Pellet coating system and control method based on dual PID regulation

Abstract

This invention discloses a pellet coating system based on dual PID control, relating to the field of slurry preparation. The system includes a stirred tank connected to a water inlet pipe and a limestone feed pipe, with internal stirring blades connected to a stirring motor. The water inlet pipe is equipped with an electric regulating valve, the limestone feed pipe with a mass flow meter, and a concentration sensor inside the stirred tank. The main control unit includes a main PID controller and a secondary PID controller. The concentration sensor is connected to the input of the main PID controller, the output of the main PID controller is connected to the electric regulating valve, the power signal output of the stirring motor is connected to the input of the secondary PID controller, and the output of the secondary PID controller is connected to the parameter setting interface of the main PID controller. This system is used for pellet slurry mixing control. This invention also discloses a control method for this system, dynamically adjusting the proportional coefficient or integral time parameter of the main PID controller to significantly reduce concentration regulation lag caused by limestone feed fluctuations or changes in slurry characteristics.

Description

Technical Field

[0001] This invention relates to the field of slurry preparation. More specifically, this invention relates to a pellet coating system and control method based on dual PID control. Background Technology

[0002] In the pellet production process, the pellet coating stage requires thorough mixing of the pellets with a slurry formed by mixing limestone and water to uniformly coat the surface of the pellets with a layer of slurry, thereby improving their high-temperature strength and metallurgical properties. During this process, the concentration stability and mixing uniformity of the slurry are key factors affecting the coating quality, directly related to the strength, reducibility, and other indicators of the pellets after subsequent calcination. Therefore, it is necessary to control the slurry concentration in real time during production and ensure that the limestone and water in the slurry are fully mixed to avoid slurry defects caused by concentration fluctuations or uneven mixing.

[0003] Existing pellet coating control systems often employ a single PID control strategy for slurry mixing. This strategy maintains stable slurry concentration by detecting the slurry concentration and adjusting the feed water or limestone rate accordingly. However, in actual production, slurry concentration is affected by multiple parameters, including limestone feed rate, water flow rate, and stirring intensity. Single PID control relies solely on a single concentration feedback signal for adjustment, making it difficult to respond quickly to sudden concentration changes under complex operating conditions. This results in significant adjustment lag. When limestone feed rate fluctuates or slurry characteristics change, the single PID controller needs to go through multiple stages, including sensor sampling, error calculation, and actuator action, often resulting in a considerable delay. Concentration deviations are difficult to correct quickly, and the fluctuation range is large. Furthermore, traditional stirring devices often operate at fixed power or frequency, and the stirring intensity cannot be dynamically adjusted according to changes in slurry viscosity and concentration. This can easily lead to sedimentation due to gravity, resulting in significant differences in slurry concentration in different areas of the mixing vessel, forming a large concentration gradient, which severely affects the uniformity of pellet coating.

[0004] Therefore, there is an urgent need for a control system that can achieve high-precision and rapid adjustment of slurry concentration, and can dynamically optimize the stirring effect according to the slurry state to effectively suppress sedimentation and stratification. This system would solve the problems of large concentration fluctuations and uneven mixing caused by control lag in existing technologies, and meet the stringent requirements of industrial continuous production for pellet coating quality. Summary of the Invention

[0005] One objective of this invention is to provide a pellet coating system and control method based on dual PID regulation, which can achieve high-precision and rapid adjustment of slurry concentration, while dynamically optimizing the stirring effect according to the slurry state, effectively suppressing slurry deposition and stratification.

[0006] One object of the present invention is to provide a pellet coating system based on dual PID control, comprising:

[0007] The mixing vessel is connected to a water inlet pipe and a limestone feeding pipe. Inside the mixing vessel, there are stirring blades to ensure that the pellets are fully mixed with the slurry made of water and limestone. The stirring blades are connected to the stirring motor for transmission.

[0008] A mass flow meter is installed in the limestone feed pipeline;

[0009] An electric regulating valve is installed in the water inlet pipe;

[0010] A concentration sensor is installed inside the stirred tank;

[0011] The main control unit includes a main PID controller and a secondary PID controller. The output of the concentration sensor is connected to the input of the main PID controller, the output of the main PID controller is connected to the control terminal of the electric regulating valve, the power signal output of the stirring motor is connected to the input of the secondary PID controller, and the output of the secondary PID controller is connected to the parameter setting interface of the main PID controller.

[0012] Another objective of this invention is to provide a control method for a pellet coating system based on dual PID regulation, applied to the aforementioned pellet coating system based on dual PID regulation, wherein the limestone feed rate is detected in real time by a mass flow meter. M s The unit is kg / h The concentration value of the slurry is detected in real time by a concentration sensor. N The unit is percentage, according to the formula. Calculate water distribution M v The unit is t / h ;

[0013] The main PID controller is based on the concentration setpoint. N p Compared with the real-time detected slurry concentration value N Calculate concentration deviation Calculate the output value ,in K p This is the proportionality coefficient. T i For integration time, T d For differential time, The differential switching coefficient ， ∈ {0,1}, will u(t) The preset range conversion relationship is mapped to the opening command and sent to the electric regulating valve. The electric regulating valve adjusts according to the opening command to make the inlet water flow rate approach the preset range. M v ;

[0014] The secondary PID controller collects the power signal of the stirring motor in real time according to the monitoring cycle and performs power deviation calculation, and the relative power deviation... ,in Set concentration value N p The corresponding motor reference power, For the actual measured power of the motor, when ΔP When the positive significant deviation condition is continuously met over several consecutive monitoring periods, the integral time of the main PID controller is updated. ,when ΔP When the negative significant deviation condition is continuously met over several consecutive monitoring periods, the proportional coefficient of the main PID controller is updated. The main PID controller adopts the updated version and Calculate the new output.

[0015] Preferably, the following conditions are met in at least three consecutive monitoring periods. ΔP≥0.05 or ΔP≤-0.05 When the condition is met, it is determined whether the positive significant deviation condition or the negative significant deviation condition is satisfied.

[0016] Preferably, a viscosity sensor is also provided at the bottom of the mixing tank. The signal output terminal of the viscosity sensor is connected to the input terminal of the secondary PID controller, and the measured value of the viscosity sensor is... m Viscosity value is calculated by using the power of the stirring motor. ,in k The conversion coefficient was determined based on the stirring motor power-viscosity calibration test. P 0 To determine the no-load power of the stirring motor and the concentration setpoint. N p Corresponding expected viscosity m a When viscosity difference When the viscosity exceeds the permissible error threshold:

[0017] like ≥ and m≥m a If this is the case, it is determined that the viscosity has actually increased, and the integral time of the main PID controller is updated. ;

[0018] like ≤ and mm ≤ mm a If the viscosity decreases, it is determined to be a true decrease, and the proportional coefficient of the main PID controller is updated. ;

[0019] like ＜ and m>m a or ＞ and m<m a The main PID controller is frozen and updated. The stirring motor enters pulse stirring mode. In pulse stirring mode, the power input frequency of the stirring motor alternates periodically between pulse frequency and normal frequency.

[0020] Preferably, the viscosity tolerance threshold is... d =min(0.05 m p 0.05 m a ).

[0021] Preferably, several turbidity sensors and several pressure sensors are installed inside the stirred tank. The turbidity sensors record the turbidity deviation ΔNTU between each other, and the pressure sensors record the pressure gradient variance. After the main PID controller is frozen and updated, the measured parameters collected in real time by various sensors are input into the pre-trained pulse strategy prediction model. The measured parameters include the relative power deviation. ΔP Viscosity differences Pressure gradient variance The turbidity deviation ΔNTU, pellet volume ratio parameters, temperature inside the stirred tank, and preset pulse frequency are obtained from the turbidity sensor. The pulse strategy prediction model outputs the pulse mode duration. t i The stirring motor uses the shortest pulse mode duration. t i The pulse stirring mode is executed at the corresponding preset pulse frequency, where the preset pulse frequency is a discrete candidate frequency set in the main control unit;

[0022] Duration of pulse mode completion t i Then, if the following conditions are met simultaneously:

[0023] (a) The actual viscosity increases or the actual viscosity decreases;

[0024] (b) Pressure gradient variance ≤ d g ;

[0025] (c) Turbidity deviation ΔNTU≤ d NTU ;

[0026] Then the system resumes updating the main PID controller parameters, where d g , d NTU These are the pressure gradient variance threshold and the turbidity deviation threshold.

[0027] Preferably, n pressure sensors are arranged on the vertical axis of the side wall of the stirred tank, with all pressure sensors evenly spaced along the effective liquid level height direction. The measured pressure values ​​of each pressure sensor at a selected time point are acquired through a synchronous data acquisition unit, and the measured pressure difference between adjacent pressure sensors is calculated to obtain a pressure difference sequence. and mean pressure gradient Pressure gradient variance Two turbidity sensors were placed at the bottom of the stirred tank and at the effective liquid level, and the turbidity deviation ΔNTU between the two turbidity sensors was recorded.

[0028] Preferably, the pulse strategy prediction model is a BP neural network model including an input layer, a hidden layer, and an output layer. The input layer of the BP neural network model includes multiple input neuron nodes, and the hidden layer has multiple hidden neuron nodes. Each input neuron node is connected to each of the hidden neuron nodes, and the hidden neuron nodes are all connected to the output layer. The relative power deviation, viscosity difference, pressure gradient variance, turbidity deviation, temperature inside the stirred tank, and preset pulse frequency are used as input values ​​for the input neuron nodes, and the output result is the pulse mode duration.

[0029] Preferably, the training of the BP neural network model includes the following steps:

[0030] S1. Limestone-water slurry with different solid contents is filled into the mixing tank in layers, and pellets are added according to the target ratio. The mixing motor is started and runs at the normal mixing frequency for a preset time, and the measured parameters of each sensor are recorded synchronously.

[0031] S2. Switch the stirring motor to the preset pulse frequency and continue stirring until the slurry stratification in the mixing vessel disappears. Record the duration of the pulse mode corresponding to this process.

[0032] S3. Repeat steps S1 to S2 to conduct experiments on various hierarchical states and different preset pulse frequency combinations to form a sample set and train the BP neural network model.

[0033] The present invention has at least the following beneficial effects:

[0034] First, the main PID controller adjusts the influent flow rate based on the concentration deviation, and the secondary PID controller analyzes the power signal of the stirring motor in real time. When the stirring motor power is detected to be continuously deviating from the reference value, the proportional coefficient or integral time parameter of the main PID controller is dynamically adjusted. This collaborative mechanism significantly reduces the concentration regulation lag caused by limestone feed fluctuations or changes in slurry characteristics, and reduces the concentration fluctuation range to within the industrial allowable threshold.

[0035] Secondly, by comparing and analyzing the measured values ​​of the viscosity sensor with the viscosity values ​​calculated from the power of the stirring motor, the actual viscosity changes can be accurately distinguished. When there is a logical conflict between the two, the system freezes the main PID controller parameters and starts the pulse stirring mode. By periodically switching the frequency, the slurry deposition chain is broken, effectively suppressing the stratification and clumping phenomena in the stirring tank and ensuring the uniformity of the slurry.

[0036] Third, in pulse stirring mode, the system inputs multi-source sensor data such as power deviation, viscosity difference, pressure gradient variance, turbidity deviation and temperature into the pre-trained pulse strategy prediction model, and outputs the optimal pulse frequency and pulse mode duration. The pulse strategy prediction model is generated based on a large amount of experimental data of stratified working conditions, and can adaptively match the current slurry state, eliminate stratification in the shortest time and restore the PID controller, reducing the production interruption time.

[0037] Other advantages, objectives and features of the present invention will become apparent in part from the following description, and in part from those skilled in the art through study and practice of the invention. Attached Figure Description

[0038] Figure 1 This is a schematic diagram of the overall system in one technical solution of the present invention. Detailed Implementation

[0039] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can implement it based on the description.

[0040] It should be understood that terms such as “having,” “comprising,” and “including” as used herein do not exclude the presence or addition of one or more other elements or combinations thereof.

[0041] It should be noted that, unless otherwise specified, the experimental methods described in the following embodiments are conventional methods, and the reagents and materials mentioned are commercially available. In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "setting" should be interpreted broadly. For example, they can refer to fixed connection or setting, detachable connection or setting, or integral connection or setting. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The terms "lateral," "longitudinal," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing this invention and simplifying the description. They do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of this invention.

[0042] like Figure 1 As shown, the present invention provides a pellet coating system based on dual PID control, comprising:

[0043] A mixing vessel is connected to a water inlet pipe and a limestone feed pipe. Inside the mixing vessel are stirring blades to ensure thorough mixing of the pellets with the slurry made from water and limestone. The stirring blades are connected to a stirring motor. A mass flow meter is installed in the limestone feed pipe. An electric regulating valve is installed in the water inlet pipe. A concentration sensor is installed inside the mixing vessel. A main control unit includes a main PID controller and a secondary PID controller. The output of the concentration sensor is connected to the input of the main PID controller, the output of the main PID controller is connected to the control terminal of the electric regulating valve, the power signal output of the stirring motor is connected to the input of the secondary PID controller, and the output of the secondary PID controller is connected to the parameter setting interface of the main PID controller.

[0044] In this technical solution, during system operation, limestone enters the mixing tank through the limestone feed pipeline. The mass flow meter detects and outputs the limestone feed rate signal to the main control unit in real time. Simultaneously, water enters the mixing tank through the water inlet pipeline. The electric regulating valve adjusts its opening according to the received control signal to control the water flow rate. After the stirring motor starts, it drives the stirring blades inside the mixing tank to rotate through the transmission connection, so that the water and limestone entering the tank are fully mixed to form a slurry. The concentration sensor detects the slurry concentration value in the mixing tank in real time and transmits the detected concentration signal to the input of the main PID controller. The main PID controller compares the slurry concentration value with the preset concentration setpoint, calculates the deviation between the two, and then processes the deviation according to the proportional-integral-derivative control algorithm. The system generates a corresponding control signal, which is transmitted to the control terminal of the electric regulating valve. The electric regulating valve adjusts its opening according to the control signal, so that the influent flow rate approaches the target value that meets the slurry concentration requirements, thereby realizing the influent flow rate regulation based on concentration deviation. During operation, the power signal of the stirring motor is continuously output to the input terminal of the secondary PID controller. After receiving the power signal, the secondary PID controller compares and analyzes it with the preset reference power. If a continuous deviation is found in the power signal, a parameter adjustment signal is generated and input through the parameter setting interface of the main PID controller to realize the dynamic adjustment of key parameters such as the proportional coefficient and integral time of the main PID controller, so that the regulation performance of the main PID controller can adapt to the changes in the slurry state.

[0045] In another technical solution, for the control method of the pellet coating system based on dual PID regulation mentioned above, the limestone feed rate is detected in real time by a mass flow meter. M s The unit is kg / h The concentration value of the slurry is detected in real time by a concentration sensor. N The unit is percentage, according to the formula. Calculate water distribution M v The unit is t / h, Specifically, based on the amount of limestone fed. M s With real-time slurry concentration value N The proportion determines the water distribution volume M v This corresponds to the required influent mass flow rate per unit time, and the electric regulating valve adjusts its opening accordingly.

[0046] The main PID controller is based on the concentration setpoint. N p Compared with the real-time detected slurry concentration value N Calculate concentration deviation Calculate the output value ,in Kp This is the proportionality coefficient. T i For integration time, T d For differential time, The differential switching coefficient ， ∈ {0,1}, will u(t) The preset range conversion relationship is mapped to the opening command and sent to the electric regulating valve. The electric regulating valve adjusts according to the opening command to make the inlet water flow rate approach the preset range. M v Specifically, These are the differential term switching coefficients, used when the differential term is enabled. =1， The differential term makes predictive adjustments based on the rate of change of concentration deviation, which can suppress potential overshoot. However, in this system, the slurry concentration signal is easily affected by stirring turbulence and uneven limestone particle distribution, resulting in high-frequency random fluctuations. Furthermore, there is a process delay from influent adjustment to concentration feedback, which may cause the differential action to over-lead and induce oscillations. Therefore, the differential term switching coefficient... It can take the value 0;

[0047] proportionality coefficient K p The adjustment effect used to amplify the current deviation, integral time. T i The weights that determine the cumulative historical deviation are used by the main PID controller to dynamically output based on the real-time status of the concentration deviation. u (t) It converts the preset range into an opening command for the electric regulating valve, ultimately realizing the relationship between the inlet water flow and the distribution water volume. M v The precise approximation allows for stable control of the slurry concentration.

[0048] The secondary PID controller collects the power signal of the stirring motor in real time according to the monitoring cycle and performs power deviation calculation, and the relative power deviation... ,in Set concentration value N p The corresponding motor reference power, For the actual measured power of the motor, when ΔP When the positive significant deviation condition is continuously met over several consecutive monitoring periods, the integral time of the main PID controller is updated. ,when ΔP When the negative significant deviation condition is continuously met over several consecutive monitoring periods, the proportional coefficient of the main PID controller is updated. The main PID controller adopts the updated version and Calculate the new output;

[0049] Specifically, the monitoring cycle of the secondary PID controller is a fixed time interval preset by the system for collecting power signals from the stirring motor and data from various sensors, with the motor's reference power... During the system commissioning phase, the slurry concentration can be stably controlled at the set value. N p Under ideal steady-state operating conditions, the power value of the stirring motor is recorded as... The positive and negative significant deviation conditions indicate that the stirring motor is in a continuous high-power or low-power state, respectively. Under these conditions, the slurry viscosity concentration increases or decreases, respectively. When the positive significant deviation condition is met, the shortened integral time is used. It can enhance the cumulative response speed of the main PID controller to historical deviations, accelerate influent compensation, suppress the concentration increase trend, and use an increased proportional coefficient when under conditions of significant negative deviation. It can improve the sensitivity of the main PID controller to the current concentration deviation, quickly reduce the influent flow rate, and prevent the slurry concentration from decreasing further. Deviation in the power of the stirring motor in a single monitoring cycle may be caused by the operating noise of the stirring motor, power grid voltage fluctuations, instantaneous slurry turbulence, etc. However, this technical solution uses multiple consecutive monitoring cycles for judgment, which can filter out occasional interference.

[0050] In another technical solution, the following conditions are met in at least three consecutive monitoring periods: ΔP≥0.05 or ΔP≤- 0.05 When the system determines whether the positive or negative significant deviation conditions are met, it can effectively filter the instantaneous fluctuations in the power of the stirring motor, ensuring that the main PID controller parameter adjustment is triggered only when the power deviation is continuous and reaches a significant level. This reduces erroneous adjustments caused by accidental factors and improves the accuracy of parameter updates.

[0051] In another technical solution, a viscosity sensor is also installed at the bottom of the stirring tank. The signal output terminal of the viscosity sensor is connected to the input terminal of the secondary PID controller, and the measured value of the viscosity sensor is... m Viscosity value is calculated by using the power of the stirring motor. ,in k The conversion coefficient was determined based on the stirring motor power-viscosity calibration test. P 0 To determine the no-load power of the stirring motor and the concentration setpoint. N p Corresponding expected viscosity m a When viscosity difference When the viscosity exceeds the permissible error threshold:

[0052] like ≥ and m≥m a If this is the case, it is determined that the viscosity has actually increased, and the integral time of the main PID controller is updated. ;

[0053] like ≤ and mm ≤ mm a If the viscosity decreases, it is determined to be a true decrease, and the proportional coefficient of the main PID controller is updated. ;

[0054] like ＜ and m>m a or ＞ and m<m a The main PID controller is frozen and updated. The stirring motor enters pulse stirring mode. In pulse stirring mode, the power input frequency of the stirring motor alternates periodically between pulse frequency and normal frequency.

[0055] In this technical solution, the conversion coefficient k Based on the power-viscosity calibration test of the agitator motor, specifically, during the system commissioning phase, under fixed agitator equipment, speed, and temperature conditions, agitation experiments were conducted sequentially with various fluids of known viscosity. The agitator shaft power at which each fluid reached a steady state was precisely measured. Subsequently, the measured agitator motor power data and corresponding viscosity values ​​were plotted as a scatter plot. The relationship curve between agitator motor power and viscosity was fitted by linear regression, and the conversion coefficient was obtained. k Optionally, in some operating conditions where the slurry concentration is high and exhibits non-Newtonian fluid characteristics, calibration needs to be based on the concept of apparent viscosity and associated with a specific shear rate. This can be achieved by selecting a stirring blade speed during the system commissioning phase, using various standard non-Newtonian fluids with known rheological properties, and measuring their apparent viscosity at the shear rate corresponding to that stirring blade speed under isothermal conditions. Simultaneously, the stable stirring motor power is accurately measured, and a scatter plot of stirring motor power versus apparent viscosity is plotted and linearly fitted to obtain the relationship curve between stirring motor power and viscosity at that speed, and the conversion coefficient is then obtained. k .

[0056] When the slurry concentration is near the setpoint, but localized uneven concentration occurs due to sedimentation or agglomeration, the resistance experienced by the stirring blades varies in different areas. The measured motor power indirectly reflects the average resistance and cannot distinguish between localized anomalies such as agglomeration or sedimentation within the mixing vessel. The measured motor power will increase due to increased local resistance, but the actual overall slurry viscosity may not reach the expected concentration. If the parameters of the main PID controller are adjusted solely based on power, it may be misinterpreted as an increase in actual viscosity, leading to an incorrect enhancement of the integral action and over-adjustment of concentration. Similarly, if the motor power decreases due to excessive fluidity in some areas, it may be misinterpreted as a decrease in viscosity, leading to an enhancement of the proportional action and disrupting the normal adjustment rhythm. Therefore, directly detecting the slurry viscosity using a viscosity sensor, combined with the measured motor power, allows for more accurate identification of the actual operating conditions. If the trends of the measured motor power and viscosity sensor data are consistent, it indicates a true change in overall viscosity. If the trends contradict each other, it indicates a localized anomaly requiring special handling. This avoids misjudgments based solely on power feedback, ensures the accuracy of the main PID controller parameter adjustment, and guarantees precise control of the slurry concentration.

[0057] In this technical solution, when the measured power of the stirring motor contradicts the measured value of the viscosity sensor, it indicates that the slurry may have problems such as local agglomeration, stratification, or sedimentation. Continuing to rely on PID parameter adjustment at this point may lead to misjudgment and further deviation of the slurry from the set concentration. To avoid the system falling into an adjustment dead loop due to erroneous data, it is necessary to immediately freeze the main PID controller update and adopt appropriate measures to resolve the issue. In pulse stirring mode, by periodically alternating the power input frequency of the stirring motor between pulse frequency and normal frequency, local agglomeration and stratification can be effectively broken up. The strong convection generated by stirring at the pulse frequency can tear apart agglomerates, sediments, and mixed stratified areas, while the lower normal frequency maintains a basic mixing state to avoid excessive energy consumption and quickly restores the homogeneity of the slurry. Once the trends between the measured value of the viscosity sensor and the measured power of the motor are consistent and the slurry state is stable, the pulse stirring mode is exited and the main PID controller parameter update is resumed. This ensures the stability of the system under complex operating conditions and reduces pellet coating quality problems caused by slurry abnormalities.

[0058] In another technical solution, the viscosity tolerance threshold d =min(0.05 m p 0.05 m a The viscosity value is calculated by back-calculating the power of the stirring motor. m p With expected viscosity m aTaking the smaller of 5% of the two values ​​as the viscosity tolerance threshold can take into account both the overall trend and the local state, avoiding the accidental triggering of adjustment due to small fluctuations, and ensuring timely response when there is a significant deviation, thereby improving the accuracy and stability of the system adjustment.

[0059] In another technical solution, several turbidity sensors and several pressure sensors are installed inside the stirred tank. The turbidity sensors record the turbidity deviation ΔNTU between them, and the pressure sensors record the pressure gradient variance. After the main PID controller is frozen and updated, the measured parameters collected in real time by various sensors are input into the pre-trained pulse strategy prediction model. The measured parameters include the relative power deviation. ΔP Viscosity differences Pressure gradient variance The turbidity deviation ΔNTU, pellet volume ratio parameters, temperature inside the stirred tank, and preset pulse frequency are obtained from the turbidity sensor. The pulse strategy prediction model outputs the pulse mode duration. t i The stirring motor uses the shortest pulse mode duration. t i The pulse stirring mode is executed at a corresponding preset pulse frequency, where the preset pulse frequency is a discrete candidate frequency set in the main control unit; the pulse mode duration is completed. t i Then, if simultaneously satisfying: true viscosity increases or decreases, and pressure gradient variance... ≤ d g Turbidity deviation ΔNTU≤ d NTU Then the system resumes updating the main PID controller parameters, where d g , d NTU These are the pressure gradient variance threshold and the turbidity deviation threshold. d g , d NTU It is determined artificially through experiments.

[0060] Specifically, relying solely on conventional high-frequency interval pulses is not suitable for all slurry conditions. For example, for slight stratification of low-viscosity slurries, high-frequency stirring may cause excessive shearing, leading to particle breakage and exacerbating local concentration fluctuations, while also increasing unnecessary energy consumption. For hard clumps of high-viscosity slurries, relatively low-frequency stirring is insufficient to provide enough shear force to effectively break up the clumps. Therefore, for special cases of slurries in the stirred tank, a pulse strategy prediction model is needed to determine the pulse frequency.

[0061] In this technical solution, the preset pulse frequency is a discrete candidate frequency set in the main control unit, such as 30Hz, 35Hz, ... 50Hz. In the pulse stirring mode, the power input frequency of the stirring motor alternates periodically between the pulse frequency and the normal frequency. The duration of both the pulse period and the normal period is a fixed duration set manually to reduce the dimensionality of variables. When the main control unit processes multiple sensor data such as turbidity, pressure, and temperature and runs a neural network model at the same time, it avoids the additional computational load caused by the dynamic adjustment of time parameters. At the same time, in the pulse stirring mode, the fixed design of the pulse period and the normal period also improves the working stability of the stirring motor. The mechanical stress and heat loss of the stirring motor are controllable during the cycle.

[0062] In this technical solution, considering that the content of pellets in the stirred tank also affects the sedimentation and stratification of the slurry, a pellet volume ratio parameter is introduced to quantify this effect. The selection principle is based on the effective pellet content in the slurry. For example, the parameter calculation formula can be set as "pellet delivery volume per unit time / total slurry volume in the stirred tank," where the unit time is determined according to the actual rhythm of the production line. The pellet volume ratio parameter itself does not directly correspond to the specific proportion or concentration of pellets in the slurry; its value only serves as a characteristic quantity representing the intensity of pellet delivery in neural network calculations. The pellet delivery volume can be measured by setting a mass sensor on the pellet conveyor belt to record the delivery mass of the pellets, and then calculating it in conjunction with the pellet density.

[0063] In this technical solution, two turbidity sensors can be installed inside the stirred tank, one at the top near the liquid surface and the other at the bottom near the stirring blades, to detect the turbidity of the slurry at different depths. The difference between the two, ΔNTU, can reflect the mixing uniformity of the slurry. Pressure sensors are distributed at different depths on the inner wall of the stirred tank. By calculating the pressure values ​​at different locations, the pressure gradient variance is obtained to quantify the spatial distribution uniformity of the slurry density. The larger the gradient variance, the more likely it is that there is local agglomeration or deposition in the slurry. Temperature is measured using existing temperature sensors. These parameters together serve as input values ​​for the pulse strategy prediction model, which can comprehensively reflect the dynamic state of the slurry and provide multi-dimensional basis for the pulse strategy prediction model.

[0064] The pulse strategy prediction model can employ a BP neural network model, which can efficiently handle the nonlinear coupling relationship between multiple parameters. The slurry state is affected by multiple factors such as power, viscosity, temperature, turbidity, and pressure. Traditional rule-based control cannot exhaust all operating conditions, while neural networks can capture the implicit correlation between parameters through training. When various parameters and preset pulse frequencies are input into the pulse strategy prediction model, the output is the pulse mode duration corresponding to this preset pulse frequency. The pulse mode duration with the shortest duration is selected to execute the pulse stirring mode, ensuring the stirring effect while minimizing the slurry state recovery time and reducing interference with the normal production rhythm. Compared with traditional rule-based control, this method can avoid the inefficiency caused by empirical frequency selection and avoid the risk of misjudgment of a single parameter through parameter coupling analysis. This allows pulse stirring to maintain high efficiency and specificity under complex operating conditions, significantly improving the system's ability to respond quickly to abnormal slurry states.

[0065] In another technical solution, n pressure sensors are arranged on the vertical axis of the side wall of the stirred tank. All pressure sensors are evenly spaced along the effective liquid level height direction. The measured pressure values ​​of each pressure sensor at a selected time point are acquired through a synchronous data acquisition unit, and the measured pressure difference between adjacent pressure sensors is calculated to obtain a pressure difference sequence. and mean pressure gradient Pressure gradient variance Two turbidity sensors are arranged at the bottom of the mixing vessel and at the effective liquid level. The turbidity deviation ΔNTU between the two turbidity sensors is recorded. Specifically, by using pressure sensors that are equally spaced in the vertical direction, the pressure changes of the slurry at different depths can be accurately captured. The pressure gradient variance can intuitively reflect the spatial uniformity of the slurry density distribution.

[0066] In another technical solution, the pulse strategy prediction model employs a BP neural network model comprising an input layer, a hidden layer, and an output layer. The input layer of the BP neural network model includes multiple input neurons, and the hidden layer has multiple hidden neurons. Each input neuron is connected to all the hidden neurons, and all hidden neurons are connected to the output layer. The relative power deviation, pellet volume ratio parameter, viscosity difference, pressure gradient variance, turbidity deviation, temperature inside the stirred tank, and preset pulse frequency are used as input values ​​for the input neurons. The output is the pulse mode duration. Specifically, when establishing the training sample set for the BP neural network model, it is necessary to formally add samples to the stirred tank... Before production, targeted testing and multi-condition data collection were conducted. For slurries of different concentrations and viscosities, various conditions such as normal mixing, slight stratification, and severe agglomeration were artificially created. The relative power deviation, pellet volume ratio parameters, viscosity differences, pressure gradient variance, turbidity deviation, temperature, and corresponding preset pulse frequencies were recorded. At the same time, the actual time required from starting pulse stirring to the slurry returning to normal state under each preset pulse frequency in each condition was marked, forming a "parameter-pulse time" mapping pair. These data were divided into training and testing sets in a ratio of 7:3 or 8:2. During training, the hidden layer weight parameters of the BP neural network model can be continuously adjusted according to the PSO algorithm to optimize the model, so that the pulse time predicted by the model gradually approaches the actual value.

[0067] For example, optimizing a BP neural network model using the PSO algorithm includes the following steps:

[0068] A1. The factors influencing the output layer are relative power deviation, pellet volume ratio parameter, viscosity difference, pressure gradient variance, turbidity deviation, temperature, and the corresponding preset pulse frequency. The total number of influencing factors is used as the number of neuron nodes *m*, and the pulse mode duration is used as the unique output value, i.e., the number of nodes in the output layer. c =1, number of hidden layer nodes c 1 for , a A random constant between 1 and 10;

[0069] A2. The mathematical expression for normalizing the sample set data is:

[0070]

[0071] in, x i Sample data representing influencing factors. x min , x max These are the minimum and maximum values ​​in the sample data, respectively. The data on influencing factors are after dimensionless processing;

[0072] A3. Initialize the mapping relationship between the population particles and the weights and thresholds of the BP neural network, including particle dimension, initial velocity, population size, learning factor, and inertia weight;

[0073] A4. Input the normalized input and output variables into the BP neural network model, calculate the particle's fitness function value, and obtain the particle's historical best fitness and global fitness. The particle's fitness function value is the mean square error of the calculation result, and its function expression is:

[0074]

[0075] in, Let be the predicted value of the i-th sample. y i Let be the true value of the i-th sample, and n be the total number of calculation results of the neural network;

[0076] A5. Iteratively calculate the particle fitness, update the historical best fitness and the global fitness according to the preset update conditions, until the preset iteration end condition is met;

[0077] A6. Update the weights and thresholds of the preset BP neural network model to obtain the optimized BP neural network model for calculating the current impulse strategy prediction model.

[0078] In another technical solution, the training of the BP neural network model includes the following steps:

[0079] S1. Limestone-water slurry with different solid contents is layered and filled into the mixing vessel, and pellets are added according to the target ratio. After the stirring motor is turned on and runs at the normal stirring frequency for a preset time, the measured parameters of each sensor are recorded simultaneously. For example, limestone-water slurry with different solid contents is first prepared outside the mixing vessel, such as two slurries with solid contents of 20% and 35%. The concentration after the two slurries are completely mixed is set as the concentration set value. The layered filling method is used to simulate the natural stratification state in actual production. At this time, the upper layer of slurry has a lower density and the lower layer has a higher density, forming obvious stratification. Then the stirring is started. The motor runs at a preset regular stirring frequency for a preset time, and then pellets are added according to the target pellet volume ratio parameters to ensure that the pellets are evenly dispersed in the stratified slurry. The preset time needs to be controlled within a short range to avoid over-stirring and eliminating the stratification phenomenon. The characteristics of the initial stratification state are retained in the stirring vessel. During this process, the measured parameters of each sensor are recorded in real time through the synchronous data acquisition unit: including the pressure values ​​of several pressure sensors on the vertical axis in the stirring vessel, the turbidity deviation ΔNTU of the turbidity sensor at the bottom and the liquid level, the relative deviation of the stirring motor power, the slurry temperature, and the pellet volume ratio parameters.

[0080] S2. Switch the stirring motor to the preset pulse frequency and continue stirring until the slurry stratification in the mixing vessel disappears. Record the pulse mode duration corresponding to this process. For example, after the conventional stirring in step S1, keep the slurry stratification unchanged. Since the mixing vessel is in a clear stratification state at this time, the overall characteristics of the slurry will show contradictory detection results. Switch the stirring motor to one of the preset discrete candidate pulse frequencies and run the pulse stirring mode. At the same time, continuously monitor the slurry state. Determine whether the stratification in the mixing vessel has completely disappeared by the pressure gradient variance and turbidity deviation ΔNTU. For example, when the pressure gradient variance is 0, the liquid layers in the mixing vessel are in a uniform state. Record the pulse mode duration at the corresponding time.

[0081] S3. Repeat steps S1 to S2 to conduct experiments on various stratification states, various pellet volume ratio parameters and different preset pulse frequency combinations to form a sample set and train the BP neural network model. For example, by adjusting key variables, diverse samples can be generated: for example, by changing the stratification state, pellet volume ratio parameters and preset pulse frequency. Each combination is repeated 3 to 5 times to reduce random errors, and finally a sample set containing thousands of data sets is formed.

[0082] The sample set was divided into a training set and a test set in an 8:2 ratio. The training set data was used to input the BP neural network model, and the hidden layer weights were iteratively adjusted so that the duration predicted by the model gradually approached the actual pulse pattern duration. After training, the generalization ability of the model was verified using the test set, and finally a pulse strategy prediction model that can be directly deployed was formed.

[0083] It should be noted that although the steps are described in a specific order above, this does not mean that they must be performed in that order. In fact, some of these steps can be executed concurrently, or even in a different order, as long as the required functionality is achieved. The number of devices and processing scale described herein are for simplification of the invention; applications, modifications, and variations of this invention will be readily apparent to those skilled in the art.

[0084] Although embodiments of the present invention have been disclosed above, they are not limited to the applications listed in the specification and embodiments. They can be applied to various fields suitable for the present invention. For those skilled in the art, other modifications can be easily made. Therefore, without departing from the general concept defined by the claims and their equivalents, the present invention is not limited to the specific details and illustrations shown and described herein.

Claims

1. A control method for a pellet coating system based on dual PID regulation, characterized in that, The dual PID-controlled pellet coating system includes: The mixing vessel is connected to a water inlet pipe and a limestone feeding pipe. Inside the mixing vessel, there are stirring blades to ensure that the pellets are fully mixed with the slurry made of water and limestone. The stirring blades are connected to the stirring motor for transmission. A mass flow meter is installed in the limestone feed pipeline; An electric regulating valve is installed in the water inlet pipe; A concentration sensor is installed inside the stirred tank; The main control unit includes a main PID controller and a secondary PID controller. The output of the concentration sensor is connected to the input of the main PID controller, the output of the main PID controller is connected to the control terminal of the electric regulating valve, the power signal output of the stirring motor is connected to the input of the secondary PID controller, and the output of the secondary PID controller is connected to the parameter setting interface of the main PID controller. The limestone feed rate is monitored in real time using a mass flow meter. M s The unit is kg / h The concentration value of the slurry is detected in real time by a concentration sensor. N The unit is percentage, according to the formula. Calculate water distribution M v The unit is t / h ; The main PID controller is based on the concentration setpoint. N p Compared with the real-time detected slurry concentration value N Calculate concentration deviation Calculate the output value ,in K p This is the proportionality coefficient. T i For integration time, T d For differential time, The differential switching coefficient ， ∈ {0,1}, will u(t) The preset range conversion relationship is mapped to the opening command and sent to the electric regulating valve. The electric regulating valve adjusts according to the opening command to make the inlet water flow rate approach the preset range. M v ; The secondary PID controller collects the power signal of the stirring motor in real time according to the monitoring cycle and performs power deviation calculation, and the relative power deviation... ,in Set concentration value N p The corresponding motor reference power, For the actual measured power of the motor, when Δ P When the positive significant deviation condition is continuously met over several consecutive monitoring periods, the integral time of the main PID controller is updated. ,when ΔP When the negative significant deviation condition is continuously met over several consecutive monitoring periods, the proportional coefficient of the main PID controller is updated. The main PID controller adopts the updated version and Calculate the new output.

2. The control method for a pellet coating system based on dual PID regulation as described in claim 1, characterized in that, Meets the requirements in at least three consecutive monitoring periods. ΔP≥0.05 or ΔP≤-0.05 When the conditions are met, it is determined whether the positive significant deviation condition or the negative significant deviation condition is satisfied.

3. The control method for a pellet coating system based on dual PID regulation as described in claim 2, characterized in that, A viscosity sensor is also installed at the bottom of the mixing vessel. The signal output terminal of the viscosity sensor is connected to the input terminal of the secondary PID controller. The measured value of the viscosity sensor is... μ Viscosity value is calculated by using the power of the stirring motor. ,in k The conversion coefficient was determined based on the stirring motor power-viscosity calibration test. P 0 To determine the no-load power of the stirring motor and the concentration setpoint. N p Corresponding expected viscosity μ a When viscosity difference When the viscosity exceeds the permissible error threshold: like ≥ and μ≥μ a If this is the case, it is determined that the viscosity has actually increased, and the integral time of the main PID controller is updated. ; like ≤ and μ≤μ a If the viscosity decreases, it is determined to be a true decrease, and the proportional coefficient of the main PID controller is updated. ; like ＜ and μ > μ a or ＞ and μ < μ a The main PID controller is frozen and updated. The stirring motor enters pulse stirring mode. In pulse stirring mode, the power input frequency of the stirring motor alternates periodically between pulse frequency and normal frequency.

4. The control method for a pellet coating system based on dual PID regulation as described in claim 3, characterized in that, Viscosity tolerance threshold δ =min(0.05 μ p 0.05 μ a ) 。 5. The control method for a pellet coating system based on dual PID regulation as described in claim 3, characterized in that, Several turbidity sensors and several pressure sensors are installed inside the stirred tank. The turbidity sensors record the turbidity deviation ΔNTU between them, and the pressure sensors record the pressure gradient variance. After the main PID controller is frozen and updated, the measured parameters collected in real time by various sensors are input into the pre-trained pulse strategy prediction model. The measured parameters include the relative power deviation. ΔP Viscosity differences Pressure gradient variance The turbidity deviation ΔNTU, pellet volume ratio parameters, temperature inside the stirred tank, and preset pulse frequency are obtained from the turbidity sensor. The pulse strategy prediction model outputs the pulse mode duration. t i The stirring motor uses the shortest pulse mode duration. t i The pulse stirring mode is executed at the corresponding preset pulse frequency, where the preset pulse frequency is a discrete candidate frequency set in the main control unit; Duration of pulse mode completion t i Then, if the following conditions are met simultaneously: The actual viscosity increases or decreases; Pressure gradient variance ≤ δ g ; Turbidity deviation ΔNTU≤ δ NTU ; Then the system resumes updating the main PID controller parameters, where δ g , δ NTU These are the pressure gradient variance threshold and the turbidity deviation threshold.

6. The control method for a pellet coating system based on dual PID regulation as described in claim 5, characterized in that, n pressure sensors are arranged on the vertical axis of the side wall of the stirred tank. All pressure sensors are evenly spaced along the effective liquid level height. The measured pressure values ​​of each pressure sensor at selected time points are acquired through a synchronous data acquisition unit, and the measured pressure difference between adjacent pressure sensors is calculated to obtain the pressure difference sequence. and mean pressure gradient Pressure gradient variance ; Two turbidity sensors are placed at the bottom of the stirred tank and at the effective liquid level, and the turbidity deviation ΔNTU between the two turbidity sensors is recorded.

7. The control method for a pellet coating system based on dual PID regulation as described in claim 6, characterized in that, The pulse strategy prediction model employs a BP neural network model comprising an input layer, a hidden layer, and an output layer. The input layer of the BP neural network model includes multiple input neuron nodes, and the hidden layer contains multiple hidden neuron nodes. Each input neuron node is connected to each of the hidden neuron nodes, and all hidden neuron nodes are connected to the output layer. The relative power deviation, pellet volume ratio parameter, viscosity difference, pressure gradient variance, turbidity deviation, temperature inside the stirred tank, and preset pulse frequency are used as input values ​​for the input neuron nodes, and the output result is the pulse mode duration.

8. The control method for a pellet coating system based on dual PID regulation as described in claim 7, characterized in that, The training of the BP neural network model includes the following steps: S1. Limestone-water slurry with different solid contents is filled into the mixing tank in layers, and pellets are added according to the target ratio. The mixing motor is started and runs at the normal mixing frequency for a preset time, and the measured parameters of each sensor are recorded synchronously. S2. Switch the stirring motor to the preset pulse frequency and continue stirring until the slurry stratification in the mixing vessel disappears. Record the duration of the pulse mode corresponding to this process. S3. Repeat steps S1 to S2 to conduct experiments on various stratification states, various pellet volume ratio parameters and different preset pulse frequency combinations to form a sample set and train the BP neural network model.

Images