CN116602435A - Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine - Google Patents
Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine Download PDFInfo
- Publication number
- CN116602435A CN116602435A CN202310805923.3A CN202310805923A CN116602435A CN 116602435 A CN116602435 A CN 116602435A CN 202310805923 A CN202310805923 A CN 202310805923A CN 116602435 A CN116602435 A CN 116602435A
- Authority
- CN
- China
- Prior art keywords
- data
- cut tobacco
- water content
- moisture
- model
- 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.)
- Pending
Links
- 241000208125 Nicotiana Species 0.000 title claims abstract description 72
- 235000002637 Nicotiana tabacum Nutrition 0.000 title claims abstract description 72
- 238000000034 method Methods 0.000 title claims abstract description 56
- 230000008859 change Effects 0.000 title claims abstract description 26
- 238000010801 machine learning Methods 0.000 title claims abstract description 19
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims abstract description 50
- 239000000463 material Substances 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims abstract description 9
- 230000007613 environmental effect Effects 0.000 claims abstract description 8
- 235000019504 cigarettes Nutrition 0.000 claims abstract description 7
- 238000004519 manufacturing process Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 12
- 238000009826 distribution Methods 0.000 claims description 10
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 6
- 238000011156 evaluation Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 6
- 238000012795 verification Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 5
- 238000011217 control strategy Methods 0.000 claims description 5
- 238000005457 optimization Methods 0.000 claims description 5
- 238000007476 Maximum Likelihood Methods 0.000 claims description 4
- 238000004891 communication Methods 0.000 claims description 4
- 238000012731 temporal analysis Methods 0.000 claims description 4
- 238000000700 time series analysis Methods 0.000 claims description 4
- 238000001276 Kolmogorov–Smirnov test Methods 0.000 claims description 3
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 238000004140 cleaning Methods 0.000 claims description 3
- 238000013135 deep learning Methods 0.000 claims description 3
- 230000004927 fusion Effects 0.000 claims description 3
- 230000010354 integration Effects 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 3
- 230000002123 temporal effect Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 238000011426 transformation method Methods 0.000 claims description 3
- 238000010521 absorption reaction Methods 0.000 abstract description 6
- 238000001035 drying Methods 0.000 description 6
- 238000004422 calculation algorithm Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000011551 log transformation method Methods 0.000 description 4
- 238000002156 mixing Methods 0.000 description 3
- 230000003068 static effect Effects 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 241000932075 Priacanthus hamrur Species 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 230000007774 longterm Effects 0.000 description 2
- 230000002787 reinforcement Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000035772 mutation Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000001105 regulatory effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- A—HUMAN NECESSITIES
- A24—TOBACCO; CIGARS; CIGARETTES; SIMULATED SMOKING DEVICES; SMOKERS' REQUISITES
- A24B—MANUFACTURE OR PREPARATION OF TOBACCO FOR SMOKING OR CHEWING; TOBACCO; SNUFF
- A24B9/00—Control of the moisture content of tobacco products, e.g. cigars, cigarettes, pipe tobacco
-
- A—HUMAN NECESSITIES
- A24—TOBACCO; CIGARS; CIGARETTES; SIMULATED SMOKING DEVICES; SMOKERS' REQUISITES
- A24B—MANUFACTURE OR PREPARATION OF TOBACCO FOR SMOKING OR CHEWING; TOBACCO; SNUFF
- A24B3/00—Preparing tobacco in the factory
- A24B3/04—Humidifying or drying tobacco bunches or cut tobacco
-
- 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
Landscapes
- Manufacture Of Tobacco Products (AREA)
Abstract
The invention discloses a machine learning-based method for analyzing the moisture change of cut tobacco in a cut tobacco making single machine, which comprises the following steps: A. historical data acquisition: and reading historical 2-5 years of data from a data platform of a cigarette factory, wherein the data comprise inlet water content, outlet water content, material characteristics, operation parameters of various equipment and environmental parameters. According to the invention, a large amount of data can be automatically acquired and analyzed through training and learning, modeling prediction with robustness and generalization capability is performed, and the absorption and dissipation efficiency of tobacco moisture can be more accurately predicted.
Description
Technical Field
The invention relates to the technical field of a tobacco shred manufacturing process, in particular to a tobacco shred moisture change analysis method in a tobacco shred manufacturing single machine based on machine learning.
Background
The invention provides a modeling method for a tobacco shred drying middle process of a cut tobacco dryer, which is disclosed by the Chinese patent, and has the publication number of CN110580326B, and the modeling method fully utilizes the characteristic that VCN model parameters are separable, optimizes the model linear parameters and nonlinear parameters respectively, and improves the prediction precision, but only utilizes the data of 4 sensors, namely an inlet tobacco shred moisture content detector arranged at the inlet of a roller, a roller barrel temperature detection sensor arranged in the middle of the roller, a moisture removal air valve electric regulating valve arranged at the upper end of the tail part of the roller, and an outlet tobacco shred moisture content detection sensor arranged at the outlet of the roller; other parameters of the equipment are not utilized, namely the influence of other parameters is not considered in the model, other variables cannot be analyzed and controlled, the model is relatively fixed in form and poor in adaptability and generalization capability, and under-fitting conditions can occur when the characteristics of the cut-tobacco drier data are greatly changed or the method is applied to other single machines; in the prior art, a paper Optimization of tobacco drying process control based on reinforcement learning is disclosed, and based on a control strategy of a deep reinforcement Q learning algorithm, the accurate control of the moisture content of tobacco shreds is realized by measuring and adjusting parameters such as the degree, the humidity and the like in the drying process in real time, but the method has some defects: 1. data acquisition and processing: the method requires a large amount of data for training, but in practical application, the data acquisition and processing may be limited, such as sensing precision, sampling frequency and the like, so that it is a problem to be solved how to effectively acquire and process the data; 2. interpretation of the model: the deep Q learning algorithm is a black box model, the decision process of the model is difficult to explain, and in practical application, the decision process of the model needs to be understood so as to carry out debugging optimization; 3. robustness of the algorithm: the performance of the method can be influenced by external environment such as weather, temperature and other factors; the invention discloses a method for on-line adjusting mixed yarn perfuming moisture, which is disclosed in Chinese patent with the bulletin number of CN111248485A, and adopts a multiple linear regression method to establish a moisture control model according to experimental data: tobacco shred moisture= 9.309-0.112, tobacco shred moisture+0.215, tobacco shred moisture+0.161; according to the production process flow of cigarettes, the moisture of blended cut stems and recycled cut tobacco is determined during cut tobacco blending, the range of target values is designed according to the process of cut tobacco blending and cigarette cut tobacco moisture, the range of the cut tobacco moisture after baking is calculated according to the moisture control model, and the online linkage control adjustment of the cut tobacco blending and the cigarette cut tobacco moisture is realized, but the multiple linear regression method used by the invention cannot describe the nonlinear relation of the moisture loss of a flavoring machine, and the nonlinear relation is more in line with the actual situation; the regression coefficient needs to be fitted again according to different material characteristics and environmental parameters in the current production, and cannot be used as a conclusion; meanwhile, only modeling is carried out on the water loss of the flavoring machine, and other equipment on the whole line is not considered; the invention discloses a Chinese patent ' an intelligent prediction model for accurately controlling a sheet drying process and application ', wherein the publication number is CN112434868A ', and the invention selects historical data of a plurality of set time periods of the sheet drying process as a sample to be selected; optionally, historical data of a set time period is used as sample data; calculating to obtain a sample database by using a simple moving average method; screening out key parameters affecting the water content of the discharged material of the dried shreds by using a SCAD method; based on a sample database, a feedforward neural network method or a support vector machine big data method is used for constructing an intelligent prediction model of the sheet drying key parameters, but the method uses simple moving average to obtain a sample database which is rough and does not translate time lag, the used sample has poor possible effect on subsequent training, the characteristics screened by SCAD dimension reduction are also obviously influenced by sample data, and meanwhile, the number of layers and complexity of the feedforward neural network are required to be improved so as to improve the prediction precision and balance the fitting problem, and the method is not popularized to other single machines.
Because the moisture absorption and dissipation efficiency of the tobacco shreds has an important influence on the quality, taste and production cost of the tobacco shreds in the traditional tobacco shred manufacturing process, the existing methods often depend on manual experience and regular sampling detection, and the methods have limitations in terms of instantaneity, accuracy and stability; in addition, the absorption and dissipation of the moisture in the tobacco shreds are affected by various factors, such as raw material characteristics, environmental conditions, equipment parameters and the like, which make the monitoring and control of the moisture absorption and dissipation efficiency more complex.
Disclosure of Invention
The invention aims to provide a machine learning-based method for analyzing the moisture change of cut tobacco in a cut tobacco manufacturing single machine, so as to solve the problems in the background technology.
In order to achieve the above purpose, the present invention provides the following technical solutions: a machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine comprises the following steps:
A. historical data acquisition: reading historical 2-5 years data from a data platform of a cigarette factory, wherein the data comprise inlet water content, outlet water content, material characteristics, operation parameters of various devices and environmental parameters;
B. data preprocessing: preprocessing the collected original data, including data cleaning, missing value processing, abnormal value detection and denoising;
C. time lag alignment of features: through the combination of the service information and various correlation methods, the inlet water content and other characteristic data related to the outlet water content are respectively aligned with the outlet water content, so that the subsequent modeling can be performed, otherwise, all the characteristics of the sample are not data acquired by the same object;
D. characteristic engineering: extracting relevant characteristics from the acquired data, wherein the relevant characteristics are used for describing the change relation between the outlet water content and the inlet water content and comprise time sequence characteristics, statistical characteristics and material characteristic parameters;
E. and (3) establishing a model: establishing a prediction model by using a method based on the combination of a Temporal Fusion Transformer deep neural network and a multivariate time series analysis model Prophet, wherein training of the model uses historical data and known outlet water content and inlet water content as target values;
F. model verification and tuning: verifying the established model by using a verification data set, and tuning to improve the accuracy and generalization capability of the model;
G. and (3) real-time data acquisition: the method comprises the steps of collecting data related to tobacco shred moisture in real time through a PLC (programmable logic controller) and sending the data to a local computer where a model is deployed;
H. analysis and prediction of moisture change: analyzing and predicting data acquired in real time by using the established model to obtain the change relation between the outlet water content and the inlet water content;
I. feedback control and optimization: continuously feeding back the data to the system through real-time monitoring and acquisition, and dynamically adjusting and optimizing the model;
J. system integration and implementation: the method is applied to each single machine and auxiliary equipment of a wire making workshop, and a complete system is built;
K. outputting evaluation indexes and updating the model regularly: and when the evaluation index gives early warning or reaches the set time, retraining the model and completing automatic deployment.
Preferably, the step A further comprises cut tobacco dryer inlet moisture, tobacco flow, roller rotating speed, roller barrel temperature, hot air speed, hot air temperature, air door opening, cut tobacco dryer outlet moisture set value, actual value, outlet temperature, season, weather, environment temperature and humidity, shift, license plate, production place and year.
Preferably, in the step C, when the correlation coefficient is calculated by using different correlation methods, it is required to calculate whether the distribution of the features satisfies the assumed condition of the correlation method, for example, when Pearson or covariance is used, the sample needs to satisfy normal distribution, and when the distribution is not satisfied, the normal distribution is checked by using K-S test, and if the distribution is not satisfied, the normal distribution is enhanced by using ECDF, box-Cox or Yoe-Johnson method, so that the calculation result of the correlation is more in line with the actual situation.
Preferably, the power transformation methods of Box-Cox and Yoe-Johnson are as follows:and (2) andmeanwhile, when if lambda is not equal to 0, y>When 0, ifλ=0, y is equal to or greater than 0, ifλ+.2, y<0, ifλ=2, y<0, both of which are solutions to the parameter lambda using maximum likelihood estimation based on input data y, i.e., respective parameter sequences, such as inlet water content, outlet water content set point, outlet water content actual value.
Preferably, in the step E, two high-performance and high-interpretability deep learning and machine learning methods are integrated, and the inlet moisture, the tobacco flow, the rotating speed of the drum, the drum temperature, the hot air speed, the hot air temperature, the opening degree of the air door, the outlet moisture set value, the season, the weather, the environment temperature and humidity, the shift, the brand, the production place, the year and the outlet temperature of the cut-tobacco dryer are taken as input values, and the outlet moisture actual value of the cut-tobacco dryer is taken as an output value.
Preferably, in the step G, for the equipment controlled by the PID and the PLC, the PID control mode needs to be set to be manual, the real-time control variable value is transmitted to the PID, the PID is then communicated to the PLC, the control variable value cannot be directly transmitted to the PLC, the PID is then transmitted to the PLC after passing through the control logic thereof, and the steps of deployment are as follows: firstly, determining a network protocol, then writing a communication code, secondly, configuring PID parameters so as to receive data from an external computer, and finally, testing and debugging to ensure normal operation.
Preferably, in the step J, the system needs to include a data acquisition and processing module, a model building and optimizing module, and a control strategy generating and adjusting module, and meanwhile, needs to be integrated with an existing automatic control system of the wire making workshop.
Compared with the prior art, the invention has the following beneficial effects:
the invention is based on a machine learning algorithm, can automatically acquire and analyze a large amount of data by training and learning, carries out modeling prediction with robustness and generalization capability, can more accurately predict the absorption and dissipation efficiency of tobacco moisture, and compared with the traditional method, the technical scheme can more accurately predict the change relation between the outlet moisture content and the inlet moisture content, improves the accuracy of data analysis and prediction, effectively improves the tobacco quality, the production efficiency and the automation level, has better mobility and guidance on services, and brings practical economic and social benefits for tobacco industry.
Detailed Description
The technical solutions of the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
A machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine comprises the following steps:
A. historical data acquisition: reading historical 2-5 years data from a data platform of a cigarette factory, wherein the data comprise inlet water content, outlet water content, material characteristics, operation parameters of various devices and environmental parameters;
B. data preprocessing: preprocessing the collected original data, including data cleaning, missing value processing, abnormal value detection and denoising processing, so as to ensure the quality and reliability of the data and prepare for subsequent analysis and modeling;
C. time lag alignment of features: through the combination of the service information and various correlation methods, the inlet water content and other characteristic data related to the outlet water content are respectively aligned with the outlet water content, so that the subsequent modeling can be performed, otherwise, all the characteristics of the sample are not data acquired by the same object;
D. characteristic engineering: extracting relevant features from the acquired data, wherein the relevant features are used for describing the change relation between the outlet water content and the inlet water content, and comprise time sequence features, statistical features and material characteristic parameters, and the aim is to extract the features with the most information content so as to support the subsequent modeling of a machine learning algorithm;
E. and (3) establishing a model: establishing a prediction model by using a method based on a Temporal Fusion Transformer deep neural network (TFT) and a multivariate time series analysis model Prophet, using historical data and known outlet water content and inlet water content as target values for training the model, and accurately predicting the change relation of the outlet water content and the inlet water content by iteratively optimizing model parameters;
F. model verification and tuning: verifying the established model by using a verification data set, and tuning to improve the accuracy and generalization capability of the model;
G. and (3) real-time data acquisition: the method comprises the steps of collecting data related to tobacco shred moisture in real time through a PLC (programmable logic controller) and sending the data to a local computer where a model is deployed;
H. analysis and prediction of moisture change: analyzing and predicting the data acquired in real time by using the established model to obtain the change relation between the outlet water content and the inlet water content, and according to the analysis result, evaluating the absorption and dissipation efficiency of the water content of the cut tobacco in the single cut tobacco making machine and predicting the change trend of the outlet water content under different material characteristics, equipment operation parameters and environmental parameters;
I. feedback control and optimization: the data which are monitored and collected in real time are continuously fed back to the system, and the model is dynamically adjusted and optimized to ensure the accuracy and stability of the model, so that a control strategy can be optimized, and the outlet moisture of a single wire making machine can be rapidly and accurately controlled under different working conditions;
J. system integration and implementation: the method is applied to each single machine and auxiliary equipment of a wire making workshop, and a complete system is built;
K. outputting evaluation indexes and updating the model regularly: and when the evaluation index gives early warning or reaches the set time, retraining the model and completing automatic deployment.
The step A also comprises cut tobacco dryer inlet moisture, tobacco flow, drum rotation speed, drum temperature, hot air speed, hot air temperature, air door opening, cut tobacco dryer outlet moisture set value, actual value, outlet temperature, season, weather, environment temperature and humidity, shift, license plate number, production place and year.
In step C, when the correlation coefficient is calculated by using different correlation methods, whether the distribution of the feature meets the assumption condition of the correlation method is needed to be calculated, for example, when Pearson or covariance is used, the sample needs to meet normal distribution, K-S test is used for checking the normal property, and if not, ECDF, box-Cox or Yoe-Johnson method is used for enhancing the normal property, so that the calculation result of the correlation is more in accordance with the actual situation.
Preferably, the power transformation methods of Box-Cox and Yoe-Johnson are as follows:and (2) andmeanwhile, when if lambda is not equal to 0, y>When 0, ifλ=0, y is equal to or greater than 0, ifλ+.2, y<0, ifλ=2, y<0, both of which are obtained by solving the parameter lambda using maximum likelihood estimation based on input data y, i.e., respective parameter sequences such as inlet water content, outlet water content set point, outlet water content actual value, it can be seen that when lambda=0 and y are solved using Box-Cox power transform>When the λ=0 and y is equal to or larger than 0, the log (y) is equivalent to the log (y+1) when the Yoe-Johnson power transformation is used for solving the λ=0, so that the power transformation can be regarded as generalized expression of the log transformation, the generalized expression of the log transformation can be enhanced, but the power transformation determines which expression is adopted according to the specific input data through maximum likelihood estimation, so that the applicability of the log transformation is wider than that of the log transformation, the normalization degree of the data is improved more, the assumption condition of subsequent correlation calculation can be met more, the calculated correlation coefficient is more reliable, and the time lag calculation of each characteristic of the equipment and the actual value of outlet moisture can be more accurate.
In the step E, two high-performance and high-interpretability deep learning and machine learning methods are integrated, and the inlet moisture, the tobacco flow, the rotating speed of a roller, the temperature of the roller, the wind speed of hot air, the temperature of hot air, the opening degree of an air door, the set value of outlet moisture, seasons, weather, environmental temperature and humidity, shifts, marks, places of production, years and the outlet temperature of the cut-tobacco drier are taken as input values, the actual value of the outlet moisture of the cut-tobacco drier is taken as output values, wherein the output values comprise static (instant unchanged) covariates and other exogenous time sequences which are only observed historically, no prior information about how the output values interact with a target is provided, and the TFT is taken as an example for illustration: the neural network used in the past is usually a black box model, and various types of inputs occurring in a common scene are not considered, and the main components of the TFT are: (1) Control mechanism, skipping any unused components in the architecture, providing adaptive depth and network complexity to accommodate a wide range of data sets and scenarios, gatesThe linear control unit is widely applied to the whole system structure, and the gate control residual network is proposed as a main building module; (2) A variable selection network for selecting an associated input variable at each time step; (3) The static covariate encoder is used for merging static characteristics into the network in a conditional time dynamic mode by encoding the context vector; (4) Time processing, learning long-term and short-term time relationships while naturally processing observed and a priori known time-varying inputs, a sequence-sequence layer is used for local feature processing, and long-term dependencies are captured using a new interpretable multi-headed attention block; (5) Multi-level prediction interval prediction, quantile prediction is generated at each prediction level, and the following formula is the principle of TFT generating explanatory attention:the multivariate time series analysis model is described by taking an addition model in Prophet as an example (other multiplication models, an addition multiplication mixed model and a regress model with related variables are similar principles), wherein y (t) is the actual value of moisture of a dependent variable outlet, g (t), s (t), h (t) is the decomposition of a linear term, a periodic term and a mutation term of y (t), other related features are expressed by the regress term, and thus the relation between each partial term and the dependent variable is decomposed and expressed, so that the model has strong principle and interpretability, and training logic of the model is as follows:
model {
// Priors
k ∼ normal(0, 5);
m ∼ normal(0, 5);
epsilon ∼ normal(0, 0.5);
delta ∼ double_exponential(0, tau);
beta ∼ normal(0, sigma);
// Logistic likelihood
y ∼ normal(C ./ (1 + exp(-(k + A * delta) .* (t - (m + A * gamma))))+
X * beta, epsilon);
// Linear likelihood
y ∼ normal((k + A * delta) .* t + (m + A * gamma) + X * beta, sigma);
}。
in the step G, for equipment controlled by the PID and the PLC, the PID control mode is set to be manual, the real-time control variable value is transmitted to the PID, the PID is communicated to the PLC, and the control variable value cannot be directly transmitted to the PLC, because the PID parameter is used for training a model, the PID is also required to be transmitted to the value during the control, the PID is transmitted to the PLC after passing through the control logic, the operation of the element corresponding to the machine control variable is realized, and the deployment steps are divided into: firstly, determining a network protocol (Ethernet/IP), then writing a communication code (Socket or API function is written by Python to communicate with a network communication module of PID), secondly, configuring PID parameters so as to receive data from an external computer, and finally, testing and debugging to ensure normal operation.
In the step J, the system needs to comprise a data acquisition and processing module, a model building and optimizing module and a control strategy generating and adjusting module, and meanwhile, the system also needs to be integrated with the existing automatic control system of the wire making workshop to realize automatic control of outlet moisture.
Although embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made therein without departing from the principles and spirit of the invention, the scope of which is defined in the appended claims and their equivalents.
Claims (7)
1. A machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine is characterized by comprising the following steps: the method comprises the following steps:
A. historical data acquisition: reading historical 2-5 years data from a data platform of a cigarette factory, wherein the data comprise inlet water content, outlet water content, material characteristics, operation parameters of various devices and environmental parameters;
B. data preprocessing: preprocessing the collected original data, including data cleaning, missing value processing, abnormal value detection and denoising;
C. time lag alignment of features: through the combination of the service information and various correlation methods, the inlet water content and other characteristic data related to the outlet water content are respectively aligned with the outlet water content, so that the subsequent modeling can be performed, otherwise, all the characteristics of the sample are not data acquired by the same object;
D. characteristic engineering: extracting relevant characteristics from the acquired data, wherein the relevant characteristics are used for describing the change relation between the outlet water content and the inlet water content and comprise time sequence characteristics, statistical characteristics and material characteristic parameters;
E. and (3) establishing a model: establishing a prediction model by using a method based on the combination of a Temporal Fusion Transformer deep neural network and a multivariate time series analysis model Prophet, wherein training of the model uses historical data and known outlet water content and inlet water content as target values;
F. model verification and tuning: verifying the established model by using a verification data set, and tuning to improve the accuracy and generalization capability of the model;
G. and (3) real-time data acquisition: the method comprises the steps of collecting data related to tobacco shred moisture in real time through a PLC (programmable logic controller) and sending the data to a local computer where a model is deployed;
H. analysis and prediction of moisture change: analyzing and predicting data acquired in real time by using the established model to obtain the change relation between the outlet water content and the inlet water content;
I. feedback control and optimization: continuously feeding back the data to the system through real-time monitoring and acquisition, and dynamically adjusting and optimizing the model;
J. system integration and implementation: the method is applied to each single machine and auxiliary equipment of a wire making workshop, and a complete system is built;
K. outputting evaluation indexes and updating the model regularly: and when the evaluation index gives early warning or reaches the set time, retraining the model and completing automatic deployment.
2. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: the step A also comprises cut tobacco dryer inlet moisture, tobacco flow, roller rotating speed, roller temperature, hot air speed, hot air temperature, air door opening, cut tobacco dryer outlet moisture set value, actual value, outlet temperature, season, weather, environment temperature and humidity, shift, brand, production place and year.
3. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: in the step C, when the correlation coefficient is calculated by using different correlation methods, whether the distribution of the feature meets the assumption condition of the correlation method is required to be calculated, for example, when Pearson or covariance is used, the sample needs to meet normal distribution, the normal property is checked by using K-S test, if not, the normal property is enhanced by using ECDF, box-Cox or Yoe-Johnson method, so that the calculation result of the correlation is more in accordance with the actual situation.
4. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: the power transformation method of Box-Cox and Yoe-Johnson comprises the following steps:and (2) andmeanwhile, when if lambda is not equal to 0, y>When 0, ifλ=0, y is equal to or greater than 0, ifλ+.2, y<0, ifλ=2, y<0, both of which are solutions to the parameter lambda using maximum likelihood estimation based on input data y, i.e., respective parameter sequences, such as inlet water content, outlet water content set point, outlet water content actual value.
5. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: in the step E, two high-performance and high-interpretability deep learning and machine learning methods are integrated, and the inlet moisture, tobacco flow, drum rotation speed, drum temperature, hot air speed, hot air temperature, air door opening, outlet moisture set value, season, weather, environment temperature and humidity, shift, brand, production place, year and outlet temperature of the cut tobacco dryer are taken as input values, and the outlet moisture actual value of the cut tobacco dryer is taken as an output value.
6. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: in the step G, for the equipment controlled by the PID and the PLC, the PID control mode needs to be set to be manual, the real-time control variable value is transmitted to the PID, the PID is then communicated to the PLC, the control variable value cannot be directly transmitted to the PLC, after passing through the control logic, the PID is then transmitted to the PLC, and the steps of deployment are as follows: firstly, determining a network protocol, then writing a communication code, secondly, configuring PID parameters so as to receive data from an external computer, and finally, testing and debugging to ensure normal operation.
7. The machine learning-based method for analyzing the moisture change of cut tobacco in a single cut tobacco making machine, which is characterized in that: in the step J, the system needs to comprise a data acquisition and processing module, a model building and optimizing module and a control strategy generating and adjusting module, and meanwhile, the system also needs to be integrated with the existing automatic control system of the wire making workshop.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310805923.3A CN116602435A (en) | 2023-07-03 | 2023-07-03 | Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310805923.3A CN116602435A (en) | 2023-07-03 | 2023-07-03 | Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116602435A true CN116602435A (en) | 2023-08-18 |
Family
ID=87676697
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310805923.3A Pending CN116602435A (en) | 2023-07-03 | 2023-07-03 | Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116602435A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117077550A (en) * | 2023-10-17 | 2023-11-17 | 首域科技(杭州)有限公司 | Modeling method of tobacco cut-tobacco dryer based on distributed hysteresis model |
-
2023
- 2023-07-03 CN CN202310805923.3A patent/CN116602435A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117077550A (en) * | 2023-10-17 | 2023-11-17 | 首域科技(杭州)有限公司 | Modeling method of tobacco cut-tobacco dryer based on distributed hysteresis model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112101480B (en) | Multivariate clustering and fused time sequence combined prediction method | |
| CN111813084B (en) | Mechanical equipment fault diagnosis method based on deep learning | |
| Wang et al. | An improved ARIMA model for precipitation simulations | |
| CN111045326B (en) | Tobacco shred drying process moisture prediction control method and system based on recurrent neural network | |
| CN116757534B (en) | Intelligent refrigerator reliability analysis method based on neural training network | |
| CN111144667A (en) | Tobacco conditioner discharged material water content prediction method based on gradient lifting tree | |
| CN116602435A (en) | Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine | |
| CN115169737A (en) | Process quality prediction method based on CNN-LSTM hybrid neural network model | |
| CN112273696B (en) | Method, device and equipment for controlling moisture after shredding | |
| CN113780420B (en) | GRU-GCN-based method for predicting concentration of dissolved gas in transformer oil | |
| CN105334185A (en) | Spectrum projection discrimination-based near infrared model maintenance method | |
| CN114115393A (en) | Method for controlling moisture and temperature at outlet of cut tobacco dryer for sheet cut tobacco making line | |
| CN117193222A (en) | Intelligent quality control system based on industrial Internet of things and big data and control method thereof | |
| CN111950795A (en) | Random forest based method for predicting water adding proportion of loosening and conditioning | |
| CN112651173B (en) | Agricultural product quality nondestructive testing method based on cross-domain spectral information and generalizable system | |
| CN111838744B (en) | Continuous real-time prediction method for moisture in tobacco shred production process based on LSTM (localized surface plasmon resonance) environment temperature and humidity | |
| CN117977536A (en) | Smart power grid load identification method | |
| CN117073360A (en) | Method and equipment for controlling water content of Chinese herbal medicine drying based on LSTM model | |
| CN118014115A (en) | Production line loss monitoring method based on long-short-term memory network | |
| CN107220705B (en) | Atmospheric tower top dry point prediction method for atmospheric and vacuum device | |
| CN106647274A (en) | Steady-state discrimination method for operation conditions in continuous production process | |
| CN116662751A (en) | Tobacco leaf moisture content detection method for removing abnormal samples based on principal component analysis and lever value method | |
| CN115685906A (en) | Tobacco flake moisture regain control method and storage medium | |
| CN115017700A (en) | Cigarette shred making process variable selection and weighting method based on SCAD algorithm | |
| Hong et al. | Drainage network flow anomaly classification based on XGBoost |
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 |