CN111024557B - Soft measurement method for water absorption of paper sheet - Google Patents
Soft measurement method for water absorption of paper sheet Download PDFInfo
- Publication number
- CN111024557B CN111024557B CN201911330641.2A CN201911330641A CN111024557B CN 111024557 B CN111024557 B CN 111024557B CN 201911330641 A CN201911330641 A CN 201911330641A CN 111024557 B CN111024557 B CN 111024557B
- Authority
- CN
- China
- Prior art keywords
- model
- water absorption
- data
- soft measurement
- paper
- 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.)
- Active
Links
- 238000010521 absorption reaction Methods 0.000 title claims abstract description 84
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 84
- 238000000691 measurement method Methods 0.000 title claims abstract description 9
- 239000000835 fiber Substances 0.000 claims abstract description 108
- 238000005259 measurement Methods 0.000 claims abstract description 73
- 238000000034 method Methods 0.000 claims abstract description 34
- 238000004519 manufacturing process Methods 0.000 claims abstract description 13
- 238000003066 decision tree Methods 0.000 claims abstract description 11
- 238000007781 pre-processing Methods 0.000 claims abstract description 10
- 238000012706 support-vector machine Methods 0.000 claims abstract description 9
- 238000010219 correlation analysis Methods 0.000 claims description 15
- 238000012795 verification Methods 0.000 claims description 10
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 230000015572 biosynthetic process Effects 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000007405 data analysis Methods 0.000 claims description 5
- 238000007689 inspection Methods 0.000 claims description 5
- 238000012545 processing Methods 0.000 claims description 4
- 238000012549 training Methods 0.000 claims description 4
- 238000004537 pulping Methods 0.000 claims description 3
- 238000004451 qualitative analysis Methods 0.000 claims description 3
- 239000002002 slurry Substances 0.000 claims description 3
- 239000000126 substance Substances 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000001514 detection method Methods 0.000 abstract description 6
- 238000010801 machine learning Methods 0.000 abstract description 3
- 230000005856 abnormality Effects 0.000 abstract description 2
- 238000005516 engineering process Methods 0.000 abstract description 2
- 238000012544 monitoring process Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 37
- 230000004992 fission Effects 0.000 description 6
- 230000007547 defect Effects 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 241000202567 Fatsia japonica Species 0.000 description 1
- 238000013480 data collection Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 238000003908 quality control method Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N13/00—Investigating surface or boundary effects, e.g. wetting power; Investigating diffusion effects; Analysing materials by determining surface, boundary, or diffusion effects
Abstract
The invention discloses a soft measurement method for water absorption of paper sheets, which comprises the following steps: the method comprises the steps of collecting data in a database and on-site data, preprocessing acquired related data, establishing a fiber form soft measurement model by applying a support vector machine algorithm, establishing a paper sheet water absorption soft measurement model by applying a gradient lifting decision tree algorithm and verifying the model. The invention has the advantages that: the soft measurement model for the water absorption of the paper sheets is established by a machine learning method, the model prediction speed is high, the accuracy is good, and the real-time soft measurement of the water absorption of all the paper sheets can be accurately realized. The technology not only solves the problem that the water absorption of all base paper products cannot be measured in production, thereby reducing the labor detection cost, but also can timely acquire the water absorption quality of the finished paper, and the paper machine process is adjusted by monitoring the product quality in real time and finding out abnormality, thereby improving the product percent of pass.
Description
Technical Field
The invention relates to the technical field of paper sheet physical index prediction of paper making enterprises, in particular to a paper sheet water absorption soft measurement method based on a support vector machine and a gradient lifting decision tree combined algorithm.
Background
The paper has hydrophilic fibers and capillaries with porous structures among the fibers, so that the paper has water absorption performance. The water absorption of paper is also a concern for paper product manufacturers, and is directly related to the quality of paper products. The requirement for the paper for daily use is that it has good water absorption performance, and the higher the grade of the paper for daily use, the higher the requirement for water absorption performance.
At present, paper quality production in paper plants for daily use has the problem of high product reject ratio, detection of paper water absorption is based on an offline quality detection instrument, high manpower and material resources are required to be consumed, and detection of all products cannot be realized. The enterprise needs to consume a large amount of labor and equipment cost to detect the water absorption of the paper, and the paper factory also has the problem that the quality of the paper which is not detected is unqualified when the raw paper is subjected to spot inspection.
Therefore, it is very important for paper mills to perform real-time quality control of all products on the water absorption of paper. The quality inspection instrument based detection has hysteresis, the quality of paper cannot be detected in real time, effective information is fed back to production to guide production of actual products, and stability and high efficiency are maintained. Meanwhile, the waste of raw materials is caused by the large fluctuation of the product quality. Therefore, when internet + and intelligent manufacturing are called for, a mathematical model is constructed by using a machine learning algorithm, and accurate real-time prediction of an important index of water absorption of the household paper is a problem to be solved urgently.
Disclosure of Invention
Aiming at the defects in the prior art, the invention provides a soft measurement method for the water absorption of paper sheets, which solves the defects in the prior art.
In order to realize the purpose, the technical scheme adopted by the invention is as follows:
a method for soft measurement of water absorption of paper sheets is characterized by comprising the following steps:
s1, collecting key modeling data based on the production process, wherein the key modeling data comprises a database and field data variables; the variables include pulp fiber morphology measurements, refiner power, throughput, and paper machine reel speed;
s2, analyzing the data of the collected variables, qualitatively analyzing the correlation between the collected variables and the water absorption, and preprocessing the data of the variables to meet the requirements of fiber morphology soft measurement modeling and paper sheet water absorption soft measurement modeling;
s3, establishing a fiber morphology soft measurement model by combining a support vector machine algorithm according to data meeting modeling requirements;
s4, performing feature selection on the data meeting the modeling requirement, establishing a paper sheet water absorption soft measurement model by combining a gradient lifting decision tree algorithm, and then performing model validity verification by using on-site data.
Further, the correlation between the qualitative analysis collection variable and the water absorption in S2 is specifically: the formation of the water absorption of the paper sheet is greatly related to the average length of the fibers, the content of the fine fibers and the bonding force among the fibers, the average length of the fibers is short, the content of the fine fibers is increased, the water absorption performance of the paper sheet is reduced, the bonding force of the fiber quality inspection is low, and the water absorption of the finished paper is strong; the average fiber length and the fine fiber content are related to the fiber shape of an original pulp board, the power and the throughput of a refiner in a pulping process, and the formation of the bonding force among fibers is related to the slurry flow in a papermaking stage and the vehicle speed and the vacuum cylinder mould speed in a forming process, so that the shape indexes of the average fiber length, the average fiber thickness, the percentage of kinked fibers and the average fiber width are collected, and the key parameters of the vehicle speed and the vacuum cylinder mould speed in the paper machine process are used as selection variables input by a paper sheet water absorption model;
the data analysis is used for checking whether dirty data and data which cannot be directly analyzed exist in original data, wherein the dirty data comprises missing values, abnormal values and data containing special characters; deleting the missing value, the abnormal value and the character of the special symbol;
data preprocessing in S2 is performed to meet modeling requirements, and input variables of the model are normalized, so that indexes of different units or orders can be compared and weighted conveniently, and influence on modeling caused by different characteristic dimensions and orders of different orders is eliminated; the processing mode is to compress the selected characteristic variables to obtain a new number series with the mean value of 0 and the standard deviation of 1; the specific mode is as follows:
wherein i is 1,2, …, n is the number of samples, j is 1,2, …, and p is the dimension of the samples;is the mean value of the samples in the j dimension, xijIs the j-dimensional value of the ith sample,for the j-dimensional normalized value of the i-th sample, SjThe standard deviation of the sample in the j dimension is taken; sj 2Is the variance of the sample in the j dimension.
Further, in the step S4, selecting characteristics of the soft measurement model for the water absorption of the paper sheet, performing correlation analysis on the preprocessed characteristics, and selecting a variable having an obvious influence on the water absorption to perform modeling, wherein a pearson correlation coefficient is used in the correlation analysis, and a specific calculation formula is as follows;
wherein the content of the first and second substances,is y andof (a) covariance, Var [ y [)],Are respectively y andthe variance of (a), E (y),are respectively y and(iii) a desire;
the correlation coefficient can reflect the degree of correlation between two variables, the value of the correlation coefficient is in the interval < -1,1 >, a positive value represents positive correlation, a negative value represents negative correlation, the closer the absolute value is to 1, the stronger the correlation between the two is, and the value of 0 represents no correlation; through the correlation analysis of the paper machine process and the water absorption of paper sheets, a variable with the correlation of the water absorption of more than 0.4 is selected as the input of a water absorption soft measurement model.
Further, the sub-steps of S3 are as follows:
s31, establishing a fiber morphology soft measurement model based on the principle of a support vector machine regression model; the kernel function can map low-dimensional data to a high-dimensional space and convert a nonlinear problem into a linear problem; the principle of kernel function selection is that the linear indivisible original data is mapped to the high-dimensional characteristic space; for a given data set, the solution formula for the SVM model is as follows
Where ω is a weight coefficient and x is an input variable [ x ]1,…xp]n×pB is an offset term, αi,Is Lagrange multiplier, K (x)i,xj) Is a kernel function;
s32, aiming at the fiber form soft measurement model and the data set, the model input is the original pulp fiber form, pulp board ratio, refiner power, throughput and concentration, and the model output is the ground pulp fiber form; adopting a cross verification method, randomly selecting 80% of training sets to train and adjust model parameters each time, and verifying the precision of the parameters by 20% of testing sets;
and S33, establishing a fiber form soft measurement model by using the parameters with better performance in the step S32, and carrying out soft measurement on the subsequent fiber form.
Further, the sub-steps of S4 are as follows:
s41, establishing a soft measurement model of paper sheet water absorption based on the model principle of the gradient lifting decision tree, and recording Fm(x) The specific formula of the overall model function is as follows:
wherein, Fm(x) As a function of the global model, Fm-1(x) Is the (M-1) th basic regression tree function, M is 1,2 … M is the regression tree, betamIs the mth regression tree weight, L is the loss function of the model, alphamIs a parameter in the mth regression tree;
s42, aiming at the vehicle speed and the vacuum cylinder mould speed in the paper machine process selected by the correlation analysis, combining the average length of the refined fibers, the average thickness of the fibers, the percentage of kinked fibers and the average width characteristics of the fibers output by the fiber form soft measurement model, and establishing a paper sheet water absorption model by utilizing a gradient lifting regression decision tree.
Further, in S4, validation is to collect the production data again for another period of time, analyze the data, pre-process the data, predict the data by using the built model, and compare the predicted data with the actual data to validate the model for the new data.
Compared with the prior art, the invention has the advantages that:
a paper sheet water absorption soft measurement model is established by a machine learning method, the model prediction speed is high, the accuracy is good, and real-time soft measurement of water absorption of all paper sheets can be accurately realized. The problem that water absorption of all base paper products cannot be measured in production is solved, so that the manual detection cost is reduced, the water absorption quality of finished paper can be timely obtained, the paper machine process is adjusted by monitoring the product quality in real time and finding out abnormality, and the product percent of pass is improved.
Drawings
FIG. 1 is a flow chart of a sheet absorbency soft measurement model based on an intelligent combination algorithm of the present invention;
FIG. 2 is a graph of the results of a model for establishing soft measurements of fiber morphology in an embodiment of the present invention;
FIG. 3 is a graph of the results of correlation analysis in an embodiment of the present invention;
FIG. 4 is a graph showing the results of a water absorption soft measurement model in an example of the present invention;
FIG. 5 is a graph showing the results of the soft measurement model for sheet absorbency in the example of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings by way of examples.
The modeling method and the steps of the paper sheet water absorption soft measurement model are shown in figure 1.
And S1, collecting key modeling data based on the production process, including database and field data collection. Including pulp fiber morphology measurements, refiner power, throughput, paper machine reel speed, etc.
And S2, qualitatively analyzing and collecting the correlation between the variables and the water absorption, and preprocessing the data of the variables to meet the requirements of fiber morphology soft measurement modeling and paper water absorption soft measurement modeling.
And S3, establishing a fiber morphology soft measurement model by combining a support vector machine algorithm according to the data meeting the modeling requirement.
S4, performing feature selection on the data meeting the modeling requirement, establishing a paper sheet water absorption soft measurement model by combining a gradient lifting decision tree algorithm, and then performing model validity verification by using on-site data.
And S5, carrying out validity verification on the built paper sheet water absorption soft measurement model. And collecting the production data of the enterprise at another period of time again, carrying out data analysis and data preprocessing, predicting by using the built model, and comparing with an actual value to verify the effectiveness of the model on new data.
Further, the correlation between the qualitative analysis collection variable and the water absorption in S2 is specifically: the formation of the water absorption of the paper sheet is greatly related to the average length of the fibers, the content of the fine fibers and the bonding force among the fibers, the average length of the fibers is short, the content of the fine fibers is increased, the water absorption performance of the paper sheet is reduced, the bonding force of the fiber quality inspection is low, and the water absorption of the finished paper is strong. The average fiber length and the fine fiber content are related to the fiber shape of an original pulp board, the power and the throughput of a refiner in the pulping process, and the formation of the bonding force among fibers is related to parameters such as slurry flow in the papermaking stage, the vehicle speed in the forming process, the vacuum cylinder mould speed and the like, so that the fiber shape indexes such as the average fiber length and the like are collected, and the key parameters such as the vehicle speed, the vacuum cylinder mould speed and the like in the paper machine process are used as selection variables input by a paper sheet water absorption model.
The data analysis is used for checking whether dirty data and data which cannot be directly analyzed exist in the original data, wherein the dirty data comprises missing values, abnormal values and data containing special characters. The missing values, abnormal values, and characters of special symbols are deleted.
The data preprocessing in the S2 is to perform normalization processing on the input variables of the model, so that the indexes of different units or orders can be compared and weighted conveniently, thereby eliminating the influence of different characteristic dimensions and orders on modeling. The processing mode is to compress the selected characteristic variables to obtain a new number array with the average value of 0 and the standard deviation of 1. The specific mode is as follows:
where i is 1,2, …, n is the number of samples, j is 1,2, …, and p is the dimension of the samples. x is the number ofjIs the mean value of the samples in the j dimension, xijIs the j-dimensional value of the ith sample,for the j-dimensional normalized value of the i-th sample, SjIs the standard deviation of the sample in the j dimension. Sj 2Is the variance of the sample in the j dimension.
And S4, selecting the characteristics of the paper sheet water absorption soft measurement model, performing correlation analysis on the preprocessed characteristics, and selecting variables with obvious influence on water absorption to model, wherein Pearson correlation coefficients are adopted in the correlation analysis, and a specific calculation formula is as follows.
Wherein the content of the first and second substances,is y andof (a) covariance, Var [ y [)],Respectively y andthe variance of (a), E (y),are respectively y andthe expectation is that.
The correlation coefficient can reflect the degree of correlation between two variables, and the value is in the interval of [ -1,1], positive value represents positive correlation, negative value represents negative correlation, the closer to 1 the absolute value is, the stronger the correlation between the two is, and the value of 0 represents no correlation. The method is characterized in that correlation analysis is carried out on paper machine technology and paper water absorption, and a variable with correlation of water absorption larger than 0.4 is selected as water absorption soft measurement model input.
The substeps of S3 are as follows:
and S31, establishing a fiber morphology soft measurement model based on the principle of the regression model of the support vector machine. The kernel function may map low-dimensional data to a high-dimensional space, converting a non-linear problem to a linear problem. The kernel function is chosen such that the problem maps to a linear separable inside the high-order feature space. For a given data set, the solution formula for the SVM model is as follows
Where ω is a weight coefficient and x is an input variable [ x ]1,…xp]n×pB is an offset term, αi,Is Lagrange multiplier, K (x)i,xj) Is a kernel function.
And S32, aiming at the fiber form soft measurement model and the data set, the model input is the original pulp fiber form, the pulp board ratio, the power, the throughput and the concentration of the refiner, and the model output is the ground pulp fiber form. And (3) adopting a cross verification method, randomly taking 80% of training sets to train and adjust model parameters each time, and verifying the precision of the parameters by 20% of testing sets.
And S33, establishing a fiber form soft measurement model by using the parameters with better performance in the step S32, and carrying out soft measurement on the subsequent fiber form.
The substeps of S4 are as follows:
s41, establishing a soft measurement model of the water absorption of the paper sheet based on the model principle of the gradient lifting regression tree, and recording Fm(x) The specific formula of the overall model function is as follows:
wherein, Fm(x) As a function of the global model, Fm-1(x) Is the (M-1) th basic regression tree function, M is 1,2 … M is the regression tree, betamIs the mth regression tree weight, L is the loss function of the model, alphamAre parameters in the mth regression tree.
S42, aiming at the variables such as the speed of the paper machine and the speed of the vacuum cylinder mould in the paper machine process selected by the correlation analysis, combining the indexes such as the weight average length of the defibrinated fibers output by the fiber form soft measurement model, and establishing a paper sheet water absorption model by utilizing a gradient lifting regression decision tree.
In step S41, the principle of the gradient lifting regression tree is specifically as follows:
s51, the gradient is used for lifting the regression tree, the regression tree is trained through multiple iterations, the final result is the weighted result of all the regression trees, and the formula is as follows:
wherein F (x, P) is the overall model function, h (x, alpha)m) Is the mth basic regression tree function, M is 1,2 … M is the regression tree, betamIs the weight of the mth regression tree, alphamAre parameters in the mth regression tree.
S52, the gradient boosting regression tree, for each regression tree, assuming that fission feature j and fission node S define a pair of half-planes in a binary search: r1(j,s)={X|Xj≤s}and R2(j,s)={X|Xj> s }, j is fission characteristic, s is fission node, R1、R2Is a region space. The objective function formula for finding fission characteristics j and fission nodes s is thus as follows:
wherein, yiAs actual output variable, c1、c2The variables are output for fitting.
S53, the kernel of the gradient lifting regression tree lies in that the m-th new regression tree is used for fitting the residual error of the previous (m-1) regression trees, and the assumption is thatIn order to train the sample to be trained,representing a set of parameters, β is the weight of each regression tree, α is a parameter within the regression tree, so the overall model function is as follows,
s54, the gradient lifting enhancement tree introduces a loss function (loss function) in order to describe the accuracy degree of the model. After adding the loss function, the sample setSolving the minimum parameters of the model, and the formula is shown as follows.
Wherein F (x, P) is the overall model function, h (x, alpha)m) Is the mth basic regression tree function, M is 1,2 … M is the regression tree, betamAs the m regression treeWeight, alphamAre parameters in the mth regression tree.
S55, the gradient lifting enhancement tree, in order to obtain the overall model function F (x, P), the formula (8) needs to be solvedAnd (4) collecting. Weights, internal parameters and starting model function F of the first regression tree1(x) As shown in formula (9), and further, since Fm(x) Is according to Fm-1(x) New function resulting from calculation of model deviation, therefore, Fm(x) Can be prepared fromm-1(x) Expressed, the formula is shown in formula (10).
Fm(x)=Fm-1(x)+βmh(x,αm) (10)
Wherein M is 1,2 … M is regression tree, alpha1Is a parameter in the 1 st regression tree, βmIs the mth regression tree weight, h (x, α)m) As the mth basic regression tree function, Fm(x) For the first m model functions, Fm-1(x) Is a function of the first m-1 functions.
S56, the gradient lifting enhancement tree is used, the minimum value is solved by adopting a gradient descent method, and the direction of the maximum descending gradient is that the loss function is in the current model Fm-1The following negative gradient direction, the calculation formula is shown in equation (11).
Then by least squaresFind Fm(x) Internal parameter α ofmCirculation ofAnd M is M +1 (M is not less than 2 and not more than M) to obtain betamCarrying in (7), finally obtaining the overall model function FM(x)。
In step S3, the parameter characteristics of the soft measurement of the fiber morphology are as follows:
the method comprises the steps that a fiber form soft measurement model based on regression of a support vector machine is established, as the collected fiber form data volume samples are small, and the accuracy of the model is influenced by overhigh and overlow parameters, model default values are adopted for penalty coefficients and regression boundaries of the model, and a linear kernel function is adopted for the kernel function, and the method is specifically shown as follows;
TABLE 1 Soft measurement model parameters of fiber morphology
The method comprises the steps of establishing a water absorption soft measurement model based on a gradient lifting regression tree, wherein the gradient lifting regression tree comprises a regression tree, tree depth, learning rate and loss functions. The number of the regression trees is the combination of the regression trees; the tree depth is the depth of the maximum tree (>2 levels), all selected here as 3 levels; the learning rate is the step length of each descending of the loss function in the gradient direction, the calculation time is increased when the learning rate is too small, and the accuracy of the predicted value is influenced when the learning rate is too large, so that the learning rate of the subject is 0.01; the loss function is usually least square, minimum absolute value difference and the like, and the least square method is adopted by default.
TABLE 2 Water absorption Soft measurement model parameters
Obtaining an input of a sample to be tested, comprising: the method comprises the following steps of (1) obtaining original pulp board fiber morphology, pulp ratio, pulp grinder power, pulp grinder concentration, pulp grinder flow and pulp fiber morphology after pulp grinding; and preprocessing the variable into a fiber form soft measurement model input variable, training and modeling the input variable, and performing soft measurement on the fiber form after pulp grinding. And then, carrying out variable selection on parameters in the paper machine process by utilizing correlation analysis, obtaining variables such as vehicle speed, vacuum cylinder mould speed and the like, and combining the average length of the ground fibers, the average thickness of the fibers, the percentage of kinked fibers and the average width characteristics of the fibers which are predicted by the fiber form soft measurement model and the gradient lifting decision tree model in the step S72 to carry out soft measurement on the water absorption of the base paper.
In the embodiment, the soft measurement model of the fiber morphology is established based on the SVM algorithm by utilizing the steps, and the output is the fiber morphology after pulp grinding. The results of the soft measurement model of fiber morphology are shown in fig. 2. As can be seen from FIG. 2, the fitting accuracy of the fiber soft measurement model is good, and the evaluation index adopts the average relative error in the specification. The average fiber length fitting average relative error is 2.84%, the average kinked fiber percentage relative error is 3.12%, the average fiber width average relative error is 2.91%, and the average fiber thickness average relative error is 3.09%, all of which are good in performance, after the fiber morphology is soft-measured, a paper sheet water absorption soft measurement model is established through the variable selection method based on a GBRT algorithm and simultaneously in combination with paper machine processes such as vehicle speed, coiling rate and the like selected by correlation analysis (the result is shown in figure 3). The water absorption soft measurement model results are shown in fig. 4. As can be seen from FIG. 4, the average relative error of water absorption was 3.45%, and the model accuracy was good.
After the fiber form soft measurement model and the water absorption soft measurement model are established, the fiber form soft measurement model and the water absorption soft measurement model are combined into a paper sheet water absorption soft measurement model, the original pulp board fiber form, pulp board proportion, pulp grinding process, paper machine process and the like are input, the model outputs the water absorption of paper sheets, production data in another time interval are collected again to carry out universality verification on the established paper sheet water absorption model, and the model verification result is shown in fig. 5. As can be seen from FIG. 5, the average relative error of the sheet water absorption soft measurement model was 3.27%, and the model accuracy was good.
It will be appreciated by those of ordinary skill in the art that the examples described herein are intended to assist the reader in understanding the manner in which the invention is practiced, and it is to be understood that the scope of the invention is not limited to such specifically recited statements and examples. Those skilled in the art can make various other specific changes and combinations based on the teachings of the present invention without departing from the spirit of the invention, and these changes and combinations are within the scope of the invention.
Claims (5)
1. A method for soft measurement of water absorption of paper sheets is characterized by comprising the following steps:
s1, collecting key modeling data based on the production process, wherein the key modeling data comprises a database and field data variables; the variables include pulp fiber morphology measurements, refiner power, throughput, and paper machine reel speed;
s2, analyzing the data of the collected variables, qualitatively analyzing the correlation between the collected variables and the water absorption, and preprocessing the data of the variables to meet the requirements of fiber morphology soft measurement modeling and paper sheet water absorption soft measurement modeling;
s3, establishing a fiber morphology soft measurement model by combining a support vector machine algorithm according to data meeting modeling requirements;
s4, performing feature selection on data meeting modeling requirements, establishing a paper sheet water absorption soft measurement model by combining a gradient lifting decision tree algorithm, and then performing model validity verification by using on-site data;
the substeps of S4 are as follows:
s41, establishing a soft measurement model of paper sheet water absorption based on the model principle of the gradient lifting decision tree, and recording Fm(x) The specific formula of the overall model function is as follows:
wherein, Fm(x) As a function of the global model, Fm-1(x) Is the (M-1) th basic regression tree function, M is 1,2 … M is the regression tree, betamIs the weight of the mth regression tree, L is the modulusLoss function of type, αmIs a parameter in the mth regression tree;
s42, aiming at the vehicle speed and the vacuum cylinder mould speed in the paper machine process selected by the correlation analysis, combining the average length of the refined fibers, the average thickness of the fibers, the percentage of kinked fibers and the average width characteristics of the fibers output by the fiber form soft measurement model, and establishing a paper sheet water absorption model by utilizing a gradient lifting regression decision tree.
2. A sheet absorbency soft measurement method as set forth in claim 1, wherein: the correlation between the qualitative analysis acquisition variable and the water absorption in the step S2 is specifically as follows: the formation of the water absorption of the paper sheet is greatly related to the average length of the fibers, the content of the fine fibers and the bonding force among the fibers, the average length of the fibers is short, the content of the fine fibers is increased, the water absorption performance of the paper sheet is reduced, the bonding force of the fiber quality inspection is low, and the water absorption of the finished paper is strong; the average fiber length and the fine fiber content are related to the fiber shape of an original pulp board, the power and the throughput of a refiner in a pulping process, and the formation of the bonding force among fibers is related to the slurry flow in a papermaking stage and the vehicle speed and the vacuum cylinder mould speed in a forming process, so that the shape indexes of the average fiber length, the average fiber thickness, the percentage of kinked fibers and the average fiber width are collected, and the key parameters of the vehicle speed and the vacuum cylinder mould speed in the paper machine process are used as selection variables input by a paper sheet water absorption model;
the data analysis is used for checking whether dirty data and data which cannot be directly analyzed exist in original data, wherein the dirty data comprises missing values, abnormal values and data containing special characters; deleting the missing value, the abnormal value and the character of the special symbol;
data preprocessing in S2 is performed to meet modeling requirements, and input variables of the model are normalized, so that indexes of different units or orders can be compared and weighted conveniently, and influence on modeling caused by different characteristic dimensions and orders of different orders is eliminated; the processing mode is to compress the selected characteristic variables to obtain a new number series with the mean value of 0 and the standard deviation of 1; the specific mode is as follows:
wherein i is 1,2, …, n is the number of samples, j is 1,2, …, and p is the dimension of the samples;is the mean value of the samples in the j dimension, xijIs the j-dimensional value of the ith sample,for the j-dimensional normalized value of the i-th sample, SjThe standard deviation of the sample in the j dimension is taken; sj 2Is the variance of the sample in the j dimension.
3. A sheet absorbency soft measurement method as set forth in claim 1, wherein: selecting characteristics of a paper sheet water absorption soft measurement model in S4, performing correlation analysis on the preprocessed characteristics, and selecting variables with obvious influence on water absorption to model, wherein a Pearson correlation coefficient is adopted in the correlation analysis, and a specific calculation formula is as follows;
wherein the content of the first and second substances,is y andof (a) covariance, Var [ y [)],Are respectively y andthe variance of (a), E (y),are respectively y and(iii) a desire;
the correlation coefficient can reflect the degree of correlation between two variables, the value of the correlation coefficient is in the interval < -1,1 >, a positive value represents positive correlation, a negative value represents negative correlation, the closer the absolute value is to 1, the stronger the correlation between the two is, and the value of 0 represents no correlation; through the correlation analysis of the paper machine process and the water absorption of paper sheets, a variable with the correlation of the water absorption of more than 0.4 is selected as the input of a water absorption soft measurement model.
4. A sheet absorbency soft measurement method as set forth in claim 1, wherein: the substeps of S3 are as follows:
s31, establishing a fiber morphology soft measurement model based on the principle of a support vector machine regression model; the kernel function can map low-dimensional data to a high-dimensional space and convert a nonlinear problem into a linear problem; the principle of kernel function selection is that the linear indivisible original data is mapped to the high-dimensional characteristic space; for a given data set, the solution formula for the SVM model is as follows
Where ω is a weight coefficient and x is an input variable [ x ]1,…xp]n×pB is an offset term, αi,Is Lagrange multiplier, K (x)i,xj) Is a kernel function;
s32, aiming at the fiber form soft measurement model and the data set, the model input is the original pulp fiber form, pulp board ratio, refiner power, throughput and concentration, and the model output is the ground pulp fiber form; adopting a cross verification method, randomly selecting 80% of training sets to train and adjust model parameters each time, and verifying the precision of the parameters by 20% of testing sets;
and S33, establishing a fiber form soft measurement model by using the parameters with better performance in the step S32, and carrying out soft measurement on the subsequent fiber form.
5. A sheet absorbency soft measurement method as set forth in claim 1, wherein: and in the step S4, the validity verification is to collect the production data again in another period of time, perform data analysis and data preprocessing, predict by using the built model and compare the model with the actual value so as to verify the validity of the model on the new data.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911330641.2A CN111024557B (en) | 2019-12-20 | 2019-12-20 | Soft measurement method for water absorption of paper sheet |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911330641.2A CN111024557B (en) | 2019-12-20 | 2019-12-20 | Soft measurement method for water absorption of paper sheet |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111024557A CN111024557A (en) | 2020-04-17 |
CN111024557B true CN111024557B (en) | 2021-06-04 |
Family
ID=70212400
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911330641.2A Active CN111024557B (en) | 2019-12-20 | 2019-12-20 | Soft measurement method for water absorption of paper sheet |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111024557B (en) |
Family Cites Families (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US4823008A (en) * | 1987-11-05 | 1989-04-18 | Process Automation Business, Inc. | Apparatus and methods employing infrared absorption means to measure the moisture content of heavy grades of paper |
DE10326489A1 (en) * | 2003-06-10 | 2005-01-05 | Igt Emus Gmbh | Water absorption and wetting properties measuring device for use with absorbent paper, e.g. hand towels or diapers, comprises computer controlled optical imaging equipment together with frame-grabbing and image processing personal computer |
TWM403653U (en) * | 2010-09-29 | 2011-05-11 | Univ Chinese Overseas | A new instrument of determining the water absorption of paper was proposed. |
CN102564362B (en) * | 2011-12-23 | 2014-04-16 | 陕西科技大学 | Online soft measuring method of paper smoothness |
CN104881715B (en) * | 2015-05-26 | 2018-11-02 | 华南理工大学 | A kind of method that paper plant predicts pulp property according to waste paper proportioning |
CN205120527U (en) * | 2015-11-25 | 2016-03-30 | 天津市中科健新材料技术有限公司 | Paper that absorbs water capability test device that absorbs water |
KR20180010755A (en) * | 2016-07-22 | 2018-01-31 | 주식회사 새롬 | Moisture absorption functional food packaging |
CN106503745A (en) * | 2016-10-26 | 2017-03-15 | 郑州埃文计算机科技有限公司 | Contextual information Forecasting Methodology based on social picture SIFT feature |
CN209167249U (en) * | 2018-05-02 | 2019-07-26 | 广东互爱健康产业科技有限公司 | A kind of water imbibition detection device of paper diaper |
CN109036556B (en) * | 2018-08-29 | 2021-06-04 | 王雁 | Method for diagnosing keratoconus case based on machine learning |
CN109440515B (en) * | 2018-11-30 | 2020-06-19 | 华南理工大学 | Soft measurement method for beating degree in papermaking process based on gradient enhanced regression tree |
-
2019
- 2019-12-20 CN CN201911330641.2A patent/CN111024557B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN111024557A (en) | 2020-04-17 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103033362B (en) | Gear fault diagnosis method based on improving multivariable predictive models | |
CN110110768A (en) | Fault Diagnosis of Roller Bearings based on Concurrent Feature study and multi-categorizer | |
CN116167640B (en) | LCP film production quality detection data analysis method and system | |
CN107885928A (en) | Consider the stepstress acceleration Degradation Reliability analysis method of measurement error | |
CN115147208A (en) | Supply chain financial credit risk evaluation method and system based on artificial intelligence | |
CN112446597B (en) | Storage tank quality assessment method, storage tank quality assessment system, storage medium, computer equipment and application | |
CN105868164A (en) | Soft measurement modeling method based on monitored linear dynamic system model | |
CN113240527A (en) | Bond market default risk early warning method based on interpretable machine learning | |
CN114429152A (en) | Rolling bearing fault diagnosis method based on dynamic index antagonism self-adaption | |
CN112904810A (en) | Process industry nonlinear process monitoring method based on effective feature selection | |
CN111402236A (en) | Hot-rolled strip steel surface defect grading method based on image gray value | |
CN110910528B (en) | Method and device for predicting tensile strength of paper sheet | |
CN110320337A (en) | A kind of automatic evaluating system of strip surface quality and method | |
CN111024557B (en) | Soft measurement method for water absorption of paper sheet | |
CN108665090B (en) | Urban power grid saturation load prediction method based on principal component analysis and Verhulst model | |
CN111091153B (en) | Soft measurement method for bulk of paper sheet | |
CN114492614A (en) | Method and device for classifying faults in hot rolling process of strip steel based on ensemble learning | |
CN112069621B (en) | Method for predicting residual service life of rolling bearing based on linear reliability index | |
CN107122609A (en) | A kind of electromechanical product quality evaluation method based on mass property gene theory | |
CN111079344A (en) | Method and system for measuring softness of paper sheet | |
CN116127398B (en) | Hydraulic pump fault diagnosis method based on mechanism model and multi-source data fusion | |
CN116500456B (en) | Ternary lithium battery capacity detection method based on machine learning | |
CN113324937B (en) | Biomass fuel quality inspection method and system based on terahertz spectrum | |
CN115455627A (en) | Method for evaluating generating capacity of wind turbine generator | |
CN113239021B (en) | Data migration method for predicting residual life of similar products |
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 | ||
PE01 | Entry into force of the registration of the contract for pledge of patent right |
Denomination of invention: A Soft Measurement Method for Water Absorption of Paper Pages Granted publication date: 20210604 Pledgee: Societe Generale Bank Limited by Share Ltd. Guangzhou branch Pledgor: GUANGZHOU POI INTELLIGENT INFORMATION TECHNOLOGY CO.,LTD. Registration number: Y2024980007912 |
|
PE01 | Entry into force of the registration of the contract for pledge of patent right |