Gold ore flotation full-process intelligent monitoring and optimal control method and system

Abstract

The invention relates to the technical field of gold ore flotation full-process monitoring and control, and discloses a gold ore flotation full-process intelligent monitoring and optimal control method and system. The method comprises the following steps: collecting flotation process parameters, and carrying out abnormal value detection and grading treatment; the graded data are fused, and a digital twinborn model is established; designing a negotiation control framework based on the model, and adjusting parameters by using a contract network protocol; and constructing a soft measurement model and an optimization strategy, and optimizing control parameters through a self-learning mechanism. The technical problems that global optimization is difficult to achieve due to unit independent control in an existing gold ore flotation control system, control decision is inaccurate due to simple data processing, key quality indexes are difficult to estimate due to lack of soft measurement means, and adaptability is poor due to limited system learning ability are solved.

Description

Technical Field

[0001] This application relates to the technical field of full-process monitoring and control of gold flotation, and particularly to an intelligent monitoring and optimization control method and system for the full process of gold flotation. Background Art

[0002] The gold flotation process is an important link in gold ore dressing, and its process flow includes multiple stages such as crushing, grinding and classification, flotation, and concentrate treatment. Traditional gold flotation control systems usually adopt unit-level PID control methods to adjust parameters for individual control points, such as pH value controllers, pulp concentration controllers, and reagent addition amount controllers. With the development of automation technology, intelligent control methods based on fuzzy control, expert systems, and neural networks have been gradually introduced in the gold flotation process, improving the control accuracy. Some advanced flotation plants also adopt foam image analysis technology based on visual recognition to indirectly evaluate the flotation effect by monitoring the characteristics of flotation foam, and begin to try to apply digital twin technology to conduct virtual simulation and optimization of the gold flotation process.

[0003] However, existing systems mostly adopt a unit-independent control mode, lacking a full-process collaborative control mechanism, insufficient information interaction between control units, and it is difficult to achieve global optimization. Secondly, the existing systems have a simple way of processing process parameters, low data quality, and limited ability to handle outliers and noise, affecting the accuracy of control decisions. Thirdly, the existing systems have insufficient adaptability to complex working conditions. When the ore properties change, the equipment ages, etc., the control effect significantly decreases and requires frequent manual intervention. Fourthly, the existing systems lack effective soft measurement means and it is difficult to accurately estimate key quality indicators such as concentrate grade and recovery rate, resulting in unclear control objectives. Finally, the existing systems have limited learning ability, cannot accumulate optimization experience from historical operation data, and the control strategy lacks a self-improving mechanism.

[0004] In traditional control systems, the internal structure design of each functional unit agent is not perfect enough, lacking effective data processing modules and decision-making reasoning modules, resulting in the inability to accurately perceive the environmental state and make reasonable decisions; at the same time, the loading methods of expert knowledge bases, control algorithm libraries, and fuzzy rule libraries in the existing technology are not flexible enough and it is difficult to adapt to the requirements of different working conditions; in addition, the boundary between the local optimization control layer and the collaborative optimization control layer is not clear, and the two-layer control logics cross and are chaotic, causing conflicts between local goals and global goals. Summary of the Invention

[0005] The present application provides an intelligent monitoring and optimization control method and system for the entire gold flotation process, which solves the technical problems in the existing gold flotation control system, including difficult global optimization due to unit independent control, inaccurate control decisions due to simple data processing, difficult estimation of key quality indicators due to lack of soft measurement means, and poor adaptability due to limited system learning ability.

[0006] In the first aspect, the present application provides an intelligent monitoring and optimization control method for the entire gold flotation process. The intelligent monitoring and optimization control method for the entire gold flotation process includes: collecting process parameter data in the entire gold flotation process through a distributed sensor network, detecting and classifying outliers in the collected data to obtain classified data; constructing a multi-source heterogeneous data fusion engine based on the classified data, standardizing and multi-level fusing the data to obtain a digital twin model for the entire gold flotation process; designing a multi-agent negotiation control architecture based on the digital twin model, and using a negotiation mechanism based on the contract net protocol to adjust control parameters to obtain a global optimization control scheme; constructing an adaptive soft measurement model and a multi-objective optimization control strategy according to the global optimization control scheme, and realizing parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

[0007] In the second aspect, the present application provides an intelligent monitoring and optimization control system for the entire gold flotation process. The intelligent monitoring and optimization control system for the entire gold flotation process includes: a classification module, configured to collect process parameter data in the entire gold flotation process through a distributed sensor network, detect and classify outliers in the collected data to obtain classified data; a fusion module, configured to construct a multi-source heterogeneous data fusion engine based on the classified data, standardize and multi-level fuse the data to obtain a digital twin model for the entire gold flotation process; an adjustment module, configured to design a multi-agent negotiation control architecture based on the digital twin model, and use a negotiation mechanism based on the contract net protocol to adjust control parameters to obtain a global optimization control scheme; a construction module, configured to construct an adaptive soft measurement model and a multi-objective optimization control strategy according to the global optimization control scheme, and realize parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

[0008] In the third aspect, there is provided an intelligent monitoring and optimization control device for the entire gold flotation process, including: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the intelligent monitoring and optimization control device for the entire gold flotation process executes the above-mentioned intelligent monitoring and optimization control method for the entire gold flotation process.

[0009] Fourthly, a computer-readable storage medium is provided. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is enabled to execute the above-mentioned intelligent monitoring and optimization control method for the whole process of gold ore flotation.

[0010] In the technical solution provided by this application, process parameter data is collected through a distributed sensor network, and outlier detection and hierarchical processing are performed on the data, effectively improving the data quality and solving the problems of simple data processing method and low quality in traditional systems; a multi-source heterogeneous data fusion engine is constructed based on the hierarchically processed data, and the data is standardized and multi-level fused to obtain a digital twin model for the whole process of gold ore flotation, realizing the real-time mapping between the physical flotation system and the virtual system, and providing an accurate information basis for subsequent decision-making; a multi-agent negotiation control architecture is designed based on the digital twin model, and a negotiation mechanism based on the contract net protocol is adopted to adjust control parameters, effectively solving the problems of insufficient information interaction among control units and difficulty in achieving global optimization in traditional systems. The negotiation mechanism significantly improves the collaborative performance of the overall system, enabling each functional unit to jointly achieve the global optimal goal while maintaining local stability; an adaptive soft-sensor model and a multi-objective optimization control strategy are constructed according to the global optimization control scheme, and parameter optimization is realized through a self-learning and knowledge accumulation mechanism, which not only solves the problem of the lack of effective soft-sensor means in traditional systems but also overcomes the defect of the lack of self-improving mechanism in control strategies. In particular, the contract net protocol applied in the negotiation control architecture, as a multi-agent negotiation mechanism, fully considers the characteristics and mutual influences of each functional unit in the flotation process. Through a structured message interaction and task decomposition negotiation process, it effectively balances multiple conflicting goals such as grade, recovery rate, and energy consumption. The introduction of the negotiation mechanism enables the system to have distributed decision-making capabilities, significantly reducing the computational burden and improving the system response speed; while the adaptive soft-sensor model realizes the accurate estimation of parameters that are difficult to directly measure through multi-model integration and transfer learning methods, and the self-learning mechanism enables the system to accumulate optimization experience from historical operation data, improving the adaptability to complex working conditions, and reducing energy consumption and reagent consumption while increasing the concentrate grade and recovery rate. Description of the Drawings

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for description in the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0012] Figure 1 It is a schematic diagram of an embodiment of the intelligent monitoring and optimization control method for the whole process of gold ore flotation in an embodiment of this application; Figure 2It is a schematic diagram of an embodiment of the intelligent monitoring and optimization control system for the entire gold flotation process in the embodiments of the present application; Figure 3 It is a structural schematic block diagram of the intelligent monitoring and optimization control device for the entire gold flotation process in the embodiments of the present invention. Specific embodiments

[0013] The embodiments of the present application provide a method and system for intelligent monitoring and optimization control of the entire gold flotation process. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims, and above-mentioned drawings of the present application are used to distinguish similar objects and do not necessarily need to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products, or devices.

[0014] For ease of understanding, the specific process of the embodiments of the present application is described below. Please refer to Figure 1 An embodiment of the intelligent monitoring and optimization control method for the entire gold flotation process in the embodiments of the present application includes: Step S101: Collect process parameter data in the entire gold flotation process through a distributed sensor network, perform outlier detection and grading processing on the collected data, and obtain graded processing data; Step S102: Build a multi-source heterogeneous data fusion engine based on the graded processing data, perform standardization and multi-level fusion on the data, and obtain a digital twin model of the entire gold flotation process; Step S103: Design a multi-agent negotiation control architecture based on the digital twin model, adopt a negotiation mechanism based on the contract net protocol to adjust control parameters, and obtain a global optimization control scheme; Step S104: Build an adaptive soft sensor model and a multi-objective optimization control strategy according to the global optimization control scheme, and realize parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

[0015] It can be understood that the execution subject of the present application can be an intelligent monitoring and optimization control system for the entire gold flotation process, or a terminal or a server. Specifically, it is not limited here. The embodiments of the present application take the server as the execution subject as an example for illustration.

[0016] Specifically, process parameter data in the entire process of gold ore flotation is collected through a distributed sensor network, and outlier detection and classification processing are performed on the data collected by the sensor network. In specific implementation, process parameter data such as pH value, pulp concentration, grinding fineness, and foam characteristics are classified according to the change rate. For rapidly changing parameters such as pH value and aeration volume, high-frequency sampling is performed at 500 ms / time, and for slowly changing parameters such as pulp concentration, low-frequency sampling is performed at 5 s / time. This classification sampling method avoids the problems of data redundancy or information loss caused by traditional fixed-frequency sampling. The collected data is subjected to outlier detection by the improved Z-score method. By calculating the standardized distance between the parameter value and the mean value of the sliding window, when the distance exceeds the threshold of 4.5, the data point is marked as an outlier. For example, during a certain flotation process, the pH value suddenly jumps from the normal 8.2 to 11.5, and the Z-score calculation result is 5.3, exceeding the threshold of 4.5, and this data point is marked as an outlier. Subsequently, the detected outliers are replaced. The replacement value is calculated by taking the 20 valid data points before and after and using the weighted average method, and the weight coefficient is determined by the exponential decay function, so that the data closer to the outlier has a smaller weight. The data is subjected to noise elimination through the fourth-order Kalman filter algorithm, and the filter parameter matrix is adjusted according to different parameter characteristics. The filtered data is classified through the data importance scoring function. The scoring function comprehensively considers three factors: the influence degree of the parameter on the process, the change range, and the fluctuation frequency. Finally, the parameter data with a scoring value greater than 0.8 is classified as level one, the parameter data with a scoring value between 0.5 and 0.8 is classified as level two, and the parameter data with a scoring value lower than 0.5 is classified as level three, forming hierarchical processed data with priorities.

[0017] Build a multi-source heterogeneous data fusion engine based on hierarchical processed data to standardize and multi-level fuse the data. First, perform unit conversion on the hierarchical processed data, convert the grinding fineness data from μm unit to mm unit, convert the pulp concentration data from percentage form to decimal form, and convert the potential data to the standard potential unit to ensure unit uniformity. Perform normalization processing on the data with unified units, adjust the numerical ranges of different parameters to the 0-1 interval to eliminate the influence of dimensional differences on subsequent analysis. The normalized data undergoes time alignment processing, align the data with different sampling frequencies according to the system clock reference point, and interpolate and fill in the sampling intervals to form spatio-temporally consistent standardized data. Perform weighted average calculation on the data from multiple sensors measuring the same parameter in the standardized data, and the weight allocation is based on the historical accuracy of each sensor to obtain the data-level fusion result. The data-level fusion result is analyzed through correlation relationship analysis to identify parameter combinations with strong correlations, such as the correlation coefficient between the pulp pH value and foam stability reaching 0.87, extract the key features characterizing the flotation process state to obtain the feature-level fusion result. Combine the feature-level fusion result with the flotation process expert knowledge to establish the corresponding relationship between the physical flotation system and the virtual system to obtain the digital twin model of the entire gold flotation process.

[0018] Design a multi-agent negotiation control architecture based on a digital twin model, and adopt a negotiation mechanism based on the contract net protocol to adjust control parameters. First, divide the entire gold ore flotation process into four main functional units: the crushing control unit, the grinding and classification control unit, the flotation control unit, and the concentrate treatment unit. Each unit is equipped with an independent decision-making controller and actuator. Load an expert knowledge base, a control algorithm library, and a fuzzy rule base for each functional unit to build agents. The internal structure of the agent includes a data processing module, a decision-making and reasoning module, and an execution control module, and has a two-layer structure of a local optimization control layer and a collaborative optimization control layer. The expert knowledge base stores the process parameter control ranges, exception handling rules, and optimization strategies of each functional unit, such as control knowledge such as the crusher feed particle size in the range of 10-30mm, the pulp concentration in the range of 35%-45%, and the flotation pH value in the range of 7-9, as well as the corresponding exception handling rules. The control algorithm library includes PID control algorithms, fuzzy control algorithms, and model predictive control algorithms. The fuzzy rule base takes the process parameter deviation value and the deviation change rate as inputs and performs fuzzy reasoning to obtain the control quantity adjustment value. Based on the digital twin model, assign initial control objectives to each agent. The concentrate treatment unit agent generates a task request according to the production index requirements, including the target value of concentrate grade and the target value of recovery rate. After the task request is transmitted hierarchically, the flotation unit agent evaluates the current working conditions after receiving the request and generates a flotation plan and resource requirements including parameters such as flotation reagent dosage, pH value, and air inflow. Each agent decomposes and negotiates tasks according to the interaction rules set by the contract net protocol, and uses the Pareto multi-objective optimization method to balance conflicting objectives to reach a consistent global optimization control plan.

[0019] An adaptive soft-sensing model and a multi-objective optimization control strategy are constructed according to the global optimization control scheme, and parameter optimization is achieved through the self-learning and knowledge accumulation mechanism. First, a soft-sensing system integrating multiple models is constructed to estimate key quality indicators that are difficult to measure directly, such as concentrate grade and recovery rate. The soft-sensing model includes a theoretical model based on flotation kinetics and a regression model based on historical data, and the results of the two types of models are weighted and fused to obtain the final estimated value. Combining the quality indicators estimated by the soft-sensing system with the actual process parameters, a hierarchical control structure is constructed, including a real-time control layer, a tactical optimization layer, and a strategic decision-making layer. The control periods of each layer are 100 ms, 2 hours, and 8 hours respectively. The improved model predictive control algorithm is adopted in the real-time control layer, with the prediction horizon set to 30 steps and the control horizon set to 5 steps. The prediction model is dynamically updated according to the corresponding relationship between the process parameters and the quality indicators. The multi-objective optimization algorithm is applied in the tactical optimization layer, setting four optimization objectives: maximizing the concentrate grade, maximizing the recovery rate, minimizing the energy consumption, and minimizing the reagent consumption. The weights of the objective functions are dynamically adjusted according to the current production requirements. The control effect is evaluated in real time. By comparing the deviation between the actual production indicators and the optimization objectives, an operation case library including operating conditions characteristics, control parameter configurations, and control effects is established. Based on the operation case library and the soft-sensing feedback, the model parameters are incrementally updated, and the control strategy is adjusted according to the control effect evaluation results, forming a self-learning closed loop to obtain continuously optimized process control parameters.

[0020] In the embodiment of the present application, process parameter data is collected through a distributed sensor network, and outlier detection and hierarchical processing are performed on the data, effectively improving the data quality and solving the problems of simple data processing method and low quality in traditional systems; a multi-source heterogeneous data fusion engine is constructed based on the hierarchically processed data, and the data is standardized and multi-level fused to obtain a digital twin model of the entire gold flotation process, realizing the real-time mapping of the physical flotation system and the virtual system, and providing an accurate information basis for subsequent decision-making; a multi-agent negotiation control architecture is designed based on the digital twin model, and a negotiation mechanism based on the contract net protocol is used to adjust control parameters, effectively solving the problems of insufficient information interaction among control units and difficulty in achieving global optimization in traditional systems. The negotiation mechanism significantly improves the collaborative performance of the overall system, enabling each functional unit to jointly achieve the global optimal goal while maintaining local stability; an adaptive soft sensor model and a multi-objective optimization control strategy are constructed according to the global optimization control scheme, and parameter optimization is achieved through a self-learning and knowledge accumulation mechanism, which not only solves the problem of the lack of effective soft measurement means in traditional systems but also overcomes the defect of the lack of self-improving mechanism in control strategies. In particular, the contract net protocol applied in the negotiation control architecture, as a multi-agent negotiation mechanism, fully considers the characteristics and mutual influences of each functional unit in the flotation process. Through a structured message interaction and task decomposition negotiation process, it effectively balances multiple conflicting goals such as grade, recovery rate, and energy consumption. The introduction of the negotiation mechanism enables the system to have distributed decision-making capabilities, significantly reducing the computational burden and improving the system response speed; while the adaptive soft sensor model realizes the accurate estimation of parameters that are difficult to directly measure through multi-model integration and transfer learning methods, and the self-learning mechanism enables the system to accumulate optimization experience from historical operation data and improve the adaptability to complex working conditions, reducing energy consumption and reagent consumption while increasing the concentrate grade and recovery rate.

[0021] In a specific embodiment, the process of executing step S101 may specifically include the following steps: Classify the collected process parameter data of pH value, pulp concentration, grinding fineness, and foam characteristics according to the change rate, set the sampling frequency of 500 ms / time for rapidly changing parameters, and set the sampling frequency of 5 s / time for slowly changing parameters; Apply the improved Z-score method to detect outliers in the classified process parameter data. By calculating the standardized distance between the parameter value and the moving window mean, mark the data points exceeding the threshold of 4.5 as outliers; Replace the detected outlier data points with the weighted average of the 20 valid data points before and after, and the weight coefficient is calculated by an exponential decay function; Apply the fourth-order Kalman filter algorithm to eliminate noise from the data after replacing outliers, and adjust the filter parameter matrix according to the data type; The filtered data is graded based on the data importance scoring function, which comprehensively considers three factors: the influence of the parameters on the process, the amplitude of change, and the frequency of fluctuation; Parameter data with a score greater than 0.8 is classified as level one, parameter data with a score between 0.5 and 0.8 is classified as level two, and parameter data with a score less than 0.5 is classified as level three, forming hierarchical processing data with priority.

[0022] Specifically, the collected process parameter data are classified and processed according to the change rate. Parameters such as pH value, pulp concentration, grinding fineness and foam characteristics change at different rates during the flotation process, so the sampling frequency needs to be set specifically. The determination of the change rate is based on the time derivative value of the parameter. The change rate of the parameter over a period of time is calculated. Parameters greater than the set threshold are defined as fast-changing parameters, such as pH value and aeration volume. For such parameters, a high-frequency sampling of 500ms / time is used; parameters less than the set threshold are defined as slow-changing parameters, such as pulp concentration and grinding fineness. For such parameters, a low-frequency sampling of 5s / time is used. This classification sampling strategy solves the problem of data redundancy or information loss caused by traditional fixed-frequency sampling, and conforms to the actual change characteristics of process parameters. After the collected data is classified by the change rate, it enters the outlier detection link. The improved Z-score method is used to detect outliers in process parameter data. The traditional Z-score method calculates the global mean and standard deviation, while the improved Z-score method uses a sliding window mechanism. For each data point, it calculates the mean and standard deviation of the previous N points, and then calculates the standardized distance between the point and the window mean. The specific operation is to take the 30 data points before the current data point as a sliding window, calculate the mean and standard deviation of the data in the window, and then calculate the Z value of the current data point, that is, the current value minus the window mean and divided by the window standard deviation. When the absolute value of the Z value exceeds the threshold of 4.5, the data point is marked as an outlier. Compared with the traditional method, the improved Z-score method is more adaptable to the dynamic change characteristics of parameters in the flotation process and can effectively identify local anomalies.

[0023] For the detected outlier data points, the weighted average replacement method is used for processing. Take 10 valid data points before and after the outlier point, a total of 20 points, and calculate the weight coefficient of each point through the exponential decay function. The form of the exponential decay function is weight Wi = , where i represents the distance between the data point and the outlier, and λ is the decay rate parameter, which is usually set to 0.2, so that the farther the data point is from the outlier, the greater the weight, and vice versa. This processing method takes into account the impact that the outlier may have on the surrounding data points, and reduces this impact by reducing the weight of the adjacent points, ensuring that the replacement value is more in line with the normal trend.

[0024] After the outliers are replaced, the data still needs to be processed by noise elimination, using the fourth-order Kalman filter algorithm. Kalman filtering is a recursive estimation algorithm that obtains the optimal estimate by establishing a system state model and a measurement model and combining the prior estimate and the current measurement value. Fourth-order Kalman filtering means that the state vector is four-dimensional, including the value of the parameter, the first derivative, the second derivative, and the third derivative, which can more accurately describe the dynamic change trend of the parameter. For different data types, the filtering parameter matrices need to be adjusted, including the process noise covariance matrix Q and the measurement noise covariance matrix R. For the pH value data with a relatively fast change, the R value is set to be small to increase the confidence in the measurement value; for the pulp concentration data with high stability, the Q value is set to be small to increase the confidence in the model prediction. By appropriately adjusting these parameters, the Kalman filter algorithm can effectively eliminate the noise in various process parameter data. The filtered data enters the classification processing link, and the data is classified based on the data importance scoring function. The scoring function comprehensively considers three factors: the influence degree of the parameter on the process, the change range, and the fluctuation frequency. The influence degree of the parameter on the process is the weight coefficient determined by expert experience and statistical analysis of historical data, indicating the influence degree of the parameter on the flotation effect; the change range is the deviation degree of the parameter value from the normal working range; the fluctuation frequency is the number of times the parameter value fluctuates up and down per unit time. The scoring function weights and sums these three factors in a certain proportion to obtain the importance scoring value of the parameter data.

[0025] The parameter data is classified according to the importance scoring value. The parameter data with a scoring value greater than 0.8 is classified as the first level, indicating the parameters that have a great influence on the process and need to be processed with high priority, such as the pH value, the reagent addition amount, etc.; the parameter data with a scoring value between 0.5 and 0.8 is classified as the second level, indicating the parameters that have a certain influence on the process but are not the most critical, such as the foam height, the aeration amount, etc.; the parameter data with a scoring value lower than 0.5 is classified as the third level, indicating the auxiliary parameters with relatively small influence or slow change, such as the temperature, the pressure, etc. The classified data has different transmission priorities and processing strategies, ensuring that the key parameter data can be transmitted and processed in a timely manner, and solving the problems of undifferentiated data processing and delayed key information in the traditional system.

[0026] In a specific embodiment, the process of executing step S102 may specifically include the following steps: Perform unit conversion on the classified data, convert the grinding fineness data to millimeters, convert the pulp concentration data to decimal form, and convert the potential data to standard potential units to obtain data with unified units; Perform normalization processing on the data with unified units, adjust the numerical ranges of different parameters to the same interval, eliminate the influence of dimension differences on subsequent analysis, and obtain normalized data; Perform time alignment processing on the normalized data, align the data with different sampling frequencies according to the system clock, interpolate and fill in the sampling intervals to obtain spatio-temporally consistent standardized data; Perform weighted average calculation on the data from multiple sensors measuring the same parameter in the standardized data, and allocate weights according to the historical accuracy of each sensor to obtain the data-level fusion result; Analyze the correlation relationships between different parameters in the data-level fusion result, identify parameter combinations with strong correlations, extract the key features characterizing the flotation process state to obtain the feature-level fusion result; Combine the feature-level fusion result with the flotation process expert knowledge, judge the current flotation process state, establish the corresponding relationship between the physical flotation system and the virtual system to obtain the digital twin model of the entire gold flotation process.

[0027] Specifically, perform unit conversion processing on the data. The data units collected by different sensors are often not unified. For example, the grinding fineness data may be expressed in micrometers (μm) and needs to be converted to millimeters (mm). The conversion method is to divide the original value by 1000; the pulp concentration data is usually expressed as a percentage and needs to be converted to a decimal form. The conversion method is to divide the percentage value by 100; the potential data may have deviations depending on the reference electrode and needs to be converted to the standard potential unit. The conversion method is to add or subtract the corresponding reference electrode potential difference. The purpose of unit conversion is to establish a unified measurement standard to avoid data interpretation errors and calculation deviations caused by different units. After unit unification, there are still dimensional differences in the data, and the numerical ranges of different parameters vary greatly. For example, the pH value is usually in the range of 0-14, while the grinding fineness may be in the range of dozens to hundreds of micrometers. Therefore, it is necessary to perform normalization processing on the data with unified units to adjust the numerical ranges of different parameters to the same interval. The normalization processing adopts the maximum-minimum normalization method. For each parameter, find the maximum and minimum values of its historical data, and then map the current value to the interval [0,1]. The calculation process is: subtract the historical minimum value from the current value, and then divide by the difference between the historical maximum value and the minimum value. For those parameters without clear upper and lower limits, the sliding window method is used to take the maximum and minimum values in the recent period for normalization. The normalization processing eliminates the influence of dimensional differences on subsequent analysis, enabling direct comparison and calculation between different parameters.

[0028] The normalized data also faces the problem of time inconsistency because the sampling frequencies of different parameters are different. For example, the pH value is sampled every 500 ms, while the pulp concentration is sampled every 5 s, and the data timestamps are not aligned. To solve this problem, a time alignment processing method is adopted. First, determine the system reference clock. Usually, the highest sampling frequency is selected as the reference, and then all parameter data are aligned according to the reference clock. For parameters with a sampling frequency lower than the reference clock, interpolation needs to be performed to fill in the gaps. Interpolation methods include linear interpolation, cubic spline interpolation, and nearest neighbor interpolation. An appropriate interpolation method is selected according to the parameter change characteristics. Linear interpolation is suitable for parameters with relatively gentle changes. Calculate the linear relationship between two adjacent sampling points and deduce the parameter values at intermediate times; cubic spline interpolation is suitable for parameters that need to maintain curvature continuity; nearest neighbor interpolation is suitable for discrete parameters. After time alignment processing, all parameter data are consistent on the time axis, forming spatio-temporally consistent standardized data.

[0029] There may be cases where multiple sensors measure the same parameter in the standardized data. For example, multiple pH sensors are distributed at different locations, or there are two main and backup sensors installed at one location. In this case, these data need to be weighted and averaged to obtain the unique accurate value of this parameter. The weight assignment is based on the historical accuracy of each sensor. The specific calculation method is as follows: Analyze the deviation between the historical data of each sensor and the measured value of the standard sample. The smaller the deviation, the higher the accuracy of the sensor, and the greater the weight is assigned; at the same time, consider factors such as the running time and failure rate of the sensor. The weight of a sensor with stable operation is increased accordingly. The weighted average calculation formula is: The final parameter value is equal to the sum of the measured values of each sensor multiplied by the corresponding weight coefficient. In this way, multi-source data are effectively fused, the accuracy of parameter estimation is improved, and a data-level fusion result is obtained. The data-level fusion result contains a large number of parameters. How to extract key features from it is the next problem to be solved. The method is to analyze the correlation relationship between different parameters and identify parameter combinations with strong correlation. The Pearson correlation coefficient between parameters is calculated using the correlation analysis method. A correlation coefficient close to 1 or -1 indicates strong correlation, and close to 0 indicates weak correlation. For highly correlated parameter groups, only one of them can be retained as a representative, or combined features can be constructed; for weakly correlated but important parameters, they are retained separately. In this way, the data dimension is reduced, and the key features representing the flotation process state are extracted, forming a feature-level fusion result.

[0030] Combine the feature-level fusion results with the expert knowledge of the flotation process to establish a digital twin model for the entire gold flotation process. First, represent the expert knowledge in a structured manner, including the causal relationships between parameters, the mapping relationships between parameters and process states, and the judgment rules for abnormal states. Then, based on the feature-level fusion results, apply the expert knowledge rules to judge the state and determine the current flotation process state. The process states include three main categories: normal, sub-healthy, and abnormal, and each category is further divided into multiple specific states, such as excessive foaming, insufficient foaming, insufficient reagents, etc. The state judgment results are used to establish the corresponding relationship between the physical flotation system and the virtual system. The virtual system is updated in real time according to the state of the physical system to form a digital twin model. The digital twin model not only includes the current state but also contains historical trends and the ability to predict the future.

[0031] Taking a gold flotation production line as an example, when the pH sensor collects a pH value of 8.2, the pulp concentration sensor collects a concentration of 42%, the grinding fineness sensor collects a fineness of 65 μm, and the foam image analysis obtains a foam stability of 0.75, first perform unit conversion: the pH value remains unchanged, the pulp concentration is converted to 0.42, and the grinding fineness is converted to 0.065 mm. Then normalize the data: the historical range of the pH value is [6.5, 9.5], and the normalized value is (8.2 - 6.5) / (9.5 - 6.5) = 0.57; the historical range of the pulp concentration is [0.35, 0.50], and the normalized value is (0.42 - 0.35) / (0.50 - 0.35) = 0.47; the grinding fineness and foam stability are normalized in the same way. Since the sampling frequencies of these parameters are different, perform time alignment processing. Based on 500 ms as the benchmark, linearly interpolate and fill in the low-frequency sampling data such as pulp concentration. For the case of multiple pH sensors, allocate weights according to the historical accuracy of each sensor. For example, the weight of sensor A is 0.6, and the weight of sensor B is 0.4. The fused pH value is 8.2×0.6 + 8.1×0.4 = 8.16. Correlation analysis finds that the pH value is strongly correlated with the foam stability, and the correlation coefficient is 0.87. Therefore, combine the two into one feature; there is also an obvious correlation between the grinding fineness and the recovery rate. Combine these features with the expert knowledge rules to judge that the current process state is the "normal - ideal foam" state, corresponding to the normal operation area in the digital twin model, and predict the flotation state within the next hour based on the parameter change trend. This data fusion method effectively solves problems such as data heterogeneity, information silos, and inaccurate state judgment in traditional flotation control systems.

[0032] In a specific embodiment, the process of executing step S103 may specifically include the following steps: The entire gold flotation process is divided into four main functional units: the crushing control unit, the grinding and classification control unit, the flotation control unit, and the concentrate treatment control unit. Each unit is equipped with an independent decision-making controller and actuator; An expert knowledge base, a control algorithm library, and a fuzzy rule library are loaded into each functional unit to construct an agent corresponding to each functional unit. Each agent has a local optimization control layer and a collaborative optimization control layer; Based on the digital twin model, an initial control target is assigned to each agent. The agent of the concentrate treatment unit generates a task request according to the production index requirements, including the target value of concentrate grade and the target value of recovery rate; The task request is hierarchically transmitted. After receiving the request, the agent of the flotation unit evaluates the current working conditions and generates a flotation plan and resource requirements including the dosage of flotation reagents, pH value, and aeration volume; According to the interaction rules set by the contract net protocol, each agent decomposes and negotiates tasks through a message passing mechanism, and uses the Pareto multi-objective optimization method to balance conflicting objectives. Each agent submits a plan in turn and evaluates the plan; The plan in the negotiation process is iteratively optimized. When the plan converges or reaches the maximum number of iterations, the negotiation is terminated. Each agent signs an execution contract to determine the control parameter configuration and obtain a globally optimized control plan.

[0033] Specifically, the entire gold flotation process is divided into control units with clear functions. According to the characteristics of the gold flotation process, the entire process is divided into four main functional units: the crushing control unit, the grinding and classification control unit, the flotation control unit, and the concentrate treatment control unit. The crushing control unit is responsible for the preliminary treatment of the raw ore, including the control of equipment such as jaw crushers and cone crushers; the grinding and classification control unit is responsible for further grinding and classifying the crushed ore, including the control of equipment such as ball mills and hydrocyclones; the flotation control unit separates valuable minerals from gangue, including the control of equipment such as flotation machines and reagent addition systems; the concentrate treatment control unit is responsible for the post-treatment of the concentrate obtained by flotation, such as dehydration and drying, including the control of equipment such as thickeners and filters. Each functional unit is equipped with an independent decision-making controller and actuator. The decision-making controller is responsible for receiving sensor data, performing decision calculations, and outputting control instructions, and the actuator is responsible for converting the control instructions into equipment actions, such as adjusting the valve opening and changing the motor speed.

[0034] Load the expert knowledge base, control algorithm library, and fuzzy rule base for each functional unit to construct the corresponding intelligent agent. The expert knowledge base is a structured representation of empirical knowledge, including the normal range of parameters, exception handling strategies, and optimization experiences, such as rules like "When the foam of the flotation machine is dull, the dosage of the collector needs to be increased". The control algorithm library includes PID control algorithms, fuzzy control algorithms, and model predictive control algorithms. The PID control algorithm is suitable for objects with high linearity, the fuzzy control algorithm is suitable for objects with strong nonlinearity but with clear expert experience, and the model predictive control algorithm is suitable for multi-input multi-output objects with a clear mathematical model. The fuzzy rule base contains decision rules based on fuzzy logic, such as rules like "If the pH value is low and the degree of lowness is large, then significantly increase the dosage of the alkali agent". After loading these three libraries, an intelligent agent with perception, decision-making, and execution capabilities is formed. Inside each intelligent agent, there is a two-layer structure of a local optimization control layer and a collaborative optimization control layer. The local optimization control layer handles the stable control of internal parameters of this unit and local target optimization, using a closed-loop feedback control method; the collaborative optimization control layer is responsible for interacting with other unit intelligent agents, coordinating the consistency between this unit and the global target, using predictive control and negotiation mechanisms.

[0035] Based on the digital twin model, assign the initial control target to each intelligent agent. The digital twin model contains the mathematical model and historical operation data of the entire gold ore flotation process, and can quantitatively describe the relationships between various parameters. During the process of initial control target assignment, the intelligent agent of the concentrate processing unit generates a task request according to the production index requirements. This request usually comes from the production plan or operator settings, and includes the target values of concentrate grade and recovery rate. The concentrate grade represents the percentage of valuable metals in the concentrate, and the recovery rate represents the percentage of valuable metals recovered from the raw ore in the total amount of the raw ore. These two indicators usually have an inverse relationship. The intelligent agent of the concentrate processing unit packs these target values into a task request message and prepares to transmit it to the upstream unit. The task request is transmitted hierarchically and is first transmitted to the intelligent agent of the flotation unit. After receiving the request, the intelligent agent of the flotation unit needs to evaluate whether the current working condition can meet the target requirements. The evaluation process includes calculating the expected concentrate grade and recovery rate under the current working condition, comparing them with the target values, and analyzing the size and reasons for the gap. If there is a gap, the intelligent agent of the flotation unit needs to calculate the parameter adjustment plan required to achieve the target according to the digital twin model, and generate a flotation plan including parameters such as the dosage of flotation reagents, pH value, and aeration volume. At the same time, calculate the resources required to implement this plan, such as reagent consumption, energy consumption, etc., to form a resource demand list. The flotation plan and resource demand are transmitted as a whole to the intelligent agent of the grinding and classification unit.

[0036] According to the interaction rules set by the Contract Net Protocol, each agent conducts task decomposition and negotiation. The Contract Net Protocol is a protocol for task allocation in multi-agent systems, which includes four main stages: task announcement, bidding, awarding, and result reporting. In gold flotation control, the protocol stipulates the message types and interaction processes. For example, the concentrate processing unit sends task request messages, and the flotation unit replies with solution response messages, etc. Each agent negotiates through this message passing mechanism. When encountering conflicting goals, the Pareto multi-objective optimization method is used for trade-off. Pareto multi-objective optimization is a method for dealing with multiple conflicting goals. A solution is called Pareto optimal if no other solution can increase the value of one objective function without reducing the value of at least one other objective function. In the flotation process, it is usually necessary to trade off multiple goals such as concentrate grade and recovery rate, production efficiency and energy consumption. Each agent submits its own optimization solution in turn and evaluates the solutions of other agents. The evaluation criteria include solution feasibility, rationality of resource requirements, and goal satisfaction. The solutions in the negotiation process are iteratively optimized. Each iteration adjusts and evaluates the solution until the termination condition is reached. The termination conditions include two cases: solution convergence and reaching the maximum number of iterations. Solution convergence means that the change amplitude of the solution in several consecutive iterations is less than the set threshold, indicating that a stable solution has been found; reaching the maximum number of iterations is to avoid infinite loops, usually set to 20 - 30 times. After the negotiation terminates, each agent "signs" the execution contract, that is, confirms to accept the final solution and clarifies its respective execution responsibilities and goals. The execution contract includes specific control parameter configurations, such as the set value of the crusher feeding speed, the set value of the ball mill rotation speed, the set value of the flotation reagent addition amount, etc., to form a global optimization control solution.

[0037] In a specific embodiment, the process of executing step S104 may specifically include the following steps: Separate data processing modules, decision-making and reasoning modules, and execution control modules are constructed for the crushing control unit, grinding and classification control unit, flotation control unit, and concentrate processing control unit respectively to form their respective agent infrastructures; Compile the process parameter control ranges, exception handling rules, and optimization strategies for each functional unit into structured knowledge entries, store them in the expert knowledge base of the corresponding agent, and establish a knowledge retrieval and matching mechanism; Build a control algorithm library inside each agent, including PID control algorithm, fuzzy control algorithm, and model predictive control algorithm, and set the algorithm selection logic to automatically switch the optimal control algorithm according to the change of working conditions; Load a fuzzy rule base for each agent. By using the process parameter deviation value and deviation change rate as inputs, after fuzzy processing, they are matched with the rule base, and a fuzzy conclusion of the control quantity adjustment value is obtained through fuzzy reasoning; Build a local optimization control layer inside each agent. The local optimization control layer receives the sensor data of this unit, executes the local control loop, maintains the stability of the parameters within this unit, and performs optimization calculations on the local objective function; Build a collaborative optimization control layer for each agent. The collaborative optimization control layer receives messages from other agents, processes the global objective constraints, coordinates the working relationships between this unit and adjacent units, and participates in the negotiation process among multiple agents.

[0038] Specifically, build independent data processing modules, decision-making and reasoning modules, and execution control modules for the crushing control unit, grinding and classification control unit, flotation control unit, and concentrate processing control unit respectively. The data processing module is responsible for receiving the raw sensor data and performing preprocessing, including operations such as outlier detection, data filtering, and data standardization; the decision-making and reasoning module performs state evaluation and control calculations based on the processed data and outputs control decisions; the execution control module then converts the control decisions into specific execution commands and sends them to the corresponding execution devices. Taking the flotation control unit as an example, the data processing module receives sensor data such as pH value, redox potential, flotation reagent flow rate, and pulp concentration, and performs format conversion, unit unification, and time alignment on the data; the decision-making and reasoning module calculates the current flotation working condition based on this data and determines whether parameters such as reagent addition amount and stirring speed need to be adjusted; the execution control module then converts the calculated reagent addition amount into a flow control signal for the reagent pump and converts the stirring speed into a frequency control signal for the frequency converter. This modular design enables each functional unit to operate independently and work together as a whole to form the agent infrastructure. Compile the process parameter control ranges, exception handling rules, and optimization strategies for each functional unit into structured knowledge entries, and knowledge representation and organization are required. The structured knowledge entries adopt the form of "condition-action". The condition part describes the working condition characteristics, and the action part describes the corresponding control measures. For example, for the flotation unit, a knowledge entry can be "Condition: The pH value is lower than 7.5 and the foam amount decreases; Action: Increase the lime addition amount to make the pH value return to the range of 8.0±0.2". Assign a unique identifier to each knowledge entry and label it according to attributes such as applicable scope and priority. Store the compiled knowledge entries in the expert knowledge base of the corresponding agent, and at the same time establish a knowledge retrieval and matching mechanism. Knowledge retrieval uses the inverted index technology to quickly locate relevant knowledge entries according to keywords; knowledge matching uses a similarity calculation method to match the current working condition with the conditions described in the knowledge entries to find the most similar entry. The similarity calculation considers the proximity of parameter values, the importance weights of parameters, and the matching degree of working condition characteristics. When some working condition parameters change or anomalies occur, the knowledge retrieval and matching mechanism will be automatically triggered to find the appropriate processing strategy for the current situation and provide a basis for decision-making and reasoning.

[0039] Build a control algorithm library inside each agent, which requires implementing multiple control algorithms and establishing a selection mechanism. The control algorithm library includes three main algorithms: PID control algorithm, fuzzy control algorithm, and model predictive control algorithm. The PID control algorithm calculates the control output based on the error and its rate of change, and is suitable for control objects with high linearity and simple models; the fuzzy control algorithm derives control decisions based on fuzzy rule reasoning, and is suitable for control objects with strong nonlinearity, difficult to establish an accurate mathematical model but with rich empirical knowledge; the model predictive control algorithm predicts the future output based on the system model and solves the optimization problem to obtain the control sequence, and is suitable for complex control objects with multiple variables, strong coupling, and constraints. Set the algorithm selection logic to automatically switch the optimal control algorithm according to the changes in working conditions. The algorithm selection logic is based on the characteristics of the current control object, the requirements of the control target, and the historical performance of each algorithm. The specific implementation method is through an evaluation function to calculate the applicability of each algorithm under the current working conditions and select the algorithm with the highest applicability. The evaluation function considers factors such as control accuracy, response speed, and robustness, and dynamically adjusts the weights of each factor according to historical data.

[0040] Fuzzy rule bases need to be loaded into each agent, and a fuzzy control process needs to be designed. First, the process parameter deviation value (E) and the deviation change rate (EC) are used as the inputs of the fuzzy controller. These two values are obtained by calculating the difference between the current parameter value and the set value and its change rate over time. Then, E and EC are fuzzified, mapping the exact numerical values onto fuzzy sets, such as fuzzy linguistic variables like "negative large", "negative small", "zero", "positive small", "positive large", etc. Fuzzification uses membership functions, and commonly used ones include triangular functions, trapezoidal functions, and Gaussian functions. The appropriate function form is selected according to the parameter characteristics. The fuzzified inputs are matched with the rules in the fuzzy rule base. The rules in the rule base are in the "If-Then" form, such as "If E is negative large and EC is positive small Then U is negative medium", where U is the control quantity adjustment value. The fuzzy set of the control quantity is calculated through fuzzy inference. The inference methods include Mamdani inference and Sugeno inference, which are selected according to specific requirements. Finally, through defuzzification, the fuzzy set is converted back to an exact numerical value. Commonly used defuzzification methods include the centroid method, the maximum membership degree method, etc. In this way, the fuzzy controller can give an appropriate control adjustment amount according to the parameter deviation and change trend, realizing intelligent control. A local optimization control layer is constructed inside each agent so that it can independently handle the control tasks of this unit. The local optimization control layer receives sensor data. The data first undergoes preprocessing, including filtering, outlier detection, and compensation, etc., and then enters the state estimation link. The state variables that are difficult to directly measure are estimated through Kalman filtering or an observer. The state estimation result is compared with the set value to calculate the control deviation, and an appropriate control algorithm is selected according to the deviation for calculation to obtain the control output. The control output also needs to undergo constraint processing to ensure within the physical limits of the actuator. The local optimization control layer also includes the optimization calculation of the local objective function. The objective function usually includes indicators such as control accuracy, stability, and resource consumption, and optimization algorithms such as gradient descent and simulated annealing are used to solve for the optimal control parameters. The local optimization control layer continuously monitors the control effect through a closed-loop feedback control mechanism, adjusts the control strategy, and maintains the stable operation of the parameters within this unit.

[0041] Build a collaborative optimization control layer for each agent to achieve coordination and cooperation among units. The collaborative optimization control layer is responsible for interacting with other agents and handling global issues involving multiple units. The collaborative optimization control layer receives messages from other agents, including status information, demand requests, negotiation proposals, etc., and parses and processes them. For global target constraints, the collaborative optimization control layer decomposes them into local constraints executable by this unit and checks their compatibility with the unit's goals. If there are conflicts, the local objective function needs to be adjusted to be consistent with the global goal. The collaborative optimization control layer is also responsible for coordinating the working relationships between this unit and adjacent units, such as ensuring the matching of material flow rates and the connection of process parameters. When participating in multi-agent negotiation, the collaborative optimization control layer generates response plans based on the capabilities and constraints of this unit, evaluates the plans of other units, and participates in the iterative negotiation process until an agreement is reached. The decision results of the collaborative optimization control layer are passed to the local optimization control layer for execution, forming a hierarchical control structure.

[0042] In a specific embodiment, the process of executing step S105 may specifically include the following steps: Encode the flotation expert experience knowledge in the form of condition-action. The condition part includes the parameter value range and status description, and the action part includes control operations and adjustment strategies to form structured knowledge entries; Assign unique identifiers to the structured knowledge entries of each functional unit, and perform multi-dimensional annotation according to the functional unit, knowledge type, and applicable conditions to construct a knowledge index table; Build a knowledge retrieval system based on the inverted index technology, and use the process parameter name, value range, and abnormal type as retrieval keywords to map to the identifier set of relevant knowledge entries; Design a knowledge matching algorithm based on weighted cosine similarity, convert the current working condition parameters into feature vectors, calculate the similarity with the condition feature vectors in the knowledge entries, and obtain the matching degree ranking result; Build a knowledge entry activation mechanism. When a certain working condition parameter deviates from the normal range or an abnormal state occurs, automatically trigger the retrieval process of relevant knowledge entries, and mark the knowledge entries with a similarity exceeding the threshold of 0.85 as the activated state; Apply conflict resolution rules to multiple activated knowledge entries. When different knowledge entries give conflicting suggestions, select the optimal knowledge entry as the decision basis based on the priority, applicable range accuracy, and historical success rate of the knowledge entries.

[0043] Specifically, the flotation expert experience knowledge is the practical experience accumulated by flotation engineers and operators over the years, including process parameter adjustment skills, abnormal state handling methods, process optimization strategies, etc. To convert this experience knowledge into structured knowledge entries that can be processed by a computer, it is necessary to encode it in the form of condition-action. The condition part describes the operating conditions characteristics that trigger the application of knowledge, including parameter value ranges and state descriptions, such as "the pH value is between 8.2 - 8.5 and the pulp concentration is lower than 35%", "the foam color is dull and the foam layer thickness is less than 5 cm", etc.; the action part describes the control operations and adjustment strategies to be taken for this condition, such as "increase the collector dosage by 15% while keeping the pH value unchanged", "reduce the air inflow rate to 0.6 m³ / min and increase the stirring intensity to 280 rpm", etc. This encoding method intuitively reflects the expert's thinking in dealing with problems, converts tacit knowledge into explicit expression, and forms structured knowledge entries.

[0044] A unique identifier is assigned to the structured knowledge entries of each functional unit and multi-dimensional annotation is carried out. The unique identifier adopts a coding rule such as "FU-KT-SN", where FU represents the functional unit code (e.g., CR represents the crushing unit, GR represents the grinding and classification unit, FL represents the flotation unit, CC represents the concentrate treatment unit), KT represents the knowledge type code (e.g., PR represents the parameter range knowledge, EH represents the exception handling knowledge, OS represents the optimization strategy knowledge), and SN represents the serial number. For example, "FL-EH-023" represents the 23rd exception handling knowledge of the flotation unit. The multi-dimensional annotation is to describe the attributes of the knowledge entries, including the functional unit attribute, the knowledge type attribute, and the applicable condition attribute. The functional unit attribute indicates the process unit to which the knowledge applies; the knowledge type attribute is divided into parameter range knowledge, exception handling knowledge, and optimization strategy knowledge; the applicable condition attribute describes the specific working conditions to which the knowledge applies, such as the raw ore grade range, equipment type, etc. Through this multi-dimensional annotation, a structured knowledge index table is constructed to facilitate subsequent knowledge retrieval. A knowledge retrieval system is constructed based on the inverted index technology, taking the process parameter name, value range, and exception type as retrieval keywords, and corresponding to the identifier set of relevant knowledge entries. The inverted index is an important technology in information retrieval, which maps the words in the document to the document list containing the word. In the expert knowledge base, the inverted index maps the keywords to the identifier set of the knowledge entries containing the keywords. Specifically, when implementing, first extract the keywords from the knowledge entries. The keywords include the process parameter name (such as "pH value", "collector"), value range (such as "below 7.0", "greater than 20 g / t"), and exception type (such as "unstable foam", "decrease in concentrate grade"), etc.; then establish a mapping table from the keywords to the knowledge entry identifiers. A keyword may correspond to multiple knowledge entries; finally, construct a term index to record in which knowledge entries each keyword appears, and its position and weight in the condition part. When relevant knowledge needs to be retrieved, input the keywords, and quickly find the identifier set of the knowledge entries containing the keywords through the inverted index, and then return the results sorted by relevance. Design a knowledge matching algorithm based on weighted cosine similarity to calculate the similarity between the current working conditions and the knowledge entries. The weighted cosine similarity is an extension of the cosine similarity, which takes into account the weight differences of different features. First, convert the current working condition parameters into a feature vector, and the dimension of the feature vector is the same as the number of condition features defined in the knowledge entries, and each dimension represents the value of a parameter or state feature; similarly, convert the condition part in the knowledge entries into a condition feature vector. Then set the weight vector according to the parameter importance, and assign higher weights to more critical parameters, such as higher weights for critical parameters such as pH value and collector dosage, and lower weights for auxiliary parameters such as temperature and pressure. Calculate the weighted cosine similarity of the two vectors to obtain the matching degree value. The closer the matching degree value is to 1, the more similar the current working conditions are to the conditions described in the knowledge entries.To improve the computational efficiency, a vector space compression technique is adopted to reduce the original high-dimensional feature vectors to a lower dimension through principal component analysis, reducing the computational amount while retaining the main feature information. For continuous-valued parameters, fuzzy interval mapping is used to handle numerical differences; for discrete state features, equality judgment or similarity measurement is adopted. Finally, the matching degree ranking results between the current working condition and each knowledge item are obtained.

[0045] A knowledge item activation mechanism is constructed to actively trigger the retrieval of relevant knowledge when the working condition changes. The knowledge item activation mechanism is based on two triggering methods: parameter overlimit triggering and state change triggering. Parameter overlimit triggering means that when a certain process parameter deviates from the normal range, the retrieval of relevant knowledge items is triggered. For example, when the pH value drops from the normal range of 7.5 - 8.5 to 7.2, the retrieval of knowledge items related to low pH value is triggered; state change triggering means that when the process state changes significantly, such as the foam quality changes from stable to unstable, the retrieval of relevant knowledge items is triggered. The retrieval process adopts a two-stage strategy: in the first stage, a candidate set of potentially relevant knowledge items is quickly screened out through an inverted index; in the second stage, for each knowledge item in the candidate set, its similarity with the current working condition is calculated, and the knowledge items with a similarity exceeding the threshold of 0.85 are marked as activated states. The threshold of 0.85 is an empirical value, which can ensure finding sufficiently relevant knowledge items and filtering out less relevant items. The knowledge items in the activated state enter the next conflict resolution link to provide a basis for the final decision. Conflict resolution rules are applied to multiple activated knowledge items to resolve the contradictions between knowledge items. In complex working conditions, it is common for multiple knowledge items to be activated simultaneously and give different or even contradictory suggestions. For example, one knowledge item suggests increasing the dosage of the collector, while another suggests reducing the dosage of the collector. The conflict resolution rules are based on three main factors: the priority of the knowledge item, the accuracy of the applicable range, and the historical success rate. The priority is an inherent attribute of the knowledge item, assigned by the knowledge engineer during coding, reflecting the importance and reliability of the knowledge; the accuracy of the applicable range indicates the degree of match between the conditions described by the knowledge item and the current working condition, and a more accurate match obtains a higher weight; the historical success rate is the proportion of successful applications of this knowledge item in the past, and the higher the success rate, the greater the weight. Considering these three factors, a comprehensive score is calculated for each activated knowledge item, and the knowledge item with the highest score is selected as the decision-making basis. If the scores of multiple knowledge items are close and the suggestions are compatible, a compromise strategy is adopted, such as taking the weighted average of each suggestion; if the suggestions are incompatible, the knowledge item with the highest score is persistently selected.

[0046] In a specific embodiment, the process of executing step S106 may specifically include the following steps: Construct a soft sensor system integrating multiple models based on a global optimization control scheme to estimate five key quality indicators, namely concentrate grade, recovery rate, mineral surface hydrophobicity, ore particle-bubble adhesion probability, and soft sensor models include a flotation kinetics theoretical model and a historical data regression model; Combine the quality indicators estimated by the soft sensor system with the actual process parameters to construct a hierarchical control structure, including a real-time control layer, a tactical optimization layer, and a strategic decision-making layer. The control periods of each layer are set to 100 ms, 2 hours, and 8 hours respectively; Adopt an improved model predictive control algorithm for the real-time control layer, set the prediction horizon to 30 steps, the control horizon to 5 steps, and update the prediction model according to the corresponding relationship between process parameters and quality indicators; Apply a multi-objective optimization algorithm to the tactical optimization layer, set four optimization objectives: maximizing concentrate grade, maximizing recovery rate, minimizing energy consumption, and minimizing reagent consumption, and dynamically adjust the objective function weights according to current production requirements; Evaluate the control effect in real time, establish an operation case base including operating conditions characteristics, control parameter configurations, and control effects by comparing the deviation between actual production indicators and optimization objectives; Based on the operation case base and soft sensor feedback, incrementally update the model parameters through transfer learning methods, and adjust the control strategy according to the control effect evaluation results to form a self-learning closed loop and obtain process control parameters.

[0047] Specifically, soft sensing refers to a method of indirectly estimating key quality indicators that are difficult to measure directly through measurable process parameters. In gold ore flotation, five key quality indicators, namely concentrate grade, recovery rate, mineral surface hydrophobicity, ore particle-bubble adhesion probability, and concentrate quality, are difficult to measure in real time online, but are crucial for control decisions. The soft sensor system adopts a multi-model integration strategy, including two major categories: a flotation kinetics theoretical model and a historical data regression model. The flotation kinetics theoretical model establishes mathematical equations based on flotation physical and chemical principles, describes the interaction process between ore particles and bubbles, and obtains the estimated value of the quality indicator by solving differential equations; the historical data regression model is based on historical operation data and establishes a mapping relationship between process parameters and quality indicators, including multiple linear regression models, support vector regression models, and neural network models. Multi-model integration improves the accuracy and robustness of the estimation by weighted combination of the outputs of each model, and the weight allocation is based on the performance of each model on the validation data set. The better the performance, the greater the weight.

[0048] The quality indicators estimated by the soft-sensing system are combined with the actual process parameters to construct a hierarchical control structure, which includes three control levels. The real-time control level is responsible for the stable control of process parameters within each unit, such as the closed-loop control of pH value, reagent addition amount, stirring speed, etc. The control period is set to 100 ms to ensure a rapid response to parameter disturbances; the tactical optimization level is responsible for the process optimization on a medium time scale, such as the dynamic adjustment of grinding fineness, flotation time, and reagent formula. The control period is set to 2 hours, and multiple optimizations are carried out within one shift; the strategic decision-making level is responsible for the production plan on a long time scale, such as throughput allocation and energy consumption plan. The control period is set to 8 hours, and it is usually adjusted once per shift. This hierarchical control structure reduces system complexity through time-scale separation, enabling each level to focus on control tasks suitable for its time scale. An improved model predictive control algorithm is adopted for the real-time control level. This algorithm has been improved in many aspects based on the standard model predictive control. First, the prediction horizon is set to 30 steps, that is, the behavior of the system within 30 sampling periods in the future is predicted; the control horizon is set to 5 steps, that is, the control input sequence for the next 5 sampling periods is calculated. The prediction model is the core of the algorithm, which describes the correspondence between process parameters and quality indicators and is represented by a state-space model. The prediction model is not fixed but is dynamically updated according to real-time data. The update methods include the recursive least squares method and the moving window batch processing method. The recursive least squares method makes small adjustments to the model parameters using newly acquired data to maintain the accuracy of the model; the moving window batch processing method regularly re-estimates the model parameters using data from a recent period of time to adapt to changes in working conditions. The improved model predictive control algorithm also introduces a constraint handling mechanism to ensure that the control output is within the physical limits of the actuator, as well as a soft constraint mechanism that allows conditional violation of some constraints when necessary.

[0049] A multi-objective optimization algorithm is applied to the tactical optimization level, and four optimization objectives are set: maximizing concentrate grade, maximizing recovery rate, minimizing energy consumption, and minimizing reagent consumption. These objectives usually conflict with each other. For example, increasing the concentrate grade often reduces the recovery rate, and reducing energy consumption may affect the throughput. The multi-objective optimization algorithm handles these conflicts through the dynamic adjustment of the objective function weights. The objective function weights change according to the current production requirements. For example, when the market price is high, the weight of the grade objective is increased; when the raw ore grade is low, the weight of the recovery rate objective is increased; during peak electricity consumption periods, the weight of the energy consumption objective is increased. The weight adjustment process takes into account market factors, resource constraints, and equipment status, forming an adaptive weight allocation mechanism. The optimization algorithm uses an improved genetic algorithm. The coding scheme is designed as real-number coding, the population size is 100, the crossover probability is 0.85, and the mutation probability is 0.15. The optimal control parameter combination is solved through multiple generations of evolution.

[0050] The control effect is evaluated in real time, and an operation case library is established. The evaluation process calculates the control performance indicators, including concentrate grade deviation, recovery rate deviation, energy consumption excess rate, and reagent use efficiency, by comparing the deviation between the actual production indicators and the optimization target. When the deviation exceeds the set threshold, the control strategy adjustment is triggered. At the same time, the current operating condition characteristics, control parameter configuration, and control effect evaluation results are recorded in the operation case library. The data structure of the operation case library includes the operating condition characteristics part (original ore properties, equipment status, etc.), the control parameter part (the set value of each control parameter), and the effect evaluation part (key indicators such as concentrate grade, recovery rate, and energy consumption). The operation case library not only records successful cases, but also failed cases, providing a comprehensive reference for subsequent optimization. A case similarity calculation method is established. When encountering new conditions, similar cases can be found in the case library and their control experience can be learned. Based on the operation case library and soft measurement feedback, the model parameters are incrementally updated through the transfer learning method, and the control strategy is adjusted according to the control effect evaluation results. Transfer learning refers to fine-tuning an existing model on new data instead of retraining it. This method retains the knowledge in the model that is relevant to the new data and only updates the parts with larger deviations. The incremental update process first calculates the prediction error of the current model, then adjusts the update step size according to the error size and pattern, and finally updates the model parameters. Parameters are updated using gradient descent or recursive least squares to ensure that the accuracy of the model continues to improve over time. Control strategy adjustments are based on the results of control effect evaluation, such as increasing the robustness parameters of the control algorithm when the grade fluctuates greatly, and adjusting the optimization target weights when the recovery rate is insufficient. The entire update and adjustment process forms a self-learning closed loop, and the system can continuously accumulate experience from practice and improve the control effect.

[0051] The above describes the method for intelligent monitoring and optimization control of the entire process of gold flotation in the embodiment of the present application. The following describes the intelligent monitoring and optimization control system for the entire process of gold flotation in the embodiment of the present application. Figure 2 In the embodiment of the present application, an embodiment of the intelligent monitoring and optimization control system for the entire process of gold flotation includes: The classification module 201 is used to collect process parameter data in the whole process of gold ore flotation through a distributed sensor network, perform abnormal value detection and classification processing on the collected data, and obtain classification processing data; The fusion module 202 is used to build a multi-source heterogeneous data fusion engine based on hierarchical processing data, standardize and multi-level fuse the data, and obtain a digital twin model of the entire process of gold mine flotation; An adjustment module 203 is used to design a multi-agent negotiation control architecture based on the digital twin model, and to adjust control parameters using a negotiation mechanism based on the contract network protocol to obtain a global optimization control solution; A building block 204 is configured to construct an adaptive soft sensor model and a multi-objective optimization control strategy according to the global optimization control scheme, and achieve parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

[0052] Through the collaborative cooperation of the above-mentioned various components, process parameter data is collected through a distributed sensor network, and outliers in the data are detected and classified, effectively improving the data quality and solving the problems of simple data processing method and low quality in traditional systems; a multi-source heterogeneous data fusion engine is constructed based on the classified data, and the data is standardized and multi-level fused to obtain a digital twin model of the entire gold flotation process, realizing the real-time mapping between the physical flotation system and the virtual system, and providing an accurate information basis for subsequent decision-making; a multi-agent negotiation control architecture is designed based on the digital twin model, and a negotiation mechanism based on the contract net protocol is used to adjust control parameters, effectively solving the problems of insufficient information interaction among control units and difficulty in achieving global optimization in traditional systems. The negotiation mechanism significantly improves the collaborative performance of the overall system, enabling each functional unit to jointly achieve the global optimal goal while maintaining local stability; an adaptive soft sensor model and a multi-objective optimization control strategy are constructed according to the global optimization control scheme, and parameter optimization is achieved through a self-learning and knowledge accumulation mechanism, which not only solves the problem of the lack of effective soft sensor means in traditional systems but also overcomes the defect of the lack of self-improving mechanism in control strategies. In particular, the contract net protocol applied in the negotiation control architecture, as a multi-agent negotiation mechanism, fully considers the characteristics and mutual influences of each functional unit in the flotation process. Through a structured message interaction and task decomposition negotiation process, it effectively balances multiple conflicting goals such as grade, recovery rate, and energy consumption. The introduction of the negotiation mechanism endows the system with distributed decision-making ability, significantly reducing the computational burden and improving the system response speed; while the adaptive soft sensor model realizes the accurate estimation of parameters that are difficult to directly measure through multi-model integration and transfer learning methods, and the self-learning mechanism enables the system to accumulate optimization experience from historical operation data, improving the adaptability to complex working conditions, reducing energy consumption and reagent consumption while increasing the concentrate grade and recovery rate.

[0053] Above Figure 2 The intelligent monitoring and optimization control system for the entire gold flotation process in the embodiments of the present invention is described in detail from the perspective of modular functional entities. Next, the intelligent monitoring and optimization control device for the entire gold flotation process in the embodiments of the present invention is described in detail from the perspective of hardware processing.

[0054] Figure 3It is a schematic structural diagram of an intelligent monitoring and optimization control device for the entire gold flotation process provided by an embodiment of the present invention. The intelligent monitoring and optimization control device 300 for the entire gold flotation process may vary greatly due to different configurations or performances, and may include one or more processors (central processing units, CPU) 310 (for example, one or more processors) and a memory 320, and one or more storage media 330 for storing application programs 333 or data 332 (for example, one or more mass storage device terminals). Among them, the memory 320 and the storage media 330 may be transient storage or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the figure), and each module may include a series of instruction operations on the intelligent monitoring and optimization control device 300 for the entire gold flotation process. Further, the processor 310 may be set to communicate with the storage media 330 and execute a series of instruction operations in the storage media 330 on the intelligent monitoring and optimization control device 300 to implement the steps of the above-mentioned intelligent monitoring and optimization control method for the entire gold flotation process.

[0055] The intelligent monitoring and optimization control device 300 for the entire gold flotation process may further include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Serve, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art can understand that Figure 3 The shown structural diagram of the intelligent monitoring and optimization control device for the entire gold flotation process does not limit the intelligent monitoring and optimization control device for the entire gold flotation process provided by the present invention, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements.

[0056] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. Instructions are stored in the computer-readable storage medium. When the instructions run on a computer, the computer is caused to execute the steps of the intelligent monitoring and optimization control method for the entire gold flotation process.

[0057] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0058] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions to enable a full-process intelligent monitoring and optimization control device for gold ore flotation (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs that can store program codes.

[0059] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of various embodiments of the present invention.

Claims

1. An intelligent monitoring and optimization control method for the whole process of gold ore flotation, characterized in that, The method includes: Collecting process parameter data in the whole process of gold ore flotation through a distributed sensor network, detecting and classifying outliers in the collected data to obtain classified data; Constructing a multi-source heterogeneous data fusion engine based on the classified data, standardizing and multi-level fusing the data to obtain a digital twin model of the whole process of gold ore flotation; Designing a multi-agent negotiation control architecture based on the digital twin model, and using a negotiation mechanism based on the contract net protocol to adjust control parameters to obtain a globally optimized control scheme; Constructing an adaptive soft sensor model and a multi-objective optimization control strategy according to the globally optimized control scheme, and realizing parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

2. The intelligent monitoring and optimization control method for the whole process of gold ore flotation according to claim 1, wherein, The collecting process parameter data in the whole process of gold ore flotation through a distributed sensor network, detecting and classifying outliers in the collected data to obtain classified data includes: Classifying the collected process parameter data of pH value, pulp density, grinding fineness, and foam characteristics according to the change rate, setting the sampling frequency of the rapidly changing parameters to 500 ms / time, and setting the sampling frequency of the slowly changing parameters to 5 s / time; Applying the improved Z-score method to detect outliers in the classified process parameter data, and marking the data points exceeding the threshold of 4.5 as outliers by calculating the standardized distance between the parameter value and the mean value of the sliding window; Replacing the detected outlier data points with the weighted average of the first 20 and the last 20 valid data points, and calculating the weight coefficient through an exponential decay function; Applying the fourth-order Kalman filter algorithm to the data after replacing outliers to eliminate noise, and adjusting the filter parameter matrix according to the data type; Classifying the filtered data based on a data importance scoring function, and the scoring function comprehensively considers three factors: the influence degree of the parameter on the process, the change amplitude, and the fluctuation frequency; Classifying the parameter data with a score value greater than 0.8 into the first level, the parameter data with a score value between 0.5 and 0.8 into the second level, and the parameter data with a score value less than 0.5 into the third level to form classified data with priorities.

3. The intelligent monitoring and optimization control method for the entire gold flotation process according to claim 1, characterized in that The constructing a multi-source heterogeneous data fusion engine based on the classified data, standardizing and multi-level fusing the data to obtain a digital twin model of the whole process of gold ore flotation includes: Performing unit conversion on the classified data, converting the grinding fineness data to millimeters, converting the pulp density data to decimal form, and converting the potential data to standard potential units to obtain data with unified units; Performing normalization processing on the data with unified units, adjusting the numerical ranges of different parameters to the same interval, and eliminating the influence of dimension differences on subsequent analysis to obtain normalized data; Performing time alignment processing on the normalized data, aligning the data with different sampling frequencies according to the system clock, and interpolating and filling the sampling intervals to obtain spatio-temporally consistent standardized data; Performing weighted average calculation on the data of the same parameter measured by multiple sensors in the standardized data, and allocating weights according to the historical accuracy of each sensor to obtain a data-level fusion result; Analyze the correlation relationships between different parameters in the data-level fusion results, identify parameter combinations with strong correlations, extract key features characterizing the flotation process state, and obtain the feature-level fusion results; Combine the feature-level fusion results with the expert knowledge of the flotation process, judge the current flotation process state, establish the corresponding relationship between the physical flotation system and the virtual system, and obtain the digital twin model of the entire gold flotation process.

4. The intelligent monitoring and optimization control method for the entire gold ore flotation process according to claim 1, characterized in that Design a multi-agent negotiation control architecture based on the digital twin model, and use a negotiation mechanism based on the contract net protocol to adjust control parameters to obtain a globally optimized control solution, including: Divide the entire gold flotation process into four main functional units: the crushing control unit, the grinding and classification control unit, the flotation control unit, and the concentrate treatment control unit. Each unit is equipped with an independent decision-making controller and actuator; Load an expert knowledge base, a control algorithm library, and a fuzzy rule library for each functional unit, construct an agent corresponding to each functional unit, and each agent has a local optimization control layer and a collaborative optimization control layer; Based on the digital twin model, assign an initial control target to each agent. The agent of the concentrate treatment unit generates a task request according to the production index requirements, including the target value of concentrate grade and the target value of recovery rate; Transmit the task request hierarchically. After receiving the request, the agent of the flotation unit evaluates the current working conditions and generates a flotation plan and resource requirements including the dosage of flotation reagents, pH value, and aeration volume; According to the interaction rules set by the contract net protocol, each agent decomposes and negotiates tasks through a message passing mechanism, uses the Pareto multi-objective optimization method to balance conflicting objectives, and each agent submits a plan in turn and evaluates the plan; Iteratively optimize the plan during the negotiation process. When the plan converges or reaches the maximum number of iterations, terminate the negotiation. Each agent signs an execution contract to determine the control parameter configuration and obtain a globally optimized control solution.

5. The intelligent monitoring and optimization control method for the whole process of gold ore flotation according to claim 4, characterized in that, The step of loading an expert knowledge base, a control algorithm library, and a fuzzy rule library for each functional unit, constructing an agent corresponding to each functional unit, and each agent having a local optimization control layer and a collaborative optimization control layer includes: Construct independent data processing modules, decision-making inference modules, and execution control modules for the crushing control unit, the grinding and classification control unit, the flotation control unit, and the concentrate treatment control unit respectively to form their respective agent infrastructure; Compile the process parameter control ranges, exception handling rules, and optimization strategies for each functional unit into structured knowledge entries, store them in the expert knowledge base of the corresponding agent, and establish a knowledge retrieval and matching mechanism; Build a control algorithm library inside each agent, including PID control algorithm, fuzzy control algorithm, and model predictive control algorithm, and set the algorithm selection logic to automatically switch the optimal control algorithm according to the change of working conditions; Load a fuzzy rule library for each agent. By using the process parameter deviation value and deviation change rate as inputs, after fuzzy processing, match with the rule library, and obtain a fuzzy conclusion of the control quantity adjustment value through fuzzy inference; Build a local optimization control layer inside each agent. The local optimization control layer receives the sensor data of this unit, executes the local control loop, maintains the stability of the parameters within this unit, and performs optimization calculations on the local objective function; Build a collaborative optimization control layer for each agent. The collaborative optimization control layer receives messages from other agents, processes the global objective constraints, coordinates the working relationships between this unit and adjacent units, and participates in the negotiation process among multiple agents.

6. The intelligent monitoring and optimization control method for the entire gold ore flotation process according to claim 5, characterized in that, Compile the process parameter control ranges, exception handling rules, and optimization strategies for each functional unit into structured knowledge entries, store them in the expert knowledge base of the corresponding agent, and establish a knowledge retrieval and matching mechanism, including: Encode the flotation expert experience knowledge in the form of condition-action. The condition part includes the parameter value range and status description, and the action part includes control operations and adjustment strategies to form structured knowledge entries; Assign unique identifiers to the structured knowledge entries of each functional unit, perform multi-dimensional annotation according to the functional unit, knowledge type, and applicable conditions, and construct a knowledge index table; Build a knowledge retrieval system based on the inverted index technology, use the process parameter name, value range, and exception type as retrieval keywords, and map them to the identifier set of relevant knowledge entries; Design a knowledge matching algorithm based on weighted cosine similarity, convert the current working condition parameters into feature vectors, calculate the similarity with the condition feature vectors in the knowledge entries, and obtain the matching degree ranking result; Build a knowledge entry activation mechanism. When a certain working condition parameter deviates from the normal range or an abnormal state occurs, automatically trigger the retrieval process of relevant knowledge entries, and mark the knowledge entries with a similarity exceeding the threshold of 0.85 as the activated state; Apply conflict resolution rules to multiple activated knowledge entries. When different knowledge entries give conflicting suggestions, select the optimal knowledge entry as the decision-making basis based on the priority, applicable range accuracy, and historical success rate of the knowledge entries.

7. The intelligent monitoring and optimization control method for the whole process of gold ore flotation according to claim 1, wherein Build an adaptive soft sensor model and a multi-objective optimization control strategy according to the global optimization control scheme, and realize parameter optimization through the self-learning and knowledge accumulation mechanism to obtain process control parameters, including: Build a soft sensor system integrating multiple models based on the global optimization control scheme to estimate five key quality indicators, namely concentrate grade, recovery rate, mineral surface hydrophobicity, and mineral particle-bubble adhesion probability. The soft sensor model includes a flotation kinetics theory model and a historical data regression model; Combine the quality indicators estimated by the soft sensor system with the actual process parameters to construct a hierarchical control structure, including a real-time control layer, a tactical optimization layer, and a strategic decision-making layer. The control periods of each layer are set to 100 ms, 2 hours, and 8 hours respectively; Adopt an improved model predictive control algorithm for the real-time control layer, set the prediction time domain to 30 steps, the control time domain to 5 steps, and update the prediction model according to the corresponding relationship between process parameters and quality indicators; Apply a multi-objective optimization algorithm to the tactical optimization layer, set four optimization goals of maximizing concentrate grade, maximizing recovery rate, minimizing energy consumption, and minimizing reagent consumption, and dynamically adjust the objective function weights according to the current production requirements; Conduct real-time evaluation of the control effect, and establish an operation case library including operating condition characteristics, control parameter configuration, and control effect by comparing the deviation between the actual production index and the optimization target. Based on the operation case library and soft measurement feedback, incrementally update the model parameters through transfer learning, and adjust the control strategy according to the control effect evaluation results to form a self-learning closed loop and obtain the process control parameters.

8. An intelligent monitoring and optimization control system for the whole process of gold ore flotation, characterized in that, For implementing the intelligent monitoring and optimization control method for the entire gold flotation process as described in any one of claims 1-7, the intelligent monitoring and optimization control system for the entire gold flotation process includes: A grading module for collecting process parameter data in the entire gold flotation process through a distributed sensor network, detecting and grading outliers in the collected data to obtain graded data. A fusion module for constructing a multi-source heterogeneous data fusion engine based on the graded data, standardizing and multi-level fusing the data to obtain a digital twin model of the entire gold flotation process. An adjustment module for designing a multi-agent negotiation control architecture based on the digital twin model, and using a negotiation mechanism based on the contract net protocol to adjust the control parameters to obtain a globally optimized control scheme. A construction module for constructing an adaptive soft measurement model and a multi-objective optimization control strategy according to the globally optimized control scheme, and realizing parameter optimization through a self-learning and knowledge accumulation mechanism to obtain process control parameters.

9. An intelligent monitoring and optimization control device for the whole process of gold ore flotation, characterized in that, It includes a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the intelligent monitoring and optimization control method for the entire gold flotation process as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is run by the processor, the processor is caused to execute the intelligent monitoring and optimization control method for the entire gold flotation process as described in any one of claims 1 to 7.

Images