Loosening moisture regaining machine inlet hot air temperature prediction control method in combination with pattern recognition PID (Proportion Integration Differentiation) control

Abstract

The invention discloses a loose moisture regaining machine inlet hot air temperature prediction control method combined with pattern recognition PID control, which comprises the following steps of: acquiring historical data in a production process of a loose moisture regaining machine, and constructing a training data set; identifying a plurality of working condition modes in the operation process of the loose moisture regaining machine by utilizing the mode identification model, and extracting dynamic characteristics related to PID (Proportion Integration Differentiation) parameter setting; constructing a dynamic PID parameter mapping model; carrying out online real-time optimization on PID parameters output by the dynamic PID parameter mapping model by adopting an optimization algorithm; and performing predictive control on the hot air temperature at the inlet of the loose moisture regaining machine by utilizing the optimized parameters, and correcting the control model. Through the PID control method combining mode recognition and PSO optimization, the control precision of the hot air temperature at the inlet of the loose moisture regaining machine is remarkably improved, the temperature deviation is reduced to + / -0.5 DEG C, the method has high adaptability to complex working conditions, and the feasibility and economic value of the method in the tobacco primary processing technology are verified.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tobacco primary processing, more specifically to loose conditioning inlet hot air temperature prediction control technology, and particularly to a loose conditioning inlet hot air temperature prediction control method combining pattern recognition PID control. BACKGROUND

[0002] In the tobacco primary processing, loose conditioning is a crucial temperature and humidity increasing link, which aims to increase the temperature and moisture content of tobacco sheets, so as to make them soft and loose to meet the requirements of subsequent processing. The stability and control accuracy of loose conditioning inlet hot air temperature directly determine the processing quality, sensory quality of tobacco sheets and production efficiency of subsequent processes.

[0003] At present, the traditional PID control method is generally used for the control of loose conditioning inlet hot air temperature. However, the production process has the characteristics of large thermal inertia, pure lag, nonlinearity and variable working conditions, which makes the traditional control method face severe challenges, which are embodied in the following aspects: Control lag and insufficient dynamic response: The loose conditioning process is a complex heat and mass exchange process. There is a large time delay from adjusting the steam valve opening to the change of inlet hot air temperature. The traditional PID controller relies on the current and past errors for adjustment and cannot predict the future temperature change trend, which leads to a "slow half step" response to disturbances, resulting in a large overshoot and a long adjustment time. Especially in the working condition switching stage of material head and tail, the temperature fluctuation is particularly obvious.

[0004] Fixed model and poor adaptability to working conditions: The parameters of the existing PID control are usually fixed or depend on experience for a limited number of preset switches. However, complex factors such as differences in raw material batches (such as tobacco leaves from different places and grades), changes in environmental temperature and humidity, fluctuations in steam pressure, and equipment aging will change the dynamic characteristics of the controlled object. Fixed PID parameters cannot adapt to these changes, resulting in a sharp decline in control performance under another working condition and poor robustness.

[0005] Single control strategy and insufficient use of process information: The existing control method is mostly single-loop feedback control, and its control quality depends heavily on the timeliness and accuracy of the feedback signal. Although some research has tried to introduce advanced methods such as fuzzy control and decoupling control, in actual application, there is a lack of deep mining and utilization of rich historical data and real-time running data. For example, the dynamic mode characteristics of different production stages (such as material head accelerated heating, steady-state heating, and material tail cooling) are not effectively identified and utilized, and multiple related variables such as steam flow, environmental humidity, and material flow are not systematically integrated and forward adjusted, and the intelligent degree of the control strategy is insufficient. SUMMARY

[0006] The present application aims at the above-mentioned problems existing at present, and provides a loose moisture regaining machine inlet hot air temperature prediction control method combining pattern recognition PID control.

[0007] The technical solution of the present application is as follows: A loose moisture regaining machine inlet hot air temperature prediction control method combining pattern recognition PID control comprises the following steps: Collecting historical data in the production process of the loose moisture regaining machine and pre-processing to construct a training data set; the historical data includes inlet hot air temperature, steam flow, inlet material moisture content and scale instantaneous flow; Based on the training data set, a pattern recognition model is used to identify multiple working condition modes including head, steady state and tail of the loose moisture regaining machine in the running process, and dynamic characteristics related to PID parameter setting are extracted; A dynamic PID parameter mapping model is constructed, which takes dynamic characteristics and real-time process variables as input and outputs real-time optimized PID parameter groups (Kp, Ki, Kd); , , ); An optimization algorithm is used to perform online real-time optimization on the PID parameters output by the dynamic PID parameter mapping model, and the optimization target is to minimize the error between the hot air temperature set value and the actual value; The optimized PID parameters are used to perform prediction control on the inlet hot air temperature of the loose moisture regaining machine, and the control model is corrected based on the real-time feedback temperature data.

[0008] Through the above method, the control precision is significantly improved: through pattern recognition, the working condition change is predicted in advance, and the dynamic PID model is used for accurate parameter setting, which can stabilize the steady state control deviation of the inlet hot air temperature within ±0.5℃, which is much better than the traditional PID control.

[0009] Further, the optimization algorithm is a particle swarm optimization PSO algorithm; the objective function of the optimization target is: , Wherein, the hot air temperature set value, the actual value of the hot air temperature, the steam flow, the weight coefficient.

[0010] Through the above method, the robustness is significantly enhanced: due to the adoption of real-time data-based pattern recognition and online parameter optimization (PSO) mechanism, the application can automatically adapt to different brands of tobacco leaves, changes in environmental humidity, equipment state drift and other complex working conditions, and shows strong self-adaptive ability and robustness.

[0011] Further, the pattern recognition model is a long short-term memory network (LSTM) model; the dynamic features include the instantaneous slope of the inlet hot air temperature, the peak time of the temperature change curve, and the temperature fluctuation amplitude within a time window.

[0012] Further, the pattern recognition model is also used to identify abnormal working condition patterns caused by steam pressure fluctuations or material flow mutations.

[0013] Further, the dynamic PID parameter mapping model is a multi-layer perception (MLP) neural network model, and the input further includes environmental humidity and hot air fan frequency.

[0014] Further, the target function further introduces a constraint term related to the quality of the tobacco sheet, and the constraint term is the variance of the outlet tobacco sheet moisture content predicted based on process data.

[0015] Further, the trigger condition for correcting the control model is that the deviation between the actual value and the model predicted value of the hot air temperature continuously exceeds the preset threshold for a set time length.

[0016] The application also includes an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements a loose conditioning machine inlet hot air temperature prediction control method combined with pattern recognition PID control when executing the program.

[0017] The application also includes a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement a loose conditioning machine inlet hot air temperature prediction control method combined with pattern recognition PID control.

[0018] The application also includes a loose conditioning machine control system comprising a sensor group, an actuator, and a controller, wherein the controller is configured to execute a loose conditioning machine inlet hot air temperature prediction control method combined with pattern recognition PID control for real-time adjustment of the opening degree of the steam valve to control the inlet hot air temperature.

[0019] Compared with the existing technology, the application has the following advantages: 1. The system overshoot is greatly reduced: the prediction control strategy can compensate for the thermal inertia of the system in advance, effectively avoiding the temperature overshoot phenomenon when the working condition is switched (such as the beginning of the material head) or disturbed, making the system transition process more stable. 2. System response speed is accelerated: compared with the traditional PID control, the response time of the application to the set value tracking and external disturbance (such as steam pressure fluctuation, material flow change) is shortened by more than 30%, and the set temperature range can be quickly restored; 3. Intelligent setting of data driving is realized: the dependence of traditional PID on accurate mathematical model is overcome, the internal mapping relationship between the characteristic mode in the historical data and the optimal PID parameter is mined, the "self-setting" and "self-learning" of the controller parameter are realized, and the dependence on expert experience is reduced; the model has strong generalization ability and is easy to maintain: the dynamic PID model is trained based on a large amount of historical data, and can cover various production scenes. When the control performance decreases due to equipment aging and other reasons, the model can be retrained and updated by injecting new production data, and the maintenance cost is low; 4. Energy saving effect is obvious: the objective function in the optimization algorithm considers the temperature error and steam energy consumption, and can find the optimal energy-saving operating point under the premise of ensuring control accuracy, effectively reducing the steam consumption in the production process. The product quality consistency is fundamentally improved: the high stability of the inlet hot air temperature provides a uniform and controllable temperature and humidity environment for the tobacco sheet, which directly improves the bulkiness and back permeability of the tobacco sheet, reduces quality defects such as "water stained tobacco", and ensures the sensory quality and batch uniformity of the final product (tobacco) BRIEF DESCRIPTION OF DRAWINGS Figure 1 The flow chart of data processing and model construction of the application.

[0020] Figure 2 The schematic diagram of the PID parameter optimization algorithm of the application. DETAILED DESCRIPTION

[0021] It should be noted that the terms "first" and "second" and the like relational terms only serve to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or inherent to such process, method, article or equipment. Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or equipment including the element.

[0022] The features and performances of the application will be further described in detail below in combination with embodiments.

[0023] Please refer to Figure 1 andFigure 2 A loose moisture regaining machine inlet hot air temperature prediction control method combined with pattern recognition PID control, comprising the following steps: Data collection and preprocessing: Collecting historical data of loose moisture regaining machine inlet hot air temperature, including temperature value, steam flow, environmental humidity, etc. Normalize the data and divide the training set and test set.

[0024] Pattern recognition and feature extraction: Using time series pattern recognition algorithm (such as long short-term memory network LSTM) to train historical data, identify hot air temperature change pattern under different working conditions (head, steady state, tail); Extract key features (such as temperature slope, peak time, etc.), establish mapping relationship with PID parameters (P, I, D) , , P, I, D are proportional parameter, integral parameter, differential parameter respectively. , , P, I, D are proportional parameter, integral parameter, differential parameter respectively.

[0025] Dynamic PID model construction: Based on feature pattern, construct a multiple-input single-output (MISO) model, the output is PID parameter; The model form is: , Where, is a nonlinear function obtained by machine learning training.

[0026] Optimization algorithm parameter setting: Use particle swarm optimization (PSO) algorithm to optimize PID parameters in real time; The objective function is to minimize the weighted sum of temperature error and energy consumption: , Where, is the set value of hot air temperature, is the actual value of hot air temperature, is the steam flow, is the weight coefficient.

[0027] Update particle position (i.e. PID parameters) by iteration to find the optimal solution.

[0028] Prediction control and real-time feedback: Use dynamic PID model to predict the temperature change trend after action response, and adjust steam flow in advance; Collect actual temperature data in real time, compare with predicted value, and correct model parameters to adapt to new working conditions.

[0029] Dynamic modeling based on historical data: Extract the characteristic pattern of hot air temperature change using pattern recognition technology, and build a dynamic PID parameter model. Optimization algorithm driven parameter self-tuning: Combine particle swarm optimization (PSO) algorithm to adjust PID parameters in real time to adapt to the change of working conditions. Predictive control strategy: Through the prediction model, the influence of thermal inertia is compensated in advance, and the stability and response speed of temperature control are improved.

[0030] The application also includes an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor implements a loose conditioning machine inlet hot air temperature predictive control method combined with pattern recognition PID control when executing the program.

[0031] The application also includes a computer-readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to implement a loose conditioning machine inlet hot air temperature predictive control method combined with pattern recognition PID control.

[0032] The application also includes a loose conditioning machine control system comprising a sensor group, an actuator, and a controller, wherein the controller is configured to execute a loose conditioning machine inlet hot air temperature predictive control method combined with pattern recognition PID control for real-time adjustment of the opening degree of the steam valve to control the inlet hot air temperature.

[0033] The beneficial effects of the method of the application are verified by specific examples.

[0034] Data acquisition and preprocessing: Data sources: Collect at least 80 continuous batches of production data from a cigarette loose conditioning machine, covering the following input features and output variables: Input features: inlet material moisture content (%), instantaneous flow rate of the scale (kg / s), main steam flow rate (kg / h), inlet hot air flow rate (kg / h), steam flow rate (kg / h), atomized steam pressure (MPa), hot air fan frequency (Hz), fresh air flow rate (kg / h).

[0035] Output variables: actual value of inlet hot air temperature (℃).

[0036] Data preprocessing: data cleaning: Remove missing values (such as invalid data caused by sensor failure) and outliers (such as outliers of sudden changes in hot air temperature), and use interpolation methods (such as linear interpolation) to complete the missing values.

[0037] Perform moving average filtering on features with high noise (such as instantaneous flow rate).

[0038] Feature Normalization: Min-Max normalization is applied to all input features and output results, mapping the range to [0, 1] to eliminate dimensional differences.

[0039] Data Partitioning: The dataset is divided by time series: 70% for training (56 batches), 15% for validation (12 batches), and 15% for testing (12 batches).

[0040] Model Construction and Parameter Optimization: Pattern Recognition and Feature Extraction: Model Selection: Long Short-Term Memory Network (LSTM) is used to train three types of working condition data in historical data separately, identifying the hot air temperature change pattern under different working conditions.

[0041] Feature Extraction: Extract Key Dynamic Features: Temperature Slope (℃ / min), Peak Time (min), and Fluctuation Frequency (Hz).

[0042] Construct Feature Matrix: Combine input features and dynamic features to form a multi-dimensional input vector.

[0043] Dynamic PID Model Construction: Model Training: Input: Normalized feature matrix (temperature pattern, steam flow, environmental humidity, etc.).

[0044] Output: PID Parameters ( , , ).

[0045] Training Goal: Minimize the Mean Squared Error (MSE) between predicted PID parameters and true values.

[0046] PSO Optimization Algorithm Implementation: Optimization Process: Initialize Particle Swarm (randomly generate PID parameter combinations).

[0047] Evaluate the fitness of each particle (objective function value).

[0048] Update individual optimum ( ) and global optimum ( ).

[0049] Update particle velocity and position according to the PSO formula: , Iterate to convergence, output the optimal PID parameters.

[0050] Deployment and Control Strategy: Real-time Prediction and Control: Prediction model deployment: Deploy the trained LSTM-PID model to receive sensor data (inlet material moisture content, steam flow, etc.) in real time.

[0051] Use the dynamic PID model to predict the hot air temperature trend after the control action response and adjust the steam flow in advance.

[0052] Feedback correction mechanism: Collect the actual hot air temperature value every 10 seconds and compare it with the predicted value.

[0053] If the deviation exceeds the threshold value (±0.5℃), trigger the model parameter correction: re-run the PSO algorithm to optimize the PID parameters.

[0054] Control logic optimization: Steam flow regulation: When the predicted temperature is higher than the set value, reduce the steam flow; when it is lower than the set value, increase the steam flow.

[0055] Introduce a fuzzy logic compensation module to handle nonlinear disturbances.

[0056] By combining pattern recognition and PSO optimized PID control method, the control accuracy of loose moisture machine inlet hot air temperature is significantly improved, the temperature deviation is reduced to ±0.5℃, and it has strong adaptability to complex working conditions, verifying its feasibility and economic value in the tobacco primary processing technology.

[0057] The above-described embodiments only express the specific implementation of the present application, and the description is more specific and detailed, but it should not be interpreted as a limitation on the scope of protection of the present application. It should be noted that for ordinary skilled persons in the art, without departing from the technical concept of the present application, a number of modifications and improvements can be made, which are within the scope of protection of the present application.

Claims

1. A method for predictive control of inlet hot air temperature in a loose rehumidifier, combining pattern recognition and PID control, characterized in that, Includes the following steps: Historical data from the production process of the loosening and rehumidifying machine were collected and preprocessed to construct a training dataset. The historical data included inlet hot air temperature, steam flow rate, inlet material moisture content, and instantaneous flow rate of the scale. Based on the training dataset, a pattern recognition model is used to identify multiple operating modes during the operation of the loose rehydration machine, including at least the material head, steady state, and material tail, and to extract dynamic features related to PID parameter tuning. Construct a dynamic PID parameter mapping model that takes dynamic characteristics and real-time process variables as input and outputs a set of PID parameters optimized in real time. , , ); An optimization algorithm is used to optimize the PID parameters output by the dynamic PID parameter mapping model online in real time. The optimization objective is to minimize the error between the hot air temperature setpoint and the actual value. The inlet hot air temperature of the loosening and rehumidifying machine is predicted and controlled using optimized PID parameters, and the control model is corrected based on real-time temperature feedback data.

2. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 1, characterized in that, The optimization algorithm is a particle swarm optimization (PSO) algorithm; the objective function of the optimization objective is: , in, Hot air temperature setting value This is the actual value of the hot air temperature. For steam flow rate, These are the weighting coefficients.

3. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 1, characterized in that, The pattern recognition model is a Long Short-Term Memory (LSTM) network model; the dynamic features include the instantaneous slope of the inlet hot air temperature, the peak time of the temperature change curve, and the temperature fluctuation amplitude within a time window.

4. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 3, characterized in that, The pattern recognition model is also used to identify abnormal operating conditions caused by fluctuations in steam pressure or sudden changes in material flow.

5. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 1, characterized in that, The dynamic PID parameter mapping model is a multilayer perceptron (MLP) neural network model, and its inputs also include ambient humidity and hot air fan frequency.

6. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 2, characterized in that, The objective function also introduces a constraint term related to the quality of the tobacco sheets, which is the variance of the moisture content of the exported tobacco sheets predicted based on process data.

7. The method for predictive control of inlet hot air temperature of a loose rehumidifier combined with pattern recognition PID control according to claim 1, characterized in that, The trigger condition for correcting the control model is: the deviation between the actual value of the hot air temperature collected in real time and the model prediction value continuously exceeds a preset threshold for a set time.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements a method for predictive control of the inlet hot air temperature of a loose rehumidifier that combines pattern recognition PID control as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements a method for predictive control of the inlet hot air temperature of a loose rehumidifier, which combines pattern recognition PID control as described in any one of claims 1-7.

10. A control system for a loose rehumidifier, comprising a sensor group, an actuator, and a controller, characterized in that, The controller is configured to perform a predictive control method for the inlet hot air temperature of a loose rehumidifier, which combines pattern recognition PID control as described in any one of claims 1-7, for real-time adjustment of the steam valve opening to control the inlet hot air temperature.

Images