Temperature prediction control method and system for materials at outlet of loosening and damping machine

Abstract

The invention discloses a temperature prediction control method and system for materials at an outlet of a loosening and dampening machine. The method comprises the steps of establishing a thermodynamic equation model, collecting and preprocessing data, optimizing and verifying model parameters, predicting and controlling the temperature in real time and the like, and comprises the following steps of: splitting a prediction model into a quantifiable linear part and a nonlinear residual term which is difficult to directly measure through correlation analysis and time delay evaluation on system input variables; and the interpretability of a physical model and the nonlinear fitting capability of machine learning are fully utilized, so that high-precision and low-calculation-complexity temperature prediction control is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tobacco primary processing equipment, and particularly relates to a loose conditioning machine outlet material temperature prediction control method and system. BACKGROUND

[0002] In the tobacco primary processing process, the loose conditioning machine is used to adjust the moisture content and temperature of the tobacco to meet the process requirements of subsequent processes. Currently, the control of the loose conditioning machine outlet material temperature mainly adopts the traditional PID control method. However, PID control has obvious deficiencies in practical application due to the following reasons:

[0003] (1) System time lag and overshoot: The loose conditioning machine has large thermal inertia, resulting in significant time lag in the system. PID controllers are difficult to respond in time, and are prone to overshoot and lag, affecting the precision and stability of temperature control.

[0004] (2) Complex environmental influence: The outlet material temperature is affected by various environmental factors that cannot be directly measured, such as heat radiation loss, material heat absorption difference, etc. Traditional PID control is difficult to accurately model and compensate for these nonlinear factors.

[0005] (3) Limitations of pure data-driven methods: Some studies attempt to use machine learning methods (such as neural networks) for temperature prediction, but due to the lack of understanding of the physical mechanism, the model has insufficient interpretability and generalization ability, and the model training requires a large amount of data and has high computational complexity. SUMMARY

[0006] In order to overcome the problems in the prior art, the present application provides a loose conditioning machine outlet material temperature prediction control method and system combining thermodynamic equations and machine learning. This method, through correlation analysis and time lag evaluation of system input variables, splits the prediction model into a quantifiable linear part and a nonlinear residual term that is difficult to measure directly, fully utilizing the interpretability of the physical model and the nonlinear fitting ability of machine learning, thereby achieving high-precision, low-complexity temperature prediction control.

[0007] In the present application, correlation analysis is a key step in prediction control. Since there are many input features in the loose conditioning machine system, and the influence of each feature on the outlet material temperature has complex correlations, it is necessary to conduct in-depth correlation analysis on each input variable to ensure that the relationship between the input features and the system output (outlet material temperature) is fully revealed. Specifically, Spearman rank correlation analysis is used to quantitatively evaluate the correlation between input variables and outlet temperature. This method can capture nonlinear correlations and identify which input features have the greatest impact on temperature control.

[0008] Through correlation analysis, not only can the input features that have little effect on temperature changes be effectively excluded, simplifying the model and reducing computational complexity, but also the time delay effects between different input features can be determined. Time delay effect refers to the fact that the influence of some input features on the system temperature is not immediate, but has a certain delay. Through time delay evaluation, the input data can be accurately aligned, allowing the model to capture the key moments that best reflect the system dynamics. Such time delay processing greatly improves the accuracy of temperature prediction, ensuring the timeliness and reliability of the control strategy.

[0009] Therefore, the role of correlation analysis in control is not only to optimize feature input, but also to improve the overall response capability and prediction accuracy of the system. It provides a more streamlined and efficient input variable set for subsequent temperature control, making temperature regulation more accurate and stable.

[0010] In real production environments, many input features cannot be obtained stably and accurately due to technical limitations, measurement conditions, or cost reasons. For example, material heat radiation loss, energy loss in air duct systems, and heat absorption differences in cylinder walls are all difficult to measure directly and change dynamically. In addition, the production environment of the loose humidifier is complex and affected by external environmental temperature, humidity, equipment aging, and other factors, making it difficult for the system to be accurately predicted and controlled through a linear physical model.

[0011] To deal with these complex and unquantifiable or difficult-to-obtain factors, the invention introduces the modeling of residual term A. Residual A represents the system energy loss and influence that cannot be accurately described by the physical model. In residual modeling, we use the XGBoost regression model, which has strong non-linear fitting ability and can effectively capture potential complex relationships in the production process. The parallel processing and optimization algorithm of XGBoost not only allows it to accurately model non-linear factors in the residual term, but also has high computational efficiency, allowing it to quickly respond to system changes.

[0012] In actual production processes, due to uncontrollable factors such as equipment state and environmental fluctuations, the existence of residual term A can make up for the shortcomings of the physical model and ensure the stability of the system in complex production environments. Residual modeling incorporates these difficult-to-quantify external factors into the prediction and control process through machine learning, allowing the model to maintain high prediction accuracy even when dealing with changing actual production conditions. In addition, the introduction of residual A also reduces the system's dependence on large amounts of data, enhancing the model's robustness. Even in cases where some data is incomplete or inaccurate, it can still provide relatively accurate predictions.

[0013] In summary, modeling the residual term A is not only an effective supplement to existing physical models but also a key measure for addressing the uncertainty and complexity of real-world production environments. By accurately modeling the residual term A through machine learning algorithms, this invention significantly improves the system's prediction and control capabilities in complex environments, making temperature regulation more flexible and reliable.

[0014] The technical solution adopted in the present invention is:

[0015] A method for predicting and controlling the temperature of materials at the outlet of a loose rehumidifying machine comprises the following steps:

[0016] S1 Establishing thermodynamic equation model

[0017] Based on the physical characteristics of the loose conditioning machine, a thermodynamic equation describing the outlet material temperature is established as the linear part of the model. This equation captures the main energy exchange processes in the system, including the following quantifiable influencing factors:

[0018] (1) Feed water flow (Q_water): used to adjust the moisture content of the material and affect the thermal balance of the system.

[0019] (2) Steam flow (Q_steam): Heat is transferred through steam to increase the material temperature.

[0020] (3) Hot air volume (Q_hot): provides heat energy to the system and affects the heating effect of the material.

[0021] Since the circulating air circulates within the system, its impact on the total system energy is minimal and can be ignored. Energy losses that cannot be directly measured, such as energy carried away by dehumidification air, thermal radiation losses from the drum wall, and differences in heat absorption by the materials, are summarized as the residual term A.

[0022] Its thermodynamic differential equation expression is:

[0023]

[0024] Parameter Description:

[0025] T_sys: system temperature, that is, the temperature of the material after absorbing heat

[0026] t: time variable

[0027] T_hot: hot air temperature

[0028] T_steam: saturated steam temperature, which can be calculated from the atomization steam pressure

[0029] T_water: feed water temperature

[0030] C_air: specific heat capacity of air

[0031] C_steam: specific heat capacity of saturated steam

[0032] C_water: specific heat capacity of water

[0033] Q_hot: hot air flow rate

[0034] Q_steam: steam flow rate

[0035] R: steam moisture gain coefficient

[0036] L_steam: steam latent heat coefficient

[0037] C: unknown term, total heat capacity of the system

[0038] A: residual term (compensation term) to represent the unquantifiable effects of the system

[0039] The differential equation is constructed as follows: the left side of the equation is the total heat capacity of the system multiplied by the system change rate, with the unit of KJ / S, and the right side of the equation is the change rate of the hot air energy, the change rate of the sensible heat energy of steam and water in the steam, the change rate of the latent heat energy of the steam, the change rate of the energy of the incoming water flow, and the residual amount A that cannot be measured and calculated, with the unit of KJ / S.

[0040] By using thermodynamic equations, the main physical mechanisms of the system can be effectively described, providing an accurate and interpretable description of the system behavior. According to the experimental results, the biggest advantage of this physical model is to reduce the dependence on a large amount of data and improve the reliability of the model, which has a significant advantage compared with purely data-driven methods.

[0041] S2 Modeling of residual term A

[0042] For nonlinear effects that cannot be directly quantified, they are summarized as residual term A, and machine learning methods are used to model them. Specifically, an XGBoost regression model is selected to predict A.

[0043] XGBoost is good at capturing complex nonlinear relationships and can effectively handle nonlinear effects in the residual term. In addition, XGBoost has high computational power and speed. This algorithm uses parallel processing and optimized loss functions to improve the training speed and prediction accuracy of the model. In addition, XGBoost has strong anti-overfitting ability through its built-in regularization mechanism, improving the generalization performance of the model. These characteristics make XGBoost an ideal choice for modeling the residual term, balancing high prediction performance and computational efficiency.

[0044] S3 Data collection and preprocessing, including:

[0045] S3.1 Data collection

[0046] Collecting the operating data of loose conditioning machine, including input features and output targets;

[0047] The input features include: inlet material moisture content, scale instantaneous flow, feed water flow, inlet hot air temperature and air volume, steam flow, atomized steam pressure, moisture exhaust damper opening, circulating air damper opening, drum speed, etc.

[0048] The output targets include: actual value of outlet material temperature.

[0049] S3.2 Data preprocessing, including:

[0050] (1) Moving average filtering: smoothing time series data, reducing noise interference, ensuring the accuracy and reliability of model input.

[0051] (2) Spearman rank correlation analysis: determine the correlation between each input feature and outlet temperature, calculate the optimal time delay, and align the data. This helps to extract effective features and enhance the prediction performance of the model.

[0052] S4 Model parameter optimization and verification, including:

[0053] S4.1 Determine the total heat capacity C of the system

[0054] The total heat capacity C of the system is a key parameter in thermodynamic equations, representing the ability of the system to store heat. By constructing an optimization problem with the difference between predicted temperature and actual temperature as the loss function, using BFGS algorithm to minimize the loss function, the optimal C value is found to ensure the reliability of the thermodynamic model.

[0055] S4.2 Calculate and model residual A

[0056] After determining C, calculate the actual residual A_calculated based on known data. Then, use XGBoost model to model A with input features as independent variables. The high prediction performance and computational efficiency of XGBoost enable it to effectively capture the nonlinear effects in the residual, improving the overall prediction accuracy.

[0057] S5 Real-time prediction and control

[0058] In practical application, deploy the trained model for real-time temperature prediction. According to the current collected input features, predict the residual A and outlet temperature T_sys_predicted. Compare the predicted temperature with the target temperature, calculate the control variables (such as steam flow) that need to be adjusted, and input them into the PID controller in advance.

[0059] A computer-readable storage medium having stored thereon a computer program executable by a processor to implement the steps of a loose conditioning machine outlet material temperature prediction control method described herein.

[0060] A loose conditioning machine outlet material temperature prediction control system, the system comprising:

[0061] Sensors for acquiring inlet material moisture content, scale instantaneous flow, feed water flow, inlet hot air temperature and air volume, steam flow, atomized steam pressure, moisture removal damper opening, circulating damper opening, drum speed, etc.

[0062] A computer-readable storage medium described herein;

[0063] A computer acquires inlet material moisture content, scale instantaneous flow, feed water flow, inlet hot air temperature and air volume, steam flow, atomized steam pressure, moisture removal damper opening, circulating damper opening, drum speed, etc. data by communicating with the above-mentioned sensors; execute the program of the computer-readable storage medium.

[0064] The beneficial effects of the present application include:

[0065] (1) Improved prediction accuracy

[0066] Combining thermodynamic model and XGBoost model, effectively capturing the linear and nonlinear effects of the system. Experimental results show that the proportion of samples with prediction error less than 0.3℃ reaches more than 80%, significantly improving the accuracy of temperature prediction.

[0067] (2) Reduce the time lag and overshoot:

[0068] Through correlation analysis we can know the time of the model to predict the outlet temperature in advance, the PID controller can adjust the control quantity in time, reduce the time lag and overshoot of the system, improve the control performance.

[0069] (3) Reduce the complexity of model calculation

[0070] The model is divided into linear and nonlinear parts, which simplifies the model structure. Compared with pure data-driven machine learning models, it reduces the amount of calculation and model complexity while maintaining high prediction accuracy.

[0071] (4) Enhance the interpretability of the model

[0072] The thermodynamic model describes the main energy exchange process of the system, which has clear physical meaning. Machine learning model handles nonlinear residuals, and the combination of the two makes the model easier to understand and maintain.

[0073] (5) Improve product quality and production efficiency

[0074] Accurate temperature control helps to improve the product quality of tobacco primary processing, reduce energy consumption, and improve production efficiency, which has important economic and social benefits. BRIEF DESCRIPTION OF DRAWINGS

[0075] Figure 1 Method flow chart of the present application.

[0076] Figure 2 Model training flow chart.

[0077] Figure 3 Real-time prediction control flow chart.

[0078] Figure 4 Input correlation heat map.

[0079] Figure 5 Actual value and predicted value comparison chart of test set residual A.

[0080] Figure 6 Actual value and predicted value comparison chart of test set steady-state stage outlet material temperature. DETAILED DESCRIPTION

[0081] Example 1

[0082] 1. Data collection and preprocessing

[0083] 1.1 Data collection

[0084] From YX cigarette factory The super-loose conditioning machine collected 30 batches of data as the training set and 5 batches of data as the test set. The data includes input features such as inlet material moisture content, scale instantaneous flow, feed water flow, inlet hot air temperature and air volume, steam flow, atomized steam pressure, and outlet material temperature actual value.

[0085] 1.2 Data preprocessing

[0086] (1) Use moving average filter to smooth input features and output targets, reduce noise, and improve data quality.

[0087] (2) Spearman rank correlation analysis was performed, 21 batches were used to analyze the correlation coefficient, 300s was used as the maximum time delay range, the correlation between each input feature and the outlet temperature was determined, the best time delay was calculated, and it was ensured that the model could accurately capture the relationship between the input and the output.

[0088] (3) Perform time alignment based on the analysis results. Take the amount with the smallest time delay among the strongly correlated inputs of the differential equation as the advance prediction control amount, uniformly advance the output of the training set by the time lag of the input point, and then calculate and align the time lags that need to be shifted back for other strongly correlated points in the equation. Timestamp alignment means that the specific time delay of each input point relative to the material flow at the entrance will be obtained, that is, when the material element enters the equipment, the predicted output value can eventually be obtained at the latest input point. This directly determines how far in advance the model can make predictions and the effectiveness of the model's predictions.

[0089] 1.3 Model parameter optimization

[0090] (1) Determine the total heat capacity C of the system

[0091] Construct an optimization problem with the difference between the predicted temperature and the actual temperature as the loss function:

[0092]

[0093] The BFGS algorithm is used to find the optimal C value by minimizing the loss function to ensure the reliability of the thermodynamic model.

[0094] (2) Calculation and modeling of residual A

[0095] After determining C, we calculate the actual residuals A of the training set. We use the XGBoost regression model to model A, using the input features as independent variables. XGBoost excels at handling nonlinear relationships and can efficiently capture complex patterns in the residuals, improving the model's predictive accuracy.

[0096] 2. Model Validation and Evaluation

[0097] 2.1 Residual Prediction

[0098] On the test set, the trained XGBoost model is used to predict the residual A_predicted, which is then substituted into the thermodynamic equation to calculate the predicted value of the outlet material temperature T_sys_predicted.

[0099] 2.2 Performance Evaluation

[0100] like Figure 4 As shown in the figure, correlation analysis reveals the three most correlated points: hot air flow, steam flow, and feed water flow. Their time delays are 100s, 220s, and 150s, respectively. Therefore, we can conclude that the model can predict the outlet temperature 100s in advance. This allows the PID controller to adjust the control variable in a timely manner, reducing system time delay and overshoot, thereby improving control performance.

[0101] The mean square error (MSE) and absolute error are calculated to evaluate the prediction performance of the model. The results show that in the steady state phase, the proportion of samples with a prediction error less than 0.3°C is more than 80%, indicating that the model has high precision.

[0102] 2.3 Visualization

[0103] A comparison chart of the actual and predicted values of the residual error A and the outlet temperature T_sys is drawn to visually demonstrate the prediction ability of the model.

[0104] 3. Temperature prediction control implementation

[0105] 3.1 Real-time prediction and control

[0106] In actual production, the trained model is deployed for real-time prediction. According to the real-time collected input features, the residual error A and the outlet temperature T_sys_predicted are predicted. The prediction results are used as the PV value of the PID controller to adjust the control variable (such as steam flow) and input the PID controller in advance.

[0107] 3.2 Improved control effect

[0108] By adjusting the control variable in advance, the time delay and overshoot of the system are reduced, the response speed and stability of the control system are improved, and the performance of the temperature control is improved.

[0109] 4. Experimental results

[0110] As shown in Figure 5 , the vertical axis is the value of the residual error A, with a unit of 10^5 KJ, and the horizontal axis is the time index, with each point separated by one second; the blue curve is the actual value of A, and the yellow dashed line is the model predicted value of A. From the image, we can clearly see that the predicted value of A has a good fitting effect.

[0111] As shown in Figure 6 , the vertical axis is the value of the outlet material temperature, with a unit of °C, and the horizontal axis is the time index, with each point separated by one second; the blue curve is the actual value of the outlet material temperature, and the yellow dashed line is the corrected value obtained after compensating the predicted value of the outlet material temperature with the residual error A. From the image, we can see that the predicted value and the actual value have a good fitting effect in terms of both the value and the trend of the value.

[0112] 5. Model performance

[0113] On the test set, the model shows high prediction accuracy. The XGBoost model effectively predicts the residual A, and the overall outlet temperature prediction value is highly consistent with the actual value. In practical applications, the use of this model to optimize the PID controller can reduce the fluctuation of the outlet material temperature, speed up the response speed of the control system, and improve the stability, verifying the effectiveness of the method.

[0114] 6. Conclusion

[0115] The present application innovatively splits the loose moisture regaining machine outlet material temperature prediction model into a linear (quantifiable) part and a nonlinear (residual) part by combining thermodynamic equations and machine learning methods. The main energy exchange process of the system is captured by the physical model, and the machine learning model handles complex nonlinear effects, improving the accuracy of the prediction, reducing the computational complexity, and enhancing the interpretability of the model. In practical applications, the time lag and overshoot problems in PID control are effectively solved, the accuracy and stability of temperature control are improved, and it has important popularization value.

Claims

1. A method for predicting and controlling the temperature of materials at the outlet of a loose rehumidifying machine, characterized in that: The following steps are involved: S1 Establishing thermodynamic equation model Based on the physical characteristics of the loose conditioning machine, a thermodynamic equation describing the outlet material temperature is established as the linear part of the model. For nonlinear effects that cannot be directly quantified, they are summarized as residual terms A and modeled using machine learning methods; S2 data collection and preprocessing; S3 model parameter optimization and verification, including: S3.1 Determine the total heat capacity C of the system; S3.2 Modeling the residual A After determining the total heat capacity C of the system, the actual residual A_calculated is calculated based on the known data. Then, the residual A is modeled using the XGBoost model with the input features as independent variables. S4 real-time temperature prediction and control Based on the currently collected input features, the residual A and outlet temperature T_sys_predicted are predicted; the predicted temperature is compared with the target temperature, the control variables that need to be adjusted are calculated, and input into the PID controller in advance for control.

2. The method according to claim 1, characterized in that The thermodynamic equation model described in step S1 includes the following quantifiable influencing factors: (1) Feed water flow Q_water: used to adjust the moisture content of the material and affect the thermal balance of the system. (2) Steam flow Q_steam: Heat is transferred through steam to increase the material temperature. (3) Hot air volume Q_hot: provides heat energy to the system and affects the heating effect of the material; For the energy that cannot be measured directly, it is summarized as the residual term A; Its thermodynamic differential equation expression is: in: T_sys: temperature of the material after absorbing heat, t: time variable, T_hot: hot air temperature, T_steam: saturated steam temperature, T_water: feed water temperature, C_air: specific heat capacity of air, C_steam: specific heat capacity of saturated steam, C_water: specific heat capacity of water, Q_hot: hot air volume, Q_steam: steam flow rate, R: steam moisture gain coefficient, L_steam: steam latent heat coefficient.

3. The method according to claim 2, characterized in that The energy that cannot be directly measured includes the energy lost by the dehumidification wind, the heat radiation loss of the cylinder wall, and the difference in heat absorption of the material.

4. The method according to claim 1, wherein In step S1, the XGBoost regression model is selected to predict A.

5. The method according to claim 1, wherein Step S2 further includes: S2.1 Data Collection Collecting operation data of the loosening and rehydrating machine, wherein the operation data includes input characteristics and output targets; The input characteristics include inlet material moisture content, scale instantaneous flow rate, feed water flow rate, inlet hot air temperature and air volume, steam flow rate, atomizing steam pressure, dehumidification damper opening, circulation damper opening and drum speed; The output target includes the actual value of the outlet material temperature; S2.2 Data preprocessing, including: (1) Moving average filtering; (2) Spearman rank correlation analysis: determine the correlation between each input feature and the outlet temperature, calculate the optimal delay, and align the data.

6. The method according to claim 5, characterized in that The moving average filtering includes smoothing the time series data to reduce noise interference.

7. The method according to any one of claims 1 to 6, characterized in that Determining the total heat capacity C of the system in step S4 also includes: By constructing an optimization problem with the difference between the predicted temperature and the actual temperature as the loss function, the BFGS algorithm is used to find the optimal total heat capacity C value of the system by minimizing the loss function Loss(C).

8. The method according to claim 7, characterized in that The minimization loss function Loss(C) includes:

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: The computer program can be executed by a processor to implement the steps of a method for predicting and controlling the temperature of materials at the outlet of a loose conditioning machine as described in any one of claims 1 to 8.

10. A temperature prediction and control system for materials at the outlet of a loose rehumidification machine, the system comprising: Used to obtain the inlet material moisture content sensor, scale instantaneous flow sensor, feed water flow sensor, inlet hot air temperature and air volume sensor, steam flow sensor, atomizing steam pressure sensor, dehumidification damper opening sensor, circulation damper opening sensor and drum speed sensor; A computer-readable storage medium as claimed in claim 9; The computer obtains the moisture content of the inlet material, the instantaneous flow rate of the scale, the feed water flow rate, the inlet hot air temperature and air volume, the steam flow rate, the atomizing steam pressure, the dehumidification damper opening, the circulation damper opening, and the drum speed by communicating with the sensor; and executes the computer program stored on the computer-readable storage medium.

Images