Automatic closed-loop control method and system for electric separation production line

Abstract

The invention relates to the technical field of automatic control, in particular to an automatic closed-loop control method and system for an electric separation production line. The method comprises the following steps: acquiring a stock bin level and equipment operation state parameters; the obtained material level of the stock bin and the equipment operation state parameters are preprocessed; according to the preprocessed stock bin material level and the preprocessed equipment operation state parameters, the equipment operation parameters are adjusted through a step-by-step feedback control strategy; and the balance of the material level of the material bin is maintained by monitoring and adjusting the operation state parameters of the equipment in real time. The blanking amount is accurately controlled, the subjectivity of manual adjustment and the inadaptability of a fixed blanking mode are eliminated, the stability and consistency of material supply are guaranteed, and stable operation of subsequent procedures is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic control, and particularly to an automatic closed-loop control method and system for an electrostatic separation production line. Background Art

[0002] In industrial production scenarios such as mineral resource processing and material separation, the electrostatic separation production line, as a key material processing equipment, undertakes the important task of effectively separating materials with different electric properties. With the continuous expansion of industrial production scale and the increasing requirements for production efficiency and product quality, the automatic control level of the electrostatic separation production line has become one of the key factors affecting the competitiveness of enterprises. Traditional electrostatic separation production lines rely on the combination of manual operation and simple electrical control. Operators need to constantly monitor the material level in the silo and manually adjust the feeding amount of the equipment. This method is unable to cope when faced with complex and changeable production conditions.

[0003] Specifically, the disadvantages of the prior art include:

[0004] Low material level control accuracy: The prior art usually only relies on simple limit switches or single-point level gauges to monitor the material level in the silo, and cannot comprehensively and accurately obtain the actual height and distribution of materials in the silo. This results in difficulty in ensuring that the materials always completely cover the heating tubes during production. If the material coverage is insufficient, the heating tubes may be damaged due to local overheating, affecting the continuity of production; at the same time, material overflow is also likely to occur, causing material waste and pollution of the production environment.

[0005] High equipment energy consumption: The traditional control method lacks effective means to optimize the operating power of the equipment. The equipment often operates under fixed working parameters and cannot adjust the power in real time according to actual production requirements. For example, when the material level in the silo is low and the feeding amount requirement is small, the equipment still operates at high power, resulting in a large amount of electric energy waste and increasing the production cost of the enterprise.

[0006] Poor regulation stability: Manual regulation or simple time-based and quantitative control methods cannot respond to various interference factors in the production process in a timely and accurate manner, such as fluctuations in material properties and subtle changes in the operating state of the equipment. This makes the adjustment of the feeding amount frequent and large in amplitude, which not only affects the service life of the equipment but also may cause unstable material supply in subsequent process links, thereby affecting the consistency of product quality.

[0007] Lack of global collaborative control: The control of each device in the existing electrostatic separation production line is relatively independent, lacking an effective information interaction and collaboration mechanism. When the material level of a certain-level silo is abnormal, it is difficult to quickly and accurately control the material level of this level by adjusting the feeding amount of the upper-level equipment, and it is impossible to optimize the operating efficiency and stability of the production line as a whole.

[0008] Therefore, there is an urgent need to propose an automatic closed-loop control method and system for an electrostatic separation production line. Summary of the Invention

[0009] To solve the above-mentioned problems, the present invention provides an automatic closed-loop control method and system for an electrostatic separation production line.

[0010] In a first aspect, an automatic closed-loop control method for an electrostatic separation production line provided by the present invention adopts the following technical solutions:

[0011] An automatic closed-loop control method for an electrostatic separation production line includes:

[0012] Obtain the silo level and equipment operation status parameters;

[0013] Preprocess the obtained silo level and equipment operation status parameters;

[0014] Adjust the equipment operation parameters through a hierarchical feedback control strategy for the preprocessed silo level and equipment operation status parameters;

[0015] Maintain the balance of the silo level by real-time monitoring and adjusting the equipment operation status parameters.

[0016] Further, the obtaining of the silo level and equipment operation status parameters includes deploying Internet of Things sensors and actuators at the equipment layer. Among them, the Internet of Things sensors include ultrasonic level gauges, infrared coverage sensors, vibration sensors, and temperature sensors, which collect the level height, material coverage status, and equipment operation parameter data in real time; the actuators include feeding motors, solenoid valves, and alarm devices, and are connected to the edge controller through the Modbus / TCP protocol.

[0017] Further, the preprocessing of the obtained silo level and equipment operation status parameters includes configuring edge computing nodes at the edge control layer to perform local data preprocessing on the obtained silo level and equipment operation status parameters. Among them, the level data of the ultrasonic sensor is processed by the extended Kalman filter to solve the state estimation problem of the non-linear silo shape; and the coverage signals of the infrared sensors are fused. When it is detected that the heating tube is not covered, the low-level warning logic of the silo level is forcibly triggered, and the effectiveness of the level height is corrected by the weighted average algorithm.

[0018] Further, for the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy includes performing self-tuning of fuzzy PID parameters based on the adaptive fuzzy PID control method. When the deviation of the current-level silo level exceeds the threshold, indirect control is achieved by adjusting the feeding amount of the upper-level equipment; based on the current deviation, deviation change rate, and historical adjustment times, dynamic adaptation to the change of material fluidity is achieved through the fuzzy rules of enhancing the differential effect when the material level suddenly changes and increasing the integral weight when there is a long-term deviation.

[0019] Furthermore, for the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy further includes using historical discharging data stored in the edge node, including time stamps, discharging motor speed, and level change rate, predicting the discharging amount trend in a future set time interval by training an LSTM neural network, and adjusting the speed of the upper-level equipment in advance according to the discharging trend to offset the material transmission lag effect.

[0020] Furthermore, for the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy further includes establishing a digital twin model of material flow in the silo based on computational fluid dynamics (CFD), with input parameters including material particle size distribution, moisture content, and discharging port size; analyzing the pulling force through the digital twin model to simulate the material flow state in the silo, identifying abnormal conditions such as bridging and segregation, and constructing a global control strategy using the deep deterministic policy gradient (DDPG) algorithm.

[0021] Furthermore, constructing a global control strategy using the deep deterministic policy gradient (DDPG) algorithm includes constructing a DDPG network model by defining the state space, action space, and reward function, including a policy model and a value model. Among them, the policy model takes the state space as input and outputs actions in the action space; the value space takes the state and action as input and outputs a value estimate; updating the value network by interacting with the environment and using the collected samples, expressed as:

[0022] Among them,

[0023] y i = R i + γQ(S i+1 , μ(S i+1 |θ μ )|θ Q ), where γ is the discount factor, used to balance the importance of the current reward and future rewards, generally taking values between 0 and 1. Update the value network parameter θ μ using the gradient descent method to minimize the loss function. Then, according to the gradient of the policy network

[0024]

[0025] , use the policy gradient ascent method to update the policy network parameter θ μ to maximize the cumulative reward.

[0026] Further, maintaining the balance of the silo level by real-time monitoring and adjusting the operating state parameters of the equipment includes local monitoring of edge nodes by displaying the level curve and equipment status through the HMI interface; remote cloud monitoring by visualizing the production line status in real time through the Web platform; retraining the LSTM prediction model and DDPG policy model regularly using production data in the cloud, and remotely updating the edge node firmware through OTA technology; the edge node runs the sensor self-calibration program daily, compares the consistency of multi-sensor data, and triggers a manual calibration work order if the limit is exceeded.

[0027] In a second aspect, an automated closed-loop control system for an electrostatic separation production line includes:

[0028] A data acquisition module configured to acquire the silo level and the operating state parameters of the equipment;

[0029] A preprocessing module configured to preprocess the acquired silo level and the operating state parameters of the equipment;

[0030] A parameter adjustment module configured to adjust the operating parameters of the equipment through a hierarchical feedback control strategy for the preprocessed silo level and the operating state parameters of the equipment;

[0031] A balance module configured to maintain the balance of the silo level by real-time monitoring and adjusting the operating state parameters of the equipment.

[0032] In a third aspect, the present invention provides a computer-readable storage medium storing multiple instructions adapted to be loaded and executed by a processor of a terminal device for the automated closed-loop control method of an electrostatic separation production line.

[0033] In a fourth aspect, the present invention provides a terminal device including a processor and a computer-readable storage medium, the processor being used to implement each instruction; the computer-readable storage medium being used to store multiple instructions adapted to be loaded and executed by the processor for the automated closed-loop control method of an electrostatic separation production line.

[0034] In summary, the present invention has the following beneficial technical effects:

[0035] 1. Based on the comprehensive analysis of the level, material characteristics, and production process parameters, the opening degree and the blanking time interval of the blanking valve are automatically adjusted through intelligent algorithms, achieving precise control of the blanking volume, eliminating the subjectivity of manual adjustment and the inadaptability of the fixed blanking method, ensuring the stability and consistency of the material supply, and being beneficial to the stable operation of subsequent processes.

[0036] 2. Automatically adjust the electric field strength of the electrostatic separator, the working parameters of the arc plate machine and the sieve plate machine, etc., according to the changes in the material properties, the operating status of the equipment, and the fluctuations in the environmental parameters monitored in real time, so that the production process can adapt to the changes in various complex working conditions, maintain stable production effects and product quality, and avoid production fluctuations caused by untimely or inaccurate manual adjustment.

[0037] 3. Through the real-time monitoring and intelligent regulation of the operating power of the equipment, automatically adjust the equipment power according to the actual production load, avoid the high-power operation of the equipment under low load, and achieve energy conservation and consumption reduction. At the same time, comprehensively monitor and analyze the energy consumption of the entire production line, which helps to formulate more scientific and reasonable energy-saving strategies and further reduce production costs.

[0038] 4. Information is shared and interacted in real time among various equipment through the industrial Internet, and equipment such as electrostatic separators, arc plate machines, and sieve plate machines can work together according to each other's operating status and production requirements. For example, when the sieve mesh of the sieve plate machine is blocked, the system will automatically adjust the working parameters of the electrostatic separator and the arc plate machine to avoid problems such as material accumulation and insufficient processing, and improve the operating efficiency and coordination of the entire production line. Description of the Drawings

[0039] Figure 1 is a schematic diagram of an automatic closed-loop control system for an electrostatic separation production line according to Embodiment 1 of the present invention.

[0040] Figure 2 is a schematic diagram of the associated control logic of an automatic closed-loop control system for an electrostatic separation production line according to Embodiment 1 of the present invention. Detailed Description of the Invention

[0041] The present invention will be further described in detail below with reference to the accompanying drawings.

[0042] Embodiment 1

[0043] Refer to Figure 1 , an automatic closed-loop control method for an electrostatic separation production line in this embodiment includes:

[0044] Obtain the bin level and equipment operating status parameters;

[0045] Preprocess the obtained bin level and equipment operating status parameters;

[0046] Adjust the equipment operating parameters through a step-by-step feedback control strategy for the preprocessed bin level and equipment operating status parameters;

[0047] Maintain the balance of the bin level by real-time monitoring and adjusting the operating status parameters of the equipment.

[0048] Specifically:

[0049] Detailed Implementation Steps of Automatic Closed - loop Control for Electrostatic Separation Production Line

[0050] I. Data Acquisition Steps

[0051] (1) Sensor Arrangement and Data Collection

[0052] Ultrasonic level gauge: Install multiple ultrasonic level gauges evenly at different height positions in the silo. Its working principle is based on the propagation characteristics of ultrasonic waves in the air. By measuring the time difference Δt between the transmitted and received ultrasonic signals, the level height h can be calculated. The calculation formula is where υ is the propagation speed of ultrasonic waves in the air (affected by temperature, generally about 340 m / s at room temperature). The level gauge collects data every 50 ms to obtain real - time and continuous level change information.

[0053] Infrared coverage sensor: Install the infrared coverage sensor on top of the heating tube. When the material covers the heating tube, the intensity of the infrared signal received by the sensor changes. By setting a suitable threshold, the material coverage state can be converted into a digital signal (0 represents uncovered, 1 represents covered). This sensor continuously monitors the material coverage situation in real - time. Once the state changes, it immediately transmits the change signal to the edge computing node.

[0054] Vibration sensor and temperature sensor: The vibration sensor is installed at the discharge port. By detecting the vibration acceleration a generated during the discharging process, it reflects the smoothness of the discharging. The voltage signal V output by the sensor has a linear relationship with the vibration acceleration α, that is, V = k×a, where k is the sensitivity coefficient of the sensor. The temperature sensor is arranged on the silo wall and uses temperature - sensitive elements such as thermistors. Its resistance value R T varies with the temperature T. By measuring the resistance value and according to the calibration curve R T = f(T), the real - time temperature inside the silo can be obtained. These two sensors also collect data at a certain frequency (such as every 100 ms) to comprehensively monitor the equipment operation parameters.

[0055] (2) Actuator Access and Control Preparation

[0056] Discharging motor: Select a discharging motor with variable - frequency control function. Its rotational speed n and the input control voltage U satisfy a certain functional relationship, expressed as n = k1×U + b, where k1 and b are the characteristic parameters of the motor. The motor is connected to the edge controller through the Modbus / TCP protocol. The edge controller can send corresponding control voltage commands to the motor according to the discharging volume adjustment requirements calculated by the control algorithm, so as to achieve precise control of the discharging volume.

[0057] Solenoid valve: The solenoid valve is installed at the material discharge port, and the material discharge flow is adjusted by controlling its opening degree α. The opening degree of the solenoid valve is related to the input control current I. Generally, an empirical model α = k2×I + c can be established, where k2 and c are the characteristic parameters of the solenoid valve. Similarly, through the Modbus / TCP protocol, the edge controller can send a control current command to the solenoid valve according to the system control requirements to achieve flexible adjustment of the material discharge flow.

[0058] Alarm device: The alarm device includes an audible and visual alarm, which is connected to the edge controller through the Modbus / TCP protocol. When the edge controller determines that the system has an abnormal situation (such as over-limit material level, equipment failure, etc.) based on the sensor data, it sends an alarm signal to the alarm device to trigger the audible and visual alarm and remind the on-site staff.

[0059] II. Data preprocessing steps

[0060] (1) Extended Kalman filter for processing material level data

[0061] State equation and observation equation: For the ultrasonic material level data, since the shape of the silo may be non-linear, the extended Kalman filter (EKF) algorithm is used for processing. Assume the material level state vector X k , and its state equation is X k = f(X k-1 , U k-1 ) + W k-1 , where f is the non-linear state transition function, U k-1 is the control input (in material level monitoring, it can be assumed to be zero because the main concern is the natural change of the material level), and W k-1 is the process noise, which follows the Gaussian distribution N(0, Q k-1 ). The observation equation is Z k = h(X k ) + V k , where h is the non-linear observation function, and V k is the observation noise, which follows the Gaussian distribution N(0, R k ).

[0062] Prediction step: At time k, first make a prediction. The predicted state estimate The predicted covariance where F k-1 is the Jacobian matrix of the state transition function f with respect to the state X at .

[0063] Update step: When new observation data Z k is received, an update is made. The Kalman gain where H k is the Jacobian matrix of the observation function h with respect to the state X at Jacobian matrix at. Updated state estimate Updated covariance P k|k =(I - K k H k )P k|k-1 , where I is the identity matrix. By continuously iterating the above prediction and update steps, the measurement error caused by the non - linear silo shape can be effectively corrected, and a more accurate material level estimate can be obtained.

[0064] (II) Fusion of infrared signal and weighted average correction

[0065] Infrared signal fusion: When the infrared sensor detects that the heating tube is not covered by the material, i.e., the coverage signal S = 0, the low - level warning logic of the material level is immediately triggered. At this time, the material level height h is forced to be set to the safety lower limit height h min close to the top of the heating tube to ensure that the material always covers the heating tube.

[0066] Weighted average correction: Under normal circumstances, the coverage signal S (taking values of 0 or 1) of the infrared sensor is weighted - averaged with the material level height \(h_{EKF}\) after being processed by h EKF . The corrected material level height h final = w1×h EKF + w2×S× hmi n, where w1 and w2 are weighting coefficients, and w1 + w w2 = 1. Generally, when S = 1 (the material covers the heating tube), w1 takes a larger value, such as w1 = 0.9 and w2 = 0.1; when S = 0, the weight corresponding to w2 increases to highlight the importance of the lower limit of the material level. In this way, the reliability and effectiveness of the material level data are ensured.

[0067] III. Parameter adjustment steps

[0068] (I) Adaptive fuzzy PID control

[0069] Inputs and outputs of the fuzzy controller: The inputs of the fuzzy controller are the current material level deviation e k = h set - h final,k , where h set is the preset material level set value, and h final，k is the material level height after pre - processing at time k; the deviation change rate where Δt is the data acquisition time interval; and the historical adjustment times N. The outputs are the proportional coefficient K p，k , integral coefficient K i,k and derivative coefficient K d,k .

[0070] Fuzzy subset division: The material level deviation \(e\) is divided into seven fuzzy subsets: "Negative Big (NB)", "Negative Medium (NM)", "Negative Small (NS)", "Zero (ZE)", "Positive Small (PS)", "Positive Medium (PM)", and "Positive Big (PB)"; the deviation change rate is also divided in a similar way. For the output PID parameters, they are also divided into different fuzzy subsets. For example, \(K_p\) is divided into "Very Small (VS)", "Small (S)", "Medium (M)", "Large (L)", "Very Large (VL)", etc.

[0071] Fuzzy control rules: According to expert experience and actual operating conditions, fuzzy control rules are formulated. For example, when the material level deviation \(e\) is "Positive Big (PB)" and the deviation change rate is "Positive Small (PS)", in order to quickly reduce the material level, the proportional coefficient \(K\) p should be increased. It can be set that \(K\) p is "Very Large (VL)"; when the material level deviation \(e\) is "Zero (ZE)" and the deviation change rate is "Zero (ZE)", the current PID parameters are kept unchanged, that is, \(K\) p is "Medium (M)", \(K\) i is "Medium (M)", \(K\) d is "Medium (M)". Another example is that when there is a long-term deviation (which can be judged by the number of historical adjustment times \(N\)), the integral coefficient \(K\) i is increased to eliminate the steady-state error; when there is a sudden change in the material level (the deviation change rate is large), the derivative coefficient \(K\) d is enhanced to quickly suppress the fluctuation of the material level.

[0072] Fuzzy inference and defuzzification: The Mamdani inference method is adopted. According to the input fuzzy values and the formulated fuzzy control rules, fuzzy inference is carried out to obtain the output fuzzy values. Finally, defuzzification is performed by the centroid method to calculate the specific values of \(W\) p，k , \(K\) i，k and \(K\) d，k . For example, for the proportional coefficient \(K\) p , its defuzzification calculation formula is where \(\mu_K\) p (\(v\) i ) is the membership degree of the fuzzy set \(K\) p at the element \(v\) i , and \(n\) is the number of elements in the fuzzy set.

[0073] Adjustment of the feeding quantity: According to the calculated PID parameters, the speed adjustment amount of the upper-level equipment (such as the feeding motor) is calculated. Assuming that the PID control algorithm is used to calculate the motor speed adjustment amount \(\Delta n\), its formula is

[0074]

[0075] Then, the rotational speed adjustment instruction is sent to the motor controller of the upper-level device through the Modbus / TCP protocol to achieve precise control of the material discharge volume.

[0076] (2) LSTM Prediction of Material Discharge Volume

[0077] Data Preparation: Utilize the historical material discharge data stored in the edge node, including the timestamp \(t\), the rotational speed \(n\) of the material discharge motor, and the material level change rate Normalize this data. For example, for the rotational speed \(n\) of the material discharge motor, the normalization formula is where \(n\) min and \(n\) max are respectively the minimum and maximum values in the historical rotational speed data. Similar normalization processing is also performed on the material level change rate.

[0078] LSTM Model Construction: Construct an LSTM neural network model, which includes an input layer, multiple LSTM hidden layers, and an output layer. The input layer inputs the normalized historical data in a time series, and the input vector at each time step is The LSTM hidden layer processes the time series data through the gating mechanism (input gate, forget gate, and output gate) to learn the long-term dependence relationships in the data. Assume the output of the hidden layer is \(H_t\), and its calculation process is relatively complex, involving operations of multiple weight matrices and bias vectors. Taking the input gate as an example, its calculation formula is \(i\) t \(=\sigma(W\) ii \times X\) t \(+W\) hi \times H\) t-1 \(+b\) i ), where \(\sigma\) is the sigmoid activation function, \(W\) ii and \(W\) hi are weight matrices, \(b\) i is the bias vector, and \(X\) t is the input vector at time \(t\). The output layer predicts the trend of the material discharge volume in a future set time interval (such as the next 10 minutes) based on the output of the hidden layer.

[0079] Model Training: Train the LSTM model with a large amount of historical data, set appropriate loss functions (such as the mean squared error loss function where \(y\) i is the true value, is the predicted value, and \(N\) is the number of samples) and optimization algorithms (such as the Adam optimization algorithm), and continuously adjust the weights and biases of the model to enable the model to accurately predict the trend of the material discharge volume.

[0080] Stocking Quantity Prediction and Adjustment: After the model training is completed, predict the stocking quantity trend for the next 10 minutes based on the current historical data. For example, predict the rotational speed of the stocking motor at a certain future moment. Adjust the rotational speed of the upper-level equipment in advance according to the prediction result. If it is predicted that the material level in the current bin will exceed the normal range, such as when it is predicted that the material level will be too high, reduce the stocking speed of the upper-level equipment 5 minutes in advance, that is, according to the difference between the predicted rotational speed and the current rotational speed. Send the rotational speed adjustment instruction to the motor controller of the upper-level equipment through the Modbus / TCP protocol to achieve the advance adjustment of the stocking quantity and offset the material transfer lag effect.

[0081] (III) Digital Twin and DDPG Algorithm

[0082] Digital Twin Model Construction: Based on the principle of computational fluid dynamics (CFD), establish a digital twin model of the material flow in the bin. In the model, consider the continuity equation of the material where ρ is the material density, is the material flow velocity vector; the momentum equation where p is the pressure, τ is the viscous stress tensor, is the gravitational acceleration vector. Input parameters such as the material particle size distribution (represented by the particle size distribution function), moisture content (affecting material properties such as viscosity), and the size of the discharge port, and use numerical methods such as the finite element method to discretely solve the above equations to simulate the material flow state in the bin. Through this digital twin model, abnormal working conditions such as "bridging" (where the material forms an arch in the bin to block the discharge channel) and "segregation" (where the material particles are separated according to characteristics such as particle size or density) can be effectively identified.

[0083] Construction of the Global Control Strategy of the DDPG Algorithm:

[0084] Definition of the State Space: The state space S contains all the information of the material levels in the bins [h1, h2, h3,...], equipment powers [P1, P2, P3,...], and material flow rates [Q1, Q2, Q3,...], that is, S = [h1, h2,..., P2, P2,..., Q1, Q2,....].

[0085] Definition of the Action Space: The action space A is the adjustment amounts of the stocking quantities at all levels [ΔQ1, ΔQ2, ΔQ3,...], where ΔQ i represents the adjustment value of the stocking quantity of the i-th level equipment, which is achieved by adjusting the rotational speed of the stocking motor or the opening degree of the solenoid valve.

[0086] Design of the Reward Function: Design the reward function \(R\) to comprehensively balance factors such as the accuracy of the material level, equipment energy consumption, and adjustment stability. For example

[0087]

[0088] , where ɑ, β, and γ are weight coefficients, and h set,i is the setpoint of the material level of the i-th silo, and h max，i is the maximum allowable material level of the i-th silo, P i is the power of the i-th device, and ΔQ i is the adjustment amount of the material discharge rate.

[0089] Network structure construction: The DDPG algorithm consists of two deep neural networks, namely the policy network (Actor network) and the value network (Critic network). The policy network takes the state space S as the input and outputs an action a in the action space A, that is, a = μ(S|θ μ ), where θ μ are the parameters of the policy network. The value network takes the state S and the action a as the input and outputs a value estimate Q(S, a|θ Q ), where θ Q are the parameters of the value network.

[0090] Training process: During the training process, by interacting with the environment (i.e., the actual operation of the electrostatic separation production line), samples (S, a, R, S′) are continuously collected, where S is the current state, a is the action taken, R is the reward obtained, and S′ is the next state. Using these samples, the value network is first updated, and its loss function is

[0091] where

[0092] y i = R i + γQ( Si+ 1, μ(S i+1 |θ μ )|θ Q ), γ is the discount factor, which is used to balance the importance of the current reward and the future reward, and generally takes values between 0 and 1. The parameters θ μ of the value network are updated by the gradient descent method to minimize the loss function. Then, according to the gradient of the policy network

[0093]

[0094] , the parameters θ μ of the policy network are updated by the policy gradient ascent method to maximize the cumulative reward.

[0095] Policy Update and Application: After extensive training, the policy network has learned a policy that can select optimal actions in different states. The cloud updates the DDPG algorithm every hour using the latest collected data and transmits the optimized control policy to the edge nodes via the network to update the control rule library of the edge nodes. The edge nodes adjust the feeding amounts of each level of equipment according to the new control rules to achieve coordinated optimization control of multiple devices and maintain the stable operation of the entire electrostatic separation production line.

[0096] IV. Steps for Maintaining Level Balance

[0097] (I) Local and Remote Monitoring

[0098] Local Monitoring of Edge Nodes:

[0099] (1) HMI Interface Display: Through the human-machine interface (HMI), the level curve is displayed in real time in a graphical manner, intuitively presenting the change of the level over time. At the same time, the operating status of the equipment is shown, such as the real-time speed of the feeding motor, the opening degree of the solenoid valve, and parameters such as the current and voltage of the equipment. This information is displayed in the form of numbers and instrument panels, etc., facilitating on-site staff to quickly understand the operating conditions of the equipment.

[0100] (2) Alarm Mechanism: When the edge node determines that the system has an abnormal situation based on sensor data and control algorithms, the local audible and visual alarm device is immediately triggered. For example, when the level exceeds the preset upper or lower threshold, or when the operating parameters of the equipment (such as excessive motor current) are outside the normal range, the alarm is activated. The alarm signal is also displayed in a prominent red warning box on the HMI interface and accompanied by a sound alarm to remind on-site staff to handle the abnormality in a timely manner.

[0101] 2. Cloud Remote Monitoring:

[0102] (1) Web Platform Visualization: Using the Web platform, managers can remotely visualize the operating status of the entire electrostatic separation production line in real time through a browser. The Web platform adopts data visualization technology to display information such as the level and equipment operating parameters in the form of dynamic charts, maps, etc., enabling managers to comprehensively and intuitively understand the operating conditions of the production line. At the same time, the platform supports querying and analyzing historical data and can generate trend reports to assist managers in making decisions.

[0103] (2) Remote alarm notification: The system is equipped with a complete remote alarm mechanism. When an abnormal situation occurs, the cloud automatically sends text messages or emails to notify relevant operation and maintenance personnel. The alarm information details the location where the abnormality occurred (such as the specific silo number or equipment name), the type of abnormality (such as over-limit material level, equipment failure, etc.), and the risk level prediction of the material level trend in the next 30 minutes. For example, by analyzing historical data and current trends, predict the probability that the material level may continue to rise and exceed the safe range within the next 30 minutes, so as to remind the operation and maintenance personnel to make preparations in advance and take corresponding measures, such as arranging on-site inspections by personnel and adjusting control strategies, etc.

[0104] (II) System optimization and calibration

[0105] Algorithm model update:

[0106] (1) Regular retraining: The cloud regularly retrains the LSTM prediction model and the DDPG policy model every week using a large amount of data accumulated during the production process. When retraining the LSTM model, newly collected historical material discharge data, material level change data, etc. are added to the training set, and the weights and biases of the model are readjusted to improve the accuracy of the model's prediction of the material discharge amount trend. For the DDPG policy model, using the latest state, action, and reward data, according to the training process of the DDPG algorithm, update the parameters of the policy network and the value network, so that the model can adapt to changes in material characteristics and equipment performance that may occur during the operation of the production line, and continuously optimize the global control strategy.

[0107] (2) OTA upgrade: Through Over-the-Air (OTA) technology, remotely update the optimized algorithm and control strategy after retraining to the firmware of the edge node. The OTA upgrade process uses a secure and reliable transmission protocol to ensure the integrity and accuracy of data transmission. After receiving the update instruction and data, the edge node automatically performs firmware upgrade, applying the latest control logic and algorithm to the actual control process to improve the control performance and adaptability of the entire system.

[0108] 2. Sensor self-calibration:

[0109] Automatic calibration program operation: The edge node automatically runs the sensor self-calibration program every day. For ultrasonic level gauges, by comparing the level data collected by multiple sensors of the same type at the same time point, calculate the consistency index of the data, such as the standard deviation

[0110] where x i is the data collected by the i-th sensor, is the average value of these data, and n is the number of sensors. For infrared coverage sensors, vibration sensors, and temperature sensors, a similar method is also adopted to compare the differences between their output data and reference values or the data of other sensors of the same type.

[0111] Calibration work order triggering: When the data difference exceeds the preset limit, an artificial calibration work order is automatically triggered. Information such as the location, number, type of the sensor, and the data difference situation is detailedly recorded in the work order. According to the work order information, professionals carry calibration equipment to the site to calibrate and maintain the sensors. For example, for an ultrasonic level gauge, it may be necessary to readjust its installation position, calibrate the transmitting and receiving circuits, etc.; for a temperature sensor, it may be necessary to calibrate it using a standard thermometer and adjust temperature compensation parameters, etc. Through regular self - calibration of the sensors and timely artificial calibration, the accuracy of sensor data acquisition is ensured, providing a reliable data basis for the entire closed - loop control system.

[0112] As Figure 2 shown, the figure shows the associated control logic between the discharge roller of electrostatic separation 1 and the material level in the bin of electrostatic separation 2 in the electrostatic separation system. There are two states of the material level in the bin of electrostatic separation 2: "too high material level" and "too low material level". When the material level in the bin of electrostatic separation 2 is too high, the measure taken is to reduce the running speed of the discharge roller of electrostatic separation 1; when the material level in the bin of electrostatic separation 2 is too low, the running speed of the discharge roller of electrostatic separation 1 is increased. Through such an adjustment method, the control of the material level in the bin of electrostatic separation 2 is realized, and it is maintained at an appropriate level.

[0113] Embodiment 2

[0114] This embodiment provides an automated closed - loop control system for an electrostatic separation production line, including:

[0115] A data acquisition module, configured to

[0116] A computer - readable storage medium, in which multiple instructions are stored, and the instructions are suitable for being loaded and executed by a processor of a terminal device for the described automated closed - loop control method of an electrostatic separation production line.

[0117] A terminal device, including a processor and a computer - readable storage medium, where the processor is used to implement each instruction; the computer - readable storage medium is used to store multiple instructions, and the instructions are suitable for being loaded and executed by the processor for the described automated closed - loop control method of an electrostatic separation production line.

[0118] The above are all preferred embodiments of the present invention. The protection scope of the present invention is not limited by this. Therefore, all equivalent changes made according to the structure, shape, and principle of the present invention should be covered within the protection scope of the present invention.

Claims

1. An automatic closed-loop control method for an electrostatic separation production line, characterized in that, It includes: Obtain the silo level and equipment operation status parameters; Preprocess the obtained silo level and equipment operation status parameters; For the preprocessed silo level and equipment operation status parameters, adjust the equipment operation parameters through a hierarchical feedback control strategy; Maintain the balance of the silo level by real-time monitoring and adjusting the equipment operation status parameters.

2. The automated closed-loop control method for an electrostatic separation production line according to claim 1, wherein The obtaining of the silo level and equipment operation status parameters includes deploying Internet of Things sensors and actuators at the equipment layer. Among them, the Internet of Things sensors include ultrasonic level gauges, infrared coverage sensors, vibration sensors, and temperature sensors, which collect the level height, material coverage status, and equipment operation parameter data in real time; the actuators include feeding motors, solenoid valves, and alarm devices, which are connected to the edge controller through the Modbus / TCP protocol.

3. The automatic closed-loop control method for an electrostatic separation production line according to claim 2, characterized in that, The preprocessing of the obtained silo level and equipment operation status parameters includes configuring edge computing nodes at the edge control layer to perform local data preprocessing on the obtained silo level and equipment operation status parameters. Among them, the level data of the ultrasonic sensor is processed using the extended Kalman filter to solve the state estimation problem of the non-linear silo shape; and the coverage signals of the infrared sensors are fused. When it is detected that the heating tube is not covered, the low-level warning logic is forcibly triggered, and the effectiveness of the level height is corrected through a weighted average algorithm.

4. An automatic closed-loop control method for an electrostatic separation production line according to claim 3, characterized in that, For the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy includes self-tuning of fuzzy PID parameters based on the adaptive fuzzy PID control method. When the level deviation of the current stage exceeds the threshold, indirect control is achieved by adjusting the feeding amount of the upper-level equipment; based on the current deviation, deviation change rate, and historical adjustment times, fuzzy rules for enhancing the differential effect during preset level mutations and increasing the integral weight during long-term deviations are used to dynamically adapt to changes in material fluidity.

5. An automatic closed-loop control method for an electrostatic separation production line according to claim 4, characterized in that For the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy also includes using the historical feeding data stored in the edge node, including timestamp, feeding motor speed, and level change rate, to predict the feeding amount trend in the future set time interval by training the LSTM neural network, and adjusting the speed of the upper-level equipment in advance according to the feeding trend to offset the material transmission lag effect.

6. The automatic closed-loop control method for an electrostatic separation production line according to claim 5, characterized in that, For the preprocessed silo level and equipment operation status parameters, adjusting the equipment operation parameters through a hierarchical feedback control strategy also includes establishing a digital twin model of the material flow in the silo based on computational fluid dynamics (CFD). The input parameters include material particle size distribution, moisture content, and feeding port size; the material flow state in the silo is simulated by analyzing the pulling force through the digital twin model, abnormal conditions such as bridging and segregation are identified, and a global control strategy is constructed using the deep deterministic policy gradient (DDPG) algorithm.

7. An automated closed-loop control method for an electrostatic separation production line according to claim 6, characterized in that The construction of the global control strategy using the Deep Deterministic Policy Gradient (DDPG) algorithm includes building a DDPG network model by defining the state space, action space, and reward function, including a policy model and a value model. Among them, the policy model takes the state space as input and outputs actions in the action space; the value space takes states and actions as input and outputs value estimates. By interacting with the environment and using the collected samples, the value network is updated, which is expressed as: Among them, y i = R i + γQ(S i+1 , μ(S i+1 | θ μ ) | θ Q ), where γ is the discount factor used to balance the importance of current and future rewards, typically taking values between 0 and 1. Update the value network parameter θ by gradient descent μ to minimize the loss function. Then, according to the gradient of the policy network Update the policy network parameter θ using the policy gradient ascent method μ to maximize the cumulative reward.

8. An automatic closed-loop control method for an electrostatic separation production line according to claim 7, characterized in that, The maintenance of the silo level balance by real-time monitoring and adjusting the operating state parameters of the equipment includes local monitoring of the edge node by displaying the level curve and equipment status through the HMI interface; remote cloud monitoring by real-time visualizing the production line status through the Web platform; retraining the LSTM prediction model and DDPG policy model regularly using production data in the cloud, and remotely updating the edge node firmware through OTA technology; the edge node runs the sensor self-calibration program daily, compares the consistency of multi-sensor data, and triggers a manual calibration work order if the limit is exceeded.

9. An automated closed-loop control system for an electrostatic separation production line, characterized in that, Including: A data acquisition module configured to acquire the silo level and equipment operating state parameters; A preprocessing module configured to preprocess the acquired silo level and equipment operating state parameters; A parameter adjustment module configured to adjust the equipment operating parameters through a hierarchical feedback control strategy for the preprocessed silo level and equipment operating state parameters; A balance module configured to maintain the silo level balance by real-time monitoring and adjusting the operating state parameters of the equipment.

10. A computer-readable storage medium storing multiple instructions, characterized in that, The instruction is adapted to be loaded and executed by a processor of a terminal device to perform the method according to claim 1.

Images