Unsteady-state intelligent control method for moisture at outlet of feeding equipment of cut stem production line

Abstract

The invention discloses an unsteady-state intelligent control method for moisture at an outlet of a feeding device of a cut stem production line, and provides an unsteady-state intelligent control method for the moisture at the outlet of the feeding device of the cut stem production line by utilizing long-time production data accumulation of the cut stem production line in combination with the core capability of a multiple linear regression algorithm through a machine learning mode. A linear relation between a plurality of independent variables and a dependent variable is analyzed for prediction, an outlet moisture prediction model is established according to the linear relation, outlet moisture prediction is carried out, a water supply valve is controlled in combination with a prediction result, and the method for regulating and controlling the outlet moisture under the conditions of stub bars, rapid adjustment and stability is achieved.

Description

Technical Field

[0001] The invention relates to the technical field of tobacco processing, and in particular to a non-steady-state intelligent control method for moisture at an outlet of a feeding device of a cut stem production line. Background Art

[0002] In the tobacco industry's tobacco production workshop, the stability and accuracy of the feeding equipment of the stem production line has an important impact on the quality of the final product. Traditional control methods are often unable to cope with non-steady-state changes in the production process, resulting in large fluctuations in the outlet moisture content, affecting product quality.

[0003] At present, the feeding equipment of the stem production line in the tobacco industry's silk-making workshop mainly adopts the traditional PID (proportional-integral-differential) control method. The PID controller adjusts the valve opening through the feedback mechanism to maintain the set outlet moisture. However, when facing non-steady-state changes in the production process, PID control often shows large hysteresis and instability, making it difficult to achieve precise control of outlet moisture. Summary of the invention

[0004] The purpose of the present invention is to provide a non-steady-state intelligent control method for the outlet moisture of a feeding device of a shredded stem production line. By machine learning and utilizing long-term production data accumulation, the outlet moisture is predicted, and the water supply valve is controlled in combination with the prediction result to achieve precise regulation of the outlet moisture.

[0005] In order to achieve the above object, the scheme of the present invention is: A non-steady-state intelligent control method for moisture at the outlet of a feeding device of a shredded stem production line comprises a stirring drum, wherein shredded stem raw materials are fed into the stirring drum from an inlet of the stirring drum through a conveyor belt, a steam pipe and a water pipe are connected to the stirring drum through a steam valve and a water supply valve, the shredded stem raw materials are humidified by the stirring drum and then fed to the next process from the outlet of the stirring drum, a metering scale is arranged on the conveyor belt, a shredded stem moisture content tester is respectively arranged at the inlet and outlet of the stirring drum, a temperature / humidity sensor is arranged in the stirring drum, and the stirring drum adjusts the moisture content of the shredded stems fed from the outlet of the stirring drum by controlling the steam valve and the water supply valve, wherein: the method comprises: establishing an outlet moisture prediction model and regulating the outlet moisture of the stirring drum based on the outlet moisture prediction model: The establishment of the outlet moisture prediction model comprises: obtaining past production data of the shredded stem production line, sending the past production data into a neural network model for training, the training is for the purpose of obtaining the delayed outlet moisture content, and the outlet moisture prediction model is obtained after training, wherein the production data comprises: steam valve opening, water supply valve opening, feeding unit batch, feeding unit tobacco brand, moisture content of shredded stems at the inlet and outlet of the mixing drum, hot air temperature at the inlet of the mixing drum, mixed air temperature, return air temperature of shredded stems, ambient temperature, ambient humidity, and scale accumulation; The control of the outlet moisture of the mixing drum based on the outlet moisture prediction model includes: Step 1: Determine the initial opening of the steam valve and the water supply valve according to past production data, start the conveyor belt to feed the shredded stem raw materials into the mixing drum, and gradually adjust the steam supply valve and the water supply valve to the initial opening; Step 2: Acquire production data in real time, input the real-time acquired production data into the outlet moisture prediction model to obtain the predicted moisture content of the delayed stems at the outlet of the mixing drum, compare the predicted moisture content of the delayed stems with the measured moisture content of the stems at the outlet of the mixing drum, adjust the steam valve and the water supply valve according to the comparison result, and realize the regulation of the moisture content of the stems at the outlet of the mixing drum.

[0006] The scheme is further that: the method further includes a slow adjustment stage, a fast adjustment stage and a smooth adjustment stage: The slow adjustment stage is the material head stage, which is the stage when the accumulated amount on the scale is ≤170KG. The outlet moisture control in the slow adjustment stage includes: (1) When the accumulated amount reaches 42 kg, gradually adjust the steam supply valve and water supply valve to the initial opening; (2) continuously inputting the real-time acquired production data into the outlet moisture prediction model at the first interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 25 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; The rapid adjustment stage is when the scale cumulative weight is 170KG ≤ ≤ 300KG. The outlet moisture control in the rapid adjustment stage includes: Inputting the real-time acquired production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 20 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; The stable adjustment stage is when the accumulated amount of the scale is ≥300KG. The outlet moisture control in the stable adjustment stage includes: Inputting the real-time acquired production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 25 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; Wherein: the second interval time is greater than the first interval time.

[0007] The solution is further that: when the number of intervals is equal to or greater than 2, the measured moisture content of the shredded stems at the outlet of the stirring drum for comparison is the average value of the sum of the moisture content of the shredded stems measured in the current interval and the moisture content of the shredded stems measured in each previous interval.

[0008] A further solution is: the first interval time is 5 seconds, and the second interval time is 10 seconds.

[0009] The solution is further that: the neural network model is a multivariate linear regression algorithm model.

[0010] The scheme further comprises: comparing the predicted moisture content of the delayed shredded stems with the measured moisture content of the shredded stems at the outlet of the mixing drum, and making no adjustment when the difference does not exceed 0.05 The beneficial effects of the present invention are: The present invention applies a multivariate linear regression algorithm to the intelligent control system of the feeding equipment of the shredded stem production line. Through machine learning, the outlet moisture is predicted by utilizing the long-term accumulation of production data, and the water supply valve is controlled in combination with the prediction results to achieve precise regulation of the outlet moisture under the conditions of material head, rapid adjustment, and stable conditions. The present invention can effectively cope with non-steady-state changes in the production process and provide a more stable control effect. The prediction and control by the multivariate linear regression algorithm can significantly improve the control accuracy of the outlet moisture. The model training and prediction are performed by utilizing the long-term accumulation of production data, and the system has strong robustness and adaptability.

[0011] The invention is further explained in detail below with reference to the accompanying drawings and specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 It is a schematic diagram of the structure of the feeding equipment of the shredded stem production line of the present invention; Figure 2 It is a waveform diagram of the outlet moisture content according to the present invention following the predicted value. DETAILED DESCRIPTION

[0013] A non-steady-state intelligent control method for moisture content at the outlet of a feeding device in a shredded stem production line, such as Figure 1 As shown, it includes a stirring drum 1, and the shredded stems 2 are fed into the stirring drum 1 from the inlet 101 of the stirring drum through a conveyor belt 3. The shredded stems are humidified by the stirring drum 1 and sent out from the outlet 102 of the stirring drum to the next process. A weighing scale 4 is arranged on the conveyor belt, and a shredded stems moisture content tester 5 is respectively arranged at the inlet 101 and the outlet 102 of the stirring drum 1. A temperature / humidity sensor 6 is arranged in the stirring drum. The stirring drum 1 adjusts the moisture content of the shredded stems sent out from the outlet of the stirring drum 1 by controlling a steam valve 7 and a water supply valve 8 by a controller (not shown). The steam valve 7 and the water supply valve 8 are connected to a steam pipe 9 and a water supply pipe 10, respectively. The method includes: establishing an outlet moisture prediction model and regulating the outlet moisture of the stirring drum 1 based on the outlet moisture prediction model: The establishment of the export moisture prediction model comprises: obtaining past production data of the shredded stem production line, sending the past production data into a neural network model for training, the neural network model is a multivariate linear regression algorithm model, the training is for the purpose of obtaining the export delayed moisture content, and the export moisture prediction model is obtained after training, the production data comprises: steam valve opening, water supply valve opening, feeding unit batch, feeding unit cigarette brand, moisture content of shredded stems at the inlet and outlet of the mixing drum, hot air temperature at the inlet of the mixing drum, mixed air temperature, return air temperature of shredded stems, ambient temperature, ambient humidity, and scale accumulation; The control of the outlet moisture of the mixing drum 1 based on the outlet moisture prediction model includes: Step 1: Determine the initial opening of the steam valve and the water supply valve according to past production data, start the conveyor belt to feed the shredded stem raw materials into the mixing drum, and gradually adjust the steam supply valve and the water supply valve to the initial opening; Step 2: Acquire production data in real time, input the real-time acquired production data into the outlet moisture prediction model to obtain the predicted moisture content of the delayed stems at the outlet of the mixing drum, compare the predicted moisture content of the delayed stems with the measured moisture content of the stems at the outlet of the mixing drum, adjust the steam valve and the water supply valve according to the comparison result, and realize the regulation of the moisture content of the stems at the outlet of the mixing drum.

[0014] Figure 2 In the figure, a shows the predicted waveform of the outlet moisture, and b shows the measured waveform of the outlet moisture. The vertical Y coordinate in the figure is the moisture content, and the horizontal X coordinate is the accumulated amount on the scale.

[0015] Wherein: the method further comprises a slow adjustment stage, a fast adjustment stage and a steady adjustment stage: The slow adjustment stage is the material head stage, which is the stage where the cumulative amount of the continuous weighing by the metering scale 4 is ≤170KG. The outlet moisture control in the slow adjustment stage includes: (1) When the accumulated amount reaches 42 kg, gradually adjust the steam supply valve and water supply valve to the initial opening; (2) Input the real-time production data into the outlet moisture prediction model at the first interval continuously to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 25 seconds, compare the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjust the steam valve and the water supply valve according to the comparison result (no adjustment is made if the difference does not exceed 0.05). A change of 0.1 in moisture content corresponds to a change of 1 in valve opening; The rapid adjustment stage is a stage in which the cumulative amount of the weighing scale 4 is 170KG≤the cumulative amount of the weighing scale≤300KG. The outlet moisture control in the rapid adjustment stage includes: Input the real-time production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the mixing drum outlet after a delay of 20 seconds, compare the predicted moisture content of the shredded stems at the delay with the measured moisture content of the shredded stems at the mixing drum outlet, and adjust the steam valve and the water supply valve according to the comparison result (no adjustment is made if the difference does not exceed 0.05), and a change of 0.1 in moisture content corresponds to a change of 1 in valve opening; The stable adjustment stage is the stage when the cumulative amount of the continuous weighing by the metering scale 4 is ≥300KG. The outlet moisture control in the stable adjustment stage includes: Input the real-time production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the mixing drum outlet after a delay of 25 seconds, compare the predicted moisture content of the shredded stems at the delay with the measured moisture content of the shredded stems at the mixing drum outlet, and adjust the steam valve and the water supply valve according to the comparison result; (no adjustment is made if the difference does not exceed 0.05), difference ∈ [0.08, 0.13]: a change of 0.1 in moisture content corresponds to a change of 0.4 in valve opening, and a difference greater than 0.03: a change of 0.1 in moisture content corresponds to a change of 0.9 in valve opening; Wherein: the second interval time is greater than the first interval time.

[0016] In the embodiment: when the number of intervals is equal to or greater than 2, the measured moisture content of the shredded stems at the outlet of the stirring drum for comparison is the average value of the moisture content of the shredded stems measured at the current interval and the moisture content of the shredded stems measured at each previous interval.

[0017] Wherein: the first interval time is 5 seconds, and the second interval time is 10 seconds.

[0018] In the embodiment: about establishing the export moisture prediction model: The neural network model is a multivariate linear regression algorithm model, which is the same as the traditional neural network model and is a well-known technology. Its formula is expressed as: Y = β0 + β1X1 + β2X2 + ... + βkXk + ε Where: Y is the dependent variable, which represents the outcome we wish to explain or predict.

[0019] X1, X2, ..., Xk are independent variables, representing the factors that affect the dependent variable.

[0020] β0 is the intercept term.

[0021] β1,β2,...,βk are the regression coefficients of each independent variable.

[0022] ε is the random error term, which represents the part that cannot be explained by the independent variables.

[0023] The relationship between data (including independent variables and dependent variables) and the multiple linear regression model. In practical applications, the data set usually contains multiple independent variables and one dependent variable. These data are used to train the multiple linear regression model, and the model parameters β0, β1, ..., βk are estimated by the least squares method and other methods. The training goal of the model is to predict the moisture content at the outlet of the mixing drum. The independent variables are the steam valve opening, the humidification water supply valve opening, the batch of the feeding unit, the tobacco brand of the feeding unit, the stem feeding water flow rate, the moisture content of the stem before feeding, the stem feeding material flow rate, the drum inlet hot air temperature, the mixed air temperature, the stem feeding return air temperature, the ambient temperature, the ambient humidity, and the scale accumulation.

[0024] The specific steps to obtain the forecast data through the export moisture forecast model are as follows: (1) Data collection: During the production process, data such as batch of feeding unit, tobacco brand of feeding unit, opening degree of compensating steam, opening degree of humidifying water supply valve, moisture content of shredded stems after feeding, water flow rate of shredded stems, moisture content of shredded stems before feeding, material flow rate of shredded stems, hot air temperature at drum inlet, mixed air temperature, return air temperature of shredded stems, ambient temperature, ambient humidity, and scale accumulation are collected in real time.

[0025] (2) Data preprocessing: Standardize the collected data to eliminate the influence of different magnitudes and dimensions between data. Data centralization and data dimensionless processing. Data centralization refers to the translation transformation of data so that the mean of the data and sample point set in the new coordinate system coincide. Dimensionless processing refers to normalizing the variance of different variables to achieve dimensionless processing in order to eliminate false variation and make the data model of each variable have equal weight, so as to improve the accuracy of model construction.

[0026] (3) Feature selection: outlet moisture is used as the output variable, and the controllable variables are the opening of the compensating steam valve and the opening of the humidifying water supply valve. The related variables are the batch of the feeding unit, the brand of the feeding unit, the opening of the compensating steam valve, the opening of the humidifying water supply valve, the flow rate of water added to the stem, the moisture content of the stem before feeding, the flow rate of the stem feeding material, the hot air temperature at the drum inlet, the mixed air temperature, the return air temperature of the stem feeding, the ambient temperature, the ambient humidity, and the cumulative amount of the scale. In order to effectively analyze the degree of correlation between the data, the Pearson correlation coefficient is used to measure the related data. The Pearson correlation coefficient is defined as the quotient of the covariance and the standard deviation between two variables. By calculating the correlation coefficient between the independent variables, highly correlated variables are identified and eliminated.

[0027] To establish the export moisture prediction model: First, collect the past production process data, use the above steps (2) and (3) to perform data preprocessing and feature selection, and organize the collected historical data according to input variables and output variables to form training data sets, test data sets, and verification data sets. Use the multivariate linear regression algorithm model to establish the export moisture prediction model using the training data; then: (1) Model evaluation: Use the test data set and validation data set to introduce the mean square error loss function (MSE) to the established model, and evaluate the model obtained by offline training. The smaller the MSE, the more accurate the model. Confirm the MSE evaluation standard according to the control index requirements and start the model evaluation.

[0028] (2) Model optimization: The optimization control is divided into stages, namely the initial stage, the stable initial stage, and the stable stage, and algorithm control is established separately to ensure better optimization control effects and the generalization ability of the model.

[0029] In the actual production process, the predicted value is calculated in real time, and the water supply valve opening is adjusted according to the predicted result. Through the feedback mechanism, the control strategy is continuously optimized to improve the stability and accuracy of the system.

Claims

1. A non-steady-state intelligent control method for moisture at the outlet of a feeding device of a shredded stem production line, comprising a stirring drum, wherein the shredded stem raw material is fed into the stirring drum from the inlet through a conveyor belt, a steam pipe and a water pipe are connected to the stirring drum through a steam valve and a water supply valve, and the shredded stem raw material is humidified by the stirring drum and then sent out from the outlet of the stirring drum to the next process, a metering scale is arranged on the conveyor belt, a shredded stem moisture content tester is respectively arranged at the inlet and outlet of the stirring drum, a temperature / humidity sensor is arranged in the stirring drum, and the stirring drum adjusts the moisture content of the shredded stem sent out from the outlet by controlling the steam valve and the water supply valve, wherein the stirring drum is characterized in that: The method comprises: establishing an outlet moisture prediction model and regulating the outlet moisture of the stirring drum based on the outlet moisture prediction model: The establishment of the outlet moisture prediction model comprises: obtaining past production data of the shredded stem production line, sending the past production data into a neural network model for training, the training is for the purpose of obtaining the delayed outlet moisture content, and the outlet moisture prediction model is obtained after training, wherein the production data comprises: steam valve opening, water supply valve opening, feeding unit batch, feeding unit tobacco brand, moisture content of shredded stems at the inlet and outlet of the mixing drum, hot air temperature at the inlet of the mixing drum, mixed air temperature, return air temperature of shredded stems, ambient temperature, ambient humidity, and scale accumulation; The control of the outlet moisture of the mixing drum based on the outlet moisture prediction model includes: Step 1: Determine the initial opening of the steam valve and the water supply valve according to past production data, start the conveyor belt to feed the shredded stem raw material into the mixing drum, and gradually adjust the steam supply valve and the water supply valve to the initial opening; Step 2: Acquire production data in real time, input the real-time acquired production data into the outlet moisture prediction model to obtain the predicted moisture content of the delayed stems at the outlet of the mixing drum, compare the predicted moisture content of the delayed stems with the measured moisture content of the stems at the outlet of the mixing drum, adjust the steam valve and the water supply valve according to the comparison result, and realize the regulation of the moisture content of the stems at the outlet of the mixing drum.

2. The non-steady-state intelligent control method for moisture content at the outlet of the feeding equipment of the shredded stem production line according to claim 1, characterized in that: The method further comprises a slow adjustment phase, a fast adjustment phase and a smooth adjustment phase: The slow adjustment stage is the material head stage, which is the stage when the accumulated amount on the scale is ≤170KG. The outlet moisture control in the slow adjustment stage includes: (1) When the accumulated amount reaches 42 kg, gradually adjust the steam supply valve and water supply valve to the initial opening; (2) continuously inputting the real-time acquired production data into the outlet moisture prediction model at the first interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 25 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; The rapid adjustment stage is when the scale cumulative weight is 170KG ≤ ≤ 300KG. The outlet moisture control in the rapid adjustment stage includes: Inputting the real-time acquired production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 20 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; The stable adjustment stage is when the accumulated amount of the scale is ≥300KG. The outlet moisture control in the stable adjustment stage includes: Inputting the real-time acquired production data into the outlet moisture prediction model at every second interval to obtain the predicted moisture content of the shredded stems at the outlet of the mixing drum after a delay of 25 seconds, comparing the predicted moisture content of the shredded stems at the outlet of the mixing drum with the measured moisture content of the shredded stems at the outlet of the mixing drum, and adjusting the steam valve and the water supply valve according to the comparison result; Wherein: the second interval time is greater than the first interval time.

3. The non-steady-state intelligent control method for moisture content at the outlet of the feeding equipment of the shredded stem production line according to claim 2, characterized in that: When the number of intervals is equal to or greater than 2, the measured moisture content of the shredded stems at the outlet of the mixing drum for comparison is the average value of the moisture content of the shredded stems measured in the current interval and the moisture content of the shredded stems measured in each previous interval.

4. The non-steady-state intelligent control method for moisture content at the outlet of the feeding equipment of the shredded stem production line according to claim 2, characterized in that: The first interval time is 5 seconds, and the second interval time is 10 seconds.

5. The non-steady-state intelligent control method for moisture content at the outlet of the feeding equipment of the shredded stem production line according to claim 2, characterized in that: The neural network model is a multiple linear regression algorithm model.

6. The non-steady-state intelligent control method for moisture content at the outlet of the feeding equipment of the shredded stem production line according to claim 2, characterized in that: The delayed predicted moisture content of the shredded stems is compared with the measured moisture content of the shredded stems at the outlet of the mixing drum, and no adjustment is made when the difference does not exceed 0.05.

Images