manual screening, removing impurities and defective materials that affect subsequent processes, laying the foundation for subsequent drying processes. Multimodal data acquisition starts from two dimensions: appearance characteristics and internal composition. Comprehensive information about the raw materials is obtained through various sensing devices. This data, linked with grading labels, forms a multi-dimensional initial baseline, which becomes the benchmark for subsequent process parameter setting and quality judgment, ensuring the integrity and usability of the initial data.

[0026] The core logic of the wilting stage is to achieve initial moisture loss and morphological stabilization of the raw materials. Differentiated layering thicknesses are designed based on grading results because raw materials at different maturity levels and locations respond differently to environmental parameters. Uniform layering combined with a breathable support frame ensures sufficient contact between the raw materials and the environment, preventing uneven wilting caused by localized environmental imbalances. Environmental parameters in the wilting zone are dynamically adjusted, following the physiological changes of the raw materials. The heating rate is gradually adjusted through gradient adjustments to guide moisture loss and prevent structural damage caused by sudden parameter changes. The distributed sensing units must cover the entire wilting zone to ensure comprehensive and real-time data collection. Data such as color changes, morphological shrinkage rate, and weight reduction percentage directly reflect the wilting state of the raw materials, providing real-time feedback for subsequent process adjustments.

[0027] The yellowing stage is a crucial step in the transformation and formation of aroma substances in aromatic tobacco. Differentiated temperature and humidity parameters and processing cycles are set based on the transformation characteristics of raw materials of different grades, ensuring that each raw material completes the yellowing process under optimal conditions. The addition of a transition section is designed to avoid abrupt parameter changes from the mid-yellowing stage to the early color-fixing stage, minimizing adverse effects on aroma substance transformation and achieving a smooth process transition. A multi-dimensional monitoring system comprehensively controls the yellowing state of raw materials from three core dimensions: internal chemical composition, color uniformity, and water loss. Real-time comparison with preset standards ensures that the degree of yellowing meets requirements, avoiding quality defects caused by excessive or insufficient yellowing.

[0028] The core design principle of the slicing stage is to ensure the consistency of the raw material's morphology and its compatibility with subsequent processing. Setting the slicing direction parallel to the main vein of the tobacco leaf, based on the physical structure of the raw material, reduces breakage and loss of effective components during slicing. Preset thickness control ensures proper drying and filling of heated cigarettes. Separate storage and labeling by grade are essential to maintain the graded properties of the raw materials and prevent a decline in quality uniformity caused by mixing different grades. The use of dedicated sealed containers provides temporary protection for the raw materials, preventing contamination or moisture changes from the external environment.

[0029] The core objective of the drying stage is to control the moisture content of the raw materials within a suitable range, ensuring their stability and storage safety. The design of differentiated spreading thickness, combined with the environmental friendliness and efficiency of natural light drying, and the inclusion of supplemental lighting modules, compensates for the instability of natural light, ensuring the continuity of the drying process. Switching to a backup drying mode for cloudy or rainy days, based on real-time monitoring of ambient light intensity, allows for flexible adjustments to the drying method, preventing prolonged drying cycles or quality degradation due to weather factors. Multiple indicators are used to determine the drying endpoint because a single indicator cannot fully reflect the drying status of the raw materials. Dual verification of moisture content and texture roughness ensures the accuracy of the drying endpoint determination, preventing mold growth during storage due to insufficient drying or loss of aroma substances due to over-drying.

[0030] The data modeling phase is the core support for achieving process intelligence. The collection of data across the entire process encompasses raw material characteristics, process parameters, monitoring results, and final quality evaluation, providing a comprehensive data foundation for model construction. Data processing and cleaning aim to remove invalid data and outliers, ensuring data accuracy and reliability. The calculation of comprehensive quantitative scores, through preset weighting coefficients, transforms multiple quantitative indicators into a unified evaluation standard, providing the model with a clear output target. The selection of the fusion algorithm is based on data characteristics and prediction requirements. The random forest algorithm possesses strong feature selection and basic prediction capabilities, while the gradient boosting tree algorithm can effectively correct model residuals. The four-dimensional prediction model constructed by combining these two algorithms can accurately uncover the correlation between process parameters and raw material quality, forming an optimal process parameter database that provides data support for subsequent production.

[0031] The core of the intelligent control phase is to achieve automated and precise adjustment of process parameters. The deployment of a local and cloud architecture balances the efficiency of real-time data processing with the convenience of model storage and upgrades. The local edge computing module can quickly process real-time data and issue adjustment commands, while the cloud platform is responsible for the long-term maintenance and optimization of the model. Automatic retrieval of initial parameters, based on the association between raw material grading labels and the parameter database, enables rapid matching of process parameters, reducing manual intervention. Dynamic adjustment of key nodes involves inputting real-time monitoring data into the predictive model to anticipate the raw material quality status. When deviations from preset standards occur, parameter adjustments are automatically initiated or manual intervention prompts are given, ensuring that the drying process remains in an optimal state at all times.

[0032] The core design principle of the packaging and storage stage is to ensure the stability and traceability of raw material quality. Cooling treatment prevents high temperatures from affecting the aroma compounds of the raw materials, while sealing and using desiccants effectively isolate them from external moisture and air, preventing moisture absorption or oxidation. The embedded traceability chip stores key information such as batch information, grading labels, process parameters, and quality grades, enabling full lifecycle traceability from harvesting to storage, providing technical support for quality control and accountability. The controlled temperature and humidity warehouse environment maintains the stable state of the raw materials over the long term, preventing quality changes during storage.

[0033] In a further embodiment of the present invention, the step of "harvesting aromatic tobacco at the appropriate maturity period and grading it according to two dimensions, eliminating invalid raw materials, selecting representative samples to collect multimodal data and recording initial parameters, and establishing a multidimensional initial baseline" specifically includes: First, harvested aromatic tobacco at the appropriate maturity stage is manually screened and graded according to both maturity and leaf part. Impurities, moldy and diseased leaves are manually removed. From each grade, several representative samples covering typical characteristics are randomly selected. Appearance image data is collected sequentially using an RGB camera and a multispectral camera, and internal characteristic data is collected using a hyperspectral sensor. At the same time, the total weight of the sample is recorded using an IoT weight sensor, and the ambient temperature and humidity at the time of harvest are recorded using an environmental sensor. The collected data are associated with the grading labels and integrated to form a multi-dimensional initial baseline, which is then stored in the data management system.

[0034] In this embodiment, the implementation of dual-dimensional grading relies on the core influencing mechanism of raw material quality. Maturity is a key factor determining the content and types of aroma substances in aromatic tobacco. Raw materials with different maturity levels exhibit significant differences in moisture loss rates and chemical component transformation patterns during subsequent drying processes. The tobacco leaf portion directly affects the physical structure of the raw material, such as thickness and fiber density, thereby influencing the material exchange efficiency between the raw material and the environment. Therefore, grading based on maturity and tobacco leaf portion as dual dimensions can comprehensively reflect the intrinsic chemical and extrinsic physical properties of the raw material, achieving refined classification. During the grading process, manual screening must combine visual observation and tactile perception, classifying raw materials into different grades according to a unified grading standard to ensure high consistency in characteristics within the same grade. In the invalid raw material removal stage, clear criteria for judging impurities, mold, and diseased leaves must be established. Manual screening must strictly adhere to these standards to avoid errors caused by subjective judgment. Simple physical screening methods can also be used to ensure that invalid raw materials are completely removed, guaranteeing the purity of the raw materials.

[0035] The design principle of multimodal data acquisition is to capture the characteristic information of raw materials from different dimensions using different types of sensing devices. The acquisition of appearance image data employs a combination of RGB and multispectral cameras. The RGB camera captures the raw material's color, shape, and other intuitive appearance features, while the multispectral camera captures spectral information invisible to the human eye, reflecting subtle differences on the surface and appearance features related to the internal structure. The combination of these two methods achieves a comprehensive depiction of the raw material's appearance. The acquisition of internal characteristic data relies on hyperspectral sensors. The principle is to illuminate the raw material with different wavelengths of light and analyze the absorption and reflection characteristics of the spectrum to obtain information related to the chemical composition inside the raw material, such as aroma precursors and moisture content. Weight data is recorded through IoT weight sensors, directly reflecting the initial mass state of the raw material and providing basic data for subsequent calculations of parameters such as moisture loss ratio. The acquisition of environmental temperature and humidity data is necessary because environmental conditions at harvest time can affect the initial state of the raw material. For example, high temperature and high humidity environments can cause rapid deterioration of the raw material. Recording this data provides an environmental background reference for subsequent process adjustments.

[0036] The core of data integration and initial baseline establishment is to achieve the correlation and standardization of various types of data. Collected appearance image data, internal characteristic data, weight data, and environmental temperature and humidity data need to be associated with corresponding grading labels to ensure that each set of data clearly corresponds to a specific grade of raw material. During data integration, different types of data need to be standardized in format, converting image data into quantifiable feature parameters and converting analog signals collected by sensors into digital signals to facilitate subsequent data storage and analysis. The essence of a multi-dimensional initial baseline is a standardized set of characteristics for a certain grade of raw material, covering the average level and fluctuation range of multiple dimensions such as appearance, internal characteristics, weight, and environment. By storing this baseline in the data management system, it can provide a reference for setting subsequent process parameters. For example, for a certain grade of raw material, the temperature and humidity parameters during the withering stage can be adjusted based on the moisture content data in its initial baseline to ensure the targeting and accuracy of the process.

[0037] In a further embodiment of the present invention, the step of "laying graded tobacco leaves on a breathable support frame according to their differentiated thickness, placing them in the wilting area and dynamically adjusting environmental parameters according to their maturity, and collecting data in real time through a distributed sensing unit" specifically includes: According to the grading results, the tobacco leaves are placed on breathable racks and spread evenly along the length of the racks, with different spreading thicknesses designed according to maturity. The racks are then moved into the withering area equipped with an IoT temperature and humidity control module. Basic environmental parameters are first set, and then the heating rate is gradually adjusted according to the maturity gradient. Several sets of distributed sensing units are evenly arranged in each area of ​​the withering area. Each set synchronously collects the color change, morphological shrinkage rate, and weight reduction ratio of the tobacco leaves at fixed time intervals. The collected data is uploaded to the local edge computing module in real time through the IoT gateway.

[0038] This embodiment focuses on the precise implementation of wilting treatment. The core logic is based on raw material grading results, enabling differentiated and dynamic control of the laying thickness and environmental parameters, while simultaneously constructing a comprehensive, real-time monitoring and data transmission system. Differentiated laying thickness design is based on raw material maturity, matching different moisture loss requirements; dynamic adjustment of environmental parameters follows the physiological changes during the wilting process, guiding orderly moisture loss through gradient heating; the deployment of distributed sensing units enables comprehensive monitoring of raw material status and environmental parameters; the cooperation between the IoT gateway and the local edge computing module ensures real-time data transmission and efficient processing, forming a closed loop for wilting treatment characterized by grading adaptation, dynamic control, real-time monitoring, and rapid feedback.

[0039] The design of differentiated layering thickness is based on the physical characteristics and moisture content differences of raw materials at different maturity levels. Raw materials with higher maturity have relatively loose cell structures and higher moisture content. If layered too thickly, poor air circulation within the raw material hinders moisture dissipation, easily leading to localized mold growth. Conversely, raw materials with lower maturity have denser cell structures and lower moisture content. If layered too thinly, rapid moisture loss can damage the raw material structure. Therefore, designing differentiated layering thicknesses for different grades of raw materials based on their maturity levels ensures that each layer of raw material has sufficient contact with the environment, achieving uniform and orderly moisture dissipation. During layering, the material is evenly laid along the length of the breathable support frame to avoid localized accumulation. The breathable support frame design further ensures air circulation at the bottom of the raw material, promoting uniform moisture loss and providing physical assurance for the stability of the wilting process.

[0040] Dynamic control of environmental parameters is the core of the wilting process. Its principle is to simulate the physiological process of natural wilting of raw materials. By adjusting the heating rate in a gradient manner, it guides the gradual loss of moisture from the raw materials, avoiding damage caused by sudden parameter changes. First, basic environmental parameters are set to provide a stable initial wilting environment for the raw materials, allowing them to adapt to environmental changes. Then, the heating rate is adjusted according to the maturity gradient. The heating rate can be appropriately accelerated for more mature raw materials to promote rapid and orderly moisture loss, while the heating rate needs to be slowed down for less mature raw materials to prevent excessive moisture loss. The wilting area is equipped with an IoT-based temperature and humidity control module, which has precise parameter adjustment capabilities. It can automatically execute heating and humidity control operations according to preset control logic, ensuring that environmental parameters always meet the wilting requirements of raw materials at different maturity levels. Compared to the traditional fixed parameter mode, this dynamic control method better aligns with the physiological changes of the raw materials, effectively improving the controllability of the wilting process.

[0041] The deployment of distributed sensing units and the construction of the data acquisition system aim to achieve comprehensive and real-time monitoring of the wilting process. Sensing units must be evenly distributed throughout the wilting area to ensure coverage of all raw material placement locations and avoid monitoring blind spots. Each group of sensing units simultaneously collects data on the color change, morphological shrinkage rate, and weight reduction percentage of the raw materials. These three indicators reflect the wilting state of the raw materials from different dimensions: color change indirectly reflects the initial transformation of the internal chemical components and moisture loss; morphological shrinkage rate directly reflects the physical morphological changes caused by moisture loss; and weight reduction percentage directly reflects the total amount of moisture lost. Sensing units collect data at fixed time intervals to ensure data continuity and timeliness, providing comprehensive status feedback for subsequent process adjustments.

[0042] The collected data is aggregated through an IoT gateway and then uploaded in real time to the local edge computing module using an efficient transmission protocol. The IoT gateway acts as a data aggregation and protocol conversion hub, ensuring unified transmission of data from different types of sensing units. The local edge computing module, on the other hand, possesses real-time data processing capabilities, enabling rapid analysis of collected data such as color changes, morphological shrinkage rates, and weight reduction percentages to determine whether the wilting progress of the raw materials meets preset standards, providing timely decision support for subsequent parameter adjustments. Compared to traditional offline data processing methods, this significantly improves data utilization efficiency and enables dynamic monitoring and rapid response to the wilting process.

[0043] In a further embodiment of the present invention, the step of "transferring the withered tobacco leaves to a yellowing room, setting differentiated temperature and humidity parameters and treatment cycles according to grade, adding a transition section, and confirming that the tobacco leaves meet the standards through a multi-dimensional monitoring system" specifically includes: After withering, the tobacco leaves are transferred to a temperature and humidity-controlled yellowing room according to grade and batch. The preset differentiated temperature and humidity parameters and processing cycle are entered into the control system. During the yellowing process, the basic parameters are maintained first. In the middle of the yellowing process, the color fixation transition section is connected and the corresponding temperature and wet-bulb temperature are switched. Data is continuously collected through a multi-dimensional monitoring system. The spectral sensor analyzes the internal chemical composition in real time, the imaging device captures color changes and quantifies uniformity, and the weight sensor records water loss. The data is compared with the preset standards in real time. Once the tobacco leaves are confirmed to meet the requirements, the yellowing process ends.

[0044] The differentiated process parameters are set based on the grading attributes and post-withering state of the raw materials. Different grades of raw materials differ in maturity, moisture content, and physical structure, resulting in varying temperature and humidity requirements during the yellowing process. After the graded batches are transferred to the yellowing chamber, the differentiated temperature and humidity parameters and processing cycles corresponding to the grading labels are entered into the control system according to a pre-established process parameter database, ensuring that each batch of raw materials receives suitable yellowing conditions. The design principle of the temperature and humidity parameters is to simulate the optimal environment for the natural yellowing of raw materials. Temperature directly affects the chemical reaction rate, determining the efficiency of aroma substance transformation, while wet-bulb temperature affects the moisture evaporation rate of the raw materials, indirectly regulating the transformation process of chemical components. The fixed processing cycle provides a basic time framework for the yellowing process while reserving space for subsequent dynamic adjustments, ensuring the orderly progress of the yellowing process.

[0045] The design of the transition section is to avoid the adverse effects on the raw materials caused by the sudden parameter changes from the mid-yellowing stage to the early color-fixing stage. The mid-yellowing stage is a crucial stage for the conversion of aroma substances in the raw materials. At this time, the chemical components of the raw materials are in an active state. If directly switched to the parameters of the early color-fixing stage, the drastic environmental change will interrupt the continuity of the chemical reaction, resulting in incomplete conversion of aroma substances or the generation of by-products. Therefore, a transition section to the early color-fixing stage is introduced in the mid-yellowing stage, and the corresponding temperature and wet-bulb temperature are gradually switched to enable the smooth transition of the conversion of the chemical components of the raw materials and ensure the continuity and stability of the formation of aroma substances. The parameter design of the transition section needs to follow the principle of gradual change and be adjusted gradually according to the state change of the raw materials to ensure that each parameter change is within the adaptation range of the raw materials and achieve a smooth connection from yellowing to color-fixing.

[0046] The construction of a multi-dimensional monitoring system is the core to achieve precise control of the yellowing state. Its principle is to comprehensively capture the yellowing state of the raw materials from both internal and external levels through different types of monitoring devices. The application of the spectral sensor is based on the spectral characteristics of substances. Different chemical components have characteristic absorption peaks for specific wavelengths of the spectrum. By analyzing the spectral data of the raw materials in real time, the changes in internal chemical components can be accurately obtained, such as the conversion progress of aroma precursor substances and the generation amount of harmful substances, directly reflecting the internal yellowing state of the raw materials. The image device reflects the degree of yellowing from the appearance level by capturing the color change of the raw materials and quantifying the uniformity, avoiding the overall quality fluctuation caused by uneven local yellowing. The water loss recorded by the weight sensor indirectly reflects the physiological metabolism state of the raw materials, which is closely related to the conversion of chemical components. The multi-dimensional monitoring data formed by the combination of the three can comprehensively and objectively reflect the yellowing process of the raw materials.

[0047] The determination mechanism of the yellowing end point is based on the comparison of real-time monitoring data with the preset compliance standards. The preset compliance standards are formulated according to the quality requirements of oriental tobacco for heated tobacco products, covering the thresholds of internal chemical components, the range of appearance color uniformity, and the water loss ratio standard. During the yellowing process, the control system compares the data collected by the multi-dimensional monitoring system with the preset standards in real time. When all indicators reach or meet the preset requirements, it is determined that the yellowing is qualified and the yellowing process is automatically ended; if a certain indicator does not meet the standard, the current process parameters are continued or fine-tuned until all indicators meet the requirements. This objective determination method based on data replaces the traditional manual subjective judgment, significantly improves the accuracy of the yellowing end point determination, avoids the situation of insufficient yellowing or over-yellowing, and ensures that the yellowing quality of each batch of raw materials can reach a unified standard.

[0048] In a further embodiment of the present invention, the step of "cutting the classified tobacco leaves with qualified yellowing into fragments of a preset thickness, with the slicing direction parallel to the main vein of the tobacco leaves, storing them separately by classification and marking" specifically includes: Yellowed and qualified tobacco leaves are fed into the tobacco cutting equipment according to their grade. First, the position of the tobacco leaves is fixed by the equipment clamps to ensure that the direction of the main vein of the tobacco leaf is parallel to the movement direction of the slicing blade. The tobacco cutting equipment is started and the leaves are sliced ​​according to the preset thickness. After slicing, large lumps are removed by sieve. The processed fragments are placed into special sealed containers, and labels with clear markings of the location and maturity are affixed to the outer wall of the containers to avoid mixing raw materials of different grades.

[0049] Yellowed, qualified tobacco leaves are fed into the cutting equipment according to their grade, avoiding the mixing of raw materials of different grades and ensuring the consistency of the quality of the sliced ​​fragments from the source. This design is based on the fact that different grades of raw materials differ in their degree of yellowing, chemical composition, and physical structure. Mixing the slices would lead to fluctuations in fragment quality, while individual slicing ensures that each batch of fragments maintains uniform quality characteristics, providing a basis for implementing differentiated drying processes in subsequent stages. During the slicing process, dedicated cutting equipment channels or batch isolation measures must be used for different grades of raw materials to avoid cross-contamination caused by residual raw materials in the equipment, ensuring the purity of the grading.

[0050] The design of slicing parallel to the midrib of the tobacco leaf is based on the fiber structure characteristics of tobacco leaves. The midrib is the core support of the fiber structure, and slicing parallel to the midrib reduces damage to the fiber structure, avoids excessive fragmentation or irregular shapes, and facilitates uniform moisture loss during subsequent drying. To achieve this directional control, the tobacco slicing equipment needs to be equipped with a special clamp to mechanically fix the tobacco leaf so that the midrib direction is consistent with the movement direction of the slicing blade during the slicing process. The clamp design must be adaptable to tobacco leaves of different sizes to ensure stability and reliability. Precise control of the slice thickness is to ensure the consistency of the physical morphology of the fragments. The preset thickness setting needs to be combined with the requirements of subsequent drying treatment and the filling characteristics of heated cigarettes. The tobacco slicing equipment must have a high-precision thickness adjustment function, controlling the cutting spacing of the blade through mechanical transmission or CNC system to ensure that the thickness of each fragment meets the preset standard. After slicing, excessive lumps are removed by sieving to further ensure the uniformity of the fragment morphology. The sieve aperture design must match the preset slice thickness to ensure screening effect.

[0051] The processed fragments are placed into dedicated sealed containers. These containers must be moisture-proof and contamination-proof to prevent environmental factors such as moisture changes and dust contamination from affecting the fragments during storage. Labels affixed to the outer walls of the containers must clearly indicate the part of the raw material and its maturity information. The label information should be concise, clear, and easy to identify, ensuring quick differentiation of different grades of raw materials in subsequent processes. The labels must be securely attached to the containers to prevent detachment during transport or storage. Digital labeling methods such as QR codes can be used to link detailed data such as the raw material's grade and yellowing treatment parameters, supporting subsequent traceability management. During storage, containers of different grades must be stored separately with clear partition markings to prevent mixing during handling and ensure orderly graded storage.

[0052] In a further embodiment of the present invention, the step of "spreading graded fragments according to differentiated thicknesses, drying them under natural light and adapting to a backup drying mode for rainy days, and determining the drying endpoint through multiple indicators" specifically includes: Graded fragments are evenly spread on a drying platform with supplemental lighting modules according to their varying thicknesses to ensure no accumulation. Ambient light intensity is monitored in real time by a light sensor. When the intensity is lower than the target value, the control system automatically activates the supplemental lighting module. In case of cloudy or rainy weather or when the light intensity is consistently below the target, the system switches to hot air drying mode and operates at a preset temperature and wind speed. At fixed intervals, moisture content is monitored simultaneously by a weight sensor, and texture roughness is analyzed by an image processing device. The drying process stops when both indicators reach preset thresholds.

[0053] The core of designing differentiated paving thickness is to match the differences in physical structure and moisture content of fragments of different grades. Due to variations in pre-weaning and yellowing treatment parameters, fragments of different grades exhibit differences in fiber density and initial moisture content. Fragments with high fiber density lose moisture more slowly; if paved too thickly, internal moisture cannot penetrate to the surface, easily resulting in a dry exterior and wet interior. Conversely, fragments with low fiber density lose moisture too quickly; if paved too thinly, excessive moisture loss can occur, damaging the texture. Therefore, designing differentiated paving thickness based on the grading label ensures that each layer of fragments is in full contact with the drying environment, achieving uniform moisture dissipation. The drying platform with supplemental lighting not only provides a flat paving surface but also lays the hardware foundation for subsequent light control. The platform's structural design must ensure air circulation to prevent delayed drying at the bottom of the fragments due to lack of air permeability.

[0054] The implementation of the light monitoring and supplemental lighting mechanism relies on the coordinated operation of sensors, control systems, and execution modules. The light sensor collects ambient light intensity in real time. Its monitoring principle involves converting light signals into electrical signals using photosensitive elements, reflecting the effective light energy of the current environment. Light intensity directly determines the evaporation rate of moisture on the surface of the fragments. When the intensity is lower than the target value, the moisture loss rate of the fragments decreases, which, if left unchecked, prolongs the drying cycle. At this point, the control system automatically activates the supplemental lighting module based on a preset threshold. The spectral design of the supplemental lighting module needs to simulate the wavelength range of natural light that is conducive to moisture evaporation without damaging aroma substances, avoiding localized overheating caused by direct strong light. The core principle of this mechanism is to maintain a constant effective evaporation energy, compensating for fluctuations in natural light through artificial supplemental lighting, ensuring a stable drying rate, and avoiding uneven moisture distribution due to insufficient light.

[0055] The switching logic for the standby mode on rainy days is based on the principle of environmental adaptability. When the light sensor detects that the light intensity is consistently below the threshold, or the humidity sensor indicates that the ambient humidity is too high, the control system automatically switches to the hot air drying mode. The core principle of hot air drying is to accelerate the evaporation of moisture from the surface of the fragments through a temperature-controlled airflow, while controlling the wind speed to prevent the fragments from being blown away or their physical structure from being damaged. The preset temperature and wind speed design must match the graded properties of the fragments. For fragments with high moisture content, the temperature can be appropriately increased to accelerate evaporation; for fragments with fragile fibers, the wind speed is reduced to protect the texture structure. The mode switching process must be seamless to avoid moisture absorption by the fragments due to machine downtime. The control system preheats the hot air equipment to ensure that the drying environment parameters stabilize quickly during the switch, ensuring the continuity of the drying process.

[0056] The principle behind using dual indicators to determine the drying endpoint is the synergistic verification of chemical and physical properties. Moisture content is a core chemical indicator for measuring drying quality, directly affecting the storage stability of the fragments. Excessive moisture content easily leads to mold growth, while insufficient moisture content causes the volatilization of aroma compounds. However, relying solely on moisture content cannot reflect changes in the physical morphology of the fragments. Some fragments may have acceptable moisture content but excessively high or low texture roughness due to uneven moisture loss during drying. The former affects subsequent filling properties, while the latter easily leads to fragment clumping. Therefore, by simultaneously collecting data using a weight sensor and an imaging device, the drying endpoint can be determined only when both reach preset thresholds. The advantage of this method is that it ensures both moisture content meets storage requirements and that the physical morphology of the fragments is suitable for subsequent processing, avoiding quality misjudgments caused by a single indicator.

[0057] In a further embodiment of the present invention, the step of "collecting full-process data, constructing a four-dimensional prediction model using a fusion algorithm, and forming an optimal process parameter database" specifically includes: The system collects hierarchical labels, multimodal monitoring data, quantitative scoring data, and sensory evaluation results synchronously at each step node. The data is then classified, organized, and cleaned. First, values ​​are assigned to each quantitative indicator, and a comprehensive quantitative score is calculated according to preset weighting coefficients. A combination algorithm of random forest and gradient boosting tree is used to normalize the cleaned data, screen key feature variables, build a basic prediction model, and output preliminary results. The model residuals are corrected through gradient boosting tree, and after multiple rounds of verification, a four-dimensional prediction model is formed. Based on the model, the combination of process parameters is derived and a database is established.

[0058] The data collection scope includes four key categories: First, grading and labeling data, namely the maturity and part grading information of the raw materials, which serves as the input basis for the model and is linked to the differentiated design of subsequent process parameters; second, multimodal monitoring data, covering the appearance / internal characteristics during the harvesting stage, the color / shrinkage rate during the wilting stage, the chemical composition / uniformity during the yellowing stage, and the moisture content / texture roughness during the drying stage. This data directly reflects the process status at each stage; third, quantitative scoring data, which forms a comprehensive score by assigning values ​​and weighting the monitoring indicators at each stage, quantifying the quality of the process; and fourth, sensory evaluation results, where professional evaluators assess the aroma, harmony, and off-flavors of the dried fragments, serving as the quality target for the model output. Sensory evaluation results are the final embodiment of the quality of heated cigarette raw materials and need to be correlated with objective monitoring data to achieve a mapping between objective data and subjective quality. Data collection must be recorded synchronously at each step to ensure that each set of process parameters corresponds to a unique quality result, providing input and output matching data pairs for subsequent modeling.

[0059] The principle of data cleaning is to remove outliers and fill in missing values ​​to avoid noise data affecting model accuracy. Classification and organization establish data relationships based on hierarchical labels, process steps, and quality results to ensure data logic. The principle of quantitative scoring is as follows: each monitoring indicator is assigned a value according to preset standards, and then weighting coefficients are set based on the degree of influence of each indicator on the final quality. A comprehensive quantitative score is calculated through weighted summation. This process transforms scattered objective indicators into a unified evaluation standard, providing the model with quantifiable intermediate outputs and avoiding excessive model complexity due to too many indicator dimensions.

[0060] First, the cleaned data is normalized to convert indicators of different magnitudes into the same interval, avoiding model bias caused by differences in magnitude. The feature selection process uses the Random Forest algorithm, which works by constructing multiple decision trees, calculating the importance of each feature to the model output, and selecting key feature variables. This step removes redundant features, reduces model complexity, and improves computational efficiency. The basic prediction model is built using the Random Forest algorithm, which has the advantage of strong fitting ability for nonlinear data and effective handling of high-dimensional data. Preliminary prediction results are output through a voting mechanism of multiple decision trees. Subsequently, the Gradient Boosting Tree algorithm is used to correct model residuals. Its principle is to iteratively train, focusing on the error of the previous round of model training each time, gradually optimizing prediction accuracy. Compared to a single algorithm, the fusion algorithm utilizes both the feature selection and anti-overfitting capabilities of Random Forest and the error correction capabilities of Gradient Boosting Tree, significantly improving the model's prediction accuracy. After model construction, multiple rounds of validation are required. Algorithm parameters are adjusted to ensure the model's stability on different batches of data, ultimately forming a four-dimensional prediction model.

[0061] Based on the trained four-dimensional prediction model, different combinations of grade labels and process parameters are input, and the model outputs corresponding quality prediction results. The optimal combination of process parameters is selected based on the quality prediction results, and its feasibility is verified through small-batch production practice. If the practice results are consistent with the model predictions, the parameter combination is stored in the database and associated with the corresponding grade label. If there are deviations, the data is fed back to the model for parameter correction, ensuring the usability of the parameters in the database. The database structure should be designed to store data according to grade labels for easy retrieval during subsequent production, while also supporting data updates to achieve continuous optimization of process parameters.

[0062] In a further embodiment of the present invention, the phrase "deploying the model in a local and cloud architecture, automatically calling initial parameters by inputting hierarchical labels, and dynamically adjusting process parameters based on real-time data at key nodes" specifically includes: The trained prediction model is deployed to both the local edge computing module and the cloud platform. The local module is responsible for real-time data processing, while the cloud platform is responsible for model storage and upgrades. During subsequent production, the raw material grading label is input, and the system automatically retrieves the corresponding initial process parameters from the parameter database and sends them to each control device. At critical nodes such as wilting, yellowing, and drying, the local module inputs the real-time collected multimodal data into the model to predict the current quality level. If the preset standard is not met, the system automatically sends adjustment instructions to the control devices to increase temperature, humidity, or extend the processing time, or issues manual intervention prompts to the operators.

[0063] The hardware selection for the local edge computing module must meet real-time data processing requirements. Its principle is to directly receive real-time data collected by distributed sensing units through locally deployed computing units, eliminating the need for cloud relay and significantly reducing data transmission latency. This is crucial for critical stages such as wilting and yellowing, where raw material conditions change rapidly; delayed parameter adjustments can easily lead to quality deviations. The core functions of the local module include: real-time data preprocessing, inputting data into the model for quality prediction, generating adjustment instructions and sending them to control equipment, and generating real-time monitoring reports. The cloud platform focuses on long-term management and expansion: storing end-to-end data, enabling unified model upgrades, supporting model sharing across multiple plants / production lines, and providing remote data viewing and management interfaces. The collaborative mechanism between the two is as follows: the local module synchronizes processed key data to the cloud, where the data is categorized, stored, and timestamped; when the model needs upgrading, the cloud sends optimized model parameters to each local module for unified updates. This architecture solves the problems of limited storage and difficult upgrades in local deployments while avoiding the transmission latency and poor real-time performance of cloud deployments, achieving a balance between real-time control and long-term optimization.

[0064] In subsequent production, operators only need to input the grading label of the current raw material. The system automatically matches the corresponding initial parameters in the optimal process parameter database using a preset association algorithm. The core of the association algorithm is to establish a mapping relationship between grading labels and parameter features. Each set of optimal parameters in the database has been pre-associated with a corresponding grading label. The system quickly locates the target parameters through keyword matching or similarity calculation. After the parameters are called, the system automatically sends them to the control equipment at each stage, eliminating the need for manual setting. The advantages of this process are: first, it avoids the time cost and errors of manual queries; second, it ensures accurate matching between initial parameters and raw material grading, laying the foundation for stable subsequent processes.

[0065] Taking the yellowing process as an example, when it enters the middle stage of yellowing, the local edge computing module continuously collects multimodal data: spectral sensors analyze changes in the internal chemical composition of the fragments, image equipment quantifies color uniformity, and weight sensors record the water loss ratio. This real-time data is input into the prediction model. Based on preset quality standards, the model predicts the current quality level of the fragments. If the prediction result meets the preset standard, the current process parameters are maintained; if it does not meet the standard, the model analyzes the cause of the deviation and generates adjustment instructions: for parameters that can be automatically adjusted, the system directly sends heating or humidification instructions to the temperature and humidity control module of the yellowing chamber; for situations requiring manual intervention, the system issues intervention prompts to the operator through audio-visual prompts or remote messages and provides suggested solutions. The quality result predicted by the model is used as a feedback signal and compared with the preset standard. The input is adjusted according to the direction of deviation until the deviation is eliminated. This mechanism ensures that the process parameters can adapt to changes in the state of the raw materials in real time, avoiding quality problems caused by fluctuations in the initial state of the raw materials.

[0066] In a further embodiment of the present invention, the step of "cooling the dried and qualified graded fragments, then grading and sealing them, embedding a traceability chip, and storing them in a constant temperature and humidity warehouse" specifically includes: The dried and graded fragments are first allowed to cool naturally to a set temperature at room temperature. They are then divided into portions according to fixed weight and placed into moisture-proof sealed bags with built-in desiccant. Before sealing, an RFID chip is embedded in a designated position inside the bag. Information such as batch number, grading label, full-process parameters, and quality prediction level are recorded through a reading and writing device. After sealing, the packaging bags are sorted and stacked according to the grading label in a temperature- and humidity-controlled warehouse with reserved ventilation channels. The temperature and humidity of the warehouse are continuously regulated within a preset range to achieve full life-cycle traceability of raw materials.

[0067] The dried fragments are still quite hot. If directly packaged, the high temperature of the fragments will cause the air inside the sealed bag to expand due to heat. Upon cooling, the air will contract, creating negative pressure, allowing external moisture to easily penetrate. Simultaneously, the moisture inside the fragments will condense into water droplets due to the sudden temperature drop, leading to excessive local moisture and mold growth. Therefore, natural cooling must be carried out at room temperature. The principle is to allow the fragments to slowly cool to the ambient temperature through heat exchange with the environment. The fragment temperature must be monitored during the cooling process to ensure that the preset temperature is reached before packaging. The cooling environment must be well-ventilated to prevent the fragments from absorbing moisture due to high local humidity. Direct sunlight should also be avoided to prevent ultraviolet rays from damaging the aroma substances. The advantages of standardized cooling are: it avoids excessive moisture loss caused by forced cooling and prevents post-packaging problems caused by incomplete cooling, ensuring that the moisture and temperature of the fragments are stable before packaging.

[0068] Moisture-proof sealed bags with built-in desiccants require materials with strong barrier properties. The principle is that the dense structure of the material prevents external moisture from penetrating, while the desiccant inside absorbs any small amount of moisture introduced during packaging, maintaining a low-humidity environment inside the bag. The amount of desiccant used needs to be designed based on the weight of the fragments and the expected storage time to ensure continued moisture absorption throughout the storage period. The embedding position of the RFID chip must avoid affecting the sealing performance; it is usually chosen in the sealing area at the edge of the bag. The chip selection must meet the requirements of small size and high stability. Its working principle is to interact with the reader / writer via radio frequency signals. The information that can be stored includes: raw material batch number, grading label, full-process parameters, quality prediction level, packaging time, and operator. Before packaging, the above information is entered into the chip using a reader / writer. After packaging, the information can be quickly read via a handheld terminal or warehouse access control system. If quality problems are found during subsequent use, the chip information can be used to trace back to the specific stage and locate the root cause of the problem. In addition, a visual label consistent with the chip information should be affixed to the outer wall of the sealed bag for easy manual identification of grading information, avoiding classification errors after packaging.

[0069] The warehouse's temperature and humidity control system must possess high-precision control capabilities. Its principle involves real-time monitoring of environmental parameters using temperature and humidity sensors. When the temperature exceeds a preset range, the cooling module is activated; when the humidity exceeds a preset range, the dehumidification module is activated; when the temperature or humidity falls below a threshold, the heating or humidification module is activated respectively, ensuring the warehouse environment remains within a suitable range for fragment storage. Suitable temperature and humidity slow down the volatilization rate of aroma substances and inhibit microbial growth. The design principle of reserved ventilation channels is that even if the temperature and humidity control system experiences a short-term malfunction, the ventilation channels allow for natural air circulation within the warehouse, preventing localized high-humidity and high-temperature areas and reducing the risk of raw material spoilage. The warehouse's stacking rules must be based on graded labels, with sealed bags of different grades stored separately, avoiding excessive stacking that could cause damage to the bottom bags, while also facilitating inventory management and raw material retrieval. During storage, raw material information is periodically read via RFID chips to verify inventory and quality status, ensuring the stability of raw materials throughout the storage period.

[0070] In a further embodiment of the present invention, the monitoring data specifically includes: distributed sensing units in each step synchronously collect corresponding monitoring data, which is then aggregated through an IoT gateway and synchronously sent to the local edge computing module and the cloud platform using encrypted transmission; the local module processes the data in real time and generates monitoring reports, while the cloud platform stores the data by batch and hierarchical tags and associates it with timestamps; the integrity is verified in real time during data transmission, and remote login to the cloud platform is supported to view real-time data and historical backtracking records, providing complete data support for machine learning models.

[0071] Distributed sensing units at each stage must collect data according to standardized protocols to ensure a unified data format across different sensor types. This design avoids data aggregation issues caused by incompatible sensor protocols. The IoT gateway, as the core device for data aggregation, works by receiving real-time data from each sensing unit, performing preliminary data processing, and then synchronously sending the aggregated data to the local edge computing module and the cloud platform. The gateway should be deployed close to the area where the sensing units are concentrated to reduce data transmission distance and signal attenuation; it should also have multi-interface expansion capabilities to support the connection of different types of sensing devices. The advantages of data aggregation are: it forms a complete data chain from harvesting and grading to packaging and storage; each set of data can be linked to a specific raw material batch and process stage, avoiding the problem of traditional data dispersion and laying the foundation for subsequent real-time control and historical backtracking.

[0072] When data is transmitted from the IoT gateway to the local module and the cloud platform, an encryption protocol is required. The mechanism works as follows: the data is encrypted at the sending end, and the receiving end decrypts the ciphertext using a key to restore the plaintext. Simultaneously, the protocol verifies the identities of both the sending and receiving ends using digital certificates to prevent fraudulent devices from accessing and stealing data. Furthermore, data signing is required during data transmission. The sending end signs the data using its private key, and the receiving end verifies the signature using its public key, ensuring that the data has not been tampered with during transmission. This is crucial for critical data such as process parameters and quality results; if the data is tampered with, it may lead to incorrect control instructions, affecting raw material quality.

[0073] The local edge computing module processes the received aggregated data in real time: first, data preprocessing, such as removing outliers caused by sensor malfunctions and appropriately filling in missing data; second, data transformation, converting raw data into indicators that can be directly used for control; and third, generating real-time monitoring reports to intuitively display the process status of each stage, facilitating operators' real-time monitoring of production. The cloud platform's storage design needs to be categorized according to three dimensions: batch, hierarchical tags, and timestamps. Each batch of raw materials corresponds to a unique batch number, and data within the same batch is stored in partitions according to hierarchical tags. Each data entry is associated with a collection timestamp. This classification method establishes a spatiotemporal and attribute-based relationship. When historical data needs to be traced, the target data can be quickly located through the batch number, hierarchical tag, or time range, significantly improving query efficiency. Cloud storage needs to adopt a distributed storage architecture to ensure that data is not lost due to single points of failure, while also supporting data compression to reduce storage capacity usage.

[0074] Integrity verification during data transmission employs a checksum or hash value mechanism: the sender calculates a checksum for each set of data and sends it along with the data; the receiver recalculates the checksum after receiving the data; if the checksum matches the sender's checksum, the data is considered complete; otherwise, a retransmission is requested. This mechanism ensures that data is not lost or damaged during transmission. The historical data backtracking function relies on the full data storage of the cloud platform. Users can remotely log in to the cloud platform, enter query conditions, and the system automatically retrieves and displays the corresponding historical data, while also supporting data export and visualization analysis.

[0075] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for drying aromatic tobacco used in heated cigarettes, characterized in that, include: Harvest aromatic tobacco at the appropriate maturity stage and grade it according to two dimensions. Remove invalid raw materials, select representative samples to collect multimodal data and record initial parameters, and establish a multidimensional initial baseline. Graded tobacco leaves are laid flat on a breathable rack according to their different thicknesses, placed in the withering area, and environmental parameters are dynamically adjusted according to maturity. Data is collected in real time through a distributed sensing unit. After wilting, the tobacco leaves are transferred to a yellowing room, where differentiated temperature and humidity parameters and treatment cycles are set according to grade, a transition section is added, and the tobacco leaves are confirmed to meet the standards through a multi-dimensional monitoring system. Cut the qualified yellowed tobacco leaves into pieces of a predetermined thickness, with the cutting direction parallel to the main vein of the tobacco leaf. Store each piece separately and label it according to grade. The graded fragments were spread out according to different thicknesses, dried by natural light and adapted to a backup drying mode for cloudy and rainy days, and the drying endpoint was determined by multiple indicators. Collect data from the entire process, use a fusion algorithm to build a four-dimensional prediction model, and form an optimal process parameter database; The model is deployed in a local and cloud architecture, and the initial parameters are automatically called when the hierarchical label is input. At key nodes, the process parameters are dynamically adjusted based on real-time data. After the dried and qualified graded fragments are cooled, they are graded, sealed, and packaged, embedded with traceability chips, and stored in a constant temperature and humidity warehouse.

2. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "harvesting aromatic tobacco at the appropriate maturity stage and grading it according to two dimensions, eliminating invalid raw materials, selecting representative samples to collect multimodal data and recording initial parameters, and establishing a multidimensional initial baseline" specifically includes: First, harvested aromatic tobacco at the appropriate maturity stage is manually screened and graded according to both maturity and leaf part. Impurities, moldy and diseased leaves are manually removed. From each grade, several representative samples covering typical characteristics are randomly selected. Appearance image data is collected sequentially using an RGB camera and a multispectral camera, and internal characteristic data is collected using a hyperspectral sensor. At the same time, the total weight of the sample is recorded using an IoT weight sensor, and the ambient temperature and humidity at the time of harvest are recorded using an environmental sensor. The collected data are associated with the grading labels and integrated to form a multi-dimensional initial baseline, which is then stored in the data management system.

3. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "laying graded tobacco leaves on a breathable rack according to their varying thicknesses, placing them in the wilting area, dynamically adjusting environmental parameters according to maturity, and collecting data in real time through a distributed sensing unit" specifically includes: According to the grading results, the tobacco leaves are placed on breathable racks and spread evenly along the length of the racks, with different spreading thicknesses designed according to maturity. The racks are then moved into the withering area equipped with an IoT temperature and humidity control module. Basic environmental parameters are first set, and then the heating rate is gradually adjusted according to the maturity gradient. Several sets of distributed sensing units are evenly arranged in each area of ​​the withering area. Each set synchronously collects the color change, morphological shrinkage rate, and weight reduction ratio of the tobacco leaves at fixed time intervals. The collected data is uploaded to the local edge computing module in real time through the IoT gateway.

4. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "transferring the withered tobacco leaves to a yellowing room, setting differentiated temperature and humidity parameters and treatment cycles according to grade, adding a transition section, and confirming that the tobacco leaves meet the standards through a multi-dimensional monitoring system" specifically includes: After withering, the tobacco leaves are transferred to a temperature and humidity-controlled yellowing room according to grade and batch. The preset differentiated temperature and humidity parameters and processing cycle are entered into the control system. During the yellowing process, the basic parameters are maintained first. In the middle of the yellowing process, the color fixation transition section is connected and the corresponding temperature and wet-bulb temperature are switched. Data is continuously collected through a multi-dimensional monitoring system. The spectral sensor analyzes the internal chemical composition in real time, the imaging device captures color changes and quantifies uniformity, and the weight sensor records water loss. The data is compared with the preset standards in real time. Once the tobacco leaves are confirmed to meet the requirements, the yellowing process ends.

5. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "cutting the qualified yellowed tobacco leaves into fragments of a predetermined thickness, with the slicing direction parallel to the main vein of the tobacco leaf, and storing and labeling them separately according to grade" specifically includes: Yellowed and qualified tobacco leaves are fed into the tobacco cutting equipment according to their grade. First, the position of the tobacco leaves is fixed by the equipment clamps to ensure that the direction of the main vein of the tobacco leaf is parallel to the movement direction of the slicing blade. The tobacco cutting equipment is started and the leaves are sliced ​​according to the preset thickness. After slicing, large lumps are removed by sieve. The processed fragments are placed into special sealed containers, and labels with clear markings of the location and maturity are affixed to the outer wall of the containers to avoid mixing raw materials of different grades.

6. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "spreading graded fragments according to differentiated thicknesses, drying under natural light with a backup drying mode for rainy days, and determining the drying endpoint through multiple indicators" specifically includes: Graded fragments are evenly spread on a drying platform with supplemental lighting modules according to their varying thicknesses to ensure no accumulation. Ambient light intensity is monitored in real time by a light sensor. When the intensity is lower than the target value, the control system automatically activates the supplemental lighting module. In case of cloudy or rainy weather or when the light intensity is consistently below the target, the system switches to hot air drying mode and operates at a preset temperature and wind speed. At fixed intervals, moisture content is monitored simultaneously by a weight sensor, and texture roughness is analyzed by an image processing device. The drying process stops when both indicators reach preset thresholds.

7. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "collecting full-process data, using a fusion algorithm to construct a four-dimensional prediction model, and forming an optimal process parameter database" specifically includes: The system collects hierarchical labels, multimodal monitoring data, quantitative scoring data, and sensory evaluation results synchronously at each step node. The data is then classified, organized, and cleaned. First, values ​​are assigned to each quantitative indicator, and a comprehensive quantitative score is calculated according to preset weighting coefficients. A combination algorithm of random forest and gradient boosting tree is used to normalize the cleaned data, screen key feature variables, build a basic prediction model, and output preliminary results. The model residuals are corrected through gradient boosting tree, and after multiple rounds of verification, a four-dimensional prediction model is formed. Based on the model, the combination of process parameters is derived and a database is established.

8. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "deploying the model in a local and cloud architecture, automatically recalling initial parameters upon inputting hierarchical labels, and dynamically adjusting process parameters based on real-time data at key nodes" specifically includes: The trained prediction model is deployed to both the local edge computing module and the cloud platform. The local module is responsible for real-time data processing, while the cloud platform is responsible for model storage and upgrades. During subsequent production, the raw material grading label is input, and the system automatically retrieves the corresponding initial process parameters from the parameter database and sends them to each control device. At critical nodes such as wilting, yellowing, and drying, the local module inputs the real-time collected multimodal data into the model to predict the current quality level. If the preset standard is not met, the system automatically sends adjustment instructions to the control devices to increase temperature, humidity, or extend the processing time, or issues manual intervention prompts to the operators.

9. The method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The phrase "cooling and then classifying and sealing the dried and qualified graded fragments, embedding a traceability chip, and storing them in a temperature- and humidity-controlled warehouse" specifically includes: The dried and graded fragments are first allowed to cool naturally to a set temperature at room temperature. They are then divided into portions according to fixed weight and placed into moisture-proof sealed bags with built-in desiccant. Before sealing, an RFID chip is embedded in a designated position inside the bag. Information such as batch number, grading label, full-process parameters, and quality prediction level are recorded through a reading and writing device. After sealing, the packaging bags are sorted and stacked according to the grading label in a temperature- and humidity-controlled warehouse with reserved ventilation channels. The temperature and humidity of the warehouse are continuously regulated within a preset range to achieve full life-cycle traceability of raw materials.

10. A method for drying aromatic tobacco for heated cigarettes according to claim 1, characterized in that, The monitoring data specifically includes: distributed sensing units in each step synchronously collect corresponding monitoring data, which is then aggregated through an IoT gateway and synchronously sent to the local edge computing module and the cloud platform using encrypted transmission; the local module processes the data in real time and generates monitoring reports, while the cloud platform stores the data by batch, hierarchical tags, and associates it with timestamps; the integrity is verified in real time during data transmission, and remote login to the cloud platform is supported to view real-time data and historical backtracking records, providing complete data support for machine learning models.

Images