Software control-based pulping equipment collaborative operation method

Abstract

The invention relates to the technical field of equipment collaborative operation, in particular to a pulping equipment collaborative operation method based on software control, which comprises the following steps of: constructing a total factor sensing network, initializing a digital twinborn body, establishing real-time data acquisition and high-fidelity virtual mapping covering a pulping whole process, and establishing a software control network; a digital mirror image is provided for subsequent collaboration; and generating a multi-objective collaborative optimization strategy, and generating a global optimal process parameter combination under the three core indexes of quality, energy consumption and efficiency. According to the method, a three-stage control framework of digital mirror image-intelligent decision-dynamic cooperation is constructed, so that essential innovation of the whole pulping process is realized, high-precision virtual mapping is established for digital twins, a process optimization strategy can be pre-verified in a virtual space, and the field debugging risk is avoided; the multi-objective optimization algorithm generates an optimal solution with balanced quality, energy consumption and efficiency in real time, and dynamically responds to order demand changes.

Description

Technical Field

[0001] This invention relates to the field of equipment collaborative operation technology, specifically a method for collaborative operation of pulping equipment based on software control. Background Technology

[0002] Traditional pulping production employs a layered, independent control model, where each process's equipment executes only preset, fixed process parameters, lacking a global coordination mechanism. This operating mode suffers from three key bottlenecks: First, fragmented data between processes leads to a "passive remedial" problem. For example, when the residual alkali rate in the cooking process exceeds the standard, the downstream bleaching process can only compensate by increasing the amount of chemicals used, resulting in resource waste. Second, the lack of dynamic coordination in equipment start-up and shutdown sequences leads to capacity losses due to delayed start-up of key equipment such as grinding mills and cooking pots. Third, anomaly handling relies on manual experience; equipment failures require shutdown for troubleshooting, and recovery time often exceeds 60 minutes. This results in higher unit energy consumption in traditional systems compared to collaborative models, and greater fluctuations in product quality.

[0003] In summary, a software-controlled collaborative operation method for pulping equipment needs to be proposed to solve the above problems. Summary of the Invention

[0004] The purpose of this invention is to provide a method for the coordinated operation of pulping equipment based on software control, so as to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: This invention proposes a method for the coordinated operation of pulping equipment based on software control, comprising the following steps: S1. Construction of a full-element perception network and initialization of a digital twin: Establishing real-time data acquisition and high-fidelity virtual mapping covering the entire pulping process, providing a "digital mirror" for subsequent collaboration; S2. Multi-objective collaborative optimization strategy generation: Under the three core indicators of quality, energy consumption, and efficiency, the global optimal combination of process parameters is generated. S3. Dynamic process parameters are intelligently distributed and equipment is started in tandem. Based on the optimization strategy, the parameters of multiple equipment are accurately matched and started synchronously, eliminating the efficiency loss caused by "start-stop asynchrony". S4. The execution layer coordinates operation and integrates multi-source feedback data to monitor equipment operating status in real time. It corrects process deviations through multi-source data fusion to ensure a closed loop of process parameters "execution-feedback-correction". S5. Cross-process anomaly detection and self-healing control: identify equipment anomalies and process fluctuations in advance, and automatically trigger repair measures to avoid downtime losses; S6. Cross-process quality coupling control solves the inefficiency of "compensating for defects in the previous process in the subsequent process" in traditional control, and achieves proactive compensation across processes through causal relationship identification; S7. Energy consumption-quality Pareto frontier dynamic optimization: Under the premise of meeting quality constraints, it finds the operating point with the lowest energy consumption in real time and supports three operating strategies: "quality priority", "energy consumption priority" and "balance mode". S8. Full-process knowledge accumulation and self-evolution: Through the deep integration of knowledge graphs and digital twins, the system achieves self-learning and self-optimization, enabling collaborative efficiency to continuously improve over time.

[0006] Preferably, the implementation process of step S1 is as follows: S1.1. Deploy 50+ data collection points in the raw material silo (humidity sensor ±0.5%), pulper (vibration sensor 0-10mm / s), cooking pot (pressure sensor 0-2MPa), mixing tank (online concentration meter response <1s), slag remover (solid content sensor ±0.1%), bleaching tower (residual chlorine sensor ±1ppm), and finished product tank (whiteness meter ≤0.5ΔE). Step S2.1 calls the standardized sensor data stream generated in this step. S1.2. Perform data denoising (sliding window mean filtering for 10s) and outlier removal (3σ principle) through the edge gateway to generate a standardized MQTT data stream (1Hz). Step S1.3 uses the cleaned data preprocessed in this step. S1.3. Construct a three-dimensional model based on the material balance equation. After training with historical data, the simulation error is ≤3%. Step S2.2. Call the model in this step to verify the feasibility of the strategy.

[0007] Preferably, the implementation process of step S2 is as follows: S2.1. Establish three-dimensional indicators: quality (whiteness ≥ 85% ISO), energy consumption (≤ 1200 kWh / ton), and efficiency (production capacity ≥ 15 tons / hour). The comprehensive weight can be dynamically adjusted. Step S2.2 calls the target constraints defined in this step. S2.2. Use the improved NSGA-III algorithm to output the Pareto front solution set (e.g., strategy A: grinding speed 550 rpm / steam temperature 145℃). Step S3.1 calls the optimized strategy solution set generated in this step.

[0008] Preferably, the implementation process of step S3 is as follows: S3.1. Convert the strategy parameters into equipment instructions: Inverter frequency = 0.5 × speed setpoint, regulating valve opening = temperature setpoint / 160 × 100%, step S3.2 call the control instructions converted in this step; S3.2. Achieve device-level coordination via PROFINET: Start the digester 30s after the pulper starts (condition: vibration <2mm / s), and start the bleaching tower 120s after the digester starts (condition: concentration ≥8%). Step S4.1 calls the start-up timing parameters of this step.

[0009] Preferably, the implementation process of step S4 is as follows: S4.1. Fuse sensor data with digital twin predicted values ​​(e.g., whiteness fusion value = 0.7 × measured value + 0.3 × predicted value). Step S4.2 calls the fused data from this step to detect deviations. S4.2. Set the dynamic deviation band (whiteness ±1% ISO), and automatically compensate when the deviation is detected (if whiteness <84% ISO, increase bleach by 0.1 kg / per deviation unit). Step S5.1 calls up the abnormal data from this step.

[0010] Preferably, the implementation process of step S5 is as follows: S5.1. Extract features using the LSTM-Attention model (e.g., 120Hz harmonic vibration → bearing wear, sudden drop in viscosity → fiber breakage). Step S5.2. Call up the abnormal features identified in this step. S5.2. Level 3 self-healing response: Level 1 abnormality (vibration 4-5 mm / s): Reduce speed by 20 rpm; Level II anomaly (underflow concentration >5%): backwash 0.6 MPa / 30 s; Level 3 abnormality (pressure > 1.8 MPa): Emergency pressure relief ≤ 5 seconds; S8.1 Invoke the exception handling case for this step.

[0011] Preferably, the implementation process of step S6 is as follows: S6.1. Identify the causal relationship between processes (e.g., fiber length ↑ 0.1mm → need to add 5% bleach), step S6.2 calls the causal strength parameter of this step; S6.2. Based on the previous process indicators (such as fiber length = 0.7mm), adjust the subsequent process parameters in real time (bleach dosage = baseline value × 1.05). Step S7.1 calls up the quality data after compensation in this step.

[0012] Preferably, the implementation process of step S7 is as follows: S7.1. Calculate the total cost (energy cost + quality loss cost) every 10 minutes, and output the lowest cost running point. Step S8.2 calls the cost optimization results of this step. S7.2. Supports three strategies: quality priority (whiteness ≥ 88%), energy consumption priority (energy consumption ≤ 1300 kWh / ton), and balance mode. Step S3.1 calls the optimization parameters of this step to issue.

[0013] Preferably, the implementation process of step S8 is as follows: S8.1. Extract entities (equipment / parameters) and relationships (such as the correlation coefficient between rotation speed and fiber length of 0.85) to construct a knowledge graph with 500+ nodes. Step S2.2: Call this graph to avoid historical fault parameters. S8.2. Based on the new data, the model is fine-tuned, the simulation error is reduced from 3% to ≤1%, the false alarm rate of anomaly detection is reduced by 40%, and step S5.1 calls the updated model to improve the detection accuracy.

[0014] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention achieves a fundamental innovation in the entire pulping process by constructing a three-level control architecture of "digital mirroring - intelligent decision-making - dynamic collaboration". The digital twin establishes a high-precision virtual mapping, enabling process optimization strategies to be pre-verified in virtual space and avoiding on-site debugging risks. The multi-objective optimization algorithm generates the optimal solution that balances quality, energy consumption, and efficiency in real time, dynamically responding to changes in order demand. The cross-process linkage mechanism breaks through the traditional control hierarchy. When the pulping process detects abnormal fiber length, it simultaneously triggers chemical feedforward compensation in the bleaching process, eliminating quality risks at the nascent stage, improving the overall utilization rate of equipment, shortening abnormal downtime, and achieving a synergistic gain where "global optimization is far superior to local optimization". Attached Figure Description

[0015] Figure 1 A flowchart of a software-controlled method for the coordinated operation of a pulping device is shown. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, please refer to Figure 1 This invention proposes a method for the coordinated operation of pulping equipment based on software control, comprising the following steps: S1. Construction of a full-element perception network and initialization of a digital twin: Establishing real-time data acquisition and high-fidelity virtual mapping covering the entire pulping process, providing a "digital mirror" for subsequent collaboration; S1.1. Deploy 50+ data collection points in the raw material silo (humidity sensor ±0.5%), pulper (vibration sensor 0-10mm / s), cooking pot (pressure sensor 0-2MPa), mixing tank (online concentration meter response <1s), slag remover (solid content sensor ±0.1%), bleaching tower (residual chlorine sensor ±1ppm), and finished product tank (whiteness meter ≤0.5ΔE). Step S2.1 calls the standardized sensor data stream generated in this step. S1.2. Perform data denoising (sliding window mean filtering for 10s) and outlier removal (3σ principle) through the edge gateway to generate a standardized MQTT data stream (1Hz). Step S1.3 uses the cleaned data preprocessed in this step. S1.3. Construct a three-dimensional model based on the material balance equation. After training with historical data, the simulation error is ≤3%. Step S2.2 calls the model in this step to verify the feasibility of the strategy. S2. Multi-objective collaborative optimization strategy generation: Under the three core indicators of quality, energy consumption, and efficiency, the global optimal combination of process parameters is generated. S2.1. Establish three-dimensional indicators: quality (whiteness ≥ 85% ISO), energy consumption (≤ 1200 kWh / ton), and efficiency (production capacity ≥ 15 tons / hour). The comprehensive weight can be dynamically adjusted. Step S2.2 calls the target constraints defined in this step. S2.2. Use the improved NSGA-III algorithm to output the Pareto front solution set (e.g., strategy A: grinding speed 550 rpm / steam temperature 145℃). Step S3.1 calls the optimized strategy solution set generated in this step. S3. Dynamic process parameters are intelligently distributed and equipment is started in tandem. Based on the optimization strategy, the parameters of multiple equipment are accurately matched and started synchronously, eliminating the efficiency loss caused by "start-stop asynchrony". S3.1. Convert the strategy parameters into equipment instructions: Inverter frequency = 0.5 × speed setpoint, regulating valve opening = temperature setpoint / 160 × 100%, step S3.2 call the control instructions converted in this step; S3.2. Achieve device-level collaboration via PROFINET: Start the digester 30s after the pulper starts (condition: vibration <2mm / s), and start the bleaching tower 120s after the digester starts (condition: concentration ≥8%). Step S4.1 calls the start-up timing parameters of this step. S4. The execution layer coordinates operation and integrates multi-source feedback data to monitor equipment operating status in real time. It corrects process deviations through multi-source data fusion to ensure a closed loop of process parameters "execution-feedback-correction". S4.1. Fuse sensor data with digital twin predicted values ​​(e.g., whiteness fusion value = 0.7 × measured value + 0.3 × predicted value). Step S4.2 calls the fused data from this step to detect deviations. S4.2. Set the dynamic deviation band (whiteness ±1% ISO), and automatically compensate when the deviation is detected (if whiteness < 84% ISO, increase bleach by 0.1 kg / per deviation unit). Step S5.1 calls up the abnormal data from this step. S5. Cross-process anomaly detection and self-healing control: identify equipment anomalies and process fluctuations in advance, and automatically trigger repair measures to avoid downtime losses; S5.1. Extract features using the LSTM-Attention model (e.g., 120Hz harmonic vibration → bearing wear, sudden drop in viscosity → fiber breakage). Step S5.2. Call up the abnormal features identified in this step. S5.2. Level 3 self-healing response: Level 1 abnormality (vibration 4-5 mm / s): Reduce speed by 20 rpm; Level II anomaly (underflow concentration >5%): backwash 0.6 MPa / 30 s; Level 3 abnormality (pressure > 1.8 MPa): Emergency pressure relief ≤ 5 seconds; S8.1 Call the exception handling case for this step; S6. Cross-process quality coupling control solves the inefficiency of "compensating for defects in the previous process in the subsequent process" in traditional control, and achieves proactive compensation across processes through causal relationship identification; S6.1. Identify the causal relationship between processes (e.g., fiber length ↑ 0.1mm → need to add 5% bleach), step S6.2 calls the causal strength parameter of this step; S6.2. Based on the previous process indicators (such as fiber length = 0.7mm), adjust the subsequent process parameters in real time (bleach dosage = baseline value × 1.05). Step S7.1 calls up the quality data after compensation in this step. S7. Energy consumption-quality Pareto frontier dynamic optimization: Under the premise of meeting quality constraints, it finds the operating point with the lowest energy consumption in real time and supports three operating strategies: "quality priority", "energy consumption priority" and "balance mode". S7.1. Calculate the total cost (energy cost + quality loss cost) every 10 minutes, and output the lowest cost running point. Step S8.2 calls the cost optimization results of this step. S7.2. Supports three strategies: quality priority (whiteness ≥ 88%), energy consumption priority (energy consumption ≤ 1300 kWh / ton), and balance mode. Step S3.1 calls and distributes the optimization parameters of this step. S8. Full-process knowledge accumulation and self-evolution: Through the deep integration of knowledge graphs and digital twins, the system can learn and optimize itself, so that collaborative efficiency can be continuously improved over time. S8.1. Extract entities (equipment / parameters) and relationships (such as the correlation coefficient between rotation speed and fiber length of 0.85) to construct a knowledge graph with 500+ nodes. Step S2.2: Call this graph to avoid historical fault parameters. S8.2. Based on the new data, the model is fine-tuned, the simulation error is reduced from 3% to ≤1%, the false alarm rate of anomaly detection is reduced by 40%, and step S5.1 calls the updated model to improve the detection accuracy.

[0018] Example 2, please refer to Figure 1 In practical applications, the present invention provides a method for the coordinated operation of software-controlled pulping equipment, specifically including the following steps: S1. Construction of a full-element perception network and initialization of a digital twin: Establishing real-time data acquisition and high-fidelity virtual mapping covering the entire pulping process, providing a "digital mirror" for subsequent collaboration; Sensor Deployment: Multi-physics sensors are deployed at key nodes such as the raw material silo (humidity sensor, accuracy ±0.5%), the pulper (vibration sensor, range 0-10mm / s, sampling frequency 1kHz), the cooking pot (pressure sensor, range 0-2MPa, accuracy 0.1%FS), the mixing tank (online concentration meter, based on near-infrared spectroscopy, response time <1s), the slag remover (solid content sensor, range 0-5%, accuracy ±0.1%), the bleaching tower (residual chlorine sensor, range 0-100ppm, accuracy ±1ppm), and the finished product tank (whiteness meter, based on CIELab standard, repeatability ≤0.5ΔE), with a total of ≥50 sampling points; Data preprocessing: Data denoising (sliding window mean filtering, window size = 10s) and outlier removal (3σ principle) are performed through edge computing gateways (including Siemens SCALANCE W788-2) to generate a standardized data stream (MQTT protocol, frequency 1Hz). Digital twin modeling: A three-dimensional dynamic model is built based on ANSYS TwinBuilder, integrating material balance equations (including mass conservation: raw material input = fiber output + slag + loss) and energy balance equations (steam heat = heat required for slurry heating + heat loss), and training model parameters (including fiber length shrinkage coefficient and thermal conductivity coefficient) using historical operating data (≥1 year), with an initial simulation error ≤3%; The digital twin acts as a "virtual controller," receiving the optimization strategy from step 2 in real time and simulating the execution results, providing a virtual test field for verifying the feasibility of the strategy (including predicting "what changes will occur in the fiber length distribution if the refiner speed is increased by 5%"). S2. Multi-objective collaborative optimization strategy generation (based on the improved NSGA-III algorithm): Under the three core indicators of quality (whiteness ≥ 85% ISO), energy consumption (kWh / ton of pulp ≤ 1200), and efficiency (capacity ≥ 15 tons / hour), the globally optimal combination of process parameters is generated. Objective function definition: Quality index Q = 0.6 × Whiteness + 0.3 × Fiber length uniformity (standard deviation ≤ 0.3 mm) + 0.1 × Dust content (particles / m) 2 ≤50); Energy consumption index E = Steam consumption per unit × 0.7 + Electricity consumption per unit × 0.3 (Steam consumption per unit = Steam consumption / Finished slurry volume, kWh / ton); Efficiency index P = Actual capacity / Designed capacity (upper limit 1.2, to prevent overload); Overall objective function: F = αQ + β(1 / E) + γP (α = 0.5, β = 0.3, γ = 0.2, weights can be dynamically adjusted according to order requirements); Constraints: Pulping machine current ≤ rated value (including 50A), cooking pot pressure ≤ safety threshold (1.8MPa), chemical reagent addition ≤ environmental protection limit (including NaOH ≤ 8kg / ton of pulp); Algorithm Execution: An improved NSGA-III algorithm is adopted, introducing an adaptive crossover probability of 0.7 + 0.2 × number of iterations / total number of iterations and a mutation probability of 0.1-0.05 × number of iterations / total number of iterations. The Pareto front solution set is searched within the parameter space, considering mill speeds of 200-800 rpm, steam temperatures of 120-160℃, and NaOH concentrations of 5-15%. The top three optimal strategies are output. Strategy A: 550 rpm, 145°C, 8% concentration; Strategy B: 600 rpm, 150°C, 7.5% concentration. The strategy solution set is sent to the execution layer of step S3 via the OPCUA protocol and synchronized to the digital twin for simulation verification, simulating the solid content change curve of the slag remover under strategy A; S3. Dynamic process parameters are intelligently distributed and equipment is started in tandem. Based on the optimization strategy, the parameters of multiple equipment are accurately matched and started synchronously, eliminating the efficiency loss caused by "start-stop asynchrony". Parameter mapping: Converting process parameters in the strategy into equipment control commands. The mill speed → frequency converter frequency (Hz) = speed × 0.5; steam flow rate → pneumatic regulating valve opening (%) = temperature setpoint / 160 × 100; chemical agent addition amount → metering pump stroke frequency (Hz) = concentration setpoint × 2 (because the concentration and stroke frequency are linearly related, the slope is determined by calibration). Timing synchronization: Based on timestamp alignment (accuracy ±10ms), set "startup buffer period": The pulper is preheated 30 seconds in advance, and the digester starts feeding after the pulper has stabilized (vibration value <2mm / s). Equipment-level coordination is achieved through synchronous commands (including PROFINET communication) of the PLC (Siemens S7-1500). Fault tolerance mechanism: If a device fails to start (the frequency converter reports fault code F002), the system automatically triggers a "degradation strategy": Switch to standby equipment and adjust related parameters: if refiner 2 fails to start, increase the speed of refiner 1 to 600 rpm to compensate for the production capacity. Real-time operating data (mill current, steam temperature) after equipment startup is uploaded in real time to the execution feedback module in step 4 via industrial Ethernet (TCP / IP) for closed-loop control; S4. The execution layer coordinates operation and integrates multi-source feedback data to monitor equipment operating status in real time. It corrects process deviations through multi-source data fusion to ensure a closed loop of process parameters "execution-feedback-correction". Operational data acquisition: Pulping mill, current value (A) = load index, normal range 30-50A; Mixing tank, online viscometer, range 10-100mPa·s, accuracy ±0.5mPa·s; Slag remover, underflow concentration (%) = fiber retention rate index, normal range 2-4%; Bleaching tower, outlet whiteness (%ISO) = quality index; Node acquisition of high-frequency data (frequency 5Hz); Data fusion algorithm: Utilizing Federated Kalman filtering (F-Kalman), the simulated values ​​from the digital twin are used as prior information to fuse on-site measurements. Whiteness measurement value = Simulation prediction value + Measurement noise (σ = 0.8% ISO); Reduce the impact of sensor noise; the whiteness measurement error after fusion is ≤0.3% ISO. Deviation Correction: Set a dynamic allowable deviation band, with a whiteness target of 85% ISO and an allowable deviation of ±1% ISO. If the whiteness after fusion is <84% ISO, the system will automatically calculate the compensation amount. ΔQ = 0.5 × (85 - current whiteness); By adjusting the amount of NaClO added to the bleaching tower: ΔAmount Added = ΔQ × 0.1kg / Δ%ISO; Provide compensation; The parameters after deviation correction (adjusted NaClO addition amount) are fed back to the distribution module in step S3 in real time, forming a dynamic loop of "distribution-execution-feedback-correction"; at the same time, abnormal data triggers the anomaly detection in step S5. S5. Cross-process anomaly detection and self-healing control (based on LSTM-Attention model) identifies equipment anomalies (blockage, scaling) and process fluctuations (sudden changes in fiber length) in advance and automatically triggers repair measures to avoid downtime losses; Anomaly feature extraction: Using the vibration signal of the pulper (1kHz sampling), the viscosity of the mixing tank (5Hz sampling), and the underflow concentration of the slag remover (5Hz sampling) as inputs, the temporal features were extracted using the LSTM-Attention model (time step = 10s, hidden layer = 64). The 120Hz harmonic in the vibration frequency corresponds to bearing wear, and the sudden drop in viscosity corresponds to fiber breakage. Anomaly Classification: Level 1 anomaly (warning): Pulper vibration value 4-5mm / s (normal <2mm / s), trigger parameter fine adjustment (reduce speed by 20rpm); Level 2 abnormality (intervention): The underflow concentration of the slag remover is >5% (normal is 2-4%), which is judged to be fiber loss. The front valve is automatically closed and backwashing is started (compressed air pressure 0.6MPa, duration 30s). Level 3 abnormality (shutdown): Cooker pressure > 2MPa (safe threshold 1.8MPa), immediately cut off the feed and start the pressure relief valve (opening time ≤ 5s); Self-healing strategy library: Predefined self-healing actions for 20 types of anomalies ("fiber length standard deviation > 0.5 mm → increase refiner speed by 30 rpm"), and continuously optimize strategies through a historical case library (≥1000 cases); The abnormal handling result (viscosity recovery value of the mixing tank after backwashing) is fed back to the digital twin to update the model parameters (correct the correlation coefficient between viscosity and fiber length), and the abnormal type and handling strategy are stored in the knowledge base (step S8). S6. Cross-process quality coupling control (dynamic compensation based on causal inference) solves the inefficiency of "compensating for defects in the previous process in the subsequent process" in traditional control, and achieves proactive compensation across processes through causal relationship identification; Causal relationship mining: Historical data was analyzed using PC algorithm (conditional independence test) to identify causal chains between processes ("slag remover underflow concentration ↑ → bleaching tower effective chlorine utilization rate ↓ → finished product whiteness ↓", causal strength = 0.7); "pulverizer fiber length ↑ → mixing tank viscosity ↑ → bleaching agent diffusion rate ↓ → need to increase NaClO addition by 5% / mm fiber length" (causal strength = 0.8). Dynamic compensation calculation: Based on the real-time monitoring of upstream process quality indicators, with fiber length = 0.7mm (normal 0.6-0.8mm), the compensation amount required for downstream processes is calculated using a causal model. NaClO addition amount = baseline value × (1 + 0.05 × (0.7 - 0.6))) = baseline value × 1.05; If the underflow concentration of the slag remover is 5% (exceeding the standard), then the effective chlorine consumption of the bleaching tower is = baseline value × (1 / (1-0.03)) = baseline value × 1.03 (to compensate for 3% loss). Cross-equipment collaborative adjustment: The compensation command is synchronized to the downstream equipment (sent to the metering pump controller of the bleaching tower via ModbusTCP) to achieve proactive control of "pre-process disturbance → downstream pre-compensation"; The compensated post-processing effect (actual whiteness value) is fed back to the causal model to correct the causal strength parameter (it was found that the effect of fiber length on whiteness is actually 0.8 instead of 0.7), thus improving the model's accuracy. S7. Energy consumption-quality Pareto front dynamic optimization (based on NSGA-II and reinforcement learning fusion): Under the premise of meeting quality constraints, it finds the running point with the lowest energy consumption in real time and supports three running strategies: "quality priority", "energy consumption priority" and "balanced mode". Real-time cost calculation: Energy cost = electricity consumption per unit × electricity price (0.8 yuan / kWh) + steam consumption per unit × coal price (0.15 yuan / kJ); Quality cost = (target whiteness - actual whiteness) × quality loss coefficient (50 yuan / %ISO); Total cost = energy cost + quality cost; Dynamic Pareto Front Generation: An initial solution set is generated using NSGA-II. Deep reinforcement learning (DQN) is introduced as the outer optimizer. The policy network is trained using a reward function (reward = -total cost + 0.1 × productivity), and the optimization is performed every 10 minutes to output the current optimal solution. Including the "balanced mode": electricity consumption 450kWh / ton, steam consumption 1.2t / ton, whiteness 86% ISO, total cost 120 yuan / ton; Strategy switching logic: If the customer order requires whiteness ≥88% ISO, switch to "quality priority" mode, constrain whiteness ≥88%, and optimize minimum energy consumption; if the electricity price is at its peak (>1.2 yuan / kWh), switch to "energy consumption priority" mode, constrain energy consumption ≤1300kWh / ton, and optimize quality compliance. The optimized energy consumption-quality strategy is distributed to the execution layer of step S3 via OPCUA, and the cost data is stored in the knowledge base (step S8) for order matching. S8. Full-process knowledge accumulation and self-evolution (continuous learning based on digital twins): Through the deep integration of knowledge graphs and digital twins, the system can achieve self-learning and self-optimization, so that collaborative efficiency can be continuously improved over time. Knowledge extraction: Extract entities (including "refiner", "brightness", "NaClO addition amount"), relationships (including "refiner speed affects fiber length (correlation coefficient 0.85)"), and rules (including "when fiber length < 0.5 mm, the speed needs to be increased by 20 rpm") from historical operation data (process parameters, abnormal cases, optimization strategies, cost data) to construct a knowledge graph in the pulping field (number of nodes ≥ 500, number of edges ≥ 1000). Digital twin iteration: Newly collected data (including characteristic data of new wood pulp raw materials) are input into the digital twin, and the model parameters are updated through transfer learning (freezing the first 8 layers of the pre-trained model and fine-tuning the last 2 layers), so that the simulation error is reduced from the initial 3% to less than 1%. Automatic strategy evolution: Historical optimization strategies are evaluated monthly (including calculating "quality compliance rate" and "energy saving rate"). High-scoring strategies (including the top 20%) are marked as "recommended strategies" and the initial solution set of NSGA-II is updated to accelerate the convergence speed of subsequent optimizations (expected to reduce the time of each iteration by 15%). Integrated with the entire process: The knowledge graph provides prior knowledge (including automatic avoidance of historical fault parameter combinations) for strategy generation in step S2. The update of the digital twin improves the perception accuracy of step S1 and the anomaly detection accuracy of step S5, forming a closed-loop evolution of "perception-optimization-execution-learning".

[0019] Through the above steps, this invention achieves a fundamental innovation in the entire pulping process by constructing a three-level control architecture of "digital mirroring - intelligent decision-making - dynamic collaboration". The digital twin establishes a high-precision virtual mapping, enabling process optimization strategies to be pre-verified in virtual space and avoiding the risks of on-site debugging. The multi-objective optimization algorithm generates the optimal solution that balances quality, energy consumption, and efficiency in real time, dynamically responding to changes in order demand. The cross-process linkage mechanism breaks through the traditional control hierarchy. When the pulping process detects abnormal fiber length, it simultaneously triggers chemical feedforward compensation in the bleaching process, eliminating quality risks at the nascent stage, improving the overall utilization rate of equipment, shortening abnormal downtime, and achieving a synergistic gain where "global optimization is far superior to local optimization".

[0020] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for the coordinated operation of a pulping device based on software control, characterized in that, Includes the following steps: S1. Construction of a full-element perception network and initialization of a digital twin: Establishing real-time data acquisition and high-fidelity virtual mapping covering the entire pulping process, providing a "digital mirror" for subsequent collaboration; S2. Multi-objective collaborative optimization strategy generation: Under the three core indicators of quality, energy consumption, and efficiency, the global optimal combination of process parameters is generated. S3. Dynamic process parameters are intelligently distributed and equipment is started in tandem. Based on the optimization strategy, the parameters of multiple equipment are accurately matched and started synchronously, eliminating the efficiency loss caused by "start-stop asynchrony". S4. The execution layer coordinates operation and integrates multi-source feedback data to monitor equipment operating status in real time. It corrects process deviations through multi-source data integration to ensure a closed loop of "execution feedback correction" for process parameters. S5. Cross-process anomaly detection and self-healing control: identify equipment anomalies and process fluctuations in advance, and automatically trigger repair measures to avoid downtime losses; S6. Cross-process quality coupling control solves the inefficiency of "compensating for defects in the previous process in the subsequent process" in traditional control, and achieves proactive compensation across processes through causal relationship identification; S7. Energy consumption-quality Pareto frontier dynamic optimization: Under the premise of meeting quality constraints, it finds the operating point with the lowest energy consumption in real time and supports three operating strategies: "quality priority", "energy consumption priority" and "balance mode". S8. Full-process knowledge accumulation and self-evolution: Through the deep integration of knowledge graphs and digital twins, the system achieves self-learning and self-optimization, enabling collaborative efficiency to continuously improve over time.

2. The method for collaborative operation of a pulping equipment based on software control according to claim 1, characterized in that, The implementation process of step S1 is as follows: S1.

1. Deploy 50+ collection points in the raw material warehouse, grinding mill, cooking pot, mixing tank, slag remover, bleaching tower, and finished product tank; S1.

2. Perform data denoising and outlier removal through the edge gateway to generate a standardized MQTT data stream; S1.

3. A three-dimensional model is constructed based on the material balance equation, and the simulation error is ≤3% after training with historical data.

3. The method for collaborative operation of a pulping device based on software control according to claim 2, characterized in that, The implementation process of step S2 is as follows: S2.

1. Establish three-dimensional indicators: quality, energy consumption, and efficiency, with the overall weights dynamically adjustable; S2.

2. The Pareto front solution set is output using the NSGA-III algorithm.

4. The method for collaborative operation of a pulping device based on software control according to claim 3, characterized in that, The implementation process of step S3 is as follows: S3.

1. Convert policy parameters into device commands: Inverter frequency = 0.5 × speed setting value; Control valve opening = Temperature setpoint / 160 × 100%; S3.

2. Device-level collaboration via PROFINET: The digester starts 30 seconds after the pulper starts, and the bleaching tower starts 120 seconds after the digester starts.

5. A method for collaborative operation of a pulping device based on software control according to claim 4, characterized in that, The implementation process of step S4 is as follows: S4.

1. Fusion of sensor data and digital twin predictions; S4.

2. Set a dynamic deviation band and automatically compensate when the deviation exceeds the limit.

6. A method for collaborative operation of a pulping device based on software control according to claim 5, characterized in that, The implementation process of step S5 is as follows: S5.

1. Extract features using the LSTM-Attention model; S5.

2. Level 3 self-healing response: Level 1 abnormality (vibration 4-5 mm / s): Reduce speed by 20 rpm; Level II anomaly (underflow concentration >5%): backwash 0.6 MPa / 30 s; Level 3 abnormality (pressure > 1.8 MPa): emergency pressure relief ≤ 5 seconds.

7. A method for collaborative operation of a pulping device based on software control according to claim 6, characterized in that, The implementation process of step S6 is as follows: S6.

1. Identify causal relationships between processes; S6.

2. Adjust the parameters of the subsequent process in real time based on the indicators of the previous process.

8. A method for collaborative operation of a pulping equipment based on software control according to claim 7, characterized in that, The implementation process of step S7 is as follows: S7.

1. Calculate the total cost every 10 minutes and output the lowest cost running point; S7.2 supports three strategies: quality priority, energy priority, and balance mode.

9. A method for collaborative operation of a pulping device based on software control according to claim 8, characterized in that, The implementation process of step S8 is as follows: S8.

1. Extract entities and relationships to build a knowledge graph with 500+ nodes; S8.

2. Based on the new data, the model was fine-tuned, and the simulation error was reduced from 3% to ≤1%, and the false alarm rate of anomaly detection was reduced by 40%.

Images