Cigarette formula maintenance synchronous prediction calculation method

Abstract

The invention discloses a cigarette formula maintenance synchronous prediction calculation method which comprises the following steps: obtaining cigarette interval yields in different production intervals according to cigarette pre-production years; on the basis of the measurement and calculation rule, obtaining a measurement and calculation sequence of different cigarette marks, and on the basis of stock strip information, obtaining the current stock weight of a cigarette mark formula leaf group and the production capacity of the cigarette mark; according to the cigarette interval yield and the cigarette brand production capacity, prediction calculation is conducted, and stock tobacco leaf resources are reasonably and effectively utilized.

Description

Technical Field

[0001] The invention relates to a cigarette formula maintenance synchronous prediction calculation method, belonging to the technical field of tobacco data analysis technology. Background Art

[0002] With the continuous development of cigarette brands, the imbalance in the structure of tobacco leaf inventory and procurement areas, parts, and categories is more prominent. Although industrial enterprises have continuously optimized leaf group formulas and processing technology in recent years, the bottlenecks in the supply and demand structure of tobacco leaf raw materials and cost pressures they face have gradually become prominent. In order to solve the structural contradictions in the supply of high-quality tobacco leaf raw materials, tobacco enterprise formula personnel make highly professional manual judgments based on objective data such as tobacco leaf origin, grade, and conventional chemical composition, combined with years of experience accumulation and sensory evaluation, to support the effective replacement of leaf group formulas. Due to the large number of indicators, tobacco leaf formula work is intensive and subjective, which is not only inefficient, but also often difficult to find the most suitable replacement tobacco leaf.

[0003] At present, the allocation of tobacco resources by tobacco industry enterprises is still mainly based on the long-term accumulated experience, combined with the judgment of senior tobacco formula personnel for fine-tuning, which has problems such as low efficiency, small adjustment space and lack of professional theoretical guidance. In recent years, the tobacco product consumption market has shown diversified characteristics, and innovative products such as medium-sized cigarettes, short-sized cigarettes, slim cigarettes, ultra-slim cigarettes, electronic cigarettes and heated cigarettes have been continuously iteratively developed, which objectively puts forward higher requirements for the product development speed, maintenance efficiency, quality control and delivery accuracy of tobacco enterprises. Especially in the current environment of tight supply of high-quality tobacco leaves, there is a serious cross-coupling in the demand for tobacco leaves by different cigarette brands. In addition, the long lag in the use of tobacco leaves by industrial enterprises due to the need for aging for 2-3 years after the purchase of tobacco leaves, urgently requires tobacco enterprises to accelerate digital transformation and upgrading, improve the digital design and maintenance of cigarette product formulas, in order to help enterprises make plans in advance, ensure the basic stability of the quality of cigarette products of various brands to be produced and the rational and effective use of inventory tobacco resources. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art, provide a method for synchronous prediction and calculation of cigarette formula maintenance, solve the problem of tobacco leaf resource measurement in cigarette formula maintenance, reasonably and effectively utilize inventory tobacco leaf resources, and maintain the stability of cigarette product quality.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] In a first aspect, the present invention provides a method for synchronous prediction and calculation of cigarette formula maintenance, comprising:

[0007] Step 1: According to the expected production year of cigarettes, obtain the cigarette interval production in different production intervals;

[0008] Step 2: Based on the calculation rules, obtain the calculation order of different cigarette brands;

[0009] Step 3: Based on the inventory leaf information, obtain the current inventory weight of the cigarette brand blend leaf group and the production capacity of the cigarette brand;

[0010] Step 4: Based on the cigarette production range and cigarette brand production capacity, a forecast is made, including the following steps:

[0011] If the production capacity of the cigarette brand is not less than the interval production of cigarettes, the prediction data of the cigarette leaf group formula is obtained according to the quality index of each leaf group in the cigarette formula table; according to the prediction data of the cigarette leaf group formula, the inventory weight of the leaf group is updated, and the step 3 is returned to complete the prediction calculation of the next cigarette brand formula leaf group;

[0012] If the production capacity of the cigarette brand is less than the cigarette interval output, obtain the insufficient tobacco data of each leaf group, perform secondary similarity substitution on the insufficient tobacco data, obtain the new leaf group formula, and return to step 3 to complete the prediction calculation of the new leaf group formula.

[0013] In combination with the first aspect, further, the secondary similar substitution includes:

[0014] Based on the near infrared spectrum-tobacco leaf quality index prediction model, the weighted values ​​of q quality indexes of cigarettes in the past several quarters are calculated to obtain the overall matrix;

[0015] According to the tobacco leaf similarity evaluation of the current tobacco leaves in stock, a similarity matrix of the tobacco leaves in stock is obtained;

[0016] By identifying s tobacco leaves to be replaced in a cigarette blend containing M tobacco leaf groups, p tobacco leaves that are most similar to the tobacco leaves to be replaced are selected from the similarity matrix to obtain p s A new leaf group formula of the cigarette;

[0017] Based on p s The M leaf group weighted values ​​of the q quality indicators of the new leaf group formula are obtained by s weighted values;

[0018] For the overall matrix, p s The weighted values ​​are reduced by principal component analysis to obtain the overall dimension reduction matrix and p s A reduced dimensionality vector;

[0019] According to the calculation p s The Mahalanobis distance from the reduced dimension vector to the overall reduced dimension matrix is ​​obtained, and the new leaf group formula corresponding to the minimum Mahalanobis distance is the cigarette formula of similar substitution.

[0020] Furthermore, by evaluating the similarity of the current tobacco leaves in stock, a similarity matrix of the tobacco leaves in stock is obtained, including:

[0021] Multi-point sampling to collect near-infrared spectral data of current tobacco leaves in stock;

[0022] The near-infrared spectral data of the current stock tobacco leaves are subjected to scattering correction, synchronous smoothing derivation, feature selection and two-dimensional similarity calculation in sequence to obtain the similarity matrix of the stock tobacco leaves.

[0023] Furthermore, calculating the two-dimensional similarity includes:

[0024] Based on the shape differences and information content differences in the spectral data, the total similarity of the spectral data of any two tobacco leaves after feature selection is calculated;

[0025] The calculation formula for the total similarity is:

[0026]

[0027] Where: S represents the total similarity of the spectrum, D shape represents the shape similarity represented by the dynamic time warping distance; D Info It represents the difference in the amount of information contained in two spectra characterized by information divergence;

[0028]

[0029]

[0030] Where: j = 1, 2, 3, .., n, x j ,y j are two different tobacco leaf sample data points after synchronous smoothing derivation of near-infrared spectra; n is the number of spectral data points;

[0031] Furthermore, the q quality indicators include:

[0032] If the cigarette is a single recipe: select q quality indicators based on the style characteristics of the cigarette brand and the actual recipe;

[0033] If the cigarette is not a single recipe: according to the principle of feature selection, use relative standard deviation to select q quality indicators; or set at least five members to vote and select q quality indicators based on the total weight.

[0034] Furthermore, at least five members are set to vote and select q quality indicators based on the total weight, including:

[0035] Each member makes a primary and secondary judgment on any two chemical indicators to obtain a comparison matrix;

[0036] Convert the comparison matrix into a judgment matrix.

[0037] The objective function is constructed by the least square method, the auxiliary function is constructed by the Lagrange conditional extreme value, and the weight is obtained by solving the judgment matrix;

[0038] The obtained weights are averaged as the total weight, and q chemical indicators that are greater than the total weight are selected as quality indicators.

[0039] Furthermore, the method further comprises:

[0040] Calculate the correlation coefficient of the weights;

[0041] The quality index selected was eliminated by eliminating members whose correlation coefficient was lower than the critical correlation coefficient at 95% confidence level.

[0042] Furthermore, based on the inventory sheet information, the current inventory weight of all leaf groups in the cigarette brand and the production capacity of the cigarette brand are obtained, including:

[0043] Based on the inventory leaf information, obtain the current inventory weight of all leaf groups in the cigarette brand;

[0044] According to the current inventory weight and the proportion of leaf groups in the cigarette brand, the production capacity of all leaf groups is obtained; among them, the smallest leaf group production capacity is the production capacity of the cigarette brand.

[0045] Furthermore, the cigarette prediction data includes interval formulation output, interval output ratio and interval quality index; the insufficient tobacco leaf data includes: insufficient tobacco leaf name, insufficient tobacco leaf weight and insufficient tobacco leaf quality index; the calculation rules include market price, interval output and number of formulation leaf group types.

[0046] In a second aspect, the present invention provides a computer device, comprising:

[0047] A memory for storing instructions;

[0048] The processor is used to execute the instructions so that the device performs operations to implement any of the above-mentioned cigarette formula maintenance synchronous prediction calculation methods.

[0049] Compared with the prior art, the present invention has the following beneficial effects:

[0050] The present invention is based on the dynamic inventory tobacco resources of industrial enterprises and the related indicator information, and carries out dynamic inventory planning and design calculation process according to the cigarette interval output and cigarette brand production capacity, so as to realize the synchronous prediction and calculation of the whole cigarette brand formula maintenance, and provide effective solutions and close to true expected calculation for tobacco industrial enterprises in tobacco procurement, cost accounting, product development, inventory tobacco resources allocation and rational utilization.

[0051] Based on near-infrared spectral data, a two-level similarity algorithm is used to complete the replacement of tobacco leaf group formula in the measurement process. The first-level replacement method first uses near-infrared spectroscopy for initial screening, and selects several tobacco leaves that are most similar to the replaced tobacco leaves as a whole to form a pre-selection pool. The second-level replacement algorithm uses Mahalanobis distance selection to optimize and screen several quality indicators to select the tobacco leaves with the smallest Mahalanobis distance from the overall first-level replacement pre-selection pool to complete the prediction of the new leaf group formula. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 A flow chart of a synchronous prediction calculation method for cigarette formula maintenance provided by an embodiment of the present invention;

[0053] Figure 2 A flow chart of an alternative method provided for an embodiment of the present invention;

[0054] Figure 3 A relationship diagram between the number of principal components and the contribution provided by an embodiment of the present invention;

[0055] Figure 4 The results of the number of small formulas of 23 cigarette brands in the entire production range provided by the embodiment of the present invention;

[0056] Figure 5 The small formulation ratios of 23 cigarette brands in the entire production range provided by the embodiment of the present invention;

[0057] Figure 6 The calculation results of the six indicators of concern in the entire production interval TB2 provided by the embodiment of the present invention;

[0058] Figure 7 The SPC control limit mean and chi-square test limit provided by the embodiment of the present invention;

[0059] Figure 8 The calculation results of the unit price indicators of 23 brands in the entire production range provided by the embodiment of the present invention. DETAILED DESCRIPTION

[0060] The present invention will be further described below in conjunction with the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present invention, and cannot be used to limit the protection scope of the present invention.

[0061] The present invention is based on the dynamic inventory tobacco resources of industrial enterprises and the near-infrared spectrum data. Through the dynamic inventory planning and design calculation process, a suitable tobacco similarity evaluation algorithm is established to realize the synchronous prediction calculation of the whole cigarette brand formula maintenance. It provides effective solutions and close to the real expected calculation for tobacco industrial enterprises in tobacco procurement, cost accounting, product development, inventory tobacco resources allocation and rational utilization, and helps enterprises to make early predictions and strategic adjustments for the subsequent actual product production.

[0062] Embodiment 1:

[0063] like Figure 1 As shown, it is a flow chart of a method for synchronous prediction of cigarette formula maintenance provided by this embodiment. The present invention establishes a tobacco leaf similarity evaluation algorithm and introduces computer-aided programming, including the following steps:

[0064] Step 1: Determine the production interval and cigarette output

[0065] Read the year of planned production Year_ph, and determine the production section Section_ph according to demand; if general cigarette industrial enterprises issue production on a quarterly basis, the annual planned production of each cigarette brand can be divided into each quarter according to demand to obtain the cigarette section production Weight_ph. The calculation starts from the first quarter in chronological order until all quarters are completed.

[0066] Step 2: Determine the order of cigarette measurement

[0067] Determine all cigarette brands in production / to be produced by industrial enterprises, and sort all cigarette brands into a certain single production interval measurement order according to market price, interval output, number of formula leaf group types, or other custom rules.

[0068] Step 3: Calculate cigarette brand production capacity

[0069] Read the name Name_py and weight Weight_py of all the stocked cigarettes, the name Name_ph_py and proportion Per_ph_py of the leaf group of cigarette L, search for the name Name_ph_py of all the leaf groups of cigarette L in Name_py, and get the corresponding current stock weight Weight_py (Name_ph_py), then the capacity Weight_ph_power calculated for each leaf group of cigarette L is:

[0070] The minimum capacity is Weight_ph_minpower: Weight_ph_minpower = min( W eight_ph_power)

[0071] Step 4: Compare production capacity with planned output

[0072] Read the L interval production of cigarettes (Weight_ph), and determine whether the current inventory production capacity meets the production of cigarettes L by comparing Weight_ph_minpower and Weight_ph.

[0073] Step 5: Capacity Sufficient Process

[0074] If Weight_ph_minpower≥Weight_ph, the stock of cigarettes is sufficient. The process is divided into three steps: ①②③:

[0075] ① Determine the cigarette L quality index (Name_index) to be output, read the quality index data Index_py of each cigarette in the cigarette L formula table, and output the interval formula output Weight_xpf, interval formula output ratio Ratio_xpf and interval formula quality index Index_xpf. "Interval formula" corresponds to the small formula mentioned in the attached figure of the specification. At this time: Weight_xpf = Weight_ph, Ratio_xpf = 1, Index_xpf = Per_ph_py*Index_py

[0076] ② Update the weight of the cigarettes in stock Weight_py, that is, deduct the weight of each cigarette involved in Weight_ph from Weight_py. Amount_used = Per_ph_py*Weight_ph

[0077] ③ Update the cigarette output Weight_ph. At this time, Weight_ph=0, and the calculation of cigarette L in this interval is completed. Return to step 3 to continue the calculation of the next cigarette brand in the production interval.

[0078] Step 6: Capacity ≥ Ratio process

[0079] Under the condition of insufficient inventory of tobacco slices, set the capacity ratio Ratio value (0≤Ratio<1). If 1>Weight_ph_minpower≥Ratio*Weight_ph, the process is divided into steps ①②③④⑤⑤:

[0080] ①Also based on Name_index, Index_ph and Index_py, output Weight_xpf, Ratio_xpf and Index_xpf first. At this time:

[0081] Weight_xpf(i)=Weight_ph_power, Ratio_xpf(i)=Weight_ph_minpower / Weight_ph, Index_xpf(i)=Per_ph_py*Index_py

[0082] Among them: i = 1, 2, 3…, t, is the interval small formula number, and t is the total number of interval small formulas.

[0083] ② Update Weight_py, that is, deduct Amount_used from Weight_py.

[0084] Amount_used = Per_ph_py * Weight_ph_minpower

[0085] ③ Update Weight_ph to the remaining output Weight_ph_remain. At this time:

[0086] Weight_ph_remain(i) = Weight_ph * Ratio_xpf(i)

[0087] ④ Calculate all underweight tobacco slices (denoted as s), that is, all tobacco slices where 1 > Weight_ph_power ≥ Ratio * Weight_ph, their quality indicators (Index_ph_py_lack), names of underweight tobacco slices (Name_ph_py_lack), and weights of underweight tobacco slices (Weight_ph_py_lack).

[0088] ⑤ Loop into the production capacity and planned output comparison process in Step 4. At this time, Weight_ph_remain replaces Weight_ph in the subsequent process.

[0089] Step 7: Production capacity < Ratio process

[0090] Under the condition of insufficient inventory of tobacco slices, if Weight_ph_minpower < Ratio * Weight_ph, this process is divided into steps ① and ②: ① Calculate all underweight tobacco slices (denoted as s), that is, all tobacco slices where Weight_ph_power < Ratio * Weight_ph, their quality indicators (Index_ph_py_lack), names of underweight tobacco slices (Name_ph_py_lack), and weights of underweight tobacco slices (Weight_ph_py_lack).

[0091] ② Loop into the production capacity and planned output comparison process in Step 4. At this time, Weight_ph_remain replaces Weight_ph in the subsequent process.

[0092] Step 8: Determine the replacement method

[0093] Based on the principle of pairwise similarity of tobacco slices, simultaneously replace the underweight tobacco slices in the cut tobacco blend of cigarette L (allin scheme), and the inventory weight of the replacement tobacco slices should be greater than Ratio of the production capacity in the cigarette brand range. Compared with the method of replacing underweight tobacco slices one by one (1by1 scheme), the allin scheme is more in line with the real scenario of cigarette blend replacement and is closer to the real expected measurement.

[0094] Example 2:

[0095] Such as Figure 2 As shown, it is a flow chart of the simultaneous substitution method provided in this embodiment, comprising the following steps;

[0096] Step 1: Collection of tobacco leaf near-infrared spectra

[0097] The tobacco leaf samples were dried at below 50°C for about 4 hours until they could be crushed with fingers. The powder was ground using a FossCyclotec 1093 cyclone mill (Foss, Denmark), and the sample powder was sieved through a 60-mesh (aperture 250 μm) sieve. An appropriate amount was placed in a sample cup, and a fixed mass weight was placed on top of the sample. After it was naturally compacted, a Fourier transform near-infrared spectrometer (Thermo Fisher Scientific China Co., Ltd.) was used to collect the response values ​​at 4000 to 10000 wave numbers (one point was taken every 6 wave numbers, the number of scans was 72 times, and each near-infrared curve had a total of 1000 data) to obtain the tobacco leaf near-infrared spectrum data.

[0098] Step 2: Collect tobacco leaf quality index data and unit price data

[0099] Tobacco leaf samples were extracted according to GB / T 19616-2004 General Principles for Sampling of Bulk Tobacco Raw Materials, and samples were prepared with reference to YC / T31-1996 Preparation of Tobacco and Tobacco Products Samples and Determination of Moisture. According to YC / T 159-2019 Determination of Water-soluble Sugar in Tobacco and Tobacco Products, YC / T468-2021 Determination of Total Alkaloids in Tobacco and Tobacco Products, YC / T 162-2011 Determination of Chlorine in Tobacco and Tobacco Products, and YC / T According to the "YC / T217-2007 Determination of potassium in tobacco and tobacco products", the six conventional chemical components of tobacco leaves, namely total sugar (TS), reducing sugar (RS), total alkaloids (TA), chlorine (Cl), total nitrogen (TN) and potassium (K), were determined respectively; according to the "YC / T222-2007 Determination of pH in tobacco and tobacco products", "YC / T202-2006 Determination of chlorogenic acid, scopoletin and rutin, polyphenolic compounds in tobacco and tobacco products", and "YC / T216-2013 Determination of starch in tobacco and tobacco products - continuous flow method", the pH, chlorogenic acid and starch of tobacco leaves were determined. Two parallel tests were performed for each sample, and the results were averaged.

[0100] Combined with the usage characteristics of tobacco industry companies and the accuracy of near-infrared spectroscopy modeling, the sensory style index was optimized and designed, including 6 items: light fragrance value, intermediate fragrance value, strong fragrance value, position value, elegance 1 and elegance 2. A 7-person professional team certified by the tobacco industry completed the evaluation and scoring of each tobacco leaf sample. The scores of the 6 indicators were all 5 to 10 points, and they were positive indicators. To ensure the reliability of the results, the two highest scores and the two lowest scores for each indicator were removed with reference to the scoring rules of competitive sports, and the average value of the remaining three was taken.

[0101] The 15 tobacco leaf quality indicators are composed of the above 9 tobacco leaf chemical indicators and 6 tobacco leaf sensory style scores.

[0102] The comprehensive cost of tobacco leaves after purchasing, sorting and re-roasting is recorded as the unit price (rmb / Kg), and the unit price of all tobacco leaves in stock is calculated.

[0103] Step 3: Construction of near infrared spectroscopy-tobacco leaf quality index prediction model

[0104] Based on the tobacco leaf near-infrared spectroscopy sample data accumulated by China Tobacco Corporation over the years and the above-mentioned 15 tobacco leaf quality index data, a relatively stable near-infrared spectroscopy-tobacco leaf quality index prediction model was established using chemometric methods. The fitting determination coefficient of the true value and predicted value of the general validation set samples reached R 2 (Generally required to be >0.9) or above can ensure that the model is usable.

[0105] Step 4: Selection of tobacco leaf indicators

[0106] Different tobacco quality indicators that different formula personnel pay attention to for different cigarette brands are generally different. According to the actual situation, the present invention is adapted to three different methods:

[0107] ① Single formulation personnel independently select q tobacco leaf quality indicators or tobacco leaf unit price indicators of concern according to the style characteristics of L cigarette brand and actual formulation experience.

[0108] ② Select q quality indicators or unit price indicators with a total weight > ratio2 as the focus indicators through expert group voting.

[0109] It is difficult for an expert group of at least 5 members to accurately weight multiple indicators based on their understanding of the importance of chemical indicators, and the confidence level is not high, but it is easier to make a primary and secondary judgment on any two indicators. The present invention uses a 0-2 three-scale-hierarchy analysis method (3-ahp) to construct a comparison matrix a ij :

[0110]

[0111] In order to reduce the arbitrariness of subjective decision-making, g ij Further transformed into the judgment matrix h ij :

[0112]

[0113] Where: If b i ≥ b j , T=1, if b i <b j , T = -1; i, j, k are the 15 tobacco leaf quality indicators or unit price indicators in step 2;

[0114] Unlike most research methods that use the eigenvector (EV) method, this invention uses the least squares (LS) method with better discrimination to calculate the weight w ahp :

[0115]

[0116] Where: When only tobacco leaf quality indicators are considered, n=15, and when price indicators are additionally considered, n=16, the same below.

[0117] The auxiliary function N is constructed using the Lagrange conditional extreme value method (where γ is an auxiliary parameter):

[0118]

[0119] By ahp ) i Find partial derivatives, joint We get n+1 linear equations with n+1 variables, and after solving them we get w ahp .

[0120] The AHP method must be tested for consistency, and must meet the following requirements:

[0121] CR=CI / RI=(λ max -n) / (n-1) / RI<0.1

[0122] Where: max is the judgment matrix q ij The maximum eigenvalue of; judgment matrix h ij When the order is 15, the parameter RI = 1.59.

[0123] In order to exclude the possibility that experts may have special "favor" or "discrimination" towards certain tobacco leaf quality indicators, an error correction mechanism is added, that is, to eliminate experts whose average Pearson correlation coefficient (PR) with others is lower than the critical correlation coefficient (LPR) at 95% confidence level.

[0124] The remaining expert group members get w ahp Find the mean, and we get Select the q items whose total weight > ratio2 as the indicators of interest.

[0125] ③ Based on the principle of feature selection, the q item with relative standard deviation>ratio3 is used as the focus indicator

[0126]

[0127] Where: m represents the number of tobacco leaf samples in stock; y j Indicates tobacco leaf quality index or unit price index.

[0128] Select RSD_ZB j >The q item of ratio3 is taken as the focus indicator, and ratio3 is generally determined between 0.01 and 0.10 according to actual conditions.

[0129] Step 5: Set target cigarette quality

[0130] Taking the quality of L cigarettes in the past two years (8 quarters) as the target, the established near-infrared prediction model is used to calculate the q-item quality index values ​​of M tobacco leaves in the tobacco leaf group formula, and the weighted values ​​of M tobacco leaves are calculated according to the proportion w of L cigarette leaf group, and the overall matrix U is obtained. 8×q :

[0131]

[0132] Where: Q 8×q×k represents the q-item quality index value of the k-th leaf group of L cigarettes in 8 quarters; w i Indicates the usage ratio of the kth leaf group.

[0133] Step 6: First-level substitution algorithm

[0134] The calculation of the similarity matrix of stock tobacco leaves is divided into five steps: near-infrared spectrum measurement of stock tobacco leaves, near-infrared spectrum scattering correction, near-infrared spectrum synchronous smoothing derivation, near-infrared spectrum feature selection, and near-infrared spectrum two-dimensional similarity algorithm.

[0135] Step 6.1: Near infrared spectroscopy of tobacco stock

[0136] Tobacco leaves, as biomass raw materials, are often not uniform enough. When sampling stock tobacco leaves, it should be ensured that they are sampled at multiple points and mixed as one sample, and then the spectrum is measured according to the method in step 1.

[0137] Step 6.2: Scattering correction for NIR spectra

[0138] Standard normal variable transformation (SNV) is mainly used to eliminate the influence of tobacco powder particle size, surface scattering and optical path change on near-infrared diffuse reflectance spectrum. The new spectral data X after SNV processing ij for:

[0139]

[0140] Where: X′ represents the original spectrum; X represents the new spectrum after processing (the same below); n represents the number of spectrum points, which is 1000 here, and X ij represents the jth spectral point of the i-th sample of the new spectrum, Xi ' j Same reason.

[0141] Step 6.3: Simultaneous smoothing of near-infrared spectra

[0142] Derivation of near-infrared spectrum can eliminate baseline drift or reduce background interference, distinguish overlapping peaks, and thus improve the resolution and sensitivity of the spectrum. However, for some high-frequency noise, derivation will further amplify these noise signals, so the original spectrum is generally smoothed before derivation. Although excessive smoothing is conducive to enhancing the signal-to-noise ratio after derivation, the spectral information loss is large. If smoothing is insufficient, it is easy to increase high-frequency noise after derivation. Derivation and smoothing are independent of each other, and it often takes multiple repeated debugging to obtain better smoothing and derivation parameters. The present invention adopts a synchronous smoothing derivation preprocessing method (SSD), which performs the two operation processes of derivation and smoothing simultaneously, effectively avoiding the cumbersome operation of repeated debugging and optimization. The method is to perform least squares fitting of the data in a moving window by a polynomial, and synchronously derive the fitting function to obtain a weighted average expression of the spectral derivative signal.

[0143] Select a cubic polynomial with a moving window size of 7, and the polynomial to be fitted is:

[0144]

[0145] in For X ij eps is the fitting error; a0, a1, a2, a3 are the parameters to be fitted; i = 1, 2, 3, ..., m, are different inventory cigarette strips (the same below); m is the number of cigarette strip samples (the same below); j = -3, -2, -1, 0, 1, 2, 3, are the 7 relative position points in the moving window, then the polynomial to be fitted can be expressed as:

[0146] in By using the least squares method, we can get the values ​​of a0, a1, a2, and a3:

[0147]

[0148] For a cubic polynomial Taking the derivative of both sides, we get

[0149]

[0150] where x ij for The derivative value of , then the above polynomial can be expressed as:

[0151]

[0152] The midpoint x of the moving window i0 As the signal value of the new spectrum after processing, then:

[0153] x i0 =0.087X i-3 -0.266X i-2 -0.230X i-1 +0.230X i1 +0.266X i2 +0.087X i3

[0154] By changing the position of the moving window, the new spectral data x can be obtained after synchronous smoothing derivative preprocessing. ij In order to keep the number of spectral points the same before and after processing, since the window size is 7, cubic spline interpolation is used to interpolate 3 points at the beginning and end of the spectrum before synchronous smoothing and derivation of the near-infrared spectrum.

[0155] Step 6.4: NIR spectroscopy feature selection

[0156] In addition to the information of the sample itself, the near-infrared spectrum also contains a lot of irrelevant and repeated information. Therefore, it is often necessary to perform feature selection (FS) on the spectrum to screen the appropriate band to improve the modeling effect. The spectral feature selection of the present invention adopts the coefficient of variation method to calculate the relative standard deviation of the absorbance at each wavelength point of all inventory sample spectra according to the following formula to obtain the variance spectrum:

[0157] Where: m represents the number of tobacco leaf samples in stock.

[0158] Choose RSD-NIR j >The spectral point of ratio4, the ratio4 is generally determined between 0.01 and 0.10 according to the actual situation.

[0159] Step 6.5: 2D similarity algorithm for near infrared spectroscopy

[0160] The difference between two near-infrared spectra can be divided into two parts: shape difference and information content difference. The present invention establishes a two-dimensional similarity evaluation method combining the two, namely,

[0161] Where: S is the total similarity of the spectra. The larger S is, the higher the similarity between the two tobacco samples is, and vice versa. D shape The shape similarity represented by the dynamic time warping distance is expressed as:

[0162] D shape (n,n)=|x n -y n |+min[D shape (n-1,n),Dshape (n,n-1),D sshape (n-1,n-1)]

[0163] Where: j = 1, 2, 3, .., n, x j ,y j are the data points of two different tobacco leaf samples after spectral preprocessing in step 6.4; n is the number of spectral data points, which is 1000 here;

[0164] D Info The information divergence is the difference in the amount of information contained in the two spectra, and its expression is:

[0165]

[0166] in: Calculate the similarity of all the spectra of the stocked tobacco strips, and arrange the names of the tobacco strips from large to small according to the similarity, and obtain the tobacco strip similarity name matrix, denoted as S mm-1 :

[0167]

[0168] Where: s i1 ,s i2 ,s i3 ,…,s im-1 It indicates the names of the tobacco leaves, excluding itself, arranged in descending order of similarity to tobacco leaf i.

[0169] Step 7: Secondary Substitution Algorithm

[0170] Step 7.1: Form a new leaf group formula for L cigarettes. For the s insufficient tobacco leaves in L cigarettes, select the first p most similar tobacco leaves of each of the s tobacco leaves according to the similarity matrix established in step 6, and get p s The near infrared model established in step 3 was used to further obtain the p s The q-item quality index of the new leaf group formula (same as in step 4) is calculated, and the weighted values ​​of the M leaf groups are obtained.

[0171]

[0172] Where: Indicates L cigarette p s The kth leaf group of a new recipe has the qth quality index value of interest; w i Indicates the usage ratio of the kth leaf group.

[0173] Step 7.2: PCA dimensionality reduction. For the overall matrix U 8×q Use principal component analysis (PCA) to reduce dimension, select the first t principal components with contribution > ratio1, and get the overall dimension-reduced matrix u8×t For step 7.1 Using the same t principal components for dimensionality reduction, we get p s The dimension-reduced vector v t .

[0174] Step 7.3: Calculate the minimum Mahalanobis distance. Calculate p s v t to u 8×t The Mahalanobis distance d (secondary optimization):

[0175]

[0176] in: for u 8×t The average of the first eight quarters, for The transpose of -1 for u 8×t The inverse matrix of the covariance matrix of . The smaller the d value, the higher the similarity. The s tobacco leaf combinations with the minimum d value are the target replacement tobacco leaf combinations.

[0177] Finally, data processing was performed, and Matlab 2020b was used to write auxiliary calculation scripts according to 5.1 to 5.2 to complete the calculation process of all processes. A high-performance computer with a 96-core i913900K processor and 128GB memory on the Windows 10 platform was used for parallel computing and solving.

[0178] Embodiment three:

[0179] This embodiment takes TB2 cigarettes as an example for illustration, and the parameter selection involved is shown in Table 1.

[0180] Table 1: Parameter selection

[0181]

[0182]

[0183] Step 1: Determine the production interval and cigarette section output. Read the expected production year (Year_ph), which is set to 3 years here. Each year is divided into 4 quarters, which are divided into 12 quarterly production intervals (Section_ph). The annual planned output of each cigarette brand is divided into 12 quarters according to demand to obtain the cigarette section output (Weight_ph). The calculation starts from the first quarter in chronological order until all quarters are completed.

[0184] Step 2: Determine the order of cigarette calculation. Determine all the cigarette brands in production / to be produced by the industrial enterprise, and sort all the cigarette brands into a single production interval calculation order (Name_ph) according to the market price. Here, there are 23 cigarette brands, which are sorted from high to low according to the market price, namely TB1 to TB23 (the following description takes TB2 as an example).

[0185] Step 3: Calculate the production capacity of cigarette brands. Read the name (Name_py) and weight (Weight_py) of all stocked cigarettes, such as the name (Name_ph_py) and proportion (Per_ph_py) of the TB2 leaf group of cigarettes, search for all Name_ph_py of TB2 in Name_py, and get the corresponding current stock weight Weight_py (Name_ph_py), then calculate the production capacity Weight_ph_power according to each leaf group of TB2 cigarettes:

[0186]

[0187] The minimum capacity is Weight_ph_minpower:

[0188] Weight_ph_minpower=min(Weight_ph_power)

[0189] Here, there are 425 cigarettes in stock, and cigarette TB2 is composed of 20 cigarettes, and the proportion is 5%. If the stock weight / 5% of a cigarette X is the smallest, the minimum production capacity of TB2 is calculated based on the cigarette X.

[0190] Step 4: Compare production capacity with planned output. Read the TB2 interval output (Weight_ph), and compare Weight_ph_minpower with Weight_ph to determine whether the current inventory capacity meets the TB2 output.

[0191] Step 5: Sufficient production capacity process. If Weight_ph_minpower≥Weight_ph, the inventory of tobacco leaves is sufficient. The process is divided into three steps: A, B, and C:

[0192] ① Determine the TB2 quality index to be output (Name_index), read the quality index data of each piece of tobacco in the TB2 formula table (Index_py), and output the interval formula yield Weight_xpf, yield ratio (Ratio_xpf) and quality index (Index_xpf). At this time: Weight_xpf = Weight_ph, Ratio_xpf = 1, Index_xpf = Per_ph_py*Index_py

[0193] ② Update the weight of the cigarettes in stock Weight_py, that is, deduct the weight of each cigarette involved in Weight_ph from Weight_py. Amount_used = Per_ph_py*Weight_ph

[0194] ③ Update the cigarette output Weight_ph. At this time, Weight_ph=0. The calculation of TB2 in this interval is completed. Return to step 3 to continue the calculation of the next cigarette brand in the production interval.

[0195] Step 6: Capacity ≥ Ratio process. Under the condition of insufficient inventory of tobacco leaves, set the capacity ratio Ratio value (0≤Ratio<1). If Weight_ph_minpower≥Ratio*Weight_ph, the process is divided into four steps:

[0196] ①Also based on Name_index, Index_ph and Index_py, output Weight_xpf, Ratio_xpf and Index_xpf first. At this time:

[0197] Weight_xpf(i)=Weight_ph_power, Ratio_xpf(i)=Weight_ph_minpower / Weight_ph, Index_xpf(i)=Per_ph_py*Index_py

[0198] Among them: i = 1, 2, 3…, t, is the interval small formula number, and t is the total number of interval small formulas.

[0199] ② Update Weight_py, that is, deduct Amount_used from Weight_py.

[0200] Amount_used=per_ph_py*Weight_ph_minpower

[0201] ③ Update Weight_ph to the remaining output Weight_ph_remain. At this time:

[0202] Weight_ph_remain(i)=Weight_ph*rRatio_xpf(i)

[0203] ④ Calculate all underweight tobacco slices (denoted as s pieces), that is, all tobacco slices with 1 > Weight_ph_power ≥ Ratio * Weight_ph, and their quality indicators (Index_ph_py_lack), names of underweight tobacco slices (Name_ph_py_lack), and weights of underweight tobacco slices (Weight_ph_py_lack).

[0204] ⑤ Loop into the production capacity and planned output comparison process in Step 4. At this time, Weight_ph_remain replaces Weight_ph in the subsequent process.

[0205] Step 7: Production capacity < Ratio process. Under the condition of insufficient inventory of tobacco slices, if Weight_ph_minpower < Ratio * Weight_ph, this process is divided into two steps:

[0206] ① Calculate all underweight tobacco slices (denoted as s pieces), that is, all tobacco slices with Weight_ph_power < Ratio * Weight_ph, and their quality indicators (Index_ph_py_lack), names of underweight tobacco slices (Name_ph_py_lack), and weights of underweight tobacco slices (Weight_ph_py_lack).

[0207] ② Loop into the production capacity and planned output comparison process in Step 4. At this time, Weight_ph_remain replaces Weight_ph in the subsequent process.

[0208] Step 8: Determine the replacement rule. Based on the principle of pairwise similarity of tobacco slices, simultaneously replace the underweight tobacco slices in the TB2 leaf group formula (all-in plan), and the inventory weight of the replacement tobacco slices should be greater than Ratio of the production capacity in the cigarette brand range. Here, Ratio is taken as 0.3.

[0209] Example 4:

[0210] Combined with Example 3, this example takes TB2 cigarettes as an example to illustrate the replacement method, which specifically includes the following steps:

[0211] Step 1: Collect near-infrared spectra of tobacco leaves. Collect near-infrared spectra of tobacco leaves according to the method in Step 1 of Example 2.

[0212] Step 2: Collect tobacco leaf quality index data and unit price data. Collect 15 quality indicators (total sugar, reducing sugar, total alkaloid, chlorine, total nitrogen, potassium, pH, chlorogenic acid, elegance 1, freshness value, intermediate fragrance value, strong fragrance value, position value, starch, and elegance 2) and unit price indicators of all inventory tobacco leaves according to the method in Step 2 of Example 2.

[0213] Step 3: Construction of near infrared spectroscopy-tobacco leaf quality index prediction model. Based on the tobacco leaf near infrared spectroscopy sample data accumulated by China Tobacco Corporation over the years and the above 15 tobacco leaf quality index data, a relatively stable near infrared spectroscopy-tobacco leaf quality index prediction model was established using chemometric methods. The true value and predicted value fitting determination coefficient of the validation set samples was required to reach 0.9 (R 2 Take 0.9) or above.

[0214] Step 4: Selection of tobacco leaf focus indicators. Here, the expert group votes to select q quality indicators with a total weight > ratio2 as focus indicators (not considering price indicators), where ratio2 = 0.75. Five tobacco leaf quality experts (A, B, C, D and E) establish a comparison matrix g according to the method in step 4. ij Table 2 lists the comparison matrix of expert A.

[0215] Table 2: Comparison matrix of expert A

[0216]

[0217]

[0218] Note: z1~z15 represent 15 tobacco leaf quality indicators: total sugar, reducing sugar, total alkaloids, chlorine, total nitrogen, potassium, pH, chlorogenic acid, elegance 1, light aroma value, intermediate aroma value, strong aroma value, part value, starch and elegance 2.

[0219] Five experts’ judgment matrix h ij The maximum eigenvalues ​​λ are 15.75, 15.21, 15.35, 16.37 and 15.21 respectively, and the consistency test CR values ​​are all less than 0.1. The objective function Z is constructed by the least square method, and the auxiliary function N is constructed by the Lagrangian conditional extreme value. The weighting results of the five experts are solved, as shown in Table 3.

[0220] Table 3: Weighting results of five experts

[0221] A B C D E <![CDATA[w ahp ]]> z1 0.076 0.09 0.089 0.034 0.058 0.069 z2 0.136 0.09 0.089 0.127 0.122 0.113 z3 0.207 0.234 0.209 0.158 0.234 0.208 z4 0.015 0.015 0.015 0.03 0.015 0.018 z5 0.045 0.046 0.041 0.043 0.041 0.043 z6 0.021 0.028 0.021 0.021 0.026 0.023 z7 0.021 0.019 0.021 0.021 0.019 0.02 z8 0.015 0.015 0.015 0.015 0.015 0.015 z9 0.015 0.015 0.015 0.015 0.015 0.015 z10 0.031 0.028 0.041 0.041 0.041 0.036 z11 0.031 0.028 0.028 0.028 0.026 0.028 z12 0.031 0.028 0.028 0.028 0.026 0.028 z13 0.076 0.09 0.089 0.049 0.122 0.085 z14 0.076 0.09 0.089 0.089 0.122 0.093 z15 0.207 0.184 0.209 0.302 0.122 0.205

[0222] The correlation coefficient of the weighted results of the five experts was further calculated. When there were five experts participating in the evaluation, the LPR was 0.88. Table 4 shows that all experts met the requirements. However, considering that the mean correlation coefficient between expert D and other experts was significantly smaller, the Grubbs outlier test showed a highly significant outlier (p<0.05), so the result of expert D was excluded, and the mean weight of the four experts A, B, C and E was used as the final weight, and W was obtained. ahp, see Table 3. Finally, the six items with relatively large weights, namely, total sugar, reducing sugar, nicotine, part value, starch and elegance 2, were selected as tobacco leaf quality indicators. At this time, the total weight = 0.773> 0.75, and q = 6.

[0223] Table 4: PR of five experts

[0224] A B C D E Mean A 0.984 0.989 0.914 0.969 0.964 B 0.984 0.983 0.889 0.977 0.958 C 0.989 0.983 0.914 0.977 0.966 D 0.914 0.889 0.914 0.911 0.907 E 0.969 0.977 0.977 0.911 0.958

[0225] Step 5: Set the target quality of cigarettes. Taking the quality of TB2 in the past eight quarters as the target, the established near-infrared prediction model is used to calculate the total sugar, reducing sugar, nicotine, part value, starch and elegant 2 six quality index values ​​of 20 tobacco leaves (M = 20) in the tobacco leaf group formula table, and the weighted values ​​of the 20 tobacco leaves are calculated according to the proportion w of the Liqun Long Mouth tobacco leaf group (all 5%), and the overall matrix U is obtained. 8×6 :

[0226]

[0227] Where: U 8×6 represents the values ​​of six quality indicators of concern for the kth leaf group of Liqun Long Mouth cigarettes in eight quarters; w i Indicates the usage ratio of the kth leaf group.

[0228] Step 6: First-level substitution algorithm

[0229] Step 6.1: Determination of near infrared spectrum of stock tobacco leaves. When sampling stock tobacco leaves, ensure that they are sampled at multiple points and mixed as one sample, and then measure the spectrum according to the method in step 1.

[0230] Step 6.2: Scattering correction of near-infrared spectrum. Perform scattering correction on the near-infrared spectrum according to the method in step 6.2.

[0231] Step 6.3: Synchronous smoothing of near infrared spectrum. According to the method in step 6.3 of embodiment 2, the midpoint x of the moving window is i0 As the signal value of the new spectrum after processing, then:

[0232] x i0 =0.087X i-3 -0.266X i-2 -0.230X i-1 +0.230X i1 +0.266X i2 +0.087X i3

[0233] By changing the position of the moving window, the new spectral data x can be obtained after synchronous smoothing derivative preprocessing. ijIn order to keep the number of spectral points the same before and after processing, since the window size is 7, cubic spline interpolation is used to interpolate 3 points at the beginning and end of the spectrum before synchronous smoothing and derivation of the near-infrared spectrum.

[0234] Step 6.4: Near infrared spectrum feature selection. Select the features of the spectrum points according to the method in step 6.4 of Example 2. Set ratio4=0.05.

[0235] Step 6.5: Near infrared spectrum two-dimensional similarity algorithm. According to the method in step 6.5 of Example 2, the similarity of all the inventory strips spectra is calculated, and the strips names are arranged from large to small according to the similarity, and the strips similarity name matrix is ​​obtained, which is recorded as S mm-1 : Where: m=425.

[0236] The tobacco similarity matrix S obtained by the present invention is 425×424 Compared with several traditional near infrared pretreatment and similarity calculation methods, the average relative deviations of the sensory quality indexes and chemical quality indexes of the 10 alternative tobacco strips closest to the target tobacco strips were statistically analyzed, and the results are shown in Table 5. As can be seen from Table 5, the relative deviations of the position, elegance 2, total sugar, reducing sugar and nicotine indexes of the pretreatment (SNV+SSD+FS) of the present invention and the similarity method (2D-CORR) of the present invention are relatively small, showing obvious advantages, and the combination of the two, i.e., SNV+SSD+FS+2D-CORR, has the most obvious advantage, which shows that the near infrared spectroscopy pretreatment and similarity evaluation method established by the present invention has better effects.

[0237] Table 5: Comparison of methods

[0238]

[0239] Taking the third quarter output of TB2 as 12000Kg as an example, its leaf group formula name, leaf group ratio, inventory quantity and leaf group demand are shown in Table 6. As shown in Table 6, Liqun Long Mouth Cigarettes are composed of 20 leaf groups, and the leaf group ratio is 5%, so the required feed quantity is 600Kg. According to the capacity ratio Ratio=0.3, the inventory requirement is at least 600×0.3=180. From the current inventory quantity, it can be seen that the inventory of six types of tobacco leaves, GZ3120CFL, HA4118CCH, HN1121CEA, SC4119CEA, SC4120XEY, and YN8119CKH, is insufficient and needs to be replaced with other similar tobacco leaves.

[0240] Table 6: Liqun Longzui formula

[0241] Recipe number Tobacco leaf name Leaf group ratio w(%) Stock quantity (Kg) Leaf group feed amount (Kg) 1 CA18GLH 5 7330 600 2 GZ20CFP 5 3404 600 3 GZ20CFL 5 48 600 4 HA18CCH 5 22 600 5 HA19XCS 5 3028 600 6 HB18CCQ 5 2288 600 7 HN19XDG 5 1327 600 8 HN21CEA 5 56 600 9 HN21CEB 5 6802 600 10 SC19CDY 5 7273 600 11 SC19CEA 5 71 600 12 SC20XEY 5 92 600 13 YN19XFL 5 1120 600 14 YN19BES 5 3340 600 15 YN19CKB 5 1637 600 16 YN21CLM 5 1020 600 17 YN21CLN 5 7263 600 18 YN19CKN 5 656 600 19 YN19CKH 5 65 600 20 ZW21ILJ 5 1650 600

[0242] Note: The first two letters of the tobacco strip code represent the place of origin, CA for Canada, GZ for Guizhou, HA for Henan, HB for Hubei, HN for Hunan, SC for Sichuan, YN for Yunnan, and ZW for Zimbabwe; the numbers in the tobacco strip code represent the year of production; the fifth digit in the tobacco strip code is the part code; the 6th and 7th digits are identification codes and have no clear meaning (the same below).

[0243] Press S 425×424 The top 10 most similar tobacco leaves of GZ3120CFL, HA4118CCH, HN1121CEA, SC4119CEA, SC4120XEY, and YN8119CKH (s=6) were selected, and the results are shown in Table 7. As can be seen from Table 7, although the similarity-based screening did not specify the consistency of the replacement tobacco leaves and the original tobacco leaves in terms of position and origin, the results still showed a certain degree of origin similarity and position similarity, among which the average matching degree of origin reached 50% and the average matching degree of position reached 70%. This is because tobacco leaves are agricultural products, which are greatly affected by the ecological environment and soil conditions, and show strong regional characteristics. However, the benefit of similarity screening is that it can help tobacco formulators provide new ideas and use the matching schemes or origins that have never been used in the past into existing formulas. For example, the matching degree of the three tobacco origins of GZ20CFL, SC19CEA and SC20XEY is not very high. Among them, GZ20CFL has 6 Sichuan tobacco leaves that are highly similar to it, SC19CEA has 7 Yunnan tobacco leaves that are highly similar to it, and SC20XEY has 3 Guizhou tobacco leaves and 3 Yunnan tobacco leaves that are highly similar to it, which expands the flexibility of the formula and facilitates the better realization of the reasonable allocation of current tobacco resources. On the other hand, the Yunnan-Guizhou-Sichuan region itself is geographically adjacent, and the climate environment and soil conditions are highly similar, which results in a high degree of similarity in tobacco leaves.

[0244] Table 7: The top 10 most similar tobacco leaves for each of the six alternative tobacco leaves

[0245]

[0246]

[0247] Note: The first two letters in the tobacco code represent the place of origin, FJ for Fujian, SD for Shandong, and GX for Guangxi.

[0248] Step 7: Secondary Substitution Algorithm

[0249] Step 7.1: Form the new leaf group formula of TB2 cigarettes. For the above 6 insufficient tobacco leaves in TB2 cigarettes, select the top 10 tobacco leaves that are most similar to each of the 6 tobacco leaves according to the similarity matrix established in step 6, and get 10 6 The near infrared model established in step 3 was used to further obtain 10 6The total sugar, reducing sugar, nicotine, part value, starch and elegance 2 of the new leaf group formula were calculated, and the weighted values ​​of 20 leaf groups were obtained.

[0250]

[0251] Where: Indicates 10 of TB2 6 The kth leaf group of a new formula has 6 quality index values ​​of concern; w i Indicates the usage ratio of the kth leaf group.

[0252] Step 7.2: PCA dimensionality reduction. For the overall matrix U 8×6 Principal component analysis (PCA) was used for dimensionality reduction, and the first t principal components with contribution > 0.9 (ratio1 = 0.9) were selected. Figure 3 The relationship between the number of principal components and the contribution degree shown in the figure shows that when t = 5, the contribution degree reaches 92%, and the overall dimension reduction matrix u is obtained 8×5 .

[0253] right Using the same 6 principal components for dimensionality reduction, we get 10 6 A reduced dimensionality vector v5.

[0254] Step 7.3: Calculate the minimum Mahalanobis distance. Calculate 10 6 indivual to u 8×5 The Mahalanobis distance (secondary optimization):

[0255]

[0256] Matlab 2020b was used to write auxiliary calculation scripts according to 5.1 to 5.2 to complete the calculation process of all processes. A high-performance computer with a 96-core i913900K processor and 128GB memory on the Windows 10 platform was used for parallel computing and solution. The leaf group formula corresponding to the minimum value is the final TB2 new cigarette leaf group formula.

[0257] Table 8 shows the TB2 calculation results. Among the top five best combinations, combination 1 (SC20CEE, HA18CCK, HN21XDP, YN19CKN, GZ20CFM, YN19CKJ) was the best replacement combination found, and none of the best combinations selected their most similar tobacco leaves, among which GZ20CFL, SC19CEA and SC20XEY did not select tobacco leaves from the same origin, and HN21CEA and SC20XEY did not select tobacco leaves from the same part.

[0258] Table 8: Top 5 best combinations

[0259] Target GZ20CFL HA18CCH HN21CEA SC19CEA SC20XEY YN19CKH Combination 1 SC20CEE HA18CCK HN21XDP YN19CKN GZ20CFM YN19CKJ Combination 2 SC20CEE HA18XCQ HN21CEC YN19CKN GZ20CFM YN19CKJ Combination 3 SC20CEE HA18CCK HN21CEC YN19CKN SC19CEK YN19CKJ Combination 4 SC20CEE HA18XCQ HN21CEC YN20CKL SC20CEK YN19CKJ Combination 5 SC20CEE HA20CCH HN21CEC YN20CKL GZ20CFM YN19CKJ

[0260] Table 9 shows the differences between the 15 quality index values ​​of the original TB2 formula and the quality indexes of the five new combinations and the original formula. The results show that the differences among the five combinations are small. From the total difference results, combination 1 and combination 2 are closest to the original formula.

[0261] Table 9: Differences in quality indicators between the top five best combinations and the original formula

[0262]

[0263] Table 10 shows the results obtained through manual sensory evaluation, which shows that combination 1 and combination 2 are close to the original formula and meet the requirements of formula maintenance.

[0264] Table 10: Manual sensory evaluation results

[0265] sample Fragrance Aroma volume Irritant Aftertaste Agglomeration Miscellaneous gas Softness total Original recipe 7.3 7.6 7.4 7.3 7.5 7.9 7.5 52.5 Combination 1 7.4 7.7 7.4 7.3 7.6 7.8 7.5 52.7 Combination 2 7.5 7.6 7.3 7.5 7.6 7.7 7.6 52.8 Combination 3 7.5 7.5 7.3 7.2 7.3 7.5 7.3 51.6 Combination 4 7.3 7.3 7.2 7.4 7.5 7.5 7.3 51.5 Combination 5 7.1 7.3 7.2 7.1 7.5 7.3 7.3 50.8

[0266] The results of the small formula number of 23 cigarette brands in the entire production range are as follows Figure 4 As shown by Figure 4 From the ratio = 0.3 (allin), we can see that the number of small formulas in a single season for all brands is between 1 and 4. This is because when the capacity ratio ratio is 0.3, 1 / ratio = 3.33, that is, at most 4 small formulas can meet the demand, and only the last small formula in a single production interval may have a ratio < 0.3. When ratio = 0 (allin), the number of small formulas for most brands can still be maintained within 4, and some brands (TB10, TB12, TB13, TB18, TB19, TB20, TB21, TB23) have significantly increased at certain quarterly nodes, resulting in a large number of extremely small formulas with ratio < 0.1, such as Figure 5As shown in the figure, the reasons for the changes are mainly from two aspects: ① Some brands have more leaf group formulas, which greatly increases the probability of replacement (such as TB10 has 26 leaf groups; TB13 and TB14 both have 34 leaf groups); ② The brands and relatively later quarters have the same type of tobacco resources being competitively consumed, resulting in relatively tight remaining stocks, which increases the probability of replacement (such as TB18, TB19, TB20, TB21, TB23). These extremely small formulas themselves have no practical significance, but to a certain extent, they can show that the leaf group formula replacement schemes of these brands are more universal, and most tobacco resources can be adapted. When ratio = 0.3 (1by1 scheme, based on the principle of pairwise similarity of tobacco leaves, the insufficient tobacco leaves in the tobacco leaf group formula are replaced one by one to form the final new leaf group formula), although the number of small formulas in a single season for all brands is also between 1 and 4, the difference in the replacement mode leads to a large difference in the number of small formulas between quarters, but there are still some brands with exactly the same small formulas in certain production intervals (such as TB6 from 1 to 7 quarters); all brands end before the calculation reaches 11 quarters, and TB12-TB23 ends in the 10th quarter. This is because under the independent replacement model, more small-capacity inventory tobacco leaves with ratio < 0.3 will be generated and cannot be used, resulting in a large amount of surplus waste (the total remaining inventory capacity ratio of the two is about 1:1.6). In addition, TB22 only has 8 quarters of data, because the company's production plan plans to eliminate TB22 in the third year. In addition, under the condition of ratio = 0.3 (allin), TB7, TB9, TB17, and TB22 have only one small formulation (i.e. ratio = 1) in some quarterly ranges, indicating that the leaf groups in the formulations of these brands may have relatively large inventory or less cross-coupling, and their formulation stability is relatively high.

[0267] The calculation results of the six indicators of concern in the entire production area TB2 are as follows Figure 6As shown. Overall, the three different measurement schemes are relatively stable in the early and middle stages, but change significantly in the later stage. In the early stage (quarters 1-4), the change is manifested as ratio = 0 (all-in) < ratio = 0.3 (all-in) ≈ ratio = 0.3 (one by one). In the middle stage (quarters 5-8), the three situations intersect with each other. In the later stage (quarters 9-12), the change is ratio = 0.3 (all-in) < ratio = 0 (all-in) < ratio = 0.3 (one by one). In the early stage, with sufficient resources, the impact of the replacement mode on the measurement results is relatively small. The situations of ratio = 0.3 (all-in) and ratio = 0.3 (one by one) are similar, while for ratio = 0 (all-in), since the inventory impact is not considered, the method will preferentially select the best tobacco leaf combination, so it is relatively more stable. In the middle stage, the cross-use of tobacco leaf resources begins to increase, and different schemes gradually diverge. The change gradient of ratio = 0.3 (one by one) increases significantly. In the later stage, with tight tobacco leaf resources, for ratio = 0 (all-in), high-quality resources are consumed too early in terms of the quantity of tobacco leaves, and for ratio = 0.3 (one by one), high-quality resources are consumed too early in terms of the variety of tobacco leaves. The quality indicators of both are not as stable as ratio = 0.3 (all-in) in the later stage, indicating that the global replacement mode is significantly better than the independent replacement mode.

[0268] Based on the SPC individual value control chart (I-MR) of the time series, the control limits of the six concerned indicators can be calculated Figure 6 among the six concerned indicators: where: LCL is the lower limit (total sugar, nicotine, reducing sugar, starch, elegance 2), and UCL is the upper limit (position); is the average value of the concerned indicator u of TB2 cigarettes in the past 8 quarters; k = 2.58, which is the coefficient under 99% confidence level; is the estimated standard deviation of the population; is the average value of the moving range; d2(2) = 1.128, which is the unbiased constant parameter when the range interval is 2. The control limits of the six concerned indicators and their average value (mean) and standard deviation (SD) results are shown in 11.

[0269] Table 11: SPC control limits and chi-square test limits

[0270]

[0271]

[0272] Note: The SPC control limits and chi-square test limits represent the number of quarters that can produce normally starting from the first quarter.

[0273] The squared Mahalanobis distance follows the chi-square distribution. Based on the chi-square distribution table, the threshold of the squared Mahalanobis distance under 99% confidence level can be determined, that is, the squared Mahalanobis distance must be satisfied. in is the chi-square statistic at 99% confidence level at df degrees of freedom. Since the Mahalanobis distance is first reduced in dimension in the present invention, the degrees of freedom are not fixed. When df = 1, 2, 3, 4, 5, 6,

[0274] As shown in Table 11, the mean and standard deviation of the six indicators of concern for different brands vary greatly, with the mean ranging from 3.83 to 11.80 and the standard deviation ranging from 0.41 to 6.20. Among them, the difference between the indicators of concern for TB17 is the smallest, except for the total sugar control limit of 11, the other indicators are all 12, and the difference of TB21 is the largest, except for the reducing sugar and part of 9, the rest are 0. Figure 7 As shown in the figure, by comparing the SPC control limit mean and the chi-square test limit, it was found that the Pearson correlation coefficient between the two was 0.80 (p<0.01), which was extremely significantly correlated. However, there were large differences between the two between different brands. The differences of 6 brands, TB1, TB2, TB9, TB11, TB21 and TB23, exceeded 2, among which TB1 had the largest difference, reaching 3.33. The single sample t test showed that there were highly significant differences between the two between 6 brands, TB1, TB2, TB7, TB12, TB17 and TB19 (p<0.05); the differences of 10 brands, TB3~TB6, TB13~TB17 and TB22, were within one quarter. The paired t test difference was used and it was found that there was no significant difference between the two (p=0.70>0.05), indicating that the use of PCA-MAHAL chi-square test limit as an independent single indicator can generally characterize the control limits of the six indicators of concern, and solve the problem that the control limits of multiple quality indicators of concern cannot be unified.

[0275] The calculation results of the unit price index of 23 brands in the entire production range are as follows: Figure 8 As shown, the upper and lower control limits are calculated using similar methods as above. From the perspective of cost control, tobacco industry enterprises pursue cost reduction and only need to pay attention to the upper limit.

[0276] Depend on Figure 8 It can be seen that the estimated unit prices of the 23 brands in most quarters are within the upper control limit. From the changes in the unit prices of each brand, the relative standard deviation (RSD) is within 5%, as shown in Table 12, which is in line with the daily product cost control range of tobacco companies. If tobacco industry companies consider further reducing costs, they can add the unit price index as a focus index and a control index into the secondary similarity calculation.

[0277] Table 12: Quarterly changes in unit price indicators of 23 brands

[0278] Brand Mean Standard Deviation RSD / % Brand Mean Standard Deviation RSD / % TB1 136.99 2.06 1.51％ TB13 124.39 3.18 2.56％ TB2 135.81 3.22 2.37％ TB14 125.16 3.46 2.77％ TB3 128.79 4.28 3.32％ TB15 125.84 2.08 1.66％ TB4 123.15 2.11 1.71％ TB16 124.49 2.87 2.31％ TB5 124.32 2.80 2.25％ TB17 121.10 3.10 2.56％ TB6 125.52 3.37 2.68％ TB18 120.41 2.78 2.31％ TB7 128.14 2.98 2.33％ TB19 116.48 2.41 2.07％ TB8 122.72 2.64 2.15％ TB20 120.51 3.59 2.98％ TB9 127.99 3.18 2.49％ TB21 115.27 1.67 1.45％ TB10 129.10 2.24 1.74％ TB22 96.34 2.02 2.09％ TB11 128.03 3.48 2.72％ TB23 94.91 3.82 4.02％ TB12 123.99 2.51 2.03％

[0279] Embodiment five:

[0280] This embodiment provides a computer device, including:

[0281] A memory for storing instructions;

[0282] The processor is used to execute the instructions so that the device performs operations to implement any of the above-mentioned cigarette formula maintenance synchronous prediction calculation methods.

[0283] The cigarette formula maintenance synchronous prediction calculation method comprises:

[0284] Step 1: According to the expected production year of cigarettes, obtain the cigarette interval production in different production intervals;

[0285] Step 2: Based on the calculation rules, obtain the calculation order of different cigarette brands;

[0286] Step 3: Based on the inventory leaf information, obtain the current inventory weight of the cigarette brand blend leaf group and the production capacity of the cigarette brand;

[0287] Step 4: Based on the cigarette production range and cigarette brand production capacity, a forecast is made, including the following steps:

[0288] If the production capacity of the cigarette brand is not less than the interval production of cigarettes, the prediction data of the cigarette leaf group formula is obtained according to the quality index of each leaf group in the cigarette formula table; according to the prediction data of the cigarette leaf group formula, the inventory weight of the leaf group is updated, and the step 3 is returned to complete the prediction calculation of the next cigarette brand formula leaf group;

[0289] If the production capacity of the cigarette brand is less than the cigarette interval output, obtain the insufficient tobacco data of each leaf group, perform secondary similarity substitution on the insufficient tobacco data, obtain the new leaf group formula, and return to step 3 to complete the prediction calculation of the new leaf group formula.

[0290] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for synchronous prediction and calculation of cigarette formula maintenance, characterized in that: include: Step 1: According to the expected production year of cigarettes, obtain the cigarette interval production in different production intervals; Step 2: Based on the calculation rules, obtain the calculation order of different cigarette brands; Step 3: Based on the inventory leaf information, obtain the current inventory weight of the cigarette brand blend leaf group and the production capacity of the cigarette brand; Step 4: Based on the cigarette production range and cigarette brand production capacity, a forecast is made, including the following steps: If the production capacity of the cigarette brand is not less than the interval production of cigarettes, the prediction data of the cigarette leaf group formula is obtained according to the quality index of each leaf group in the cigarette formula table; according to the prediction data of the cigarette leaf group formula, the inventory weight of the leaf group is updated, and the step 3 is returned to complete the prediction calculation of the next cigarette brand formula leaf group; If the production capacity of the cigarette brand is less than the cigarette interval output, obtain the insufficient tobacco data of each leaf group, perform secondary similarity substitution on the insufficient tobacco data, obtain the new leaf group formula, and return to step 3 to complete the prediction calculation of the new leaf group formula.

2. The method according to claim 1, characterized in that: The secondary similar substitutions include: Based on the near infrared spectrum-tobacco leaf quality index prediction model, the weighted values ​​of q quality indexes of cigarettes in the past several quarters are calculated to obtain the overall matrix; According to the tobacco leaf similarity evaluation of the current tobacco leaves in stock, a similarity matrix of the tobacco leaves in stock is obtained; By identifying s tobacco leaves to be replaced in a cigarette blend containing M tobacco leaf groups, p tobacco leaves that are most similar to the tobacco leaves to be replaced are selected from the similarity matrix to obtain p s A new leaf group formula of the cigarette; Based on p s The M leaf group weighted values ​​of the q quality indicators of the new leaf group formula are obtained by s weighted values; For the overall matrix, p s The weighted values ​​are reduced by principal component analysis to obtain the overall dimension reduction matrix and p s A reduced dimensionality vector; According to the calculation p s The Mahalanobis distance from the reduced dimension vector to the overall reduced dimension matrix is ​​obtained, and the new leaf group formula corresponding to the minimum Mahalanobis distance is the cigarette formula of similar substitution.

3. The method according to claim 2, characterized in that Based on the tobacco leaf similarity evaluation of the current stock tobacco leaves, the similarity matrix of the stock tobacco leaves is obtained, including: Multi-point sampling to collect near-infrared spectral data of current tobacco leaves in stock; The near-infrared spectral data of the current stock tobacco leaves are subjected to scattering correction, synchronous smoothing derivation, feature selection and two-dimensional similarity calculation in sequence to obtain the similarity matrix of the stock tobacco leaves.

4. The method according to claim 3, characterized in that Calculating two-dimensional similarity involves: Based on the shape differences and information content differences in the spectral data, the total similarity of the spectral data of any two tobacco leaves after feature selection is calculated; The calculation formula for the total similarity is: Where: S represents the total similarity of the spectrum, D shape represents the shape similarity represented by the dynamic time warping distance; D Info It represents the difference in the amount of information contained in two spectra characterized by information divergence; Where: j = 1, 2, 3, .., n, x j ,y j are two different tobacco leaf sample data points after synchronous smoothing derivation of near-infrared spectra; n is the number of spectral data points; 5. The method according to claim 2, characterized in that: The q quality indicators include: If the cigarette is a single recipe: select q quality indicators based on the style characteristics of the cigarette brand and the actual recipe; If the cigarette is not a single recipe: according to the principle of feature selection, use relative standard deviation to select q quality indicators; or set at least five members to vote and select q quality indicators based on the total weight.

6. The method according to claim 5, characterized in that At least five members are set to vote and select q quality indicators based on the total weight, including: Each member makes a primary and secondary judgment on any two chemical indicators to obtain a comparison matrix; Convert the comparison matrix into a judgment matrix. The objective function is constructed by the least square method, the auxiliary function is constructed by the Lagrange conditional extreme value, and the weight is obtained by solving the judgment matrix; The obtained weights are averaged as the total weight, and q chemical indicators that are greater than the total weight are selected as quality indicators.

7. The method according to claim 6, characterized in that The method further comprises: Calculate the correlation coefficient of the weights; The quality index selected was eliminated by eliminating members whose correlation coefficient was lower than the critical correlation coefficient at 95% confidence level.

8. The method according to claim 1, characterized in that Based on the inventory information of cigarette leaves, the current inventory weight of all leaf groups in the cigarette brand and the production capacity of the cigarette brand are obtained, including: Based on the inventory leaf information, obtain the current inventory weight of all leaf groups in the cigarette brand; According to the current inventory weight and the proportion of leaf groups in the cigarette brand, the production capacity of all leaf groups is obtained; among them, the smallest leaf group production capacity is the production capacity of the cigarette brand.

9. The method according to claim 1, characterized in that: The cigarette prediction data includes interval formulation output, interval output ratio and interval quality index; the insufficient tobacco leaf data includes: insufficient tobacco leaf name, insufficient tobacco leaf weight and insufficient tobacco leaf quality index; the calculation rules include market price, interval output and number of formulation leaf group types.

10. A computer device, characterized in that: include: A memory for storing instructions; The processor is used to execute the instructions so that the device performs the operations of implementing the cigarette formula maintenance synchronous prediction calculation method as described in any one of claims 1-9.

Images