CN112016046A - Method and device for constructing tobacco shred structure prediction model - Google Patents
Method and device for constructing tobacco shred structure prediction model Download PDFInfo
- Publication number
- CN112016046A CN112016046A CN202010710238.9A CN202010710238A CN112016046A CN 112016046 A CN112016046 A CN 112016046A CN 202010710238 A CN202010710238 A CN 202010710238A CN 112016046 A CN112016046 A CN 112016046A
- Authority
- CN
- China
- Prior art keywords
- tobacco
- tobacco shred
- leaf
- flake
- score
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 241000208125 Nicotiana Species 0.000 title claims abstract description 455
- 235000002637 Nicotiana tabacum Nutrition 0.000 title claims abstract description 455
- 238000000034 method Methods 0.000 title claims abstract description 57
- 238000004364 calculation method Methods 0.000 claims abstract description 59
- 238000000611 regression analysis Methods 0.000 claims abstract description 51
- 239000011159 matrix material Substances 0.000 claims abstract description 36
- 238000000513 principal component analysis Methods 0.000 claims abstract description 18
- 239000012634 fragment Substances 0.000 claims description 51
- 238000001514 detection method Methods 0.000 claims description 31
- 238000004458 analytical method Methods 0.000 claims description 24
- 239000000843 powder Substances 0.000 claims description 23
- 238000010219 correlation analysis Methods 0.000 claims description 22
- 230000000717 retained effect Effects 0.000 claims description 12
- 238000010276 construction Methods 0.000 claims description 5
- 230000001186 cumulative effect Effects 0.000 abstract description 5
- 230000000694 effects Effects 0.000 description 24
- 238000000540 analysis of variance Methods 0.000 description 18
- 238000012847 principal component analysis method Methods 0.000 description 7
- 238000012417 linear regression Methods 0.000 description 6
- 238000000547 structure data Methods 0.000 description 6
- 238000012795 verification Methods 0.000 description 6
- 230000006872 improvement Effects 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000005520 cutting process Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 3
- 239000000463 material Substances 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 235000019504 cigarettes Nutrition 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 239000011148 porous material Substances 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 239000000779 smoke Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 238000009423 ventilation Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The application discloses a method and a device for constructing a tobacco shred structure prediction model, which are used for detecting the leaf structure of each acquired tobacco flake sample and the cut tobacco shred structure; carrying out principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the cumulative variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components; constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component; performing stepwise multiple regression analysis on the scores of the main components of the targets and the tobacco shred structures, and outputting regression models of the scores of the main components of the targets and the tobacco shred structures; the tobacco shred structure prediction model of the leaf structure is obtained by combining the score calculation model and the regression model, and the technical problems that the tobacco shred structure of the predicted leaf structure of the conventional tobacco shred structure prediction method is greatly different from the actual tobacco shred structure of the leaf, and the prediction precision is low are solved.
Description
Technical Field
The application relates to the technical field of tobacco processing, in particular to a method and a device for constructing a tobacco shred structure prediction model.
Background
The structure of the cut tobacco, namely the distribution of the size of the cut tobacco, is a core index which influences the physical characteristics of single cigarette weight, density, suction resistance, total ventilation rate, hardness and the like of a rolled product, combustibility and smoke release behavior. The leaf structure determines the distribution of the cut tobacco size after shredding, and the determined relation is established between the cut tobacco structure and the leaf structure, so that the requirement for determining the cut tobacco structure can be converted into the requirement for determining the cut tobacco structure, the leaf structure after threshing and redrying is guided, and the cigarette rolling quality can be improved. The tobacco shred structure of the predicted blade structure of the existing tobacco shred structure prediction method is greatly different from the actual tobacco shred structure of the blade, and the problem of low prediction precision exists.
Disclosure of Invention
The application provides a method and a device for constructing a tobacco shred structure prediction model, which are used for solving the technical problems that the tobacco shred structure of a predicted blade structure of the conventional tobacco shred structure prediction method is greatly different from the actual tobacco shred structure of a blade, and the prediction precision is not high.
In view of this, the first aspect of the present application provides a method for constructing a tobacco shred structure prediction model, including:
acquiring a plurality of tobacco flake samples, and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample;
performing principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components;
constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model;
performing stepwise multiple regression analysis on the scores of the target principal components and the tobacco shred structure, and outputting regression models of the scores of the target principal components and the tobacco shred structure;
and combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
Optionally, the detecting the leaf structure of each of the smoked sheet samples includes:
and detecting the numerical value of each leaf structure index in each tobacco flake sample to obtain the leaf structure of each tobacco flake sample, wherein the leaf structure indexes comprise a large and medium flake rate, a large flake rate, a medium flake rate, a small flake rate, a fragment rate, a powder rate and a small powder rate.
Optionally, detecting the cut tobacco structure of each cut tobacco sample comprises:
and detecting the proportion of each tobacco shred structure index in the tobacco shred samples obtained by shredding the tobacco flake samples to obtain the tobacco shred structures of the tobacco flake samples.
Optionally, the obtaining a plurality of tobacco flake samples, detecting the leaf structure of each tobacco flake sample and the shredded tobacco structure, and then further comprising:
performing correlation analysis on the tobacco shred structure indexes and the leaf structure indexes in each tobacco flake sample based on a correlation analysis method of a binary distance variable to obtain correlation coefficients of each tobacco shred structure index and each leaf structure index;
when the correlation coefficient of the leaf structural index and any tobacco shred structural index is within a first preset range, the tobacco shred structural index and the leaf structural index in the tobacco flake sample are reserved;
and when the correlation coefficients of the leaf structural indexes and all the tobacco shred structural indexes are not in the first preset range, removing the leaf structural indexes.
Optionally, when the correlation coefficients of the leaf structure index and all the tobacco shred structure indexes are not within the first preset range, removing the leaf structure index, and then further including:
when detecting that a detection index and a characteristic index exist in the reserved blade structure indexes at the same time, preferentially reserving the detection index;
the detection indexes comprise a large fragment rate, a medium fragment rate, a small fragment rate, a fragment rate and a powder rate, and the characteristic indexes comprise a large fragment rate, a medium fragment rate and a small powder rate.
Optionally, when detecting that a detection index and a feature index coexist in the retained blade structure index, preferentially retaining the detection index, and then further including:
performing correlation analysis on the retained blade structure indexes by using a correlation analysis method based on a binary distance variable to obtain correlation coefficients between every two blade structure indexes;
and when the correlation coefficient of the blade structure indexes between every two is not in a second preset range, removing the tobacco flake sample corresponding to the correlation coefficient which is not in the second preset range.
Optionally, the combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure, and then further includes:
and detecting the leaf structure of the obtained tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into the tobacco shred structure prediction model to predict the tobacco shred structure, and outputting the tobacco shred structure of the tobacco flakes to be detected.
The second aspect of the present application provides a device for constructing a tobacco shred structure prediction model, including:
the acquisition unit is used for acquiring a plurality of tobacco flake samples and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample;
the first analysis unit is used for carrying out principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components;
the building unit is used for building a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model;
the second analysis unit is used for performing stepwise multiple regression analysis on the scores of the target main components and the tobacco shred structures and outputting regression models of the scores of the target main components and the tobacco shred structures;
and the combining unit is used for combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
Optionally, the obtaining unit specifically includes:
an acquisition subunit for acquiring a number of smoked sheet samples;
the first detection subunit is used for detecting each leaf structure index numerical value in each tobacco flake sample to obtain a leaf structure of each tobacco flake sample, wherein the leaf structure index comprises a large and medium flake rate, a large flake rate, a medium flake rate, a small flake rate, a fragment rate and a small fragment rate;
and the second detection subunit is used for detecting the proportion of each tobacco shred structure index in the tobacco shred samples obtained by shredding the tobacco flake samples to obtain the tobacco shred structures of the tobacco flake samples.
Optionally, the method further includes:
and the prediction unit is used for detecting the leaf structure of the acquired tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into the tobacco shred structure prediction model for tobacco shred structure prediction, and outputting the tobacco shred structure of the tobacco flakes to be detected.
According to the technical scheme, the method has the following advantages:
the application provides a method for constructing a tobacco shred structure prediction model, which comprises the following steps: acquiring a plurality of tobacco flake samples, and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample; carrying out principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the cumulative variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components; constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model; performing stepwise multiple regression analysis on the scores of the main components of the targets and the tobacco shred structures, and outputting regression models of the scores of the main components of the targets and the tobacco shred structures; and combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
According to the construction method of the tobacco shred structure prediction model, the leaf structure and the tobacco shred structure of the obtained tobacco flake sample are detected, and in order to avoid the problem of multiple collinearity among leaf structures, the leaf structures are subjected to principal component analysis; in order to ensure the accuracy of the tobacco shred structure prediction model, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95 percent as target principal components, and outputting a score coefficient matrix of the target principal components; the method comprises the steps of constructing a score calculation model of each target main component and each leaf structure based on a score coefficient matrix, simultaneously calculating to obtain scores of each target main component, further carrying out stepwise multiple regression analysis on the scores of each target main component and the tobacco shred structures to obtain regression models of the scores of the target main components and the tobacco shred structures, and combining the score calculation models and the regression models to obtain a tobacco shred structure prediction model of the leaf structures.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present application, and it is obvious for those skilled in the art that other drawings can be obtained according to the drawings without inventive exercise.
Fig. 1 is a schematic flow chart of a method for constructing a tobacco shred structure prediction model according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a device for constructing a tobacco shred structure prediction model according to an embodiment of the present application.
Detailed Description
The application provides a method and a device for constructing a tobacco shred structure prediction model, which are used for solving the technical problem of low prediction precision of the conventional tobacco shred structure prediction method.
In order to make the technical solutions of the present application better understood, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application, and it is obvious that the described embodiments are only a part of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
For easy understanding, referring to fig. 1, an embodiment of a method for constructing a tobacco shred structure prediction model provided by the present application includes:
The tobacco flake samples preferably adopt single-material tobacco flakes of different years, different producing areas, different parts, different grades and different leaf structures and formula tobacco flakes of different combinations thereof. Then, the leaf structure and the shredded tobacco structure of each tobacco flake sample are detected.
102, carrying out principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components.
In order to avoid the problem of multiple collinearity among leaf structures, in the embodiment of the present application, a principal component analysis method is adopted to analyze the leaf structures of all tobacco flake samples, and extract principal components, in order to ensure the accuracy of the established tobacco shred structure prediction model, a preset number of principal components with an accumulated variance contribution rate greater than or equal to 95% are selected as target principal components, and usually, the accumulated variance contribution rate of 3 principal components is greater than or equal to 95%, so that in the embodiment of the present application, 3 principal components with an accumulated variance contribution rate greater than or equal to 95% are preferably selected as target principal components. By extracting 3 principal components from the leaf structure, a scoring coefficient matrix of the 3 principal components is obtained.
And 103, constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model.
And constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, thereby establishing a relation model of the target principal component score and the blade structure, and calculating the score of each target principal component through the score calculation model. The score calculation model and the scores of the target principal components can be obtained by inputting the score matrix and the blade structure data into SPSS software and calculating through the SPSS software.
And 104, performing stepwise multiple regression analysis on the scores of the main components of the targets and the tobacco shred structures, and outputting regression models of the scores of the main components of the targets and the tobacco shred structures.
The scores of the 3 target main components and the tobacco shred structure data can be input into SPSS software, the scores of the 3 target main components and the tobacco shred structure are subjected to stepwise multiple regression analysis through the SPSS software, and a regression model of the scores of the target main components and the tobacco shred structure with remarkable regression effect is output, so that a relation model between the tobacco shred structure and the scores of the target main components is established.
And 105, combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
Through the steps, a relation model (score calculation model) between the score of the target main component and the leaf structure and a relation model (regression model) between the tobacco shred structure and the score of the target main component can be established, so that the relation model between the tobacco shred structure and the leaf structure is indirectly established, and the tobacco shred structure prediction model of the leaf structure can be obtained by combining the two models.
According to the construction method of the tobacco shred structure prediction model in the embodiment of the application, the leaf structure and the tobacco shred structure of the obtained tobacco flake sample are detected, and in order to avoid the problem of multiple collinearity among leaf structures, the leaf structure is subjected to principal component analysis; in order to ensure the accuracy of the tobacco shred structure prediction model, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95 percent as target principal components, and outputting a score coefficient matrix of the target principal components; the method comprises the steps of constructing a score calculation model of each target main component and each leaf structure based on a score coefficient matrix, simultaneously calculating to obtain scores of each target main component, further carrying out stepwise multiple regression analysis on the scores of each target main component and the tobacco shred structures to obtain regression models of the scores of the target main components and the tobacco shred structures, and combining the score calculation models and the regression models to obtain a tobacco shred structure prediction model of the leaf structures.
The above is an embodiment of the method for constructing the tobacco shred structure prediction model provided by the present application, and the following is another embodiment of the method for constructing the tobacco shred structure prediction model provided by the present application.
Another embodiment of a method for constructing a tobacco shred structure prediction model provided by the present application includes:
step 201, obtaining a plurality of tobacco flake samples, and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample.
The tobacco flake samples are preferably prepared from single-material tobacco and different combinations of single-material tobacco in different years, different producing areas, different parts, different grades and different leaf structures, and the specific leaf sample information refers to table 1.
TABLE 1 smoked sheet sample information
Detecting the index value of each leaf structure in each tobacco flake sample, for example, detecting the large and medium flake rate, the large flake rate, the medium flake rate, and the like of each tobacco flake sample to obtain the leaf structure of each tobacco flake sample, wherein detecting the proportion of each leaf structure index in each tobacco flake sample belongs to the prior art, and detailed descriptions of the specific process are omitted here. The structural indexes of the blades comprise large and medium flake rates, large flake rates, medium flake rates, small flake rates, fragment rates and small fragment rates.
In the embodiment of the application, 3 tobacco flakes are respectively selected from 4 producing areas in table 1 to serve as tobacco flake samples, the tobacco flake samples are subjected to leaf structure detection by adopting a loose and moisture regained image method, and the detection results are shown in table 2.
TABLE 2 blade Structure test results
After the leaf structure is detected and recorded, the tobacco flake sample is cut into threads, tobacco flakes with single components can be selected to be cut into threads, the cutting width can be 1.0mm, tobacco flake samples can be obtained, the occupied proportion of each tobacco flake structural index in the tobacco flake samples is detected, the tobacco flake structure of each tobacco flake sample is obtained, the tobacco flake structural index can be the size of each layer of tobacco shreds, each layer of tobacco shreds are obtained through screening of sieve pores with different sizes, the occupied proportion of each layer of tobacco shreds in the tobacco shreds is further obtained through calculation, wherein the sieve pores can be specifically set according to actual conditions, and specific limitation is not performed. In the embodiment of the application, the cut tobacco structure detection is performed on cut tobacco samples obtained by cutting tobacco flake samples of 4 production places according to the method specified in YC/T178-.
TABLE 3 tobacco shred Structure test results
Step 202, performing correlation analysis on the tobacco shred structure indexes and the leaf structure indexes in each tobacco flake sample to obtain correlation coefficients of the tobacco shred structure indexes and the leaf structure indexes.
In the embodiment of the application, a correlation analysis method of binary distance variables is adopted to perform correlation analysis on the tobacco shred structure indexes and the leaf structure indexes in each tobacco flake sample to obtain correlation coefficients of each tobacco shred structure index and each leaf structure index in each tobacco flake sample, wherein the correlation coefficients comprise Pearson correlation coefficients and significance P values (Sig values). In the embodiment of the present application, correlation analysis is performed on each index in the leaf structure of table 2 and the tobacco shred structure of table 3, and the obtained correlation result refers to table 4.
TABLE 4 correlation between leaf structure and tobacco shred structure
Step 203, when the correlation coefficient of the leaf structural index and any tobacco shred structural index is within a first preset range, keeping the tobacco shred structural index and the leaf structural index in the tobacco flake sample; and when the correlation coefficients of the leaf structural indexes and all the tobacco shred structural indexes are not in a first preset range, removing the leaf structural indexes.
When the absolute value of the Pearson correlation coefficient of the leaf structural index and any tobacco shred structural index is more than or equal to 0.500, or the P value of the leaf structural index and any tobacco shred structural index is less than or equal to 0.01, the correlation between the tobacco shred structural index and the leaf structural index is obvious, and the tobacco shred structural index and the leaf structural index in the tobacco flake sample are reserved; when the absolute value of the Pearson correlation coefficient of the leaf structure index and all the tobacco shred structure indexes is less than 0.500 or the P value of the leaf structure index and all the tobacco shred structure indexes is greater than 0.01, the correlation between the tobacco shred structure indexes and the leaf structure indexes is not obvious, and the leaf structure indexes corresponding to the Pearson correlation coefficient absolute value of less than 0.500 or the P value of more than 0.01 are removed. As can be seen from Table 4, the correlation between the median fraction and each index in the tobacco shred structure is not significant, so that the median fraction in the leaf structure is removed, and the remaining leaf structure indexes are reserved for subsequent analysis.
And step 204, when detecting that the detection indexes and the characteristic indexes simultaneously exist in the reserved blade structure indexes, preferentially reserving the detection indexes.
And detecting whether a detection index and a characteristic index exist in the leaf structure indexes retained in each tobacco flake sample at the same time, if so, preferentially retaining the detection index for avoiding data repeated operation, otherwise, not performing any treatment, and retaining the existing leaf structure indexes. The detection indexes comprise a large fragment rate, a medium fragment rate, a small fragment rate, a fragment rate and a powder rate, and the characteristic indexes comprise a large fragment rate, a medium fragment rate and a small powder rate (small fragment rate + powder rate). Referring to table 4, the detection indexes (large fragment rate, small fragment rate, and powder rate) and the feature indexes (large medium fragment rate and small powder rate) exist at the same time, the small fragment rate, the fragment rate, and the powder rate in the detection indexes are preferentially retained, and the small powder rate is removed.
And 205, carrying out correlation analysis on the leaf structure indexes reserved in each tobacco flake sample to obtain a correlation coefficient between every two leaf structure indexes.
The general leaf structure indexes have strong correlation, and the relation between the leaf structure indexes and the tobacco shred structure is established directly, so that the multiple collinearity problem exists. The precondition for constructing the tobacco shred structure prediction model by adopting the principal component analysis method is that the leaf structures have strong correlation, and if the principal component analysis method is directly adopted to extract the principal components of the leaf structures to construct the tobacco shred structure prediction model, the principal components among the leaf structures with weak correlation are probably extracted to construct the tobacco shred structure prediction model with low accuracy.
Therefore, in the embodiment of the application, before the main components are extracted, the fitting degree of the blade structure is checked. Specifically, a binary distance variable correlation analysis method is adopted to perform correlation analysis on every two leaf structure indexes retained in each tobacco flake sample to obtain a correlation coefficient between every two leaf structure indexes, wherein the correlation coefficient comprises a Pearson correlation coefficient and a significance P value.
And step 206, when the correlation coefficient of the leaf structure indexes is not in the second preset range, removing the tobacco flake sample corresponding to the correlation coefficient not in the second preset range.
When the absolute value of the Pearson correlation coefficient between every two leaf structure indexes in the tobacco flake sample is more than or equal to 0.500 or the P value is less than or equal to 0.01, the correlation between the leaf structures of the tobacco flake sample is stronger, and the method is suitable for extracting principal components by adopting a principal component analysis method; when the absolute value of the Pearson correlation coefficient between every two leaf structure indexes in the tobacco flake sample is less than 0.500 or the P value is more than 0.01, the tobacco flake sample is not suitable for the main component analysis method, and the tobacco flake sample is removed. In the embodiment of the application, the structural indexes of the leaves reserved for a certain tobacco flake sample are subjected to related analysis, and the analysis result is shown in table 5.
TABLE 5 correlation between vane Structure indices
As can be seen from table 5, there is a very significant correlation between the leaf structure indexes, so that the method is suitable for extracting principal components by using a principal component analysis method to avoid the multiple collinearity problem between the leaf structures.
And step 207, performing principal component analysis on the leaf structures reserved by the tobacco flake samples, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components.
In order to avoid the problem of multiple collinearity among leaf structures, the leaf structures of all tobacco flake samples are analyzed by adopting a principal component analysis method in the embodiment of the application, principal components are extracted, and in order to ensure the accuracy of the established tobacco flake structure prediction model, a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% are selected as target principal components. In the embodiment of the present application, principal component analysis is performed on the blade structure, principal components are extracted, the variance contribution rate of each principal component refers to table 6, and the cumulative variance contribution rate of 3 principal components obtained by extraction is > 95%, so that 3 principal components with the cumulative variance contribution rate greater than or equal to 95% are selected as target principal components in the embodiment of the present application, that is, 3 new variables are constructed to replace the blade structure. Tables 7 and 8 are component matrices of the extracted 3 principal components.
TABLE 6 variance contribution ratio of principal component
TABLE 7 composition matrix
TABLE 8 rotated composition matrix
By extracting principal components from the blade structure, score coefficient matrices of 3 target principal components are obtained, and the principal components can be extracted by SPSS software, and the score coefficient matrices of the principal components are output, please refer to table 9.
TABLE 9 scoring coefficient matrix for principal components
And 208, constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model.
And constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, thereby establishing a relation model of the target principal component score and the blade structure, and calculating the score of each target principal component through the score calculation model. The score calculation model and the scores of the target principal components can be obtained by inputting the score matrix and the blade structure data into SPSS software and calculating through the SPSS software. Specifically, the score calculation model of each target principal component is as follows:
wherein, YjCalculating a model for the score of the jth target principal component, j being 1, 2, 3; alpha is alphaijThe score coefficient of the jth target principal component of the blade structure index i, n is the number of the reserved blade structure indexes, XiIs the structural index i of the blade,
the mean value of all leaf structure indexes i in all tobacco flake samples,
the standard deviation of all leaf structure index i in all tobacco flake samples.
Continuing with the foregoing example, the 12 large and medium film rates X were calculated from the 12 tobacco sample samples of the 4 production areas1Mean value of
77.57 standard deviation
Is 6.37; chip rate X2Mean value of
16.78, standard deviation
Is 3.92; fraction X3Mean value of
Is 5.62, standard deviation
Is 2.67; fraction X4Mean value of
0.025, standard deviation
And was 0.012.
From the above parameters, in combination with the score coefficients of table 9, a score calculation model of the first target principal component can be obtained as follows:
Y1＝-0.213*(X1-77.57)/6.37-0.668*(X2-16.78)/3.92+1.495*(X3-5.62)/2.67-0.295*(X4-0.025)/0.012；
the score calculation model of the second target principal component is:
Y2＝-0.338*(X1-77.57)/6.37-0.365*(X2-16.78)/3.92-0.275*(X3-5.62)/2.67+1.727*(X4-0.025)/0.012；
the score calculation model of the third target principal component is:
Y3＝-0.978*(X1-77.57)/6.37+2.695*(X2-16.78)/3.92-1.630*(X3-5.62)/2.67-1.887*(X4-0.025)/0.012。
the large and medium leaf rate, the small leaf rate, the fragment rate and the powder rate in the leaf structure indexes of 12 tobacco leaf samples of the 4 producing areas are correspondingly input into each score calculation model, and the score of each target principal component is correspondingly obtained, please refer to table 10.
TABLE 10 score of target principal Components
And 209, performing stepwise multiple regression analysis on the scores of the main components of the targets and the tobacco shred structures, and outputting regression models of the scores of the main components of the targets and the tobacco shred structures.
The scores of the 3 target main components and the tobacco shred structure data can be input into SPSS software, the scores of the 3 target main components and the tobacco shred structure are subjected to stepwise multiple regression analysis through the SPSS software, the scores of the target main components with remarkable regression effect and a regression model of the tobacco shred structure are output, and therefore a relation model between the tobacco shred structure and the scores of the target main components is established, and the regression model is a linear regression model consisting of coefficients and constant coefficients of the target main components of the model with the best regression fitting effect.
According to the scores of the target principal components in table 10, stepwise multiple regression analysis was performed on the scores of the 3 target principal components and each layer of the tobacco shred structure, and the results of the regression analysis are shown in tables 11 to 22.
(1) Regression model of scores of first layer tobacco shred structure and target principal component
As can be seen from table 11, the correlation coefficient between the score of the first target principal component and the first layer tobacco shred structure is 0.918, and the determination coefficient is 0.828, which indicates that the score of the first target principal component and the analysis result of the regression analysis (model 1) of the first layer tobacco shred structure are significant; the correlation coefficient of the score of the second target main component and the first layer tobacco shred structure is 0.972, the determination coefficient is 0.944, and the regression effect of the score of the second target main component and the regression analysis (model 2) of the first layer tobacco shred structure is better than that of the model 1. Since the scores of the first, second and third target principal components and the analysis result of the tobacco shred structure regression analysis (model 3) are not significant, i.e. the P value is not significant, i.e. the model 3 is the same as the model 2, the model 3 is excluded. In the models 2 and 3, the fitting effect of the model 2 is better, so that the coefficient of the model 2 is used for establishing a final regression model.
As can be seen from the analysis of variance results in table 12, the P value of the linear regression reaches a very significant level, indicating that the fitting effect of the score of the first target principal component, the score of the second target principal component, and the first-layer tobacco shred structure is good. As can be seen from the regression model parameters in Table 13, the first target principal component score for model 2 has a coefficient of-7.733, the second target principal component score has a coefficient of-2.677, and the constant coefficient is 59.526. Therefore, the regression model of the scores of the first layer of tobacco shred structure and the target principal component is as follows:
Z1＝-7.733Y1-2.677Y2+59.526；
TABLE 11 regression analysis results of scores of target principal components and first layer tobacco shred structures
TABLE 12 score of target principal component and analysis of variance of regression analysis of first layer tobacco shred Structure
TABLE 13 score of target principal component and regression model parameters for first layer tobacco shred structure
(2) Regression model of scores of second-layer tobacco shred structure and target principal component
The score of the target principal component and the structure of the second layer of cut tobacco are subjected to stepwise regression analysis, the analysis result is shown in table 14, the correlation coefficient of the model 2 is 0.963, the determination coefficient is 0.926, and the regression effect is better than that of the model 1, so that the regression model is constructed by using the correlation coefficient of the model 2.
As can be seen from the analysis of variance results in table 15, the P value of the linear regression reached a very significant level, indicating that the fitting effect of the score of the first target principal component, the score of the second target principal component, and the second layer tobacco shred structure was good. As can be seen from the parameters of the regression model in table 16, the linear fitting effect of the score of the first target principal component and the score of the second target principal component is good, and therefore, the regression model of the score of the second layer tobacco shred structure and the target principal component is:
Z2＝4.706*Y1+1.200*Y2+21.918
TABLE 14 score of target principal component and regression analysis result of second layer tobacco shred structure
| Model (model) | R | R side | Adjusted R square | Error of standard estimation |
| 1 | 0.933a | 0.870 | 0.857 | 1.90861 |
| 2 | 0.963b | 0.926 | 0.910 | 1.51216 |
TABLE 15 score of target principal component and analysis of variance of regression analysis of second layer tobacco shred structure
TABLE 16 score of target principal component and regression model parameters for second layer tobacco shred structure
(3) Regression model of scores of third-layer tobacco shred structure and target main component
The score of the target principal component and the third-layer tobacco shred structure are subjected to stepwise regression analysis, the analysis result is shown in table 17, the correlation coefficient of the model 2 is 0.961, the determination coefficient is 0.924, and the regression effect is better than that of the model 1. As can be seen from the analysis of variance results in table 18, the P value of the linear regression reaches a very significant level, indicating that the fitting effect of the score of the first target principal component, the score of the second target principal component, and the third layer tobacco shred structure is good. As can be seen from the parameters of the regression model in table 19, the linear fitting effect of the score of the first target principal component and the score of the second target principal component is good, and therefore, the regression model of the score of the third layer tobacco shred structure and the target principal component is:
Z3＝2.867*Y1+1.368*Y2+17.716；
TABLE 17 regression analysis results of scores of target principal components and third-layer tobacco shred structures
| Model (model) | R | R side | Adjusted R square | Error of standard estimation |
| 1 | 0.868a | 0.753 | 0.728 | 1.72388 |
| 2 | 0.961b | 0.924 | 0.907 | 1.00783 |
TABLE 18 score of target principal component and analysis of variance of regression analysis of third cut tobacco structure
TABLE 19 score of target principal component and regression model parameters for third layer tobacco shred structure
(4) Regression model of scores of fourth-layer tobacco shred structure and target principal component
The score of the target principal component and the structure of the fourth layer of cut tobacco are subjected to stepwise regression analysis, the analysis result is shown in table 20, the correlation coefficient of the model 1 is 0.702, the determination coefficient is 0.493, and the regression effect is good. According to the analysis of variance results in table 21, the P value of the linear regression approaches the most significant level, indicating that the score of the first target principal component and the fitting effect of the fourth layer tobacco shred structure are good. As can be seen from the parameters of the regression model in table 22, the linear fitting effect of the score of the first target principal component is still good, and therefore, the regression model of the score of the fourth layer tobacco shred structure and the target principal component is:
Z4＝0.085*Y1+0.768；
TABLE 20 score of target principal Components and regression analysis results of tobacco shred Structure of fourth layer
| Model (model) | R | R side | Adjusted R square | Error of standard estimation |
| 1 | 0.702a | 0.493 | 0.442 | 0.08991 |
TABLE 21 score of target principal component and analysis of variance of regression analysis of tobacco shred Structure in fourth layer
Table 22 shows regression models of scores of main components and fourth-layer tobacco shred structures
And step 210, combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
Through the steps, a relation model (score calculation model) between the score of the target main component and the leaf structure and a relation model (regression model) between the tobacco shred structure and the score of the target main component can be established, so that the relation model between the tobacco shred structure and the leaf structure is indirectly established, and the tobacco shred structure prediction model of the leaf structure can be obtained by combining the two models, namely, the score calculation model is substituted into the regression model to obtain the tobacco shred structure prediction model of the leaf structure.
In the embodiment of the application, the tobacco shred structure prediction model of the obtained leaf structure is verified, two groups of tobacco shreds in Sichuan and Hunan are respectively selected as verification samples, the leaf structure of the verification samples is obtained by adopting the same leaf structure and tobacco shred structure detection method, the leaf structure of the verification samples is shown in a table 23, the true value of the tobacco shred structure and the predicted value predicted by the tobacco shred structure prediction model are shown in a table 24, and the absolute deviation range of the predicted value and the true value is 0.19-1.184% as can be known from the table 24, so that the tobacco shred structure prediction model in the embodiment of the application has high prediction accuracy.
TABLE 23 leaf Structure of Sichuan and Hunan tabacco
TABLE 24 verification results of tobacco shred structure prediction model based on leaf structure
And step 211, detecting the leaf structure of the obtained tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into a tobacco shred structure prediction model for tobacco shred structure prediction, and outputting the tobacco shred structure of the tobacco flakes to be detected.
The tobacco flake structure detection method comprises the steps of obtaining tobacco flakes needing to be detected as leaf samples to be detected, detecting leaf structures of the tobacco flakes, obtaining leaf structure data, inputting the leaf structure data into a tobacco shred structure prediction model to predict the tobacco shred structures, and outputting the tobacco shred structures of the tobacco flakes to be detected.
The above is an embodiment of the method for constructing the tobacco shred structure prediction model provided by the present application, and the following is a specific application example of the method for constructing the tobacco shred structure prediction model provided by the present application.
1. Sample preparation
The selected tobacco flake sample is the first-class tobacco module in the middle after threshing and redrying in different producing areas in 2017, and the specific information is shown in table 25.
TABLE 25 smoked sheet sample information
2. Blade detection structure
The leaf structure was examined by the image method after loosening and dampening, and the examination results are shown in table 26.
Vane structure of watch 26
3. Detect pipe tobacco structure
The tobacco sheets with single component are shredded, the shredding width is 1.0mm, the tobacco shred structure is detected by using a vibrating screen Retsch AS400 (Germany) after the shredding, the result is shown in table 27, the screen mesh size in the embodiment of the application is 8, and the corresponding tobacco shred structure indexes are 8.
Watch 27 tobacco shred structure
4. Screening structural indexes of blades
(1) The leaf structure indexes and the tobacco shred structure indexes of 4 production places are subjected to correlation analysis, the analysis results are shown in table 28, and according to the table 28, the proportion correlation between the medium piece rate and the tobacco shred structure of each layer is not obvious, so that the medium piece rate is removed, and other indexes are reserved and subjected to subsequent analysis.
TABLE 28 correlation between leaf Structure indices and tobacco shred Structure indices
(2) In the retained blade structure indexes, detection indexes (large fragment rate, small fragment rate, fragment rate and powder rate) and constructed characteristic indexes (large and medium fragment rate and small powder rate), in order to avoid data repeated operation and take the completeness of data into consideration, the large fragment rate and the small powder rate are removed, and the large and medium fragment rate, the small fragment rate, the fragment rate and the powder rate are retained to participate in subsequent analysis.
5. Testing and determining fitness of principal component analysis
The preserved blade structure is subjected to correlation analysis, the analysis results refer to tables 29 and 30, and the blade structure indexes shown by the results have extremely obvious correlation relations, so that the method is suitable for constructing new variables by a principal component analysis method, and the problem of multiple collinearity among the blade structures is solved.
TABLE 29 correlation between blade Structure indices
TABLE 30 correlation between blade Structure indices
6. Construction of New variables Using principal component analysis
The leaf structure was subjected to KMO and bartlett sphericity tests and the results are shown in table 31, showing very significant significance, indicating that the leaf structure is suitable for principal component analysis.
TABLE 31 KMO and Batterit sphericity test results
Three principal components are extracted from the blade structure, the cumulative variance contribution rate (table 32) of the three principal components reaches 100% (> 95.00%), and the modeling accuracy can be guaranteed, so that the three principal components are extracted from the blade structure, namely three new variables are constructed to replace the blade structure indexes. Tables 33 and 34 show a component matrix from which three principal components are extracted.
TABLE 32 variance contribution ratio of principal component
Table 33 component matrix
Table 34 rotated component matrix
7. Calculating a score for the target principal component
By extracting principal components from the blade structure, score coefficient matrices of 3 target principal components are obtained, and the principal components can be extracted by SPSS software, and the score coefficient matrices of the principal components are output, please refer to table 35.
Table 35 score coefficient matrix of principal components
The score calculation model for calculating the first target principal component is as follows:
Y1＝-0.213*(X1-77.57)/6.37-0.668*(X2-16.78)/3.92+1.495*(X3-5.62)/2.67-0.295*(X4-0.025)/0.012；
the score calculation model of the second target principal component is:
Y2＝-0.338*(X1-77.57)/6.37-0.365*(X2-16.78)/3.92-0.275*(X3-5.62)/2.67+1.727*(X4-0.025)/0.012；
the score calculation model of the third target principal component is:
Y3＝-0.978*(X1-77.57)/6.37+2.695*(X2-16.78)/3.92-1.630*(X3-5.62)/2.67-1.887*(X4-0.025)/0.012。
wherein, X1Large and medium slice rate, unit: percent;
X2-chip rate, unit: percent;
X3-fragmentation rate, in units: percent;
X4-end-up rate, unit: percent;
the large and medium leaf rate, the small leaf rate, the fragment rate and the powder rate in the leaf structure indexes of 12 tobacco leaf samples of the 4 producing areas are correspondingly input into each score calculation model, and the scores of each target principal component are correspondingly obtained, please refer to table 36.
TABLE 36 score of target principal Components
| Numbering | First target principal component | Second target principal component | Third target principal component |
| 1 | 1.52 | 1.08 | -1.29 |
| 2 | 1.65 | -0.45 | 0.73 |
| 3 | 1.69 | -0.43 | 0.58 |
| 4 | -0.39 | 1.30 | 1.31 |
| 5 | -0.81 | 1.40 | 1.66 |
| 6 | -0.82 | 1.22 | -1.49 |
| 7 | -0.76 | -1.16 | 0.24 |
| 8 | -0.52 | -1.29 | 0.33 |
| 9 | -0.74 | -1.05 | -0.23 |
| 10 | -0.27 | -0.13 | -0.92 |
| 11 | -0.37 | -0.10 | -0.86 |
| 12 | -0.19 | -0.38 | -0.06 |
8. Establishing regression model of tobacco shred structure and target principal component score
(1) Regression model of scores of first layer tobacco shred structure and target principal component
The scores of the 3 target principal components and the tobacco shred structure index are subjected to stepwise multiple regression analysis respectively, and according to table 37, the correlation coefficient between the score of the first target principal component and the first layer tobacco shred structure is 0.842, and the determination coefficient is 0.710, which shows that the score of the first target principal component and the analysis result of the regression analysis (model 1) of the first layer tobacco shred structure are very obvious. As can be seen from the analysis of variance results in table 38, the P value of the linear regression reached a very significant level, indicating that the fitting effect of the score of the first target principal component and the first layer tobacco shred structure was good. As can be seen from the parameters of the regression model in table 39, the score of the first target principal component and the coefficient of the constant both reach very significant levels by inspection, and therefore, the regression model of the score of the first layer tobacco shred structure and the first target principal component is:
H1＝-8.941*Y1+18.860；
TABLE 37 regression analysis results of target principal components and first layer tobacco shred structure
TABLE 38 analysis of variance of regression analysis of target principal components and first layer tobacco shred structures
TABLE 39 regression model parameters for target principal components and first layer tobacco shred structures
(2) Regression model of scores of second and third layer tobacco shred structures and target main components
And (3) performing stepwise multiple regression analysis on the scores of the 3 target main components and the second-layer and third-layer tobacco shred structures respectively, wherein the regression analysis result cannot meet the accurate regression model condition due to the fact that the correlation between the second-layer and third-layer tobacco shred structures and the blade structure is weak, and therefore, a regression model is not established for the second-layer and third-layer tobacco shred structures. When the tobacco shred structure index and the leaf structure index are subjected to correlation analysis to screen the leaf structure index, the tobacco shred structure index can be screened, when the absolute value of the Person correlation coefficient of a certain tobacco shred structure index and all the leaf structure indexes is less than 0.5 or the P value is more than 0.1, the tobacco shred structure index is removed and does not participate in subsequent modeling, meanwhile, the tobacco shred structure index can be used for checking whether the division of the tobacco shred structure index is reasonable, and when the relevance of the tobacco shred structure index and all the leaf structure indexes is not high, the tobacco shred structure index can be considered to be divided again.
(3) Regression model of scores of fourth-layer tobacco shred structure and target principal component
The scores of the 3 target principal components are subjected to stepwise multiple regression analysis with the tobacco shred structure of the fourth layer respectively, and according to tables 40-42, the correlation coefficient between the score of the first target principal component and the tobacco shred structure of the fourth layer is 0.768, and the determination coefficient is 0.590, which shows that the fitting effect between the score of the first target principal component and the tobacco shred structure of the fourth layer is good. The regression model of the scores of the fourth layer of tobacco shred structure and the first target main component is as follows:
H4＝0.480*Y1+8.063；
TABLE 40 score of target principal Components and regression analysis results of fourth layer tobacco shred Structure
TABLE 41 score of target principal Components and analysis of variance of regression analysis of tobacco shred Structure in layer four
TABLE 42 score of target principal component and regression model parameters for tobacco shred structure in layer four
(4) Regression model of scores of fifth-layer tobacco shred structure and target principal component
The scores of the 3 target main components and the fifth layer tobacco shred structure are subjected to stepwise multiple regression analysis respectively, and according to the table 43-45, the correlation coefficient of the score of the first target main component and the fifth layer tobacco shred structure is 0.841, and the determination coefficient is 0.707, which shows that the fitting effect of the score of the first target main component and the fifth layer tobacco shred structure is better. The regression model of the scores of the fifth layer tobacco shred structure and the first target main component is as follows:
H5＝1.501*Y1+14.323；
TABLE 43 regression analysis results of scores of target principal Components and fifth layer tobacco shred structures
TABLE 44 analysis of variance of regression analysis of target principal component score and fifth cut tobacco structure
TABLE 45 score of target principal component and regression model parameters for fifth cut tobacco structure
(5) Regression model of scores of tobacco shred structure of sixth layer and target principal component
The scores of the 3 target principal components are subjected to stepwise multiple regression analysis with the tobacco shred structure on the sixth layer respectively, and according to tables 46-48, the scores of the first target principal component and the second target principal component and the regression model established by the tobacco shred structure on the sixth layer have good effect, the correlation coefficient of the model is 0.945, the decision coefficient is 0.894, and the analysis of variance result shows that the regression effect is extremely obvious. The regression model of the sixth layer of tobacco shred structure, the score of the first target principal component and the score of the second target principal component is as follows:
H6＝5.244*Y1+1.602*Y2+23.646；
TABLE 46 score of target principal Components and regression analysis results of tobacco shred Structure of sixth layer
TABLE 47 score of target principal component and analysis of variance of regression analysis of tobacco shred Structure in sixth layer
TABLE 48 score of target principal component and regression model parameters of tobacco shred structure in sixth layer
(6) Regression model of scores of tobacco shred structure of seventh layer and target main component
The scores of the 3 target principal components are subjected to stepwise multiple regression analysis with the tobacco shred structure on the sixth layer respectively, and according to the table 49-51, the regression model established by the scores of the first target principal components and the tobacco shred structure on the seventh layer has a good effect, the correlation coefficient of the model is 0.891, the decision coefficient is 0.794, and the analysis of variance shows that the regression effect is extremely obvious. The regression model of the scores of the seventh layer of tobacco shred structure and the first target main component is as follows:
H7＝1.656*Y1+4.884；
TABLE 49 score of target principal component and regression analysis result of tobacco shred structure of seventh layer
TABLE 50 score of target principal component and analysis of variance of regression analysis of tobacco shred structure of layer seven
TABLE 51 score of target principal component and regression model parameters for tobacco shred structure of layer seven
(6) Regression model of scores of tobacco shred structure in eighth layer and target main component
The scores of the 3 target principal components are subjected to stepwise multiple regression analysis with the tobacco shred structure on the sixth layer respectively, and according to tables 52-54, the scores of the first target principal components and the regression model established by the tobacco shred structure on the eighth layer have good effect, the correlation coefficient of the model is 0.750, the decision coefficient is 0.563, and the analysis of variance shows that the regression effect is extremely obvious. The regression model of the scores of the eighth layer of tobacco shred structure and the first target main component is as follows:
H8＝0.045*Y1+0.151；
TABLE 52 regression analysis results of scores of target principal components and tobacco shred structure in layer eight
TABLE 53 score of target principal Components and analysis of variance of regression analysis of tobacco shred Structure of layer eight
TABLE 54 score of target principal component and regression model parameters for tobacco shred structure in layer eight
9. Tobacco shred structure prediction model for establishing leaf structure
(1) Score calculation model
The score calculation model of the first target principal component is:
Y1＝-0.213*(X1-77.57)/6.37-0.668*(X2-16.78)/3.92+1.495*(X3-5.62)/2.67-0.295*(X4-0.025)/0.012；
the score calculation model of the second target principal component is:
Y2＝-0.338*(X1-77.57)/6.37-0.365*(X2-16.78)/3.92-0.275*(X3-5.62)/2.67+1.727*(X4-0.025)/0.012；
the score calculation model of the third target principal component is:
Y3＝-0.978*(X1-77.57)/6.37+2.695*(X2-16.78)/3.92-1.630*(X3-5.62)/2.67-1.887*(X4-0.025)/0.012。
(2) regression model
A first layer: h1＝-8.941*Y1+18.860；
A second layer: none;
and a third layer: none;
a fourth layer: h4＝0.480*Y1+8.063；
And a fifth layer: h5＝1.501*Y1+14.323；
A sixth layer: h6＝5.244*Y1+1.602*Y2+23.646；
A seventh layer: h7＝1.656*Y1+4.884；
An eighth layer: h8＝0.045*Y1+0.151。
And combining the score calculation model and the regression model to obtain the tobacco shred structure prediction model of the leaf structure.
10. Model validation
Two groups of leaves in Sichuan and Hunan are respectively selected, the detected leaf structures are shown in a table 55, after the tobacco is cut according to the cutting width of 1.0mm, the vibrating screen RetschAS400 (Germany) is used for detecting the two groups of tobacco shred structures, the results are shown in a table 56, and the absolute deviation range of the predicted value and the true value is 0.02-6.87% from the verification result of a tobacco shred structure prediction model in the table 56. Therefore, the tobacco shred structure prediction model in the embodiment of the application has high accuracy.
Blade structure of watch 55
TABLE 56 tobacco shred structure prediction model verification results based on leaf structure
The above is a specific application example of the method for constructing the tobacco shred structure prediction model provided by the present application, and the following is an embodiment of the apparatus for constructing the tobacco shred structure prediction model provided by the present application.
Referring to fig. 2, an embodiment of a device for constructing a tobacco shred structure prediction model provided by the present application includes:
the obtaining unit 301 is configured to obtain a plurality of tobacco flake samples, and detect a leaf structure and a shredded tobacco structure of each tobacco flake sample.
The first analysis unit 302 is configured to perform principal component analysis on the leaf structure of each smoked sheet sample, select a preset number of principal components with an accumulated variance contribution rate greater than or equal to 95% as target principal components, and output a score coefficient matrix of the target principal components.
The constructing unit 303 is configured to construct a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculate a score of each target principal component based on the score calculation model.
And a second analysis unit 304, configured to perform stepwise multiple regression analysis on the scores of the target principal components and the tobacco shred structures, and output a regression model of the scores of the target principal components and the tobacco shred structures.
And the combining unit 305 is configured to combine the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
The above is an embodiment of a device for constructing a tobacco shred structure prediction model provided by the present application, and the following is another embodiment of a device for constructing a tobacco shred structure prediction model provided by the present application, and the device includes:
the obtaining unit 401 is configured to obtain a plurality of tobacco flake samples, and detect a leaf structure and a shredded tobacco structure of each tobacco flake sample.
The first analysis unit 402 is configured to perform principal component analysis on the leaf structure of each tobacco flake sample, select a preset number of principal components with an accumulated variance contribution rate greater than or equal to 95% as target principal components, and output a score coefficient matrix of the target principal components.
A constructing unit 403, configured to construct a score calculation model of each target principal component and the leaf structure based on the score coefficient matrix, and obtain a score of each target principal component based on the score calculation model.
And a second analysis unit 404, configured to perform stepwise multiple regression analysis on the scores of the target principal components and the tobacco shred structures, and output a regression model of the scores of the target principal components and the tobacco shred structures.
And the combining unit 405 is used for combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
As a further improvement, the obtaining unit 401 specifically includes:
the acquisition sub-unit 4011 is configured to acquire a plurality of tobacco flake samples;
the first detecting subunit 4012 is configured to detect a leaf structure index value in each tobacco flake sample to obtain a leaf structure of each tobacco flake sample, where the leaf structure index includes a large-medium flake rate, a small flake rate, a fragment rate, a small fragment rate, and a small fragment rate;
and the second detecting sub-unit 4013 is configured to detect a proportion of each structural index of the tobacco shred in the tobacco shred sample obtained by shredding the tobacco shred sample, so as to obtain the tobacco shred structure of each tobacco shred sample.
As a further improvement, the method further comprises the following steps:
a third analysis unit 406, configured to perform correlation analysis on the tobacco shred structure indexes and the leaf structure indexes in each tobacco flake sample based on a correlation analysis method of a binary distance variable, to obtain correlation coefficients of each tobacco shred structure index and each leaf structure index;
the first retaining unit 407 is configured to retain the tobacco shred structure index and the leaf structure index in the tobacco flake sample when the correlation coefficient between the leaf structure index and any tobacco shred structure index is within a first preset range;
the first removing unit 408 is configured to remove the leaf structure index when the correlation coefficients of the leaf structure index and all the tobacco shred structure indexes are not within the first preset range.
As a further improvement, the method further comprises the following steps:
a second retaining unit 409 configured to preferentially retain the detection index when detecting that the detection index and the feature index coexist in the retained blade structure index;
the detection indexes comprise a large fragment rate, a medium fragment rate, a small fragment rate, a fragment rate and a powder rate, and the characteristic indexes comprise a large fragment rate, a medium fragment rate and a small powder rate.
As a further improvement, the method further comprises the following steps:
the fourth analysis unit 410 is configured to perform correlation analysis on the retained blade structure indexes based on a correlation analysis method of a binary distance variable to obtain correlation coefficients between every two blade structure indexes;
and the second removing unit 411 is used for removing the smoked sheet samples corresponding to the correlation coefficients which are not in the second preset range when the correlation coefficients between every two leaf structure indexes are not in the second preset range.
As a further improvement, the method further comprises the following steps:
and the prediction unit 412 is used for detecting the leaf structure of the obtained tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into the tobacco shred structure prediction model for tobacco shred structure prediction, and outputting the tobacco shred structure of the tobacco flakes to be detected.
It is clear to those skilled in the art that, for convenience and brevity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present application may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The above embodiments are only used for illustrating the technical solutions of the present application, and not for limiting the same; although the present application has been described in detail with reference to the foregoing embodiments, it should be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions in the embodiments of the present application.
Claims (10)
1. A method for constructing a tobacco shred structure prediction model is characterized by comprising the following steps:
acquiring a plurality of tobacco flake samples, and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample;
performing principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components;
constructing a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model;
performing stepwise multiple regression analysis on the scores of the target principal components and the tobacco shred structure, and outputting regression models of the scores of the target principal components and the tobacco shred structure;
and combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
2. The method for constructing the tobacco shred structure prediction model according to claim 1, wherein the step of detecting the leaf structure of each tobacco shred sample comprises the following steps:
and detecting the numerical value of each leaf structure index in each tobacco flake sample to obtain the leaf structure of each tobacco flake sample, wherein the leaf structure indexes comprise a large and medium flake rate, a large flake rate, a medium flake rate, a small flake rate, a fragment rate, a powder rate and a small powder rate.
3. The method for constructing the tobacco shred structure prediction model according to claim 2, wherein the step of detecting the tobacco shred structure of each shredded tobacco flake sample comprises the following steps:
and detecting the proportion of each tobacco shred structure index in the tobacco shred samples obtained by shredding the tobacco flake samples to obtain the tobacco shred structures of the tobacco flake samples.
4. The method for constructing a tobacco shred structure prediction model according to claim 3, wherein the steps of obtaining a plurality of tobacco shred samples, detecting the leaf structure of each tobacco shred sample and the shredded tobacco shred structure further comprise:
performing correlation analysis on the tobacco shred structure indexes and the leaf structure indexes in each tobacco flake sample based on a correlation analysis method of a binary distance variable to obtain correlation coefficients of each tobacco shred structure index and each leaf structure index;
when the correlation coefficient of the leaf structural index and any tobacco shred structural index is within a first preset range, the tobacco shred structural index and the leaf structural index in the tobacco flake sample are reserved;
and when the correlation coefficients of the leaf structural indexes and all the tobacco shred structural indexes are not in the first preset range, removing the leaf structural indexes.
5. The method for constructing the tobacco shred structure prediction model according to claim 4, wherein when the correlation coefficients of the leaf structure index and all the tobacco shred structure indexes are not within the first preset range, the leaf structure index is removed, and then the method further comprises the following steps:
when detecting that a detection index and a characteristic index exist in the reserved blade structure indexes at the same time, preferentially reserving the detection index;
the detection indexes comprise a large fragment rate, a medium fragment rate, a small fragment rate, a fragment rate and a powder rate, and the characteristic indexes comprise a large fragment rate, a medium fragment rate and a small powder rate.
6. The method for constructing the tobacco shred structure prediction model according to claim 5, wherein when detecting that the detection index and the characteristic index exist in the retained leaf structure index at the same time, the detection index is preferentially retained, and then the method further comprises the following steps:
performing correlation analysis on the retained blade structure indexes by using a correlation analysis method based on a binary distance variable to obtain correlation coefficients between every two blade structure indexes;
and when the correlation coefficient of the blade structure indexes between every two is not in a second preset range, removing the tobacco flake sample corresponding to the correlation coefficient which is not in the second preset range.
7. The method for constructing the tobacco shred structure prediction model according to claim 1, wherein the tobacco shred structure prediction model of the leaf structure is obtained by combining the score calculation model and the regression model, and then the method further comprises the following steps:
and detecting the leaf structure of the obtained tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into the tobacco shred structure prediction model to predict the tobacco shred structure, and outputting the tobacco shred structure of the tobacco flakes to be detected.
8. A construction device of a tobacco shred structure prediction model is characterized by comprising the following steps:
the acquisition unit is used for acquiring a plurality of tobacco flake samples and detecting the leaf structure and the shredded tobacco structure of each tobacco flake sample;
the first analysis unit is used for carrying out principal component analysis on the leaf structure of each tobacco flake sample, selecting a preset number of principal components with the accumulated variance contribution rate of more than or equal to 95% as target principal components, and outputting a score coefficient matrix of the target principal components;
the building unit is used for building a score calculation model of each target principal component and the blade structure based on the score coefficient matrix, and calculating to obtain the score of each target principal component based on the score calculation model;
the second analysis unit is used for performing stepwise multiple regression analysis on the scores of the target main components and the tobacco shred structures and outputting regression models of the scores of the target main components and the tobacco shred structures;
and the combining unit is used for combining the score calculation model and the regression model to obtain a tobacco shred structure prediction model of the leaf structure.
9. The device for constructing the tobacco shred structure prediction model according to claim 8, wherein the obtaining unit specifically comprises:
an acquisition subunit for acquiring a number of smoked sheet samples;
the first detection subunit is used for detecting each leaf structure index numerical value in each tobacco flake sample to obtain a leaf structure of each tobacco flake sample, wherein the leaf structure index comprises a large and medium flake rate, a large flake rate, a medium flake rate, a small flake rate, a fragment rate and a small fragment rate;
and the second detection subunit is used for detecting the proportion of each tobacco shred structure index in the tobacco shred samples obtained by shredding the tobacco flake samples to obtain the tobacco shred structures of the tobacco flake samples.
10. The tobacco shred structure prediction model construction device according to claim 8, further comprising:
and the prediction unit is used for detecting the leaf structure of the acquired tobacco flakes to be detected, inputting the leaf structure of the tobacco flakes to be detected into the tobacco shred structure prediction model for tobacco shred structure prediction, and outputting the tobacco shred structure of the tobacco flakes to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010710238.9A CN112016046B (en) | 2020-07-22 | 2020-07-22 | Construction method and device of tobacco shred structure prediction model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010710238.9A CN112016046B (en) | 2020-07-22 | 2020-07-22 | Construction method and device of tobacco shred structure prediction model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112016046A true CN112016046A (en) | 2020-12-01 |
| CN112016046B CN112016046B (en) | 2023-11-28 |
Family
ID=73498792
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010710238.9A Active CN112016046B (en) | 2020-07-22 | 2020-07-22 | Construction method and device of tobacco shred structure prediction model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112016046B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113576016A (en) * | 2021-07-15 | 2021-11-02 | 河南中烟工业有限责任公司 | Shredding adjusting method based on tobacco shred structure |
| CN114098136A (en) * | 2021-12-20 | 2022-03-01 | 河南中烟工业有限责任公司 | Method and processing device for improving tobacco shred structure on line |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107330467A (en) * | 2017-07-03 | 2017-11-07 | 云南昆船设计研究院 | A kind of Forecasting Methodology influenceed based on piece cigarette morphological feature on tobacco structure |
| WO2018040302A1 (en) * | 2016-08-31 | 2018-03-08 | 上海创和亿电子科技发展有限公司 | Method and system for measuring width of cut tobacco piece or cut stem, and server having said system |
-
2020
- 2020-07-22 CN CN202010710238.9A patent/CN112016046B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018040302A1 (en) * | 2016-08-31 | 2018-03-08 | 上海创和亿电子科技发展有限公司 | Method and system for measuring width of cut tobacco piece or cut stem, and server having said system |
| CN107330467A (en) * | 2017-07-03 | 2017-11-07 | 云南昆船设计研究院 | A kind of Forecasting Methodology influenceed based on piece cigarette morphological feature on tobacco structure |
Non-Patent Citations (1)
| Title |
|---|
| 张延军;李旭华;林锐锋;文俊;叶为民;: "主成分回归分析法在永州烟叶评吸质量与外观质量的关系模型构建中的应用", 中国农学通报, no. 18 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113576016A (en) * | 2021-07-15 | 2021-11-02 | 河南中烟工业有限责任公司 | Shredding adjusting method based on tobacco shred structure |
| CN114098136A (en) * | 2021-12-20 | 2022-03-01 | 河南中烟工业有限责任公司 | Method and processing device for improving tobacco shred structure on line |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112016046B (en) | 2023-11-28 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112016046A (en) | Method and device for constructing tobacco shred structure prediction model | |
| CN106897570B (en) | Chronic obstructive pulmonary disease testing system based on machine learning | |
| CN104063599B (en) | Index screening and processing method for evaluating quality of tobacco leaves | |
| WO2019153595A1 (en) | Method for predicting risk of chronic obstructive pulmonary disease, server, and computer readable storage medium | |
| CN107796782B (en) | Redrying quality stability evaluation method based on tobacco leaf characteristic spectrum consistency measurement | |
| CN108181263A (en) | The extraction of tobacco leaf genius loci and method of discrimination based near infrared spectrum | |
| CN107767079A (en) | A kind of objective integrated evaluating method of tobacco style feature | |
| CN110013048B (en) | Method and device for determining tobacco shred structure suitable for fine cigarette | |
| CN110132880B (en) | Tobacco leaf overall sensory quality evaluation method based on near infrared spectrum | |
| CN110122915A (en) | A kind of beating and double roasting processing quality integrated evaluating method | |
| CN107909264A (en) | The objective integrated evaluating method of stability between a kind of tobacco style feature year | |
| CN109324016A (en) | A kind of determination method of redried odor type style | |
| CN103892435A (en) | Method for preparing tobacco powder raw materials and stabilizing quality of reconstituted tobacco | |
| CN111640451B (en) | Maturity evaluation method and device, and storage medium | |
| CN109856139B (en) | Tree leaf quantity estimation method and system | |
| CN112067757A (en) | Automatic analysis method for sensory comfort of tobacco | |
| CN115169751B (en) | Method and system for improving maintenance quality of traditional Chinese medicine decoction pieces | |
| Saatci et al. | Determination of respiratory parameters by means of hurst exponents of the respiratory sounds and stochastic processing methods | |
| EP0012418B1 (en) | Apparatus for classifying particles contained in a blood sample | |
| CN107934712B (en) | Elevator health degree detection method and device based on statistical models | |
| CN112241820A (en) | Risk identification method and device for key nodes in fund flow and computing equipment | |
| CN115931738A (en) | Method and system for evaluating quality stability of finished tobacco flakes | |
| CN115936532A (en) | Saline-alkali soil stability assessment method and system based on BP neural network | |
| CN111513022B (en) | Breeding method of small-body meat ducks with low skin fat rate | |
| CN113075360A (en) | Method for measuring comprehensive stability of cigarette physical indexes |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |