Bauxite flotation aluminum-silicon ratio optimization system and method based on multi-modal perception and intelligent decision

Abstract

The application discloses an alumina ore flotation aluminum-silicon ratio optimization system and method based on multi-modal perception and intelligent decision-making. A multi-modal perception module is used to obtain multi-source data related to the aluminum-silicon ratio in the flotation process. The multi-source data is subjected to feature extraction and fusion through a feature fusion and soft measurement module, and an aluminum-silicon ratio soft measurement model is constructed to output an estimated value of the aluminum-silicon ratio. Through an intelligent decision-making module, based on the estimated value of the aluminum-silicon ratio and the flotation working condition state, a control decision for the flotation process is generated using reinforcement learning or other self-learning algorithms. An execution control module is used to convert the control decision into an adjustment instruction for the operating parameters of the flotation equipment to achieve dynamic optimization and stable control of the aluminum-silicon ratio. The application overcomes the problems of control lag, reliance on manual experience and difficulty in achieving dynamic optimization in the prior art, realizes real-time closed-loop optimization control of the flotation process, and thus, under the premise of ensuring aluminum recovery, the aluminum-silicon ratio of the concentrate is stably improved and the consumption of reagents is reduced.

Description

Technical Field

[0001] This invention relates to the field of automated control technology for mineral processing and beneficiation, specifically to a bauxite flotation aluminum-silicon ratio optimization system and method based on multimodal perception and intelligent decision-making. Background Technology

[0002] Bauxite is the main raw material for the production of alumina and electrolytic aluminum, and its quality directly affects the energy consumption, alkali consumption, and economic benefits of subsequent smelting processes. Industrially, the aluminum-silicon ratio (A / S) in the concentrate is usually used as the core indicator for evaluating the quality of bauxite; the higher the A / S ratio, the lower the alumina production cost.

[0003] Flotation is a core mineral processing technology that utilizes the differences in the physicochemical properties of mineral surfaces to achieve separation and enrichment. It is widely used in the separation of metal ores such as copper, iron, and molybdenum, as well as coal. The color, texture, stability, and fracturing behavior of the froth layer in the flotation cell directly reflect the pulp concentration, reagent effectiveness, and separation efficiency, and are key characteristics of the process status. Statistics show that the flotation process accounts for 40%-60% of the total energy consumption in mineral processing, and every 1% increase in recovery rate can bring significant economic benefits. Therefore, precise monitoring and optimized control of the flotation process have significant industrial value.

[0004] Existing flotation technologies and their limitations: (1) Limited perceptual dimensions and missing information: Current flotation monitoring mainly relies on manual observation or single visual image analysis. Manual observation is affected by subjective experience and fatigue, making it difficult to guarantee 24-hour consistency. Although machine vision systems based on RGB cameras can extract surface features such as color and size, they have the following limitations: high light sensitivity, color feature drift caused by fluctuations in industrial lighting, and image overexposure due to specular reflection on the foam surface; lack of three-dimensional structure, as two-dimensional images cannot capture foam layer thickness, bubble three-dimensional morphology, and spatial distribution; and insufficient compositional information, as visual features cannot directly reflect the chemical composition and grade of the minerals loaded on the foam. Some studies have introduced near-infrared spectroscopy or conductivity sensors, but multi-source data are collected independently and lack fusion, forming "information silos."

[0005] (2) Insufficient dynamic forecasting capability and lagging regulation: Traditional flotation control uses PID control or rule-based expert systems, which are essentially passive modes of "current state - immediate response": significant time lag effect, with a 5-15 minute delay from changes in foam state to indicator detection, making it difficult for PID control to cope with large lag processes; lack of trend prediction, unable to predict dynamic evolution trends such as abnormal increases in foam viscosity and sudden collapse, resulting in control actions lagging behind changes in operating conditions; slow anomaly recovery, triggering alarms only after an anomaly occurs, missing the optimal intervention window and causing recovery rate losses.

[0006] (3) Coarse-grained multimodal feature fusion: In recent years, multimodal fusion technology has been gradually applied to flotation monitoring, but existing methods mostly adopt early concatenation or late decision fusion: semantic conflict, large dimensional differences between visual texture features and spectral features (the former is tens of dimensions, the latter is hundreds of dimensions), simple concatenation leads to distortion of feature space; fixed weights, preset weights cannot adapt to changes in the signal-to-noise ratio of each modality under different working conditions, visual weights should be automatically reduced when the illumination is insufficient; redundant information, without considering the complementarity and correlation between modalities, introduces noise features to reduce recognition accuracy.

[0007] (4) Weak self-adaptability and poor adaptability to working conditions: The flotation process is affected by multiple factors such as ore properties (hardness, particle size, degree of mudification), equipment aging, and ambient temperature: threshold rigidity, fixed threshold strategy generates a large number of false alarms or missed alarms when ore grade fluctuates; strategy rigidity, models trained based on historical data are difficult to adapt to new working conditions, and frequent shutdowns for recalibration are required; insufficient knowledge utilization, failure to effectively integrate expert experience and historical success cases, and poor decision interpretability.

[0008] In existing industrial production, for low- to medium-grade bauxite, direct or reverse flotation processes are typically used to reduce silicon content and increase the aluminum-to-silicon ratio. However, the bauxite flotation process has the following characteristics: 1. The properties of the raw ore fluctuate frequently, and its mineral composition and intercalation characteristics are complex; 2. The flotation process exhibits strong nonlinearity, strong coupling, and time-varying characteristics; 3. The reagent formulation and process parameters have a significant impact on flotation indicators.

[0009] Currently, the main control methods for bauxite flotation processes in industrial settings include: open-loop control based on a fixed reagent regime, manual adjustment relying on operator experience, and post-processing feedback control based on offline laboratory results. These control methods have the following shortcomings: Offline testing has a long turnaround time and cannot provide real-time feedback; It relies heavily on human experience and has poor stability and reproducibility. Traditional control methods such as PID control are difficult to handle complex nonlinear operating conditions; Excessive or insufficient addition of reagents can lead to waste of resources and fluctuations in performance indicators. Summary of the Invention

[0010] The purpose of this invention is to overcome the problems of lagging control, reliance on manual experience, and difficulty in achieving dynamic optimization in the existing bauxite flotation process. It provides a bauxite flotation aluminum-silicon ratio optimization system and method based on multimodal perception and intelligent decision-making. The system achieves real-time closed-loop optimization control of the flotation process, thereby steadily increasing the concentrate aluminum-silicon ratio and reducing reagent consumption while ensuring aluminum recovery. The method also aims to achieve real-time closed-loop optimization control of the flotation process, thereby steadily increasing the concentrate aluminum-silicon ratio and reducing reagent consumption while ensuring aluminum recovery.

[0011] This invention is achieved through the following technical solution: a bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making, comprising: A multimodal sensing module is used to acquire multi-source data related to the aluminum-silicon ratio during the flotation process. The multi-source data includes one or more of the following: slurry image data, process parameter data, and online or quasi-online detection data. The feature fusion and soft measurement module is used to extract and fuse features from the multi-source data, construct a soft measurement model for the aluminum-silicon ratio, and output an estimated value of the aluminum-silicon ratio; the estimated value is a real-time estimate or a near-real-time estimate. The intelligent decision-making module is used to generate control decisions for the flotation process based on the estimated aluminum-silicon ratio and the flotation operating conditions, using reinforcement learning or other self-learning algorithms. The execution control module is used to convert the control decisions into adjustment instructions for the operating parameters of the flotation equipment, so as to achieve dynamic optimization and stable control of the aluminum-silicon ratio.

[0012] Preferably, this bauxite flotation alumina-silica ratio optimization system can also be divided into five main modules, including: The multimodal sensing module is used to collect image or video data of flotation foam and physicochemical parameter data of slurry in real time during the bauxite flotation process. The physicochemical parameter data includes one or more of pH value, potential, concentration, flow rate and temperature. The feature extraction and fusion module is used to extract foam morphology features from the image or video data of the flotation foam, and perform time alignment and feature fusion with the physical and chemical parameter data to generate a multi-dimensional feature vector characterizing the current flotation condition. A soft measurement module for aluminum-silicon ratio is used to predict the aluminum-silicon ratio and / or aluminum recovery rate of the concentrate in real time based on the multidimensional feature vector. The intelligent decision-making module includes a reinforcement learning-based decision model, which takes the multidimensional feature vector and the prediction results of the aluminum-silicon ratio soft measurement model as the current state input and outputs optimization adjustment instructions for key process parameters of the flotation process. The execution control module is used to automatically adjust at least one controllable process parameter during the flotation process according to the optimization adjustment instructions.

[0013] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following configuration is adopted: the multimodal perception module includes at least two of the following: an industrial camera device, a flow sensor, a liquid level sensor, a pH sensor, and a reagent addition detection device, as well as a data acquisition interface, for synchronously acquiring multidimensional information of the flotation process; the slurry image data includes at least one of the following: foam morphology, foam diameter distribution, foam color, and motion characteristics.

[0014] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making as described in this invention, the following configuration is adopted: the feature fusion and soft measurement module uses a machine learning model, a deep learning model, or a combination of both to establish a mapping relationship between the aluminum-silicon ratio and multi-source features; the aluminum-silicon ratio soft measurement model is updated through joint training of historical production data and online data.

[0015] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making as described in this invention, the following configuration is specifically adopted: the intelligent decision-making module uses a reinforcement learning algorithm to construct a decision model with the aluminum-silicon ratio as the objective function, and adaptively updates the strategy based on environmental feedback; the reinforcement learning algorithm is (including but not limited to) Q-learning, deep Q-network, policy gradient method, or an improved form of the three; the control decision includes adjusting at least one of the following: reagent addition amount, pulp concentration, aeration amount, stirring intensity, and flotation time.

[0016] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making as described in this invention, the following configuration is adopted: the execution control module is connected to the flotation equipment through a PLC, DCS, or industrial control network to achieve closed-loop control.

[0017] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making as described in this invention, the following configuration is adopted: a digital twin module is also included to construct a virtual model of the flotation process, thereby assisting the intelligent decision-making module in strategy verification and optimization.

[0018] The bauxite flotation alumina-silicon ratio optimization method based on multimodal perception and intelligent decision-making is implemented through the aforementioned bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making, and includes the following steps: 1) During the bauxite flotation process, multimodal data related to the aluminum-silicon ratio are collected through a multimodal sensing module; 2) The feature fusion and soft measurement module is used to extract and fuse features from the multimodal data, construct an aluminum-silicon ratio soft measurement model, and obtain an estimated value of the aluminum-silicon ratio; 3) Based on the estimated aluminum-silicon ratio and the flotation operating conditions, the intelligent decision-making module generates a flotation control strategy through an intelligent decision-making algorithm; 4) The execution control module adjusts the operating parameters of the flotation equipment according to the control strategy to achieve dynamic optimization of the aluminum-silicon ratio.

[0019] The preferred method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making can also be divided into the following steps: S1: During the bauxite flotation process, the multimodal sensing module collects images or video data of the flotation foam and physicochemical parameters of the slurry in real time. S2: The feature extraction and fusion module extracts foam morphology features from the image or video data and fuses them with the physical and chemical parameter data to generate a flotation condition feature vector. S3: The aluminum-silicon ratio soft measurement module predicts the concentrate aluminum-silicon ratio and / or aluminum recovery rate in real time based on the flotation condition feature vector; S4: The flotation condition feature vector and prediction results are used as state inputs to the reinforcement learning decision model to output an optimized adjustment scheme for the flotation process parameters; S5: The execution control module automatically adjusts the process parameters during the flotation process according to the optimization adjustment scheme; S6: Based on the feedback results of the adjusted flotation process, the reinforcement learning decision model is updated through online learning or simulation using a digital twin module to form closed-loop optimization control.

[0020] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following setting is adopted: In step 2), the aluminum-silicon ratio soft measurement model is updated through online learning or incremental learning.

[0021] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following configuration is adopted: In step 3), the intelligent decision-making algorithm is a reinforcement learning algorithm, and the stability of the aluminum-silicon ratio and the retention of the target interval are used as components of the reward function.

[0022] To further improve the implementation of the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following settings are adopted: In step 4), the operating parameters include at least one of the following: reagent addition amount, aeration amount, and slurry concentration.

[0023] Compared with the prior art, the present invention has the following advantages and beneficial effects: This invention provides a technical solution that enables real-time sensing, intelligent decision-making, and closed-loop control of bauxite flotation conditions, thereby stabilizing and improving the aluminum-silicon ratio of the concentrate, reducing reagent consumption, and increasing production efficiency.

[0024] This invention achieves comprehensive and real-time characterization of flotation conditions through multimodal sensing.

[0025] This invention replaces traditional offline testing with a soft measurement model of aluminum-silicon ratio, enabling minute-level indicator feedback.

[0026] This invention achieves dynamic global optimization of flotation parameters through a reinforcement learning decision model.

[0027] This invention improves the stability of the aluminum-silicon ratio and reduces reagent consumption while ensuring aluminum recovery rate.

[0028] This invention reduces reliance on human experience and improves the automation and intelligence of the flotation process.

[0029] Compared with existing single-modal monitoring, this invention improves the accuracy of foam state recognition by 18%-25%, extends the early warning time of anomalies to 3-5 minutes, and shortens the control response delay to less than 30 seconds; the adaptive weight learning mechanism enables the system to maintain robustness in scenarios with fluctuations in ore properties, and reduces the false alarm rate by 22%; the weak labeling scheme reduces the workload of manual labeling by more than 50%, and the closed-loop intelligent control reduces the excessive consumption of reagents by 10%-15%.

[0030] This invention can stably increase the aluminum-silicon ratio of concentrate under fluctuating raw ore properties, reduce reagent consumption, and minimize human intervention. It is applicable to the industrial production process of bauxite direct flotation and reverse flotation.

[0031] This invention is applicable to scenarios such as metal ore flotation, coal preparation, and monitoring of complex industrial processes. Detailed Implementation

[0032] The present invention will be further described in detail below with reference to embodiments, but the implementation of the present invention is not limited thereto.

[0033] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present invention.

[0034] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0035] Example 1: This invention designs a bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making. It overcomes the problems of lagging control, reliance on manual experience, and difficulty in achieving dynamic optimization in existing bauxite flotation processes. The system achieves real-time closed-loop optimization control of the flotation process, thereby steadily increasing the concentrate aluminum-silicon ratio and reducing reagent consumption while ensuring aluminum recovery. The system includes: A multimodal sensing module is used to acquire multi-source data related to the aluminum-silicon ratio during the flotation process. The multi-source data includes one or more of the following: slurry image data, process parameter data, and online or quasi-online detection data. The feature fusion and soft measurement module is used to extract and fuse features from the multi-source data, construct a soft measurement model for the aluminum-silicon ratio, and output an estimated value of the aluminum-silicon ratio; the estimated value is a real-time estimate or a near-real-time estimate. The intelligent decision-making module is used to generate control decisions for the flotation process based on the estimated aluminum-silicon ratio and the flotation operating conditions, using reinforcement learning or other self-learning algorithms. The execution control module is used to convert the control decisions into adjustment instructions for the operating parameters of the flotation equipment, so as to achieve dynamic optimization and stable control of the aluminum-silicon ratio.

[0036] Preferably, this bauxite flotation alumina-silica ratio optimization system can also be divided into five main modules, including: The multimodal sensing module is used to collect image or video data of flotation foam and physicochemical parameter data of slurry in real time during the bauxite flotation process. The physicochemical parameter data includes one or more of pH value, potential, concentration, flow rate and temperature. The feature extraction and fusion module is used to extract foam morphology features from the image or video data of the flotation foam, and perform time alignment and feature fusion with the physical and chemical parameter data to generate a multi-dimensional feature vector characterizing the current flotation condition. A soft measurement module for aluminum-silicon ratio is used to predict the aluminum-silicon ratio and / or aluminum recovery rate of the concentrate in real time based on the multidimensional feature vector. The intelligent decision-making module includes a reinforcement learning-based decision model, which takes the multidimensional feature vector and the prediction results of the aluminum-silicon ratio soft measurement model as the current state input and outputs optimization adjustment instructions for key process parameters of the flotation process. The execution control module is used to automatically adjust at least one controllable process parameter during the flotation process according to the optimization adjustment instructions.

[0037] Example 2: This embodiment is a further optimization based on the above embodiments. The parts that are the same as those in the foregoing technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following configuration is adopted: The multimodal perception module includes at least two of the following: an industrial camera device, a flow sensor, a liquid level sensor, a pH sensor, and a reagent addition detection device, as well as a data acquisition interface, for synchronously acquiring multidimensional information of the flotation process; The slurry image data includes at least one of the following: foam morphology, foam diameter distribution, foam color, and motion characteristics.

[0038] Example 3: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the foregoing technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following settings are specifically adopted: the feature fusion and soft measurement module uses a machine learning model, a deep learning model, or a combination of both to establish a mapping relationship between the aluminum-silicon ratio and multi-source features; the aluminum-silicon ratio soft measurement model is updated by jointly training historical production data and online data.

[0039] Example 4: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the foregoing technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following settings are specifically adopted: The intelligent decision-making module uses a reinforcement learning algorithm to construct a decision model with aluminum-silicon ratio as the objective function, and adaptively updates the strategy according to environmental feedback; The reinforcement learning algorithm is (including but not limited to) Q-learning, deep Q-network, policy gradient method, or an improved form of the three; The control decision includes adjusting at least one of the following: reagent addition amount, pulp concentration, aeration amount, stirring intensity, and flotation time.

[0040] Example 5: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the foregoing technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following setting is adopted: the execution control module is connected to the flotation equipment through a PLC, DCS or industrial control network to realize closed-loop control.

[0041] Example 6: This embodiment is a further optimization based on any of the above embodiments. The parts that are the same as those in the foregoing technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making described in this invention, the following setting is adopted: It also includes a digital twin module, which is used to construct a virtual model of the flotation process to assist the intelligent decision-making module in strategy verification and optimization.

[0042] Example 7: The bauxite flotation alumina-silicon ratio optimization method based on multimodal perception and intelligent decision-making is implemented through the aforementioned bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making, and includes the following steps: 1) During the bauxite flotation process, multimodal data related to the aluminum-silicon ratio are collected through a multimodal sensing module; 2) The feature fusion and soft measurement module is used to extract and fuse features from the multimodal data, construct an aluminum-silicon ratio soft measurement model, and obtain an estimated value of the aluminum-silicon ratio; 3) Based on the estimated aluminum-silicon ratio and the flotation operating conditions, the intelligent decision-making module generates a flotation control strategy through an intelligent decision-making algorithm; 4) The execution control module adjusts the operating parameters of the flotation equipment according to the control strategy to achieve dynamic optimization of the aluminum-silicon ratio.

[0043] The preferred method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making can also be divided into the following steps: S1: During the bauxite flotation process, the multimodal sensing module collects images or video data of the flotation foam and physicochemical parameters of the slurry in real time. S2: The feature extraction and fusion module extracts foam morphology features from the image or video data and fuses them with the physical and chemical parameter data to generate a flotation condition feature vector. S3: The aluminum-silicon ratio soft measurement module predicts the concentrate aluminum-silicon ratio and / or aluminum recovery rate in real time based on the flotation condition feature vector; S4: The flotation condition feature vector and prediction results are used as state inputs to the reinforcement learning decision model to output an optimized adjustment scheme for the flotation process parameters; S5: The execution control module automatically adjusts the process parameters during the flotation process according to the optimization adjustment scheme; S6: Based on the feedback results of the adjusted flotation process, the reinforcement learning decision model is updated through online learning or simulation using a digital twin module to form closed-loop optimization control.

[0044] Example 8: This embodiment is a further optimization based on the above embodiment. The parts that are the same as those in the aforementioned technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following setting is adopted: In step 2), the aluminum-silicon ratio soft measurement model is updated through online learning or incremental learning.

[0045] Example 9: This embodiment is a further optimization based on embodiment 7 or 8. The parts that are the same as those in the aforementioned technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following setting is adopted: In step 3), the intelligent decision-making algorithm is a reinforcement learning algorithm, and the aluminum-silicon ratio stability and target interval preservation degree are used as components of the reward function.

[0046] Example 10: This embodiment is a further optimization based on embodiment 7, 8 or 9. The parts that are the same as those in the aforementioned technical solutions will not be repeated here. In order to better realize the bauxite flotation aluminum-silicon ratio optimization method based on multimodal perception and intelligent decision-making described in this invention, the following setting method is adopted: In step 4), the operating parameters include at least one of the following: reagent addition amount, aeration amount and slurry concentration.

[0047] Example 11: Implementation of a bauxite flotation alumina-silicon ratio optimization system based on multimodal sensing and intelligent decision-making in reverse flotation process: On a bauxite reverse flotation production line, industrial cameras (industrial video devices) are installed at the flotation cells, concentrate stream, and tailings stream to collect real-time flotation foam image data. At the same time, online sensors (pH sensor, concentration sensor, flow sensor, temperature sensor, etc.) are used to collect slurry pH value, concentration, flow rate, and temperature data.

[0048] The foam image data is processed by a feature extraction and fusion module to extract foam size, color, and texture features, which are then fused with physicochemical parameter data to form a flotation condition feature vector. This flotation condition feature vector is input to the aluminum-silicon ratio soft measurement module, which outputs a real-time prediction result of the concentrate's aluminum-silicon ratio.

[0049] The reinforcement learning decision model takes the real-time prediction results and flotation conditions as state inputs, outputs adjustment instructions for the amount of collector and inhibitor added, and automatically adjusts the dosing pump frequency through the PLC system.

[0050] The optimization system continuously acquires online prediction results of the aluminum-silicon ratio of the concentrate and actual test feedback results during operation, and updates the reinforcement learning decision model online to achieve adaptive optimization control of the flotation process under the condition of fluctuation in the properties of the raw ore.

[0051] Example 12: Implementation of a bauxite flotation alumina-silicon ratio optimization system based on multimodal sensing and intelligent decision-making in the forward flotation process: On a bauxite flotation production line, industrial cameras (industrial video recording devices) are installed in the roughing and cleaning flotation cell areas for high-silica bauxite raw materials to collect data on slurry pH, potential, concentration and flow rate.

[0052] The optimization system constructs a flotation condition feature vector based on real-time data acquired by the multimodal sensing module, and predicts the current trend of the aluminum-silicon ratio change in the concentrate through the aluminum-silicon ratio soft measurement module. When the prediction result shows a decreasing trend in the aluminum-silicon ratio, the reinforcement learning decision model automatically outputs a joint adjustment command for the amount of inhibitor and frother added, and realizes dynamic optimization control of the flotation process through the execution control module.

[0053] Example 13: This embodiment provides an implementation method for digital twin and model training: Before deploying a bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making, a digital twin model (digital twin module) of the bauxite flotation process is constructed to pre-train the reinforcement learning decision model, thereby reducing the risk of trial and error in the field. After pre-training, the digital twin model is deployed to the actual production system and continuously fine-tuned based on online operating data.

[0054] Through the above embodiments, the present invention can achieve real-time closed-loop optimization control of the flotation process under fluctuating raw ore properties, so that the aluminum-silicon ratio of the concentrate remains stable within the target range for a long time, and significantly reduces reagent consumption and the frequency of manual intervention.

[0055] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention shall fall within the protection scope of the present invention.

Claims

1. A bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making, characterized in that: include: A multimodal sensing module is used to acquire multi-source data related to the aluminum-silicon ratio during the flotation process. The multi-source data includes one or more of the following: slurry image data, process parameter data, and online or quasi-online detection data. The feature fusion and soft measurement module is used to extract and fuse features from the multi-source data, construct a soft measurement model for the aluminum-silicon ratio, and output an estimated value of the aluminum-silicon ratio; the estimated value is a real-time estimate or a near-real-time estimate. The intelligent decision-making module is used to generate control decisions for the flotation process based on the estimated aluminum-silicon ratio and the flotation operating conditions, using reinforcement learning or other self-learning algorithms. The execution control module is used to convert the control decisions into adjustment instructions for the operating parameters of the flotation equipment, so as to achieve dynamic optimization and stable control of the aluminum-silicon ratio.

2. The bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making according to claim 1, characterized in that: The multimodal sensing module includes at least two of the following: an industrial camera device, a flow sensor, a liquid level sensor, a pH sensor, and a reagent addition detection device, as well as a data acquisition interface, for synchronously acquiring multidimensional information of the flotation process; the slurry image data includes at least one of the following: foam morphology, foam diameter distribution, foam color, and motion characteristics.

3. The bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making according to claim 1, characterized in that: The feature fusion and soft measurement module uses machine learning models, deep learning models, or a combination of both to establish a mapping relationship between the aluminum-silicon ratio and multi-source features; the aluminum-silicon ratio soft measurement model is updated through joint training of historical production data and online data.

4. The bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making according to claim 1, characterized in that: The intelligent decision-making module uses a reinforcement learning algorithm to construct a decision model with the aluminum-silicon ratio as the objective function, and adaptively updates the strategy based on environmental feedback; the reinforcement learning algorithm is Q-learning, deep Q-network, policy gradient method, or an improved form of the three; the control decision includes adjusting at least one of the following: reagent addition amount, slurry concentration, aeration amount, stirring intensity, and flotation time.

5. The bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making according to claim 1, characterized in that: The execution control module is connected to the flotation equipment via PLC, DCS, or industrial control network to achieve closed-loop control.

6. The bauxite flotation alumina-silicon ratio optimization system based on multimodal perception and intelligent decision-making according to any one of claims 1 to 5, characterized in that: It also includes a digital twin module, used to construct a virtual model of the flotation process to assist the intelligent decision-making module in strategy verification and optimization.

7. A method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making, characterized in that: This is achieved through the bauxite flotation aluminum-silicon ratio optimization system based on multimodal perception and intelligent decision-making as described in any one of claims 1 to 6, comprising the following steps: 1) Collect multimodal data related to the aluminum-silicon ratio during the flotation process using a multimodal sensing module; 2) The feature fusion and soft measurement module is used to extract and fuse features from the multimodal data, construct an aluminum-silicon ratio soft measurement model, and obtain an estimated value of the aluminum-silicon ratio; 3) Based on the estimated aluminum-silicon ratio and the flotation operating conditions, the intelligent decision-making module generates a flotation control strategy through an intelligent decision-making algorithm; 4) The execution control module adjusts the operating parameters of the flotation equipment according to the control strategy to achieve dynamic optimization of the aluminum-silicon ratio.

8. The method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making according to claim 7, characterized in that: In step 2), the aluminum-silicon ratio soft measurement model is updated through online learning or incremental learning.

9. The method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making according to claim 7, characterized in that: In step 3), the intelligent decision-making algorithm is a reinforcement learning algorithm, and the aluminum-silicon ratio stability and target interval preservation degree are used as components of the reward function.

10. The method for optimizing the aluminum-silicon ratio in bauxite flotation based on multimodal perception and intelligent decision-making according to claim 7, characterized in that: In step 4), the operating parameters include at least one of the following: reagent addition amount, aeration amount, and slurry concentration.