System and method for intelligently and synergistically preparing high-purity iron through iron ore alkali dissolution-electrolysis

Abstract

According to the iron ore alkali dissolution-electrolysis intelligent collaborative optimization system and the dynamic purification method provided by the invention, artificial intelligence and a metallurgical process are fused, and full-process dynamic regulation and control are realized. The system comprises: 1, an intelligent collaborative optimization module, which dynamically and collaboratively optimizes dissolution-electrolysis process parameters (NaOH concentration, liquid-solid ratio, Fe < 3 + > concentration and current density) based on reinforcement learning (DQN algorithm), predicts an impurity dissolution trend and adds a specific inhibitor in a gradient manner in combination with machine learning, and inhibits generation of Al2O3 / SiO2; the equipment health management module realizes corrosion / blockage two-factor accurate early warning (the accuracy is greater than or equal to 92%) through deep learning analysis of a corrosion image and a vibration spectrum, and is linked with a maintenance decision; the energy-resource coupling module integrates waste heat recovery (greater than or equal to 85%) and a multi-effect evaporation regeneration technology, so that the energy consumption per ton of iron is reduced to 2.9 GJ, and the alkali consumption is less than or equal to 0.95 kg; and 4, a self-adaptive process platform: dynamically updating the model based on a federated learning framework, and adapting to a complex ore source with Fe greater than or equal to 35% and Al2O3 less than or equal to 8%. The embodiment verifies that the system improves the hematite dissolution rate to 94%, reduces the non-planned shutdown of equipment by 40%, reduces the comprehensive cost by 14%, and is suitable for efficient and clean utilization of low-grade iron ores.

Description

Technical Field

[0001] The present invention belongs to the field of green intelligent metallurgy technology, specifically relating to a system and method for the efficient and clean utilization of iron ore that integrates artificial intelligence and electrolytic metallurgy. In particular, the present invention uses intelligent collaborative optimization technology to achieve dynamic regulation of the entire iron ore dissolution-electrolysis process in an alkaline environment, addressing key issues in traditional processes such as lag in parameter control, extensive impurity control, severe equipment corrosion, and low energy efficiency. This technology integrates reinforcement learning algorithms, multimodal sensor networks, and digital twin technology, and is suitable for the efficient utilization of low-grade iron ore resources and ultra-low-carbon metallurgical production scenarios. Background Art The production of high-purity iron is of great significance in the modern steel industry and the field of advanced materials. Alkaline electrolysis has attracted widespread attention due to its environmentally friendly nature (compared to acidic processes). For example, the European Union's Ultra-Low CO2 Steelmaking (ULCOS) program explored a process for direct electrolysis of Fe2O3 particles dispersed in a NaOH solution. However, the traditional alkaline electrolysis process still has the following key bottlenecks:

[0002] 1) Relying on manual experience to adjust parameters such as NaOH concentration (5-8M) and temperature (80-95°C) makes it difficult to optimize the balance between dissolution rate and energy consumption in real time, resulting in energy consumption of up to 3.5 GJ per ton of iron and alkali consumption ≥1.2 kg / ton of iron;

[0003] 2) Poor impurity control: The dissolution of impurities such as Al2O3 and SiO2 fluctuates greatly (±15%), which easily forms silica gel and blocks the filter press equipment (failure rate ≥15%), increasing maintenance costs.

[0004] 3) The strong alkaline environment accelerates the corrosion of the reactor wall (the annual maintenance cost of titanium-based materials increases by 30%), and traditional monitoring methods are difficult to provide timely warnings;

[0005] 4) Waste heat (80-95°C) is not effectively recovered, electrolytic cell preheating relies on external energy, and the overall energy efficiency is less than 60%, which restricts the development of green metallurgy. Summary of the Invention

[0006] This invention provides an intelligent collaborative optimization system and method for alkaline dissolution and electrolysis of iron ore. By deeply integrating artificial intelligence with metallurgical processes, it achieves dynamic optimization and precise control of the entire dissolution-electrolysis process. The core of this invention lies in the construction of a closed-loop intelligent system of "perception-decision-execution-optimization", which mainly includes the following innovative modules:

[0007] 1. Intelligent collaborative optimization system, including a multimodal perception layer, a dynamic optimization core, and an impurity suppression strategy.

[0008] Multimodal data perception layer: Deploys temperature sensors (±0.1°C), pH probes (±0.01), turbidity, and vibration sensors to collect real-time data from reactors, filtration systems, and electrolyzers.

[0009] Dynamic optimization decision core: 1) Using deep reinforcement learning (DQN algorithm) and digital twin technology to dynamically adjust the dissolution stage parameters: dissolution stage: NaOH concentration 5.5-7.8M, liquid-solid ratio 3.2:1-4.8:1 and electrolysis stage: Fe 3+ Concentration 105-145g / L, current density 200-350A / m 2 ; 2) Response time ≤ 50ms, dissolution rate increased to 94%.

[0010] Intelligent impurity suppression strategy: Based on the XGBoost model to predict the Al2O3 / SiO2 dissolution trend, the gradient addition of citric acid (0.05-0.15wt%) and sodium hexametaphosphate (0.1-0.3wt%) was triggered to reduce the amount of impurities dissolved by 32-55%.

[0011] 2. Equipment health interlocking management system, including corrosion early warning, blockage early warning and maintenance decision-making.

[0012] Corrosion early warning: Analyze the reactor inner wall image (resolution ≥ 200dpi) through CNN and identify the pitting depth (accuracy ± 0.1mm);

[0013] Blockage warning: Utilize the LSTM model to analyze the filter press vibration spectrum (0-5kHz) and predict the probability of blockage (accuracy ≥ 92%).

[0014] Maintenance decisions: Automatically schedule titanium-based material repairs or spare parts replacements based on risk levels, extending equipment life by 2.3 times.

[0015] 3. Energy-resource coordinated optimization system, including renewable energy, waste heat-electrolysis coordination and intelligent alkali regeneration.

[0016] Renewable energy, waste heat and electrolysis coordination: Through the enterprise's smart microgrid and the Internet of Things, renewable energy and waste heat from the dissolution process (80-95°C) are matched with the preheating requirements of the electrolytic cell, saving energy and using green electricity for ultra-low carbon smelting.

[0017] Intelligent alkali regeneration: using a multi-effect evaporator (evaporation efficiency ≥ 85%) combined with a regression model to optimize the concentration ratio (1.2-1.8), alkali consumption ≤ 0.95kg / ton iron.

[0018] 4. Adaptive process optimization platform.

[0019] Knowledge graph: Integrates historical production data (over 100,000 batches), ore image libraries, and electrolyte composition maps to discover the optimal process combination through clustering algorithms;

[0020] Model adaptation: Dynamically adapts to different grades of iron ore (Fe ≥ 35%) and impurity fluctuations (Al2O3 ≤ 8%) based on the Federated Learning framework. Specific implementation method:

[0022] The system of the present invention adopts an "edge-cloud" collaborative computing architecture, which includes the following three levels:

[0023] 1. Perception and execution layer: Temperature monitoring: temperature sensor; pH monitoring: pH probe; actuator: PLC-controlled dosing pump.

[0024] Detailed description of the hardware configuration of the perception execution layer system:

[0025] Perception layer equipment list:

[0026] (1) Temperature monitoring system:

[0027] Main sensor: Rosemount 848T (range 0-150℃, accuracy ±0.1℃);

[0028] Spare sensor: PT100 thermal resistor (Class A accuracy);

[0029] Installation location: 1 set each in the upper, middle and lower sections of the reactor, 3 sets in total.

[0030] (2) pH monitoring system:

[0031] Main probe: Mettler-Toledo InPro3250 (with automatic cleaning function);

[0032] Calibration cycle: automatic calibration every 8 hours;

[0033] Protection grade: IP68.

[0034] Actuator configuration:

[0035] (1) Dosing system:

[0036] Metering pump: Siemens Dosimax series (flow range 0.5-50L / h);

[0037] valve: Type 650 pneumatic control valve (leakage level VI).

[0038] (2) Heating system:

[0039] Main heater: ABB ACS880 inverter drive (power 0-100kW adjustable);

[0040] Backup heater: resistance heating rod (for emergency use).

[0041] 2. Edge computing layer: Deploy NVIDIA Jetson AGX Xavier to achieve real-time data preprocessing and fast response (≤50ms).

[0042] 3. Cloud analysis layer: Uses the Alibaba Cloud Industrial Brain AI platform to run digital twin models and federated learning frameworks.

[0043] The software process includes real-time control loops and human-computer interaction.

[0044] Real-time control loop: data acquisition → PCA dimensionality reduction → model inference → instruction issuance (cycle ≤ 5 minutes);

[0045] Human-computer interaction: The visual dashboard integrates a 3D digital twin model to display corrosion heat maps and AI decision-making basis.

[0046] Control algorithm implementation details:

[0047] (1) DQN algorithm parameter settings:

[0048] Neural network structure: input layer: 32 nodes (corresponding to 32 process parameters); hidden layer: 3 layers (128-64-32 nodes); output layer: 8 nodes (corresponding to 8 control variables).

[0049] Training parameters: learning rate: 0.001 (Adam optimizer); discount factor: 0.95; experience replay cache: 10,000 sets of data.

[0050] (2) Real-time control process:

[0051] Through the verification of the examples, this system realizes:

[0052] 1. Improved process efficiency: The dissolution rate increased by 9-20% to 94%.

[0053] 2. Reduced production costs: Energy consumption per ton of iron is reduced to 2.9 GJ (reduced by 17%), and alkali consumption is reduced to 0.89 kg (reduced by 26%).

[0054] 3. Improved equipment reliability: Unplanned downtime is reduced by 40%, and maintenance costs are significantly reduced.

[0055] 4. Environmental benefits: waste heat utilization rate ≥ 85%, achieving ultra-low carbon smelting. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 It is a system flow chart of the present invention;

[0057] Figure 2 It is a system module diagram of the present invention. DETAILED DESCRIPTION

[0058] The following description is intended to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are merely examples, and those skilled in the art may conceive of other obvious variations.

[0059] Example 1

[0060] The efficient dissolution and impurity control of low-grade hematite (Fe2O3 50%) specifically includes the following steps:

[0061] Raw material analysis: Analyze the composition of hematite (Fe2O350%, Al2O36%, SiO24%);

[0062] Material dissolution: Use NaOH to dissolve hematite, initial conditions: NaOH solution: 6.0M, temperature 80°C; liquid-to-solid ratio: 3.5:1.

[0063] 1) AI dynamically adjusts the NaOH concentration to 6.8M (originally fixed at 6.0M) and the temperature to 92°C (originally 85°C); dynamic adjustment method: parameters are updated every 5 minutes; adjustment range limits: NaOH concentration: ±0.2M / time; temperature: ±2°C / time.

[0064] 2) The liquid-to-solid ratio was optimized to 4.2:1 (originally 3.5:1).

[0065] Impurity suppression: Add impurity removers according to conditions. Inhibitor dosing strategy: Primary addition: 30 minutes after dissolution begins; Gradient supplementation: Automatic adjustment based on turbidity changes; Maximum dosage limit: citric acid ≤ 0.15wt%.

[0066] XGBoost triggered the gradient addition of citric acid (0.12 wt%) and sodium hexametaphosphate (0.2 wt%).

[0067] Implementation results: 1) The dissolution rate increased from 85% to 94%; 2) The amount of Al2O3 dissolved decreased by 32%, and the formation of colloidal silica decreased by 55%; 3) The energy consumption per ton of iron was reduced to 2.9GJ (originally 3.5GJ), and the alkali consumption was 0.89kg (originally 1.2kg).

[0068] Example 2

[0069] Anti-clogging of high silicon iron ore (SiO2 10%) and extension of equipment life include:

[0070] Raw material analysis: Analyze the composition of limonite (Fe2O345%, Al2O35%, SiO210%);

[0071] Device Management:

[0072] 1) CNN real-time monitoring of reactor wall corrosion (repair is triggered when pitting depth ≤ 0.2 mm);

[0073] Corrosion monitoring process:

[0074] (a) Image acquisition: Camera: Basler ace series (5 megapixels); Shooting frequency: once every 2 hours; Shooting angle: 6 fixed observation points.

[0075] (b) Image analysis: CNN model input: 256×256 pixel RGB image; output: corrosion area percentage, maximum corrosion depth; alarm threshold: pitting depth ≥ 0.2 mm.

[0076] 2) When LSTM predicts the filter press to have a blockage probability of ≥90%, it will automatically backflush.

[0077] Energy synergy: Dissolution waste heat (90°C) matches the electrolytic cell preheating requirements.

[0078] Implementation results: 1) The failure rate of filter presses dropped from 15% to 5%; 2) Unplanned equipment downtime was reduced by 40%, significantly reducing maintenance costs; 3) The waste heat utilization rate reached 85%, saving 18% of energy.

[0079] Example 3

[0080] Adaptive process optimization for complex composition iron ore (Fe 35%), specifically including:

[0081] Raw material analysis: Analyze the composition of mixed iron ore (Fe2O335%, Al2O38%, containing a small amount of phosphorus);

[0082] Data-driven optimization: 1) Federated learning dynamically adjusts model parameters to adapt to low-grade ores; 2) Knowledge graph recommends the optimal process combination: NaOH 7.2M, current density 280A / m 2 .

[0083] Alkali solution regeneration: Multi-effect evaporator concentration ratio 1.5, evaporation efficiency 88%.

[0084] Implementation results: 1) The dissolution rate is stabilized at 90% (the original process fluctuated 70-85%); 2) The alkali consumption is reduced to 0.92 kg / ton iron (originally 1.3 kg).

[0085] Comparison of Examples and Patent Advantages

[0086] System operation process:

[0087] 1. Data acquisition stage: Multi-sensor real-time acquisition of process parameters (cycle ≤ 1s); data is processed by PCA dimensionality reduction.

[0088] 2. Intelligent decision-making stage: The edge performs fast response control (≤50ms); the cloud performs in-depth analysis and model optimization.

[0089] 3. Execution feedback stage: PLC accurately executes control instructions; effect data is fed back to the learning system.

[0090] 4. Visual monitoring: The 3D digital twin interface displays real-time process parameters, equipment health status, and AI decision-making basis.

[0091] By optimizing the environmental model and intelligent agent design, the iron purity, energy consumption, hydrogen generation, and production efficiency of the high-purity iron electrolysis process have been further optimized. The intelligent agent can dynamically adjust electrolysis parameters and adapt to complex electrolysis environments, achieving efficient, energy-saving, and stable high-purity iron electrolysis production.

[0092] The above shows and describes the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The above embodiments and descriptions merely illustrate the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An intelligent collaborative optimization system for iron ore alkaline dissolution and electrolysis It is characterized by Includes the following modules: Dissolution parameter dynamic control network: Through real-time data collection by multimodal sensors, combined with reinforcement learning (DQN algorithm) and digital twin technology, it dynamically optimizes NaOH concentration (5.5-7.8M), liquid-solid ratio (3.2:1-4.8:1), current density (200-350A / m 2 )parameter; Equipment status interlocking maintenance unit: Based on the CNN-LSTM fusion model, it realizes dual-factor early warning of reactor corrosion and filter press blockage, and automatically triggers maintenance instructions; Energy cross-process scheduling engine: Utilizes digital twin technology to match dissolution waste heat (80-95°C) with electrolytic cell preheating requirements, achieving a waste heat recovery rate of ≥85%; Electrolyte quality feedback correction module: According to Fe 3+ The concentration (105-145g / L) can be adjusted in real time to adjust the electrolysis process parameters.

2. The system according to claim 1 Its characteristics are: The dynamic control network uses a multi-objective optimization algorithm to balance the dissolution rate, energy consumption and impurity control objectives; The NaOH concentration in the dissolution stage is preferably controlled at 6.5-7.2M, and the current density optimization range in the electrolysis stage is 250-320A / m 2 .

3. The system according to claim 1 Its characteristics are: In the equipment maintenance unit, the CNN model analyzes the image of the reactor inner wall (resolution ≥ 200dpi), with a corrosion identification accuracy of ±0.1mm; The LSTM model analyzes the vibration spectrum (0-5kHz), and the jam warning accuracy is ≥92%.

4. The system according to claim 1 Its characteristics are: The energy scheduling engine monitors waste heat temperature in real time through the Internet of Things, and the response time for matching the electrolyzer preheating demand is ≤30 seconds; The alkali solution is regenerated by a multi-effect evaporator with an evaporation efficiency of ≥85% and a concentration ratio of 1.2-1.

8.

5. The system according to claim 1 Its characteristics are: The data-driven process improvement platform integrates a federated learning framework to dynamically adapt to fluctuations in ore composition; The knowledge graph integrates 100,000+ batches of historical data and recommends the optimal process combination with a confidence level of ≥95%.

Images