A method for premixing reconstituted tobacco stacking raw materials

Abstract

This invention relates to a method for premixing raw materials for reconstituted tobacco stacks, comprising: constructing a near-infrared spectral model to predict the chemical composition content of reconstituted tobacco raw materials; dividing the stacked raw materials into at least 30 feeding units; collecting the near-infrared spectra of raw material samples from each feeding unit; using the constructed near-infrared spectral model to predict the chemical composition content of the raw material samples from each feeding unit; and, based on the aforementioned prediction results, using a brute-force exhaustive algorithm to divide all feeding units into at least two batches, such that the average chemical composition content of the premixed formulation in each batch is equal to or deviates from the average chemical composition content of the entire stack by no more than 1%. This method is efficient, low-cost, pollution-free, lossless, and simultaneously measures and calculates multiple chemical composition indicators of raw materials, and automatically premixes stacked raw materials, improving the uniformity of raw materials within and between batches, and achieving product quality stability.

Description

Technical Field

[0001] This invention belongs to the field of spectral detection, and specifically relates to a method for premixing reconstituted tobacco stacking raw materials. Background Technology

[0002] When reconstituted tobacco is produced using the papermaking process, the intra-batch and inter-batch stability of the finished product quality depends on the stability of the raw materials used. The raw materials for reconstituted tobacco include tobacco stems, tobacco fragments, and screened tobacco dust. Figure 1 The images show tobacco stem raw materials of different lengths and stem diameters. Figure 2 The image shows raw materials of tobacco leaves and tobacco leaf fragments in different states. Due to the wide range of sources, specifications, years, quality differences, and unstable purchase quantities of reconstituted tobacco raw materials, the chemical composition content of the raw materials fluctuates greatly, resulting in unstable quality, which directly affects the quality stability of the finished product.

[0003] In reconstituted tobacco raw material warehouses, raw materials of single specifications are piled together. Upon entry into the warehouse, only random samples are taken to determine the chemical composition content, making it impossible to accurately know the chemical composition content of raw materials at a specific location. However, current premixing methods use cross-sections of the raw materials in the warehouse for discharge, resulting in random combinations of raw materials. This makes it impossible to know the chemical composition content of the raw materials used, nor can it determine the quality stability of the premixed product.

[0004] Therefore, in order to ensure the stability of production process quality, it is urgent to establish a scientific and reasonable method to improve the premixing uniformity of stacked raw materials. Summary of the Invention

[0005] A first aspect of the present invention provides a method for premixing reconstituted tobacco stacking raw materials, comprising:

[0006] S1: Construct a near-infrared spectral model to predict the chemical composition content of reconstituted tobacco raw materials;

[0007] S2: Divide the stacked raw materials into at least 30 feeding units;

[0008] S3: Collect the near-infrared spectrum of the raw material sample of each feeding unit divided in S2, and use the near-infrared spectral model constructed in S1 to predict the chemical composition content of the raw material sample of each feeding unit. The chemical composition is selected from total nitrogen, total alkaloids, water-soluble sugar, potassium, chlorine and moisture, preferably total alkaloids.

[0009] S4: Based on the prediction results of S3, use a brute-force exhaustive algorithm to divide all feeding units into at least 2 batches, so that the average chemical composition content of each batch after premixing is equal to or deviates from the average chemical composition content of the entire stack by no more than 1%.

[0010] In some implementations, a brute-force algorithm is used to divide all feeding units into 2, 3, 4, 5, 6, 7, 8, 9, or 10 batches.

[0011] In some embodiments, S1 of the premixing method includes the following operations:

[0012] Collect the near-infrared spectrum of the modeling sample;

[0013] Detect the chemical composition content of the modeling sample;

[0014] The near-infrared spectra of the modeled samples were fitted with the chemical composition content to establish a near-infrared spectral model for predicting stacked raw materials.

[0015] In some implementations, a continuous flow method is used to detect the chemical composition content of the modeling sample.

[0016] In some implementations, partial least squares method is used to fit the near-infrared spectrum of the modeled sample to the chemical composition content.

[0017] In some embodiments, the near-infrared spectral region ranges from 4000 to 7500 cm⁻¹. -1 Near-infrared spectra of modeled samples or raw material samples to be tested are acquired using integrating sphere spectral diffuse reflectance or fiber optic diffuse reflectance acquisition modes.

[0018] In some implementations, the near-infrared spectral region ranges from 4256 to 7030 cm⁻¹. -1 Near-infrared spectra of modeled samples or raw material samples to be tested are acquired using integrating sphere spectral diffuse reflectance or fiber optic diffuse reflectance acquisition modes.

[0019] In some implementations, when acquiring near-infrared spectra, the parameters of the infrared spectrometer are set as follows: near-infrared spectral resolution of 8.0 cm⁻¹. -1 The sample was scanned 68 times, and the background scan frequency was once every 30 minutes.

[0020] In some implementations, after acquiring the near-infrared spectrum of the modeling sample, the spectrum is further preprocessed, including at least one of vector normalization, standard canonical transformation, first derivative, second derivative, multivariate signal correction, and spectral smoothing.

[0021] In some implementations, the spectral smoothing process includes Savitzky-Golay smoothing filtering and Norris derivative filtering.

[0022] In some implementations, the preprocessing includes multivariate signal correction, Norris derivative filtering, and first-order derivative processing.

[0023] In some implementations, the spectrum is smoothed using Norris derivative filtering with a segment length of 5 and a segment spacing of 5.

[0024] In some implementations, brute-force computation is performed using the Python-Pandas data processing library.

[0025] In some implementations, S4 of the premixing method includes the following operation:

[0026] a) Calculate the average chemical composition content of the entire stack of raw materials, and calculate the chemical composition content fluctuation and percentage fluctuation of each feeding unit;

[0027] b) Based on the required number of premixed batches, and according to the chemical composition content, content fluctuation or fluctuation percentage of the feeding unit, a brute-force exhaustive algorithm is used to randomly sample the feeding units in the stack to obtain a random group with a relatively average number of units in each group.

[0028] c) Calculate and record the average chemical composition content of each random group, then calculate and record the standard deviation of the average chemical composition content between random groups, or calculate and record the fluctuation or percentage fluctuation of the chemical composition content of each random group. Record the grouping results as the baseline group.

[0029] d) Repeat steps b)-c) to obtain a new random group and standard deviation, or content fluctuation or percentage fluctuation. If the new standard deviation is lower than the previous standard deviation, or the new content fluctuation or percentage fluctuation is lower than the previous content fluctuation or percentage fluctuation, then the new random group is used as the baseline group. If the new standard deviation is higher than or equal to the previous standard deviation, or the new content fluctuation or percentage fluctuation is higher than or equal to the previous content fluctuation or percentage fluctuation, then the original baseline group is maintained.

[0030] e) Repeat steps b)-d) above until the average chemical content of each group is equal to or deviates from the average chemical content of the entire stack by no more than 1%, and the chemical content of the raw materials in each premixed formulation covers high, medium and low content.

[0031] In some implementations, the predicted chemical composition data of all feeding units obtained in S3 are recorded in a DataFrame, and Pandas' random sampling function is used to randomly sample all feeding units to obtain random groups.

[0032] A second aspect of the present invention provides a stacking raw material premixing apparatus, comprising:

[0033] The sampling module is configured to extract raw material samples from the feeding unit from the stacked raw materials;

[0034] The near-infrared acquisition module is configured to acquire the near-infrared spectrum of the raw material sample from the feeding unit;

[0035] The calculation module is configured to predict the chemical composition content of the feeding unit and calculate the chemical composition content of the entire stack of raw materials using the near-infrared spectral model constructed according to the present invention. The chemical composition is selected from total nitrogen, total alkaloids, water-soluble sugars, potassium, chlorine and moisture, preferably total alkaloids.

[0036] The premixing formulation module is configured to perform operation S4 of the premixing method described in this invention based on the calculated chemical component content.

[0037] The premixing module is configured to perform premixing on the feeding unit of the stacked raw materials according to the premixing formula.

[0038] A third aspect of the present invention provides a stacking raw material premixing apparatus, comprising:

[0039] Memory, used to store instructions;

[0040] A processor is configured to execute the instructions, causing the stacked raw material premixing device to perform the operation of the premixing method described in this invention.

[0041] A fourth aspect of the present invention provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the premixing method described in the present invention.

[0042] In some embodiments, the computer-readable storage medium of the present invention can be implemented as a non-transitory computer-readable storage medium.

[0043] The beneficial effects achieved by this invention are as follows:

[0044] This invention uses a three-dimensional positioning method to divide the feeding units. The positions of the feeding units cover the top, middle, bottom, front, back, left, and right of the entire stack, eliminating the need to pre-classify the raw materials according to their chemical composition, thus improving work efficiency. This invention constructs a near-infrared spectral model, which is accurate and reliable, and can quickly, efficiently, cost-effectively, without pollution or loss, simultaneously measure and calculate the content of multiple chemical components in the raw materials. This invention uses a computer exhaustive algorithm to automatically divide the stacked raw materials into several batches of premixed formulations, achieving uniformity of raw material composition content within and between batches.

[0045] This invention enables rapid and automated multi-batch grouping of reconstituted tobacco raw materials, achieving uniform and stable chemical composition content among different batches, thereby improving raw material uniformity and enhancing product quality stability. Attached Figure Description

[0046] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and are used to explain the invention, but do not constitute an undue limitation of the invention.

[0047] Figure 1 It shows tobacco stem raw materials of different lengths and stem diameters;

[0048] Figure 2 It shows raw materials of tobacco flakes and tobacco leaf fragments in different states;

[0049] Figure 3 This is a three-dimensional schematic diagram of the stacked raw materials, where: A represents the vertical position of the stack, B represents the horizontal position of the stack, and C represents the front and back position of the stack;

[0050] Figure 4 This shows the total alkaloid content of the short tobacco stems;

[0051] Figure 5 The percentage fluctuation of total alkaloids in short tobacco stems is shown;

[0052] Figure 6 This shows the total alkaloid content of the tobacco flakes;

[0053] Figure 7 The percentage fluctuation in total alkaloids in the tobacco flakes is shown.

[0054] Figure 8 A schematic flowchart of an embodiment of the reconstituted tobacco stacking raw material premixing method provided by the present invention;

[0055] Figure 9 This is a schematic flowchart illustrating an embodiment of the present invention for constructing a near-infrared spectral model of reconstituted tobacco raw materials.

[0056] Figure 10 A schematic diagram of one embodiment of a stacking raw material premixing device;

[0057] Figure 11 This is a schematic diagram of another embodiment of a premixing device for stacked raw materials. Detailed Implementation

[0058] The present invention will now be described more fully with reference to the accompanying drawings, illustrating exemplary embodiments thereof. Obviously, the described embodiments are merely some, and not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] The terms "front," "back," "left," "right," "up," "middle," and "down" used in this document to indicate orientation or positional relationships are based on the orientation or positional relationships shown in the accompanying drawings. They are used only for the purpose of facilitating the description of the present invention and simplifying the description, and are not intended to indicate or imply that the stacked materials referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting the scope of protection of the present invention.

[0060] Part of this invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0061] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0062] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0063] The stacking raw material premixing device, model building module, sampling module, acquisition module, calculation module, premixing formula module, and premixing module described in this invention can be implemented as a general-purpose processor, programmable logic controller (PLC), digital signal processor (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, discrete hardware component, or any suitable combination thereof for performing the functions described in this invention.

[0064] Those skilled in the art will understand that all or part of the steps of this invention can be implemented by hardware or by a program instructing the relevant hardware to implement them. The program can be stored in a non-transitory computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0065] Figure 8 A schematic flowchart of an embodiment of the reconstituted tobacco stacking raw material premixing method provided by the present invention, wherein:

[0066] Step S1: Construct a near-infrared spectral model to predict the chemical composition content of reconstituted tobacco raw materials.

[0067] Modeling operation process as follows Figure 9 As shown, it includes:

[0068] Near-infrared spectra of the modeling samples were collected, preferably using integrating sphere diffuse reflectance or fiber optic diffuse reflectance acquisition modes. The spectral range of the near-infrared spectrum was 4000–7500 cm⁻¹. -1 The preferred length is 4256–7030 cm. -1 Preferably, the near-infrared spectral resolution is 8.0 cm⁻¹. -1 The sample was scanned 68 times, and the background scan frequency was once every 30 minutes.

[0069] The chemical composition content of the modeling sample is determined. The chemical composition is at least one of total nitrogen, total alkaloids, water-soluble sugar (total sugar), potassium, chlorine and moisture. It is preferred to use total alkaloids, which have relatively small changes during storage, as the main indicator. It is preferred to use a continuous flow method for determination. The determination method refers to industry standards.

[0070] The near-infrared spectrum of the modeling sample is fitted with the measured content of the corresponding chemical components to establish a near-infrared spectral model for predicting the chemical component content of the stacked raw materials. The partial least squares method is preferably used to fit the near-infrared spectrum of the modeling sample with the measured content of the corresponding chemical components.

[0071] Preferably, the modeling samples include tobacco leaf fragment samples and tobacco stem samples. Preferably, a portion of the tobacco leaf fragment samples is ground into powder and sieved, while a portion remains intact; all of the tobacco stem samples are ground into powder and sieved.

[0072] After acquiring the near-infrared spectrum of the modeling sample, the process also includes preprocessing the spectrum. The preprocessing includes at least one of vector normalization, standard canonical transformation, first derivative, second derivative, multivariate signal correction, and spectral smoothing. The spectral smoothing includes Savitzky-Golay smoothing filtering (segment length 7; segment spacing 3) and Norris derivative filtering (segment length 5; segment spacing 5).

[0073] For example, by using multivariate signal correction to eliminate differences caused by sample inhomogeneity, using Norris derivative filtering with a segment length of 5 and a segment spacing of 5 to smooth high-frequency noise in the spectrum and retain useful low-frequency information, and using first-order derivative processing to eliminate the influence of baseline drift, higher resolution and clearer contour changes are obtained than the original spectrum.

[0074] Step S2: Divide the stacked raw materials into feeding units according to three-dimensional positioning.

[0075] The raw materials in the stack are divided into at least 30 (e.g., 40, 50, 60, 70, 80, 90, 100) feeding units, with the feeding units located at the top, middle, bottom, front, back, left, and right of the entire stack. Using this method, there is no need to pre-classify the stacked raw materials according to their chemical composition; materials can be directly prepared from the raw material warehouse to the premixing area using three-dimensional positioning, improving work efficiency. The feeding unit can be one bag / box of raw material or multiple bags / boxes of raw material.

[0076] Step S3: Collect the near-infrared spectrum of the raw material sample of each feeding unit, and use the aforementioned near-infrared spectral model to predict the chemical composition content of the raw material sample of each feeding unit. The chemical composition includes, for example, total nitrogen, total alkaloids, water-soluble sugars (total sugars), potassium, chlorine and moisture, preferably total alkaloids.

[0077] Near-infrared spectra are preferably acquired using integrating sphere diffuse reflectance or fiber optic diffuse reflectance acquisition modes. Integrating sphere diffuse reflectance mode is used to acquire the spectrum of the sample after it has been ground into powder, while fiber optic diffuse reflectance mode is used to acquire the spectrum of the sample in its original state. The spectral range of the near-infrared spectrum is 4000–7500 cm⁻¹. -1 The preferred length is 4256–7030 cm. -1 Preferably, the near-infrared spectral resolution is 8.0 cm⁻¹. -1 The sample was scanned 68 times, and the background scan frequency was once every 30 minutes.

[0078] Step S4: Based on the aforementioned chemical composition content prediction results, use a brute-force exhaustive algorithm to divide all feeding units into at least 2 (e.g., 2, 3, 4, 5, 6, 7, 8, 9 or 10) batches, so that the average chemical composition content of each batch after premixing is equal to or deviates from the average chemical composition content of the entire stack by no more than 1%.

[0079] Preferably, the process is performed using a computer-based exhaustive algorithm, such as the Python-Pandas data processing library. This involves recording the spatial location and component content information of all feeding units in a DataFrame, and then using Pandas' random sampling function to randomly sample all feeding units to obtain random groups.

[0080] The specific steps are as follows:

[0081] a) Calculate the average chemical composition content of the entire stack of raw materials to be premixed, and calculate the chemical composition content fluctuation and percentage fluctuation for each feeding unit;

[0082] b) Based on the required number of premixed batches, and according to the chemical composition content, content fluctuation or fluctuation percentage of the feeding unit, a brute-force exhaustive algorithm is used to randomly sample the feeding units in the stack to obtain a random group with a relatively average number of units in each group.

[0083] c) Calculate and record the average chemical composition content of each random group, then calculate and record the standard deviation of the average chemical composition content between random groups, or calculate and record the fluctuation or percentage fluctuation of the chemical composition content of each random group. Record the grouping results as the baseline group; preferably, the number of feeding units in each random group is not less than 6.

[0084] d) Repeat steps b)-c) to obtain a new random group and standard deviation, or content fluctuation or percentage fluctuation. If the new standard deviation is lower than the previous standard deviation, or the new content fluctuation or percentage fluctuation is lower than the previous content fluctuation or percentage fluctuation, then the new random group is used as the baseline group. If the new standard deviation is higher than or equal to the previous standard deviation, or the new content fluctuation or percentage fluctuation is higher than or equal to the previous content fluctuation or percentage fluctuation, then the original baseline group is maintained.

[0085] e) Repeat steps b)-d) above until the average chemical content of each group is equal to or deviates from the average chemical content of the entire stack by no more than 1%. The grouping at this point is the final premix formulation. The average chemical content of the intra-batch and inter-batch in the premix formulation is close to the average of the entire stack, and the chemical content of the raw materials in each premix formulation covers high, medium and low content.

[0086] Figure 10 This is a schematic diagram of one embodiment of the stacking raw material premixing device provided by the present invention. Figure 10 As shown, the stacking raw material premixing device provided by the present invention may include a sampling module, a near-infrared acquisition module, a calculation module, a premixing formulation module, and a premixing module, wherein:

[0087] The sampling module is configured to extract raw material samples from the feeding unit from the stacked raw materials.

[0088] The near-infrared acquisition module is configured to acquire the near-infrared spectrum of the raw material sample from the feeding unit. It preferably uses integrating sphere diffuse reflectance or fiber optic diffuse reflectance acquisition modes, with the near-infrared spectral range being 4000–7500 cm⁻¹. -1 The preferred length is 4256–7030 cm.-1 Preferably, the near-infrared spectral resolution is 8.0 cm⁻¹. -1 The sample was scanned 68 times, and the background scan frequency was once every 30 minutes.

[0089] The calculation module is configured to predict the chemical composition content of the feeding unit using the near-infrared spectral model constructed according to the present invention, wherein the chemical composition is selected from total nitrogen, total alkaloids, water-soluble sugars, potassium, chlorine and moisture, preferably total alkaloids.

[0090] The premix formulation module is configured to perform operation S4 of the premixing method described in this invention based on the calculated average content of chemical components, to obtain the formulation sheet for the required premix batch, such that the average content of chemical components in each batch is equal to or deviates from the average content of the chemical components in the entire stack by no more than 1%.

[0091] The premixing module is configured to perform premixing on the feeding unit of the stacked raw materials according to the premixing formula.

[0092] Figure 11 This is a schematic diagram of another embodiment of the stacking raw material premixing device of the present invention. Figure 11 As shown, the stacking raw material premixing device of the present invention includes a memory 71 and a processor 72.

[0093] The memory 71 is used to store instructions, and the processor 72 is coupled to the memory 71. The processor 72 is configured to implement the stacking raw material premixing method of the present invention based on the execution of instructions stored in the memory.

[0094] like Figure 11 As shown, the stacking raw material premixing device also includes a communication interface 73 for information exchange with other devices. Additionally, the stacking raw material premixing device includes a bus 74, through which the processor 72, communication interface 73, and memory 71 communicate with each other.

[0095] The memory 71 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk drive. The memory 71 may also be a memory array. The memory 71 may also be divided into blocks, and these blocks may be combined into virtual volumes according to certain rules.

[0096] Furthermore, the processor 72 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0097] The premixing method for reconstituted tobacco stacking raw materials provided by this invention has an accurate and reliable model, and is simple and convenient to operate. It can quickly and automatically group reconstituted tobacco raw materials into multiple batches, so as to achieve uniform and stable chemical composition content of raw materials among different batches, thereby improving the uniformity of raw materials and enhancing product quality stability.

[0098] The present invention will be illustrated below with specific examples. Unless otherwise specified, conditions should be followed according to standard conditions or the manufacturer's recommendations. Reagents or instruments whose manufacturers are not specified are all commercially available products.

[0099] Example 1: Establishing a near-infrared spectral model

[0100] 1. Collect and prepare samples

[0101] Sampling was conducted on reconstituted tobacco raw materials at different locations in the stack, from top to bottom and from front to back and from left to right. A total of 305 tobacco leaf fragments and 194 tobacco stem samples were collected from different regions in 2021-2022. The tobacco leaf fragment samples were divided into two parts. One part was ground into powder and passed through a 40-mesh sieve. The tobacco stem samples were also ground into powder and passed through a 40-mesh sieve.

[0102] 2. Acquisition of spectra

[0103] Fourier transform near-infrared spectrometer (manufacturer: Thermo Fisher Scientific, model: Antaris 2) was used to acquire the spectra of the ground tobacco stems and tobacco leaf fragments using the integrating sphere diffuse reflectance method, and the spectra of the unprocessed tobacco leaf fragments were acquired using fiber optic diffuse reflectance. The acquisition parameters are as follows:

[0104] Acquisition modules: Rotating sample cup integrating sphere diffuse reflectance acquisition module, fiber optic diffuse reflectance module.

[0105] Data format: Log1 / R

[0106] Resolution: 8.0cm -1

[0107] Spectral range: 4000-10000cm -1

[0108] Number of sample scans: 68

[0109] Background scanning frequency: once every 30 minutes

[0110] Spectrum saving formats: SPA and SPC

[0111] Each sample was scanned three times consecutively, and the average value of the resulting spectral data was taken.

[0112] 3. Determination of chemical composition content

[0113] The water-soluble sugar (total sugar), total alkaloids, chlorine, potassium, total nitrogen, and moisture content of the sample were determined using a continuous flow analyzer according to the methods specified in the industry standards YC / T 159-2019 Determination of Water-Soluble Sugars in Tobacco and Tobacco Products by Continuous Flow Method, YC / T 217-2007 Determination of Potassium in Tobacco and Tobacco Products by Continuous Flow Method, YC / T 162-2011 Determination of Chlorine in Tobacco and Tobacco Products by Continuous Flow Method, YC / T 161-2002 Determination of Total Nitrogen in Tobacco and Tobacco Products by Continuous Flow Method, YC / T468-2013 Determination of Total Alkaloids in Tobacco and Tobacco Products by Continuous Flow Method (Potassium Thiocyanate), and YC / T31-1996 Preparation of Samples and Determination of Moisture in Tobacco and Tobacco Products by Oven Method. These values ​​were used as the determination values ​​for these chemical components and moisture content of the sample.

[0114] 4. Construct a near-infrared spectral model

[0115] 4.1 Select the spectral region and perform spectral data preprocessing.

[0116] To eliminate the influence of random noise and reduce systematic errors, the original spectrum of the sample can be preprocessed using one or more of the following methods: vector normalization, standard canonical transformation, first derivative, second derivative, multivariate signal correction, or spectral smoothing.

[0117] Based on the original near-infrared spectrum, the near-infrared absorption spectrum of the tobacco leaf fragment samples is between 4000 and 7500 cm⁻¹. -1 The regional signal is strong and rich in information. To eliminate the effects of high-frequency random noise, baseline drift, sample inhomogeneity, light scattering, etc., and to fully extract the effective feature information contained in the spectrum and improve the prediction accuracy of the calibration model, necessary preprocessing of the spectrum is required.

[0118] The acquired raw spectra were preprocessed using multiplicative signal correction (MSC), standard normal variation (SNV), and / or first-order derivative, second-order derivative combined with Savitzky-Golay (SG) smoothing filtering (segment length 7; segment spacing 3) or Norris derivative filtering (segment length 5; segment spacing 5). The results showed that different preprocessing methods had a certain impact on the model's prediction results. The prediction performance was better when multiplicative signal correction was used, and first-order derivative combined with Norris smoothing filtering was employed.

[0119] Taking water-soluble sugars and total alkaloids as examples, the calibration models for water-soluble sugars and total alkaloids were optimized through spectral preprocessing, and the results are shown in Table 1. Considering all parameters of the model, the following methods were used to preprocess the spectra to obtain ideal results: ① Multivariate signal correction was used to eliminate differences caused by sample inhomogeneity; ② Norris derivative filtering with a segment length of 5 and a segment spacing of 5 was used to smooth the spectrum, eliminating high-frequency noise and retaining useful low-frequency information; ③ First-order derivative processing was used to eliminate the influence of baseline drift, obtaining higher resolution and clearer contour changes than the original spectrum. The optimal spectral preprocessing parameters for other chemical components such as water-soluble sugars, chlorine, potassium, total nitrogen, and water were determined using the aforementioned methods. The final modeling spectral region was determined to be 4256.13–7029.55 cm⁻¹. -1 Between these two points, the result is optimal at this time.

[0120] Table 1 Comparison of water-soluble sugars and total alkaloids in tobacco leaf fragments under different pretreatments

[0121]

[0122] 4.2 Constructing a near-infrared spectral model

[0123] The near-infrared spectra of two tobacco leaf fragment samples obtained by two spectral acquisition modes, namely integrating sphere spectral diffuse reflectance and fiber optic diffuse reflectance, were fitted with the corresponding chemical composition values ​​measured by the continuous flow method to establish near-infrared spectral models for predicting the contents of water-soluble sugars, total alkaloids, chlorine, potassium, and total nitrogen in tobacco leaf fragments.

[0124] The near-infrared spectra of tobacco stem samples obtained by the diffuse reflectance spectroscopy acquisition mode of the integrating sphere were fitted with the corresponding chemical composition values ​​measured by the continuous flow method using the partial least squares (PLS) method to establish near-infrared spectral models for predicting the contents of water-soluble sugars, total alkaloids, chlorine, potassium, and total nitrogen in tobacco stems.

[0125] The technical specifications of the near-infrared spectral model of tobacco leaf fragments constructed based on the diffuse reflectance acquisition mode of the integrating sphere spectrum are shown in Table 2.

[0126] Table 2. Relevant technical indicators of the near-infrared spectral model of tobacco leaf fragments constructed based on the integrating sphere spectral diffuse reflectance acquisition mode.

[0127]

[0128]

[0129] The technical specifications of the near-infrared spectral model of tobacco leaf fragments constructed based on the fiber optic diffuse reflection acquisition mode are shown in Table 3.

[0130] Table 3. Relevant technical indicators of the near-infrared spectral model of tobacco leaf fragments constructed based on fiber optic diffuse reflection acquisition mode.

[0131]

[0132] The technical specifications of the near-infrared spectral model of tobacco stems constructed based on the diffuse reflectance acquisition mode of the integrating sphere spectrum are shown in Table 4.

[0133] Table 4. Relevant technical indicators of the near-infrared spectral model of tobacco stems constructed based on the diffuse reflectance acquisition mode of the integrating sphere spectrum.

[0134]

[0135]

[0136] Example 2: Calculation Method for Improving the Uniformity of Stacked Raw Materials

[0137] like Figure 3 As shown, the reconstituted tobacco raw material is divided into feeding units according to three-dimensional positioning, and the position of each feeding unit in three-dimensional space is recorded through three directions: A, B, and C.

[0138] Near-infrared spectra of the feeding units were collected, and the chemical composition content of the raw material samples in each feeding unit was predicted using the near-infrared spectral model established in Example 1, with the total alkaloids, which changed relatively little during storage, as the main indicator.

[0139] Multiple feed units with values ​​above and below the average are premixed in one batch to make the average chemical composition content of that batch close to the average of the entire stack. Furthermore, the average chemical composition content of other premixed batches in that stack is also close to the average of the entire stack, ensuring that the average chemical composition content within and between batches is close to the average of the entire stack. The average total alkaloid content of each batch of premixed formulation is controlled to be equal to or deviate from the average total alkaloid content of the entire stack by no more than 1%.

[0140] The brute-force search algorithm is used to exhaustively search every possible combination to achieve the above solution objective. The brute-force search algorithm is implemented using the Python-Pandas data processing library, and follows the workflow as follows:

[0141] 1. Create a DataFrame, input the three-dimensional spatial position of all feeding units of the entire stack of raw materials and the component content of each feeding unit predicted by the constructed near-infrared model, and calculate the average chemical component content of the entire stack of raw materials, as well as the chemical component content fluctuation and fluctuation percentage of each feeding unit.

[0142] 2. Based on the number of premixed batches, use Pandas' random sampling function to randomly sample the DataFrame to obtain random groups. Each group is a premixed batch, and the number of samples in each random group is roughly the same.

[0143] 3. Calculate and record the average chemical component content within each random group, and the standard deviation of the average chemical component content between random groups, or calculate and record the fluctuation or percentage fluctuation of the chemical component content within each random group, and record the grouping result at this time as the baseline group.

[0144] 4. Repeat steps 2-3, comparing the newly obtained random group standard deviation with the previous standard deviation. If it is lower than the previous standard deviation, the new grouping result is used as the baseline grouping. If it is higher than or equal to the previous standard deviation, the original baseline grouping is maintained. Alternatively, compare the newly obtained random grouping chemical component content fluctuation or fluctuation percentage with the previous chemical component content fluctuation or fluctuation percentage. If it is lower than the previous chemical component content fluctuation or fluctuation percentage, the new grouping result is used as the baseline grouping. If it is higher than or equal to the previous chemical component content fluctuation or fluctuation percentage, the original baseline grouping is maintained.

[0145] 5. Repeat steps 2-4 above until the average total alkaloid content of each batch of premixed formulation is equal to or deviates from the average total alkaloid content of the entire stack by no more than 1%. The grouping at this point is the final premixed formulation sheet.

[0146] This method can be used to create a premix formula sheet for several consecutive batches based on the premixing needs of the raw materials.

[0147] Example 3: Premixed Formulation of Short Tobacco Stem Stacking Raw Materials

[0148] 1. Sampling

[0149] The entire stack of raw materials was divided into 54 feeding units using a three-dimensional positioning method, with 4 bags of raw materials in each feeding unit.

[0150] 2. Detect the total alkaloid content.

[0151] Infrared spectra of the feeding units were collected, and the near-infrared spectral model established in Example 1 was used to predict the total alkaloid content of 54 feeding units, and the fluctuation of total alkaloids and the percentage fluctuation of total alkaloids were calculated.

[0152] The total alkaloid fluctuation is the difference between the total alkaloid content of raw materials at a certain location and the arithmetic mean of the total alkaloid content of raw materials in the whole stack. The total alkaloid fluctuation percentage is the percentage obtained by subtracting the arithmetic mean of the total alkaloid content of raw materials in the whole stack from the total alkaloid content of raw materials at a certain location, and then dividing by the arithmetic mean of the total alkaloid content of raw materials in the whole stack. That is, the total alkaloid fluctuation percentage = 100% * (total alkaloid content of raw materials at a certain location - arithmetic mean of the total alkaloid content of raw materials in the whole stack) / arithmetic mean of the total alkaloid content of raw materials in the whole stack. The data is shown in Table 5 below.

[0153] Table 5

[0154]

[0155]

[0156]

[0157] Figure 4 and Figure 5 The table shows the fluctuations and percentages of total alkaloid content, with the horizontal axis corresponding to the "Number" column in Table 5. It can be seen that the total alkaloid content in the short tobacco stem stacks gradually increases from front to back, with fluctuations ranging from -27% to 21%. If premixing is done using the current cross-sectional discharge method, the lower total alkaloid content at the front of the stack will be mixed together, while the higher content at the back will be mixed together, making it impossible to guarantee the uniformity and consistency of the premixing of the entire stack of raw materials.

[0158] 3. Automatic formulation

[0159] Using the raw material premixing method established in Example 2, based on the principle of matching the positive and negative fluctuations of total alkaloids, and according to the processing capacity of the production line, the whole stack of raw materials is divided into 3 or 5 batches for mixing, so that the average value of the total alkaloids in each batch is equal to or close to the average value of the total alkaloids in the whole stack of raw materials (0.82), thereby achieving uniformity and consistency of the premixing of the whole stack of raw materials.

[0160] According to the exhaustive scheme of Example 2, the formulas for dividing the whole stack into 3 batches or 5 batches were calculated respectively. The formulas for 3 batches of premixing are shown in Table 6 below, and the formulas for 5 batches of premixing are shown in Tables 7-1 and 7-2 below.

[0161] 3.1 Formula for premixing in 3 batches

[0162] Table 6

[0163]

[0164]

[0165] This formulation is an exemplary solution, which can achieve uniform premixing of stacked raw materials.

[0166] 3.2 Formula for premixing in 5 batches

[0167] Table 7-1

[0168]

[0169] Table 7-2

[0170]

[0171]

[0172] This formulation is an exemplary solution, which can achieve uniform premixing of stacked raw materials.

[0173] Example 4: Premixed Formulation of Tobacco Fragment Stacking Raw Materials

[0174] 1. Sampling

[0175] The entire stack of raw materials is divided into 60 feeding units using a three-dimensional positioning method, with 8 bags of raw materials in each feeding unit.

[0176] 2. Detect the total alkaloid content.

[0177] Some tobacco leaf fragments were ground into powder and passed through a 40-mesh sieve. Near-infrared spectra of the tobacco leaf fragment feeding unit were collected using the diffuse reflectance acquisition mode of the integrating sphere spectrum. The near-infrared spectral model established in Example 1 was used to predict the total alkaloid content of 60 feeding units. The total alkaloid fluctuation and the total alkaloid fluctuation percentage were calculated according to the method in Example 3. The data are shown in Table 8 below.

[0178] Table 8

[0179]

[0180]

[0181]

[0182] Figure 6 and Figure 7The table shows the fluctuations and percentage changes in total alkaloid content, with the horizontal axis corresponding to the "Number" column in Table 8. It can be seen that there is no significant difference in the total alkaloid content before and after the stacking of tobacco fragments. The ratio of feed units above and below the average value in the stack is approximately 1:2. The fluctuation in total alkaloid content above the average value in the stack ranges from 13.8% to 77.2%, a large fluctuation, while the fluctuation in total alkaloid content below the average value ranges from -34.1% to -1.6%, a smaller fluctuation. If the current cross-sectional premixing method is used, the average total alkaloid content within and between batches will differ due to the varying fluctuation ranges. If a 1:2 feed ratio of above and below the average value in the stack is used, although the premixing effect will be better than the current cross-sectional premixing method, the problem of differences in total alkaloid content within and between batches of reconstituted tobacco products will still exist.

[0183] 3. Automatic formulation

[0184] Using the method for improving the uniformity of raw material premixing established in Example 2, the raw materials in the whole stack are divided into 4 batches according to the principle of matching the positive and negative fluctuations of total alkaloids. Based on the processing capacity of the production line, the average value of the total alkaloids in each batch is equal to or close to the average value of the total alkaloids in the whole stack of raw materials, which is 1.23, thereby achieving uniformity and consistency of the premixing of the whole stack of raw materials.

[0185] Following the exhaustive scheme of Example 2, the formulations for premixing in four batches were calculated, and the results are shown in Tables 9-1 and 9-2 below.

[0186] Table 9-1

[0187]

[0188] Table 9-2

[0189]

[0190]

[0191] This formulation is an exemplary solution, which can achieve uniform premixing of stacked raw materials.

[0192] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them; although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications can still be made to the specific implementation of the present invention or equivalent substitutions can be made to some technical features without departing from the spirit of the technical solutions of the present invention, and all such modifications and substitutions should be covered within the scope of the technical solutions claimed in the present invention.

Claims

1. A method for premixing reconstituted tobacco stacking raw materials, comprising: S1: Construct a near-infrared spectral model to predict the chemical composition content of reconstituted tobacco raw materials; S2: Divide the stacked raw materials into at least 30 feeding units; S3: Collect the near-infrared spectra of the raw material samples from each feeding unit divided in S2, and use the near-infrared spectral model constructed in S1 to predict the chemical composition content of the raw material samples from each feeding unit. The chemical composition is selected from total nitrogen, total alkaloids, water-soluble sugars, potassium, chlorine and moisture. S4: Based on the prediction results of S3, use a brute-force exhaustive algorithm to divide all feeding units into at least two batches, ensuring that the average chemical composition content of each batch after premixing is equal to or deviates from the average chemical composition content of the entire stack by no more than 1%. S1 includes the following operations: Collect the near-infrared spectrum of the modeling sample; Detect the chemical composition content of the modeling sample; The near-infrared spectra of the modeled samples were fitted with the chemical composition content to establish a near-infrared spectral model for predicting the stacked raw materials. S4 includes the following operations: a) Calculate the average chemical composition content of the entire stack of raw materials, and calculate the chemical composition content fluctuation and percentage fluctuation of each feeding unit; b) Based on the required number of premixed batches, and according to the chemical composition content, content fluctuation or fluctuation percentage of the feeding unit, a brute-force exhaustive algorithm is used to randomly sample the feeding units in the stack to obtain a random group with a relatively average number of units in each group. c) Calculate and record the average chemical composition content of each random group, then calculate and record the standard deviation of the average chemical composition content between random groups, or calculate and record the fluctuation or percentage fluctuation of the chemical composition content of each random group. Record the grouping results as the baseline group. d) Repeat steps b)-c) to obtain a new random group and standard deviation, or content fluctuation or percentage fluctuation. If the new standard deviation is lower than the previous standard deviation, or the new content fluctuation or percentage fluctuation is lower than the previous content fluctuation or percentage fluctuation, then the new random group is used as the baseline group. If the new standard deviation is higher than or equal to the previous standard deviation, or the new content fluctuation or percentage fluctuation is higher than or equal to the previous content fluctuation or percentage fluctuation, then the original baseline group is maintained. e) Repeat steps b)-d) above until the average chemical content of each group is equal to or deviates from the average chemical content of the entire stack by no more than 1%, and the chemical content of the raw materials in each premix formula covers high, medium and low content.

2. The premixing method according to claim 1, wherein the chemical component is total plant alkaloids.

3. The premixing method according to claim 1, wherein all feeding units are divided into 2, 3, 4, 5, 6, 7, 8, 9 or 10 batches using a brute-force exhaustive algorithm.

4. The premixing method of claim 1, wherein the chemical composition content of the modeling sample is detected using a continuous flow method.

5. The premixing method of claim 1, wherein partial least squares method is used to fit the near-infrared spectrum of the modeled sample with the chemical composition content.

6. The premixing method according to any one of claims 1-5, wherein The spectral range of the near-infrared spectrum is 4000~7500 cm⁻¹ -1 ; Near-infrared spectra of modeled samples or raw material samples to be tested are acquired using integrating sphere spectral diffuse reflectance or fiber optic diffuse reflectance acquisition modes.

7. The premixing method according to claim 6, wherein the near-infrared spectral region is 4256~7030 cm⁻¹. -1 .

8. The premixing method according to any one of claims 1-5, wherein when acquiring near-infrared spectra, the parameters of the infrared spectrometer are set to: near-infrared spectral resolution of 8.0 cm⁻¹. -1 The sample was scanned 68 times, and the background scan frequency was once every 30 minutes.

9. The premixing method according to any one of claims 1-5, wherein After acquiring the near-infrared spectrum of the modeling sample, the process also includes preprocessing the spectrum, which includes at least one of vector normalization, standard canonical transformation, first derivative, second derivative, multivariate signal correction, and spectral smoothing.

10. The premixing method of claim 9, wherein the spectral smoothing process includes Savitzky-Golay smoothing filtering and Norris derivative filtering.

11. The premixing method of claim 9, wherein the preprocessing includes multivariate signal correction, Norris derivative filtering, and first derivative processing.

12. The premixing method of claim 11, wherein a Norris derivative filter with a segment length of 5 and a segment spacing of 5 is used to smooth the spectrum.

13. The premixing method according to any one of claims 1 to 5, wherein the computation is performed using a brute-force algorithm via the Python-Pandas data processing library.

14. The premixing method according to any one of claims 1-5, wherein the predicted chemical composition content of all feeding units obtained in S3 is recorded in a DataFrame, and the random sampling function of Pandas is used to randomly sample all feeding units to obtain random groups.

15. A stacking raw material premixing apparatus for performing the premixing method according to any one of claims 1-14, comprising: The sampling module is configured to extract raw material samples from the feeding unit from the stacked raw materials; The near-infrared acquisition module is configured to acquire the near-infrared spectrum of the raw material sample from the feeding unit; The calculation module is configured to predict the chemical composition content of the feeding unit based on the near-infrared spectral model, wherein the chemical composition is selected from total nitrogen, total alkaloids, water-soluble sugars, potassium, chlorine and moisture; The premixed formulation module is configured to perform the operation described in S4 based on the calculated chemical component content; The premixing module is configured to perform premixing on the feeding unit of the stacked raw materials according to the premixing formula.

16. A stacking raw material premixing device, comprising: Memory, used to store instructions; A processor is configured to execute the instructions, causing the stacking material premixing apparatus to perform the operation of the premixing method according to any one of claims 1-14.

17. A computer-readable storage medium, wherein, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the premixing method according to any one of claims 1-14.

Images