CN117993770A - Quality prediction method and system for machined product of conditioning machine - Google Patents
Quality prediction method and system for machined product of conditioning machine Download PDFInfo
- Publication number
- CN117993770A CN117993770A CN202410073951.5A CN202410073951A CN117993770A CN 117993770 A CN117993770 A CN 117993770A CN 202410073951 A CN202410073951 A CN 202410073951A CN 117993770 A CN117993770 A CN 117993770A
- Authority
- CN
- China
- Prior art keywords
- data
- loosening
- production data
- conditioning
- machine
- 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
- 230000003750 conditioning effect Effects 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims abstract description 42
- 238000004519 manufacturing process Methods 0.000 claims abstract description 146
- 238000013016 damping Methods 0.000 claims abstract description 37
- 238000012549 training Methods 0.000 claims abstract description 30
- 238000007781 pre-processing Methods 0.000 claims abstract description 23
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 35
- 238000012545 processing Methods 0.000 claims description 34
- 238000009825 accumulation Methods 0.000 claims description 28
- 230000002159 abnormal effect Effects 0.000 claims description 25
- 238000010606 normalization Methods 0.000 claims description 20
- 238000012216 screening Methods 0.000 claims description 11
- 229920003266 Leaf® Polymers 0.000 claims description 10
- 238000004364 calculation method Methods 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 9
- 239000011159 matrix material Substances 0.000 claims description 7
- 241000208125 Nicotiana Species 0.000 claims description 6
- 235000002637 Nicotiana tabacum Nutrition 0.000 claims description 6
- 238000010923 batch production Methods 0.000 claims description 5
- 230000006870 function Effects 0.000 claims description 4
- 210000002364 input neuron Anatomy 0.000 claims description 3
- 238000002156 mixing Methods 0.000 claims description 3
- 230000009467 reduction Effects 0.000 claims description 3
- 238000012360 testing method Methods 0.000 claims description 3
- 230000000875 corresponding effect Effects 0.000 description 8
- 238000007599 discharging Methods 0.000 description 4
- 235000019504 cigarettes Nutrition 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005303 weighing Methods 0.000 description 2
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000003754 machining Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method and a system for predicting the quality of a machined product of a damping machine, wherein the method comprises the following steps of 1, obtaining a basic data set according to production data of the damping machine; step 2, preprocessing a basic data set to obtain a sample data set; step 3, constructing a quality prediction model of the moisture regain machine processed product; step 4, training a model; step 5, predicting the quality of the processed product of the conditioning machine; the system comprises a real-time updating module of the production data of the damping machine, a preprocessing module of the production data of the damping machine and a quality prediction module of the processed product of the damping machine. According to the method and the system for predicting the quality of the moisture regain machine processed product, the accuracy of the prediction model is improved through feature selection, the influence caused by data time stagnation is reduced, and the accuracy of the prediction model is better ensured. According to the method, the prediction model is iteratively updated and updated in real time by combining the actual production condition, so that the accuracy of the prediction result is greatly improved.
Description
Technical Field
The invention belongs to the technical field of tobacco shred production quality monitoring, and particularly relates to a method and a system for predicting the quality of a processed product of a conditioning machine.
Background
The improvement and optimization of the cigarette shredding process can improve the qualification rate of products, reduce the waste of raw materials and greatly improve the economic benefit. The conditioning process is the first step in the tobacco shred making process, and the performance of the conditioning process is directly related to the quality of cigarette products.
Predictive model driven methods have been applied in the tobacco field, but research has found that there are not many applications in conditioning machines today. The prediction model established by taking the historical data as the training set can be used for prediction, but the characteristic parameters of the damping machine in different production batches are obviously different, so that the current prediction effect has larger deviation; time lags caused by data measured at different parts of the damping machine equipment have great influence on accurate model establishment; the trained static model can not be correspondingly adjusted for changes of production environment conditions, equipment performances and the like in batch production, and good prediction effects are difficult to obtain in all production batches. Therefore, how to design a product quality prediction method capable of selecting characteristic parameters of a damping machine and ensuring that a model is not affected by data time lag is currently needed in the art, and the model can be corrected according to current data in an actual production process.
Disclosure of Invention
The invention aims to provide a method and a system for predicting the quality of a machined product of a conditioning machine, so as to solve the technical problems.
In order to achieve the above purpose, the present invention provides the following technical solutions:
the invention discloses a quality prediction method of a moisture regain machine processed product, which comprises the following steps:
step 1, obtaining a basic data set according to production data of a damping machine, and specifically comprising the following steps:
Step 11, acquiring historical production data: comparing the current loose and remoistened tobacco leaf production grades, and acquiring historical production data of a plurality of batches with the latest grades from a database, wherein the historical production data comprise production process data and corresponding product quality data;
Step 12, initializing real-time production data: initializing and emptying current real-time production data;
step 13, accumulating real-time production data: collecting real-time production data, including production process data and corresponding product quality data, and storing the real-time production data;
Step 14, judging the quantity of the real-time production data: comparing the real-time production data quantity with a set threshold t 0, and if the real-time production data quantity is larger than the set threshold t 0, executing step 15; if the number of the real-time production data is smaller than or equal to the set threshold t 0, returning to the step 13 and performing loop execution until the current number of the real-time production data is larger than the set threshold t 0;
Step 15, merging production data: combining the acquired historical production data of a plurality of batches with the real-time production data, outputting the combined production data, returning to the step 12 to initialize the real-time production data, and executing the next circulation process; taking all the output combined production data as a basic data set;
Step 2, preprocessing the basic data set to obtain a sample data set: preprocessing a basic data set, wherein the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high noise data filtering, feature selection and normalization processing, and the preprocessed basic data set is used as a sample data set;
Step 3, constructing a quality prediction model of the moisture regain machine processed product: constructing a moisture regaining machine processing product quality prediction model based on an extreme learning machine, wherein the model comprises an input layer, a hidden layer and an output layer: the input layer has d input neurons corresponding to the input characteristic parameters of the sample data set; setting the number L of ELM hidden layer nodes, inputting layer weight omega and hidden layer bias b, wherein omega is a matrix with the size of d multiplied by L, and b is a matrix with the size of 1 multiplied by L, wherein specific numerical values are given randomly; the output layer corresponds to the product quality data;
Step 4, model training: training a constructed damping machine processed product quality prediction model by using a sample data set as training data, and equally dividing the sample data into three parts, wherein the first part is used for model coarse training, the second part is used for model self-training, and the third part is used for predictive regression training of verification tests; using the average absolute error of the model predicted value and the actual value as a model verification index, and when the average absolute error is smaller than an industry standard threshold value, passing the model verification;
Step 5, predicting the quality of the moisture regain machine processed product: and predicting the quality of the processed product of the conditioning machine to be predicted by using the trained model.
Further, the production process data in the step 1 comprises moisture after loosening and conditioning, temperature after loosening and conditioning, roller steam pressure of a loosening and conditioning machine, return air temperature of the loosening and conditioning machine, wall temperature of the loosening and conditioning machine, humidifying water accumulation amount of the loosening and conditioning machine, humidifying water flow of the loosening and conditioning machine, electronic weighing and accumulating amount before loosening and conditioning, electronic weighing and flow before loosening and conditioning, moisture before loosening and conditioning, sheet blending proportion, hot air fan frequency, blade conditioning discharging chamber temperature, blade conditioning discharging chamber steam pressure, primary water adding flow, primary water accumulation amount, secondary water adding flow, secondary water accumulation amount, moisture discharging motor frequency, roller motor frequency, loosening and compensating steam flow, loosening and compensating steam accumulation amount, environment temperature and environment humidity.
Further, in step 1, the historical production data of the plurality of batches with the latest brand, specifically, the historical production data of not less than 10 batches, are obtained from the database.
Further, the data filtering in step 2 is as follows: firstly, dividing the data into stages, specifically judging whether the production data is data of a stub bar stage, a stable stage or a tailing stage according to three characteristic parameters, namely, the moisture data after loosening and conditioning, the accumulation amount of an electronic scale before loosening and conditioning and the flow amount of the electronic scale before loosening and conditioning, and then selecting the data of the stable stage for the next processing;
The abnormal data processing is as follows: judging whether the data is abnormal data or not through an abnormal data judging formula, marking the abnormal data, and replacing the abnormal data through an interpolation method; the abnormal data judging formula is |x-mu| >3 sigma, wherein x is a certain type of production data characteristic parameters, mu is an average value of the type of production data characteristic parameters, and sigma is a standard deviation of the type of production data characteristic parameters; the calculation formulas are respectively X k is the characteristic parameter of a certain type of production data in the kth group, and n is the total number of samples of the characteristic parameter of the production data in the kth group;
the time-lag data alignment is: the data are aligned in time lag data, and the production environment is assessed to be constant temperature and humidity, namely the environment temperature and the environment humidity are constant, so that other data with the environment temperature and the environment humidity removed are aligned in time lag data;
the high noise data filtering is as follows: by sliding the mean value filter formula Denoising the data, wherein t' is the acquisition time, and M is the size of a moving average window; or utilize low pass filter formula/>Carrying out noise reduction treatment on the data, wherein omega' is the actual frequency, omega c is the normalized cut-off frequency, and n o is the filter order;
the characteristics are selected as follows: selecting characteristic parameters with high correlation with the quality of a moisture regaining machine processed product, namely product quality characteristic data, from all characteristic parameters of production process data, then adopting Catboost algorithm, taking the screened product quality characteristic data as input of a model, taking the quality of the moisture regaining machine processed product as output of the model, calculating an identification of the importance of the characteristic correlation in the algorithm as an importance PVC value of the calculated characteristic, and selecting the characteristic parameters with PVC of more than 0.02 to finish characteristic selection by taking the importance as a standard, so as to obtain the finally screened characteristic parameters;
The normalization process is as follows: according to the normalization formula And carrying out normalization processing on the characteristic parameters of the production data corresponding to the energy efficiency modeling characteristic vector x, namely the characteristic parameters finally screened after characteristic selection, wherein x * is the normalization result of the characteristic parameters of certain type of production data.
Further, the specific mode of selecting the characteristic parameters with high correlation with the quality of the product processed by the conditioning machine, namely the product quality characteristic data, from the characteristic parameters of the production process data is as follows: by pearson correlation formulaCalculating the correlation between each characteristic parameter and the product quality standard characteristic parameter, wherein x i is the product quality standard characteristic parameter,/>For the average value of the standard characteristic parameters of the product quality, y i is the characteristic parameter to be selected,An average value of the characteristic parameters to be selected; deleting constant characteristic parameters, namely characteristic parameters with r=0.3, and low-correlation characteristic parameters, namely characteristic parameters with r < 0.3, through calculation, and screening product quality characteristic data; and calculating the loose and remoistened water data as product quality standard characteristic parameters.
Further, the calculation formula of the importance degree PVC of the feature is: pvc= Σ trees,leafs(v1-avr)2·leafleft+(v2-avr)2·leafright; the higher the importance is, the greater the PVC value is, wherein leaf left and leaf right represent the weights of the left and right leaves, v 1 and v 2 represent the objective function values of the left and right leaves, respectively,Is the average predicted value of the nodes.
Further, the finally screened characteristic parameters are as follows: the method comprises the following steps of loosening moisture regaining, loosening moisture regaining temperature, loosening moisture regaining machine roller steam pressure, loosening moisture regaining machine roller wall temperature, loosening moisture regaining machine humidifying water flow, loosening moisture regaining electronic scale accumulation amount, loosening moisture regaining electronic scale flow, loosening moisture regaining moisture, hot air fan frequency, primary water adding flow, primary water adding accumulation amount, secondary water adding flow, secondary water adding accumulation amount, moisture removing motor frequency, loosening compensation steam flow and loosening compensation steam accumulation amount.
Further, the industry standard threshold in step 4 is 0.1.
The invention also discloses a system for predicting the quality of the machined product of the damping machine, which comprises a real-time updating module of the production data of the damping machine, a preprocessing module of the production data of the damping machine and a predicting module of the quality of the machined product of the damping machine;
The real-time updating module of the production data of the damping machine is used for acquiring historical batch production data, accumulating and updating the real-time production data, combining the historical batch production data with the real-time production data, and outputting the combined production data, namely a basic data set;
The damping machine production data preprocessing module is used for preprocessing the basic data set to obtain a sample data set; the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high-noise data filtering, feature selection and normalization processing;
The damping machine processing product quality prediction module is used for constructing a damping machine processing product quality prediction model, training the constructed model based on a sample data set until model training is completed, and predicting the product quality with prediction by using the damping machine processing product quality prediction model after training is completed.
The beneficial effects of the invention are as follows: according to the method and the system for predicting the quality of the moisture regain machine processed product, disclosed by the invention, the model is trained through the feature selection and the feature parameters which are most closely related to the quality of the product, so that the accuracy of the prediction model is improved, the influence caused by data time stagnation is reduced, and the accuracy of the prediction model is better ensured. According to the method, the prediction model is iteratively updated and updated in real time by combining the actual production condition, so that the accuracy of the prediction result is greatly improved, the product quality is improved, and the production standardization and intellectualization level is improved.
The invention will be described in further detail with reference to the drawings and the detailed description.
Drawings
FIG. 1 is a flow chart of a method for predicting the quality of a conditioning machine processed product;
FIG. 2 is a flow chart for obtaining a base data set based on the conditioning machine production data;
Fig. 3 is a block diagram of a system for predicting the quality of a product processed by the conditioning machine.
Detailed Description
Example 1
The embodiment discloses a method for predicting the quality of a moisture regain machine processed product, as shown in fig. 1 and 2, the method comprises the following steps:
step 1, obtaining a basic data set according to production data of a damping machine, and specifically comprising the following steps:
Step 11, acquiring historical production data: comparing the current loose and remoistened tobacco production brand, and acquiring historical production data of more than 10 batches (generally 10 batches) with the brand from a database, wherein the historical production data comprise production process data and corresponding product quality data, and the production process data comprise characteristic parameters such as moisture after loose remoistening, temperature after loose remoistening, loose remoistening roller steam pressure, loose remoistening return air temperature, loose remoistening roller wall temperature, loose remoistening humidifying water accumulation amount, loose remoistening humidifying water flow, electronic scale accumulation amount before loose remoistening, electronic scale flow before loose remoistening, moisture before loose remoistening, sheet blending proportion, hot air fan frequency, blade remoistening discharge chamber temperature, blade remoistening discharge chamber steam pressure, primary water addition amount, secondary water addition accumulation amount, moisture discharging motor frequency, roller motor frequency, loose compensation steam flow, loose compensation steam accumulation amount, environment temperature, environment humidity and the like;
Step 12, initializing real-time production data: initializing and emptying current real-time production data;
Step 13, accumulating real-time production data: collecting real-time production data including production process data and corresponding product quality data through each sensor in the hardware equipment, and storing the real-time production data;
Step 14, judging the quantity of the real-time production data: comparing the real-time production data quantity with a set threshold t 0, and if the real-time production data quantity is larger than the set threshold t 0, executing step 15; if the number of the real-time production data is smaller than or equal to the set threshold t 0, returning to the step 13 and performing loop execution until the current number of the real-time production data is larger than the set threshold t 0;
Step 15, merging production data: combining the acquired historical production data of a plurality of batches with the real-time production data, outputting the combined production data, returning to the step 12 to initialize the real-time production data, and executing the next circulation process; all the output combined production data are taken as a base data set.
Assuming that the basic dataset contains n groups of production data, wherein part of the production data are shown in table 1, x iwk is the water content data before loosening and conditioning of the k group, x ctk is the wall temperature data of the wall of the loosening and conditioning machine of the k group, x wfk is the humidifying water flow data of the loosening and conditioning machine of the k group, x wtk is the return air temperature data of the loosening and conditioning machine of the k group, and y owk is the water content data after loosening and conditioning of the k group.
TABLE 1
Step 2, preprocessing the basic data set to obtain a sample data set: the method comprises the steps of preprocessing a basic data set, wherein the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high noise data filtering, feature selection, normalization processing and the like, and the preprocessed basic data set is used as a sample data set.
The specific process is as follows:
1) Data screening: the data are firstly subjected to stage division, specifically, whether the production data are data of a stub bar stage, a stable stage or a tailing stage is judged according to three characteristic parameters, namely, the moisture data after loosening and dampening, the accumulation amount of the electronic scale before loosening and dampening, and the flow amount of the electronic scale before loosening and dampening, and then the data of the stable stage are selected for the next processing.
2) Abnormal data processing: judging whether the data is abnormal data or not through an abnormal data judging formula, marking the abnormal data, and replacing the abnormal data through an interpolation method.
The abnormal data judging formula is |x-mu| >3 sigma, wherein x is a certain type of production data characteristic parameters, mu is an average value of the type of production data characteristic parameters, and sigma is a standard deviation of the type of production data characteristic parameters; the calculation formulas are respectivelyX k is the k-th set of certain type of production data characteristic parameters, and n is the total number of samples of the type of production data characteristic parameters.
Assuming that the data in Table 1 needs to be processed for abnormal data, taking the moisture data before loose conditioning as an example, thenWherein mu iw is the average value of the moisture data before loosening and conditioning, and sigma iw is the standard deviation of the moisture data before loosening and conditioning; then calculating the size relation of |x iw-μiw | and 3 sigma iw in all data; assuming |x iwk-μiw|>3σiw, the k group of pre-loose conditioning moisture data is marked as outlier data and interpolated formula/>Processing exception data, wherein/>Is the processed k-th set of loose pre-conditioning moisture data, x iwk-1 is the k-1-th set of loose pre-conditioning moisture data, and x iwk+1 is the k+1-th set of loose pre-conditioning moisture data.
3) Time-lapse data alignment: and (3) time-lapse data alignment is carried out on the data, so that each feature is acquired to a consistent object after the data alignment, and the accuracy of subsequent prediction is improved. The production environment is assessed to be constant temperature and humidity, namely the ambient temperature and the ambient humidity are constant, so that other data of the removed ambient temperature and the removed ambient humidity are aligned in time-lag data.
Time-lag data alignment is carried out on the data, taking the moisture data before loosening and conditioning as an example, and the time of the moisture data alignment window before loosening and conditioning is t, thenWherein/>Is the moisture data before loosening and conditioning at time k after time-lag data alignment, and x iwk+t is the moisture data before loosening and conditioning at the original time k+t.
4) High noise data filtering: by sliding the mean value filter formula Denoising the data, wherein t' is the acquisition time, and M is the size of a moving average window; or utilize low pass filter formula/>And carrying out preliminary noise reduction treatment on the data to reduce the influence of noise, wherein ω' is an actual frequency, ω c is a normalized cut-off frequency, and n o is a filter order.
5) Feature selection: the characteristic parameters with high correlation with the quality of the moisture regain machine processed product, namely the product quality characteristic data, are selected from the characteristic parameters of the production process data, and the specific modes are as follows: by pearson correlation formulaCalculating the correlation between each characteristic parameter and the product quality standard characteristic parameter, wherein x i is the product quality standard characteristic parameter,/>For the average value of the standard characteristic parameters of the product quality, y i is the characteristic parameter to be selected,An average value of the characteristic parameters to be selected; deleting constant characteristic parameters (r=0.3 characteristic parameters) and low-correlation characteristic parameters (r < 0.3 characteristic parameters) through calculation, and screening out product quality characteristic data; because the moisture data after loosening and conditioning is highly correlated with the quality of the product, the moisture data after loosening and conditioning is generally calculated as a product quality standard characteristic parameter.
Then adopting Catboost algorithm, taking the product quality characteristic data screened in the preliminary step as input of a model, taking the quality of the product processed by the conditioning machine as output of the model, calculating the importance mark of the characteristic correlation in the algorithm as the importance PVC value of the calculated characteristic, and selecting the characteristic parameter with higher importance (PVC > 0.02) to finish characteristic selection by taking the importance mark as a standard, so as to obtain the finally screened characteristic parameter.
The calculation formula of the importance degree PVC of the features is as follows: pvc= Σ trees,leafs(v1-avr)2·leafleft+(v2-avr)2·leafright; the higher the importance is, the greater the PVC value is, wherein leaf left and leaf right represent the weights of the left and right leaves, v 1 and v 2 represent the objective function values of the left and right leaves, respectively,Is the average predicted value of the nodes.
After feature selection, finally screened feature parameters are as follows: the method comprises the following steps of loosening moisture regaining, loosening moisture regaining temperature, loosening moisture regaining machine roller steam pressure, loosening moisture regaining machine roller wall temperature, loosening moisture regaining machine humidifying water flow, loosening moisture regaining electronic scale accumulation amount, loosening moisture regaining electronic scale flow, loosening moisture regaining moisture, hot air fan frequency, primary water adding flow, primary water adding accumulation amount, secondary water adding flow, secondary water adding accumulation amount, moisture removing motor frequency, loosening compensation steam flow and loosening compensation steam accumulation amount.
6) Normalization: according to the normalization formulaAnd carrying out normalization processing on the production data characteristic parameters (namely the characteristic parameters finally screened after characteristic selection) corresponding to the energy efficiency modeling characteristic vector x, wherein x * is the normalization result of the characteristic parameters of certain types of production data.
Normalization of the data, taking the moisture data before loose conditioning as an example, mu iw is the average value of the moisture data before loose conditioning, sigma iw is the standard deviation of the moisture data before loose conditioning,X * iwk is the normalization of the k-th set of loose pre-conditioning moisture data.
Step 3, constructing a quality prediction model of the moisture regain machine processed product: constructing a moisture regaining machine processing product quality prediction model based on an extreme learning machine, wherein the model comprises an input layer, a hidden layer and an output layer: the input layer has d input neurons corresponding to the input characteristic parameters of the sample data set; setting the number L of ELM hidden layer nodes, inputting layer weight omega and hidden layer bias b, wherein omega is a matrix with the size of d multiplied by L, and b is a matrix with the size of 1 multiplied by L, wherein specific numerical values are given randomly; the output layer corresponds to the product quality data.
Step 4, model training: and training the constructed quality prediction model of the conditioning machining product by using the sample data set as training data, wherein the sample data is equally proportioned into three parts in model training, the first part is used for model coarse training, the second part is used for model self-training, and the third part is used for predictive regression training of verification test.
Assuming that the feature number of the sample data is N, inputting the sample data into the hidden layer, wherein X= [ X 1,x2,...,xN]T ], obtaining hidden layer output H according to a calculation formula H=sigmoid (xω+b), wherein sigmoid is an activation function, ω is an input layer weight, and b is a hidden layer bias; according to the relation between the hidden layer and the output layerAnd obtaining the magnitude of the output layer weight beta, wherein H T is a generalized inverse matrix of H, and C is a constraint coefficient.
And taking the average absolute error of the model predicted value and the actual value as a model verification index, and when the average absolute error is generally considered to be smaller than an industry standard threshold (generally 0.1) according to the industry standard threshold, passing the model verification.
Step 5, predicting the quality of the moisture regain machine processed product: and predicting the quality of the processed product of the conditioning machine to be predicted by using the trained model.
And inputting the product quality characteristic parameters screened in the production process data of the to-be-predicted damping machine processed product into a trained model, and outputting the product quality characteristic parameters to be predicted damping machine processed product quality.
Example two
The embodiment discloses a quality prediction system for a machined product of a conditioning machine, which comprises a real-time updating module for production data of the conditioning machine, a preprocessing module for the production data of the conditioning machine and a quality prediction module for the machined product of the conditioning machine, as shown in fig. 3.
The real-time updating module of the production data of the damping machine is used for acquiring the production data of the historical batch, accumulating and updating the real-time production data, combining the production data of the historical batch with the real-time production data, and outputting the combined production data, namely a basic data set.
The damping machine production data preprocessing module is used for preprocessing the basic data set to obtain a sample data set; the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high-noise data filtering, feature selection and normalization processing.
The damping machine processing product quality prediction module is used for constructing a damping machine processing product quality prediction model, training the constructed model based on a sample data set until model training is completed, and predicting the product quality with prediction by using the damping machine processing product quality prediction model after training is completed.
Finally, it should be noted that the above description is only for the purpose of illustrating the technical solution of the present invention and not for the purpose of limiting the same, and that although the present invention has been described in detail with reference to the preferred arrangement, it will be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention.
Claims (9)
1. A method for predicting the quality of a conditioning machined product, the method comprising the steps of:
step 1, obtaining a basic data set according to production data of a damping machine, and specifically comprising the following steps:
Step 11, acquiring historical production data: comparing the current loose and remoistened tobacco leaf production grades, and acquiring historical production data of a plurality of batches with the latest grades from a database, wherein the historical production data comprise production process data and corresponding product quality data;
Step 12, initializing real-time production data: initializing and emptying current real-time production data;
step 13, accumulating real-time production data: collecting real-time production data, including production process data and corresponding product quality data, and storing the real-time production data;
Step 14, judging the quantity of the real-time production data: comparing the real-time production data quantity with a set threshold t 0, and if the real-time production data quantity is larger than the set threshold t 0, executing step 15; if the number of the real-time production data is smaller than or equal to the set threshold t 0, returning to the step 13 and performing loop execution until the current number of the real-time production data is larger than the set threshold t 0;
Step 15, merging production data: combining the acquired historical production data of a plurality of batches with the real-time production data, outputting the combined production data, returning to the step 12 to initialize the real-time production data, and executing the next circulation process; taking all the output combined production data as a basic data set;
Step 2, preprocessing the basic data set to obtain a sample data set: preprocessing a basic data set, wherein the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high noise data filtering, feature selection and normalization processing, and the preprocessed basic data set is used as a sample data set;
Step 3, constructing a quality prediction model of the moisture regain machine processed product: constructing a moisture regaining machine processing product quality prediction model based on an extreme learning machine, wherein the model comprises an input layer, a hidden layer and an output layer: the input layer has d input neurons corresponding to the input characteristic parameters of the sample data set; setting the number L of ELM hidden layer nodes, inputting layer weight omega and hidden layer bias b, wherein omega is a matrix with the size of d multiplied by L, and b is a matrix with the size of 1 multiplied by L, wherein specific numerical values are given randomly; the output layer corresponds to the product quality data;
Step 4, model training: training a constructed damping machine processed product quality prediction model by using a sample data set as training data, and equally dividing the sample data into three parts, wherein the first part is used for model coarse training, the second part is used for model self-training, and the third part is used for predictive regression training of verification tests; using the average absolute error of the model predicted value and the actual value as a model verification index, and when the average absolute error is smaller than an industry standard threshold value, passing the model verification;
Step 5, predicting the quality of the moisture regain machine processed product: and predicting the quality of the processed product of the conditioning machine to be predicted by using the trained model.
2. The method according to claim 1, wherein the production process data in the step 1 includes moisture after loosening and conditioning, temperature after loosening and conditioning, loosening and conditioning drum steam pressure, loosening and conditioning return air temperature, loosening and conditioning drum wall temperature, loosening and conditioning machine humidifying water accumulation amount, loosening and conditioning machine humidifying water flow, loosening and conditioning electronic scale accumulation amount, loosening and conditioning electronic scale flow, loosening and conditioning moisture before conditioning, flake blending proportion, hot air fan frequency, blade conditioning discharge chamber temperature, blade conditioning discharge chamber steam pressure, primary water addition amount, primary water addition accumulation amount, secondary water addition accumulation amount, moisture removal motor frequency, drum motor frequency, loosening and compensating steam flow, loosening and compensating steam accumulation amount, ambient temperature and ambient humidity.
3. The method according to claim 1, wherein in step 1, the historical production data of the plurality of batches with the latest brand, in particular, the historical production data of not less than 10 batches, are obtained from a database.
4. The method for predicting the quality of a moisture regain machine processed product according to claim 1, wherein the data screening in step 2 is: firstly, dividing the data into stages, specifically judging whether the production data is data of a stub bar stage, a stable stage or a tailing stage according to three characteristic parameters, namely, the moisture data after loosening and conditioning, the accumulation amount of an electronic scale before loosening and conditioning and the flow amount of the electronic scale before loosening and conditioning, and then selecting the data of the stable stage for the next processing;
The abnormal data processing is as follows: judging whether the data is abnormal data or not through an abnormal data judging formula, marking the abnormal data, and replacing the abnormal data through an interpolation method; the abnormal data judging formula is |x-mu| >3 sigma, wherein x is a certain type of production data characteristic parameters, mu is an average value of the type of production data characteristic parameters, and sigma is a standard deviation of the type of production data characteristic parameters; the calculation formulas are respectively X k is the characteristic parameter of a certain type of production data in the kth group, and n is the total number of samples of the characteristic parameter of the production data in the kth group;
the time-lag data alignment is: the data are aligned in time lag data, and the production environment is assessed to be constant temperature and humidity, namely the environment temperature and the environment humidity are constant, so that other data with the environment temperature and the environment humidity removed are aligned in time lag data;
the high noise data filtering is as follows: by sliding the mean value filter formula Denoising the data, wherein t' is the acquisition time, and M is the size of a moving average window; or utilize low pass filter formula/>Carrying out noise reduction treatment on the data, wherein omega' is the actual frequency, omega c is the normalized cut-off frequency, and n o is the filter order;
the characteristics are selected as follows: selecting characteristic parameters with high correlation with the quality of a moisture regaining machine processed product, namely product quality characteristic data, from all characteristic parameters of production process data, then adopting Catboost algorithm, taking the screened product quality characteristic data as input of a model, taking the quality of the moisture regaining machine processed product as output of the model, calculating an identification of the importance of the characteristic correlation in the algorithm as an importance PVC value of the calculated characteristic, and selecting the characteristic parameters with PVC of more than 0.02 to finish characteristic selection by taking the importance as a standard, so as to obtain the finally screened characteristic parameters;
The normalization process is as follows: according to the normalization formula And carrying out normalization processing on the characteristic parameters of the production data corresponding to the energy efficiency modeling characteristic vector x, namely the characteristic parameters finally screened after characteristic selection, wherein x * is the normalization result of the characteristic parameters of certain type of production data.
5. The method for predicting the quality of a product by a conditioning machine according to claim 4, wherein the specific manner of selecting the characteristic parameter having high correlation with the quality of the product by the conditioning machine, i.e., the product quality characteristic data, from among the characteristic parameters of the production process data is as follows: by pearson correlation formula Calculating the correlation between each characteristic parameter and the product quality standard characteristic parameter, wherein x i is the product quality standard characteristic parameter,/>For the average value of the standard characteristic parameters of the product quality, y i is the characteristic parameter to be selected,/>An average value of the characteristic parameters to be selected; deleting constant characteristic parameters, namely characteristic parameters with r=0.3, and low-correlation characteristic parameters, namely characteristic parameters with r <0.3, through calculation, and screening out product quality characteristic data; and calculating the loose and remoistened water data as product quality standard characteristic parameters.
6. The method for predicting the quality of a moisture regain machine product of claim 4, wherein the calculation formula of the importance PVC of the feature is: pvc= Σ trees,leafs(v1-avr)2·leaflleft+(v2-avr)2·leaflright; the higher the importance is, the greater the PVC value is, wherein leafl left and leaf right represent the weights of left and right leaves, v 1 and v 2 represent the objective function values of left and right leaves, respectively,Is the average predicted value of the nodes.
7. The method for predicting the quality of a moisture regain machine product of claim 4, wherein the final filtered characteristic parameters are: the method comprises the following steps of loosening moisture regaining, loosening moisture regaining temperature, loosening moisture regaining machine roller steam pressure, loosening moisture regaining machine roller wall temperature, loosening moisture regaining machine humidifying water flow, loosening moisture regaining electronic scale accumulation amount, loosening moisture regaining electronic scale flow, loosening moisture regaining moisture, hot air fan frequency, primary water adding flow, primary water adding accumulation amount, secondary water adding flow, secondary water adding accumulation amount, moisture removing motor frequency, loosening compensation steam flow and loosening compensation steam accumulation amount.
8. The method of claim 1, wherein the industry standard threshold in step 4 is 0.1.
9. The system is characterized by comprising a real-time updating module of the production data of the damping machine, a preprocessing module of the production data of the damping machine and a quality prediction module of the machined product of the damping machine;
The real-time updating module of the production data of the damping machine is used for acquiring historical batch production data, accumulating and updating the real-time production data, combining the historical batch production data with the real-time production data, and outputting the combined production data, namely a basic data set;
The damping machine production data preprocessing module is used for preprocessing the basic data set to obtain a sample data set; the preprocessing sequentially comprises data screening, abnormal data processing, time-lag data alignment, high-noise data filtering, feature selection and normalization processing;
The damping machine processing product quality prediction module is used for constructing a damping machine processing product quality prediction model, training the constructed model based on a sample data set until model training is completed, and predicting the product quality with prediction by using the damping machine processing product quality prediction model after training is completed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410073951.5A CN117993770A (en) | 2024-01-18 | 2024-01-18 | Quality prediction method and system for machined product of conditioning machine |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410073951.5A CN117993770A (en) | 2024-01-18 | 2024-01-18 | Quality prediction method and system for machined product of conditioning machine |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117993770A true CN117993770A (en) | 2024-05-07 |
Family
ID=90886425
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410073951.5A Pending CN117993770A (en) | 2024-01-18 | 2024-01-18 | Quality prediction method and system for machined product of conditioning machine |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117993770A (en) |
-
2024
- 2024-01-18 CN CN202410073951.5A patent/CN117993770A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Ko et al. | Assessment of achievable PI control performance for linear processes with dead time | |
| CN109261724B (en) | Method for improving preset model precision in multi-variety rolling mode | |
| CN111045326B (en) | Tobacco shred drying process moisture prediction control method and system based on recurrent neural network | |
| CN108920863B (en) | Method for establishing energy consumption estimation model of robot servo system | |
| CN112273696B (en) | Method, device and equipment for controlling moisture after shredding | |
| KR101432436B1 (en) | Apparatus and method for prediction of influent flow rate and influent components using nearest neighbor method | |
| CN112273695A (en) | Method, device and equipment for predicting water content of loose moisture regain outlet | |
| CN107146035B (en) | Method for calculating batch coefficients in large-goods production of knitted clothes | |
| CN112757053B (en) | Model fusion tool wear monitoring method and system based on power and vibration signals | |
| CN115936262B (en) | Yield prediction method, system and medium based on big data environment interference | |
| CN114115393A (en) | Method for controlling moisture and temperature at outlet of cut tobacco dryer for sheet cut tobacco making line | |
| CN111950795A (en) | Random forest based method for predicting water adding proportion of loosening and conditioning | |
| US20240185115A1 (en) | Method and apparatus for early warning of dry pump shutdown, electronic device, storage medium and program | |
| CN115879369A (en) | Coal mill fault early warning method based on optimized LightGBM algorithm | |
| CN114357870A (en) | Metering equipment operation performance prediction analysis method based on local weighted partial least squares | |
| CN117993770A (en) | Quality prediction method and system for machined product of conditioning machine | |
| CN111950166A (en) | Cost optimization method for household paper making machine based on data mining | |
| CN110991723A (en) | Application method of artificial intelligence in seasonal load prediction | |
| CN116602435A (en) | Machine learning-based method for analyzing moisture change of cut tobacco in cut tobacco making machine | |
| CN113876008B (en) | Method for controlling stability of moisture content of loose and moisture regained tobacco flakes | |
| CN111582619B (en) | Adaptive design method based on correlation and dependency analysis | |
| CN113778042A (en) | Industrial control equipment instruction simulation operation communication system based on informatization modeling | |
| CN113076309A (en) | System and method for predicting water adding amount of raw tobacco shred | |
| CN117493911A (en) | Cut-tobacco dryer outlet moisture prediction method based on multiple linear regression model | |
| CN115600641A (en) | Prediction system and method for moisture content of cut stems at inlet of fluidized bed |
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 |