Intelligent mineral separation system and method based on digital brain

Abstract

The invention belongs to the technical field of beneficiation processes, and discloses an intelligent beneficiation system and method based on a digital brain. The method comprises the following steps: collecting raw ore property data, brain-like auditory data and brain-like sensory data; cleaning and verifying the collected data, evaluating the data, and when the data is evaluated to be in an abnormal state, giving out a corresponding early warning; when the evaluation is in a normal state, building an ore grinding process MPC model, a flotation process MPC model, a grading process MPC model and a dehydration process MPC model based on the MPC model for dynamically predicting the ore dressing process; based on the digital brain, combining each MPC model, outputting an intelligent adjustment strategy of the preparation equipment, and performing intelligent adjustment on the preparation equipment according to the intelligent adjustment strategy through brain-like movement equipment; according to the method, the adjusting strategy of the ore dressing equipment is continuously optimized based on the digital brain in combination with the multiple MPC models, so that the ore dressing process can be always kept in the optimal state.

Description

Technical Field

[0001] The present invention relates to the technical field of ore dressing processes. More specifically, the present invention relates to an intelligent ore dressing system and method based on a digital brain. Background Art

[0002] In traditional ore dressing processes, there is a lack of effective coordination between various processes. There is little real-time data interaction and linkage adjustment in links such as grinding, flotation, classification, and dewatering. For example, when the particle size of the grinding product changes, it is difficult for the flotation process to know in time and make corresponding adjustments, so that parameters such as the addition amount of flotation reagents and the aeration amount cannot be adapted to the new grinding product, thereby affecting the stability of the entire ore dressing process and the product quality.

[0003] Chinese Patent Application No. CN115718464A discloses a visual ore dressing production full-process process index optimization decision-making system, including a server, which is communicatively connected to a data acquisition unit, an ore dressing pre-analysis unit, an equipment process analysis unit, a process process analysis unit, an index comprehensive determination unit, a process optimization verification unit, and a display terminal; through data calibration, union analysis, and signal output, the invention comprehensively analyzes the state of the ore dressing production process of the visual ore dressing full process, and realizes the optimization control of the ore dressing full process by triggering different optimization decision operations. After the optimization operation is completed, the effect of the optimization operation of the visual ore dressing full-process production process is verified and analyzed by using data calibration and normalization analysis methods, so as to realize the optimization control of the ore dressing production full-process process index while saving the economic benefits of visual ore dressing.

[0004] Although the above method can meet most scenarios, through research and practical application of the above method and the existing technology, it is found that the above method and the existing technology have at least the following partial defects:

[0005] There is a lack of a detailed prediction mechanism for the dynamic changes in the ore dressing process; there is a lack of comprehensive consideration of equipment state changes, raw ore property fluctuations, and complex interactions between various process links; the monitoring and early warning of potential faults and abnormal situations during equipment operation are not comprehensive enough.

[0006] In view of this, the present invention proposes an intelligent ore dressing system and method based on a digital brain to solve the above problems. Summary of the Invention

[0007] In order to overcome the above defects of the prior art and to achieve the above object, the present invention provides the following technical solution: an intelligent ore dressing method based on a digital brain, including the following steps:

[0008] Collect the raw ore property data, brain-inspired auditory data, and brain-inspired sensory data; the raw ore property data includes the content of useful minerals, the content of impurity minerals, and the hardness of the raw ore in the raw ore; the brain-inspired auditory data includes the equipment decibel and vibration frequency; the brain-inspired sensory data includes the equipment operating temperature, the equipment energy consumption, and the process parameter data;

[0009] Clean and verify the collected data, and evaluate the brain-inspired auditory data, the equipment operating temperature, and the equipment energy consumption. When the evaluation is in an abnormal state, issue corresponding early warnings;

[0010] When the evaluation is in a normal state, build an MPC model for the grinding process, an MPC model for the flotation process, an MPC model for the classification process, and an MPC model for the dewatering process based on the MPC model, for dynamic prediction of the ore dressing process;

[0011] Based on the digital brain combined with the MPC model for the grinding process, the MPC model for the flotation process, the MPC model for the classification process, and the MPC model for the dewatering process, output the intelligent adjustment strategy of the ore dressing equipment, and use the brain-inspired motion equipment to intelligently adjust the ore dressing equipment according to the intelligent adjustment strategy.

[0012] Further, based on the MPC model, establish an MPC model for the grinding process according to the ball mill power, the ball mill reference power, the pulp concentration, the reference pulp concentration, the particle size, and the reference particle size;

[0013] Set the prediction horizon and the control horizon , where, ; in each control period , according to the current state , use the established MPC model for the grinding process to predict the grade of the grinding product at the moment in the future ; , Construct an optimization objective function for the grinding process ;

[0014] Solve the objective function through a nature-inspired optimization algorithm to obtain the grinding control sequence at the current moment , where, is the grinding control quantity at the current moment in the control period ; make the objective function minimum, take as the grinding control input at the current moment, act on the grinding process, and repeat the above process in the next control period;

[0015] An MPC model for the flotation process is established based on the flotation reagent addition amount, the reference reagent addition amount, the aeration amount of the flotation machine, the reference aeration amount, the grade of the grinding product, and the reference grade of the grinding product; an MPC model for the classification process is established based on the reference throughput; for the dewatering process, an MPC model for the dewatering process is established, and the MPC model for the dewatering process includes establishing a dynamic model for the thickening process of the thickener and a dynamic model for the filtration process of the filter press.

[0016] Furthermore, the method for outputting the intelligent adjustment strategy of the ore dressing equipment includes:

[0017] Step 1: Define the state information in the ore dressing process as the state space ;

[0018] Step 2: Define the adjustment actions for the MPC control parameters as space , and the adjustment actions include the prediction horizon, the control horizon, and the adjustment coefficients of the weight matrix;

[0019] Step 3: Preset the time steps of the prediction horizon , initialize various sensors, and initialize the state vector to 0;

[0020] Step 4: For each time step , according to the current state vector , iteratively update the state value function through the value iteration algorithm;

[0021] Step 5: Based on the updated state value function, select the action combination that maximizes the state value function , as the selected action for the current time step;

[0022] Step 6: Execute the selected action , and each ore dressing process synchronously updates the state to obtain the state vector at the next moment;

[0023] Step 7: Repeat Step 3 - Step 6, continuously optimize the intelligent adjustment strategy until the ore dressing process ends.

[0024] Furthermore, the state space includes the content of useful minerals , the content of impurity minerals , the hardness of the raw ore , the power of the ball mill , the temperature of the ball mill , the aeration amount of the flotation machine , the particle size , the pulp density , the addition amount of flotation reagents , Grinding product grade , Throughput , Grinding control input at the current moment , Flotation control input at the current moment , Classification control input at the current moment and dewatering control input at the current moment ; Based on the state space Obtain the corresponding state vector.

[0025] Further, the action space includes action vectors Obtained according to the prediction horizons of different ore dressing processes corresponding to the grinding process, flotation process, classification process, and dewatering process the control horizons, and the adjustment coefficients corresponding to the ore dressing processes.

[0026] Further, the state value function is updated according to the reward function value and the maximum state value under the next state vector predicted by the value iteration algorithm after executing the action.

[0027] Further, the method for obtaining the reward function includes:

[0028] Step 4.1, Calculate the concentrate grade improvement reward according to the target concentrate grade and the current concentrate grade;

[0029] Step 4.2, Calculate the recovery rate improvement reward according to the target recovery rate and the current recovery rate;

[0030] Step 4.3, Calculate the energy consumption reduction reward according to the initial energy consumption and the current energy consumption;

[0031] Step 4.4, Calculate the equipment wear reduction reward according to the initial wear degree of the equipment and the current wear degree;

[0032] Step 4.5, Calculate and obtain the reward function based on the sum of the concentrate grade improvement reward, the recovery rate improvement reward, the energy consumption reduction reward, and the equipment wear reduction reward .

[0033] Further, the method for predicting the maximum state value under the next state vector after executing the action by the value iteration algorithm includes:

[0034] The future output of the grinding process predicted by the grinding process MPC model the future output of the flotation process predicted by the flotation process MPC model the future output of the classification process predicted by the classification process MPC model and the future output of the dewatering process predicted by the dewatering process MPC model Obtain the next state vector, calculate the reward function value corresponding to the next state vector, calculate the state value function value corresponding to each action, and select the maximum value among the state value function values as the predicted maximum state value under the next state vector after the action is executed.

[0035] Further, the method for verifying the collected data includes:

[0036] Compare the collected data with a preset range threshold. If the collected data is within the preset range threshold, the data is valid; otherwise, it is marked as abnormal data and recollected.

[0037] The method for evaluating the operating temperature of the device and issuing corresponding warnings includes:

[0038] Calculate the change rate of the device operating temperature according to the ratio of the temperature change value of the device at different time nodes to the time interval.

[0039] Compare the change rate of the device operating temperature with a preset change rate threshold. When the change rate of the device operating temperature exceeds the preset change rate threshold, issue a corresponding warning of abnormal device operating temperature.

[0040] Further, the method for evaluating the vibration frequency and issuing corresponding warnings includes:

[0041] Compare the device decibel with a preset device decibel threshold. When the device decibel is higher than the preset device decibel threshold, judge that the device is operating abnormally and issue a corresponding warning of abnormal decibel.

[0042] Compare the vibration frequency with a preset vibration frequency threshold. When the vibration frequency is higher than the preset vibration frequency threshold, judge that the device is operating abnormally and issue a corresponding warning of abnormal vibration frequency.

[0043] The method for evaluating the energy consumption of the device and issuing corresponding warnings includes:

[0044] Accumulate the real-time power of the device within a preset time to obtain the actual energy consumption value, compare the actual energy consumption value with a preset energy consumption threshold. When the actual energy consumption value is higher than the preset energy consumption threshold, judge that the device energy consumption is abnormal and issue a warning of abnormal device energy consumption; otherwise, judge that the device is working normally.

[0045] An intelligent ore dressing system based on a digital brain implements an intelligent ore dressing method based on the digital brain, including:

[0046] Data acquisition module: It acquires raw ore property data, brain-like auditory data, and brain-like sensory data; the raw ore property data includes the content of useful minerals, the content of impurity minerals, and the hardness of the raw ore in the raw ore; the brain-like auditory data includes equipment decibels and vibration frequencies; the brain-like sensory data includes equipment operating temperature, equipment energy consumption, and process parameter data;

[0047] Data evaluation module: It cleans and verifies the acquired data, and evaluates the brain-like auditory data, equipment operating temperature, and equipment energy consumption. When the evaluation is in an abnormal state, it issues corresponding warnings;

[0048] Dynamic prediction module: When the evaluation is in a normal state, it builds MPC models for the grinding process, flotation process, classification process, and dewatering process based on the MPC model, and dynamically predicts the ore dressing process;

[0049] Intelligent ore dressing module: Based on the digital brain combined with the MPC models for the grinding process, flotation process, classification process, and dewatering process, it outputs an intelligent adjustment strategy for the ore dressing equipment, and uses the brain-like motion equipment to intelligently adjust the ore dressing equipment according to the intelligent adjustment strategy.

[0050] Technical effects and advantages of the intelligent ore dressing system and method based on the digital brain of the present invention:

[0051] The present invention acquires raw ore property data, can formulate precise ore dressing strategies according to the specific characteristics of the raw ore, helps to improve the concentrate grade and recovery rate, and enhances the quality of ore dressing products; it acquires brain-like auditory data and brain-like sensory data, and conducts corresponding processing and evaluation warnings, providing comprehensive guarantee for the reliable operation of the equipment; it also uses the MPC models of the grinding, flotation, classification, and dewatering processes for dynamic prediction, realizing the coordinated optimization of each link of ore dressing; it can effectively reduce the mismatch between links and improve the quality of ore dressing products; based on the digital brain combined with multiple MPC models, it continuously optimizes the adjustment strategy of the ore dressing equipment, enabling the ore dressing process to always operate close to the optimal state, stabilizing and enhancing the quality of ore dressing products. Description of the drawings

[0052] Figure 1 It is a schematic flow chart of the intelligent ore dressing method based on the digital brain of the present invention;

[0053] Figure 2 It is a schematic flow chart of the method for outputting the intelligent adjustment strategy of the ore dressing equipment of the present invention;

[0054] Figure 3 It is a schematic flow chart of the method for obtaining the reward function of the present invention;

[0055] Figure 4This is the structural diagram of the intelligent ore dressing system based on the digital brain of the present invention. Detailed implementation manners

[0056] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0057] Embodiment 1

[0058] Please refer to Figure 1 As shown, the intelligent ore dressing method based on the digital brain provided in this embodiment includes the following steps:

[0059] Collect raw ore property data, brain-like auditory data, and brain-like sensory data;

[0060] The raw ore property data includes the content of useful minerals, the content of impurity minerals, and the hardness of the raw ore in the raw ore; it is obtained through the detection of the collected raw ore. Collecting the raw ore property data can provide a reference standard for the automatic adjustment of the operation state of the equipment, enable the equipment to operate in the best state, and improve the ore dressing efficiency and quality.

[0061] The brain-like auditory data includes the equipment decibel and vibration frequency, and is obtained through the collection of brain-like auditory equipment; the brain-like auditory equipment includes shock wave detection equipment, vibration detection equipment, etc.; collecting the brain-like auditory data can provide real-time feedback on the operation state of the equipment, improve the optimization efficiency of the later ore dressing equipment, and can also ensure the stability of the ore dressing process and improve the quality of the ore dressing products.

[0062] The brain-like sensory data includes the equipment operating temperature, equipment energy consumption, and process parameter data; the process parameter data includes the reference power of the ball mill, the power of the ball mill, the reference pulp concentration, the pulp concentration, the reference particle size, the particle size, the reference reagent addition amount, the flotation reagent addition amount, the reference aeration amount, the aeration amount of the flotation machine, the reference grade of the ground product, the grade of the ground product (indicating the content of the useful component, that is, the target mineral in the ground product), the reference throughput, and the throughput; the equipment operating temperature data can monitor the working state of the equipment, ensure the efficient operation of the equipment, and improve the ore dressing efficiency; the equipment energy consumption can intuitively reflect the energy utilization situation of the equipment, so that when the equipment is optimized and adjusted later, it can not only ensure the ore dressing effect, but also reduce the energy consumption and improve the energy utilization efficiency; thereby reducing the ore dressing cost and improving the ore dressing benefit; the process parameter data directly affects the operation effect of the ore dressing equipment, and the operation parameters of the equipment can be accurately adjusted according to the process parameters later; ultimately improving the overall quality of the ore dressing products.

[0063] Clean and verify the collected data, and evaluate the brain-inspired auditory data and the device operating temperature. When the evaluation is in an abnormal state, issue corresponding warnings. Cleaning the data can remove outliers and incorrect data. Verifying the data can ensure that the data conforms to the physical laws and actual situations of the ore dressing process. The data after cleaning and verification can serve as the basis for the intelligent ore dressing system to make accurate decisions, ensuring the achievement of ideal ore dressing requirements. The device operating temperature is one of the key indicators reflecting the device operating state. Calculating the change rate of the device operating temperature can detect potential device failures in advance, prevent the device operating temperature from being too high, extend the service life of the device, and ensure the stability of the ore dressing process at the same time. Different device operating states and parameter settings will result in different sound characteristics. Evaluating the brain-inspired auditory data can help optimize device performance and reduce the impact of device failures on ore dressing production.

[0064] By evaluating the device energy consumption, the energy consumption situation of the device can be monitored in real time, and thus the ore dressing cost can be reduced through manual adjustment.

[0065] The methods for verifying the collected data include:

[0066] Compare the collected data with the preset range threshold. If the collected data is within the preset range threshold, the data is valid; otherwise, mark it as abnormal data and collect it again.

[0067] The methods for evaluating the device operating temperature and issuing corresponding warnings include:

[0068] Calculate the change rate of the device operating temperature according to the ratio of the temperature change value of the device at different time nodes to the time interval. The specific calculation formula is:

[0069] ;

[0070] Where, is the change rate of the device operating temperature; is the device operating temperature at time node ; is the device operating temperature at time node ; is the time node and the time node time interval;

[0071] Compare the change rate of the device operating temperature with the preset change rate threshold. When the change rate of the device operating temperature exceeds the preset change rate threshold, issue a corresponding warning for abnormal device operating temperature.

[0072] The methods for evaluating the vibration frequency and issuing corresponding warnings include:

[0073] Compare the device decibel with the preset device decibel threshold. When the device decibel is higher than the preset device decibel threshold, it is determined that the device is operating abnormally, and a corresponding decibel abnormality warning is issued;

[0074] Compare the vibration frequency with the preset vibration frequency threshold. When the vibration frequency is higher than the preset vibration frequency threshold, it is determined that the device is operating abnormally, and a corresponding vibration frequency abnormality warning is issued.

[0075] The method for evaluating the device energy consumption and issuing corresponding warnings includes:

[0076] Accumulate the real-time power within the preset time of the device to obtain the actual energy consumption value. Compare the actual energy consumption value with the preset energy consumption threshold. When the actual energy consumption value is higher than the preset energy consumption threshold, it is determined that the device energy consumption is abnormal, and a device energy consumption abnormality warning is issued; otherwise, it is determined that the device is working normally.

[0077] When the evaluation is in a normal state, based on the MPC model, build the MPC model of the grinding process, the MPC model of the flotation process, the MPC model of the classification process, and the MPC model of the dewatering process to dynamically predict the ore dressing process;

[0078] Based on the MPC model, the MPC model of the grinding process is established as follows:

[0079] ;

[0080] Among them, is the grade of the grinding product; is the grinding rate constant; is the power of the ball mill; is the reference power of the ball mill; is the pulp concentration; is the reference pulp concentration; is the particle size; is the reference particle size;

[0081] Set the prediction horizon and the control horizon , where, ; In each control period , according to the current state , use the established MPC model of the grinding process to predict the grade of the grinding product at the future moments, at the moment ;

[0082] Construct the optimization objective function of the grinding process :

[0083] ;

[0084] Among them, is the reference trajectory of the grinding product grade at time ; and are weight matrices used to balance the tracking error and the change of control action, is the change amount of the grinding control input; is the norm;

[0085] Solve the objective function through a nature-inspired optimization algorithm to obtain the grinding control sequence at the current time , where is the control period The grinding control amount at the current time in ; make the objective function the smallest, take as the grinding control input at the current time, act on the grinding process, and repeat the above process in the next control period;

[0086] Establish a flotation process MPC model based on the flotation reagent addition amount, the reference reagent addition amount, the aeration amount of the flotation machine, the reference aeration amount, the grinding product grade and the reference grinding product grade; establish a classification process MPC model based on the reference throughput; for the dewatering process, establish a dewatering process MPC model, and the dewatering process MPC model includes establishing a dynamic model of the thickening process for the thickener and a dynamic model of the filtration process for the filter press.

[0087] Based on the digital brain combined with the grinding process MPC model, the flotation process MPC model, the classification process MPC model and the dewatering process MPC model, output the intelligent adjustment strategy of the beneficiation equipment, and perform intelligent adjustment on the beneficiation equipment through the brain-like motion equipment according to the intelligent adjustment strategy.

[0088] The method for solving the objective function through a nature-inspired optimization algorithm includes:

[0089] The nature-inspired optimization algorithm can select optimization algorithms such as the particle swarm optimization algorithm or the genetic algorithm. Taking the particle swarm optimization algorithm as an example:

[0090] Initialize the particle swarm parameters: population size , particle dimension , maximum number of iterations , termination threshold , inertia weight, learning factor;

[0091] Initialize the initial state:

[0092] The current control period , the current state ( is the power of the ball mill, is the pulp concentration, is the particle size); particle Particle position , where is the particle grinding control amount at the current moment in the control period ; is the particle grinding control amount at the current moment in the control period; randomly generated within the allowable control range; the particle particle velocity is initialized to 0 or a randomly small value;

[0093] For each particle , input the control sequence into the grinding process MPC model to predict the grinding product grade at the next moments ; Substitute into the objective function to calculate the fitness value ; Minimize the fitness value as the objective function;

[0094] Record the position corresponding to the historical best fitness of the particle itself, and the position corresponding to the current optimal fitness of all particles; Update the particle velocity and position according to the iterative formulas of position and velocity in the particle swarm optimization algorithm;

[0095] If the number of iterations reaches or the change in the fitness value is less than the termination threshold , terminate the iteration; otherwise return to the step of calculating the fitness value; Select the first element of the global optimal position as the control input at the current moment and act on the grinding equipment (such as adjusting the power of the ball mill).

[0096] Referring to Figure 2 , the method for outputting the intelligent adjustment strategy of the ore dressing equipment includes:

[0097] Step 1. Define the state information in the ore dressing process as the state space ; The state space includes the content of useful minerals , the content of impurity minerals , the hardness of the raw ore , the power of the ball mill , the temperature of the ball mill , the aeration volume of the flotation machine , the particle size , the pulp density , the addition amount of flotation reagents , the grinding product grade , the throughput , the grinding control input at the current moment , the flotation control input at the current moment​ and the classification control input at the current moment and the dehydration control input at the current moment ; obtain the state vector included in the state space as follows:

[0098] .

[0099] Step 2, define the adjustment action of the MPC control parameters as space , and the adjustment action includes the prediction horizon, the control horizon, and the adjustment coefficients of the weight matrix; the action space includes the action vector specifically as follows:

[0100] ;

[0101] wherein, is the prediction horizon of the th ore dressing process; is the control horizon of the th ore dressing process; is the first adjustment coefficient of the weight matrix of the th ore dressing process; is the second adjustment coefficient of the weight matrix of the th ore dressing process; is the ore dressing process index, corresponding to the grinding process, the flotation process, the classification process, and the dehydration process respectively.

[0102] Step 3, preset the prediction horizon time step , initialize various sensors, and initialize the state vector to 0;

[0103] Step 4, for each time step , according to the current state vector , iteratively update the state value function through the value iteration algorithm;

[0104] The state value function is updated according to the reward function value and the maximum state value under the next state vector predicted by the value iteration algorithm after executing the action. The update formula of the state value function is as follows:

[0105] ;

[0106] wherein, is the state value at time step ; is the reward function value at time step ; is the discount factor, representing the importance of future rewards; is the predicted state vector at the next time step ; is the maximum state value under the next state vector after executing the action predicted by the value iteration algorithm.

[0107] Refer to Figure 3 to obtain the reward function The method includes:

[0108] Step 4.1, calculate the concentrate grade improvement reward , set the target concentrate grade as , and the current concentrate grade as , then: ; where is the concentrate grade reward coefficient. When , otherwise ;

[0109] Step 4.2, calculate the recovery rate improvement reward , set the target recovery rate as , and the current recovery rate as , then: ; where is the recovery rate reward coefficient. When , otherwise ;

[0110] Step 4.3, calculate the energy consumption reduction reward , set the initial energy consumption , and the current energy consumption as , then: ; where is the energy consumption reward coefficient. When , otherwise ;

[0111] Step 4.4, calculate the equipment wear reduction reward , the equipment wear degree is related to the equipment operation time and the equipment load , and the wear degree function is: ; where is the current equipment wear degree; is the wear coefficient, and the equipment wear is reduced by adjusting the action; obtain the equipment wear reduction reward : ; where is the equipment wear reward coefficient; is the initial wear degree;

[0112] Step 4.5, Based on the concentrate grade improvement reward , the recovery rate improvement reward , the energy consumption reduction reward and the equipment wear reduction reward calculate the reward function ;

[0113] The method for calculating the reward function includes: .

[0114] The method for predicting the maximum state value under the next state vector after performing an action through the value iteration algorithm includes:

[0115] The future output of the grinding process predicted by the grinding process MPC model , the future output of the flotation process predicted by the flotation process MPC model , the future output of the classification process predicted by the classification process MPC model and the future output of the dewatering process predicted by the dewatering process MPC model to obtain the next state vector, calculate the reward function value corresponding to the next state vector, calculate the state value function value corresponding to each action, and select the maximum value among the state value function values corresponding to all actions as the predicted maximum state value under the next state vector after performing the action.

[0116] Step 5, According to the updated state value function, select the action combination that maximizes the state value function , as the selected action for the current time step;

[0117] Step 6, Execute the selected action , and each ore dressing process synchronously updates the state to obtain the next moment state vector ;

[0118] Step 7, Repeat Step 3 - Step 6, continuously optimize the intelligent adjustment strategy until the ore dressing process ends.

[0119] By continuously repeating the above process, it is possible to gradually learn the optimal decision-making strategy, enabling the digital brain to make the optimal action decision according to the real-time state of the ore dressing process and controlling the brain-like motion equipment to execute actions; the brain-like motion equipment includes electric valves, frequency converters, intelligent equipment, cloud robots, production equipment, etc.

[0120] Embodiment 2

[0121] Please refer to Figure 4 as shown, the intelligent ore dressing system based on the digital brain provided in this embodiment includes:

[0122] Data acquisition module: It acquires raw ore property data, brain-like auditory data, and brain-like sensory data; the raw ore property data includes the content of useful minerals, impurity mineral content, and raw ore hardness in the raw ore; the brain-like auditory data includes equipment decibels and vibration frequencies; the brain-like sensory data includes equipment operating temperature, equipment energy consumption, and process parameter data;

[0123] Data evaluation module: It cleans and verifies the acquired data, and evaluates the brain-like auditory data, equipment operating temperature, and equipment energy consumption. When the evaluation is in an abnormal state, it issues corresponding warnings;

[0124] Dynamic prediction module: When the evaluation is in a normal state, it builds an MPC model for the grinding process, flotation process, classification process, and dewatering process based on the MPC model, and dynamically predicts the ore dressing process;

[0125] Intelligent ore dressing module: Based on the digital brain combined with the MPC models for the grinding process, flotation process, classification process, and dewatering process, it outputs an intelligent adjustment strategy for the ore dressing equipment, and the ore dressing equipment is intelligently adjusted according to the intelligent adjustment strategy through the brain-like motion equipment.

[0126] As mentioned above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

[0127] Finally: The above are only the preferred embodiments of the present invention and are not used to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. An intelligent mineral processing method based on digital brain, characterized in that: The steps include: Collecting raw ore property data, brain-like auditory data and brain-like sensory data; the raw ore property data includes the content of useful minerals in the raw ore, the content of impurity minerals and the hardness of the raw ore; the brain-like auditory data includes the decibel and vibration frequency of the equipment; the brain-like sensory data includes the equipment operating temperature, equipment energy consumption and process parameter data; Clean and verify the collected data, and evaluate the brain-like auditory data, equipment operating temperature, and equipment energy consumption. When the evaluation is in an abnormal state, issue a corresponding warning; When the assessment is in a normal state, the grinding process MPC model, flotation process MPC model, classification process MPC model and dehydration process MPC model are built based on the MPC model to dynamically predict the mineral processing process; Based on the digital brain combined with the MPC model of the grinding process, the MPC model of the flotation process, the MPC model of the classification process and the MPC model of the dehydration process, the intelligent adjustment strategy of the mineral processing equipment is output, and the mineral processing equipment is intelligently adjusted according to the intelligent adjustment strategy through the brain-like motion device.

2. The intelligent mineral processing method based on digital brain according to claim 1 is characterized in that: Based on the MPC model, an MPC model of the grinding process is established according to the ball mill power, the ball mill reference power, the slurry concentration, the reference slurry concentration, the particle size and the reference particle size; Set the prediction time domain and control time domain in, ; In each control cycle , according to the current state , using the established grinding process MPC model to predict future in a moment, in a moment Grinding product grade at ; Construct grinding process optimization objective function ; The objective function is solved by the nature-inspired optimization algorithm to obtain the grinding control sequence at the current moment ,in, To control the cycle The grinding control quantity at the current moment in the objective function minimum, take As the grinding control input at the current moment, it acts on the grinding process and repeats the above process in the next control cycle; An MPC model for the flotation process is established based on the amount of flotation reagent added, the benchmark amount of reagent added, the flotation machine aeration amount, the benchmark aeration amount, the grinding product grade and the benchmark grinding product grade. An MPC model for the classification process is established based on the benchmark processing volume. For the dehydration process, an MPC model for the dehydration process is established, and the MPC model for the dehydration process includes a dynamic model for the concentrating process for the concentrator and a dynamic model for the filtering process for the filter.

3. The intelligent mineral processing method based on digital brain according to claim 2 is characterized in that: The method for outputting the intelligent adjustment strategy of the mineral processing equipment includes: Step 1: Define the state information in the mineral processing process as the state space ; Step 2: Define the adjustment action of the MPC control parameters as the action space ,The adjustment actions include the adjustment coefficients of the prediction time domain, the control time domain and the weight matrix; Step 3: Preset the prediction time step , initialize various sensors and initialize the state vector is 0; Step 4: For each time step , according to the current state vector , iteratively update the state value function through the value iteration algorithm; Step 5: Based on the updated state value function, select the state value function The biggest action set , as the selected action for the current time step; Step 6: Execute the selected action , each beneficiation process updates its state synchronously to obtain the state vector at the next moment ; Step 7. Repeat steps 3 to 6 to continuously optimize the intelligent adjustment strategy until the mineral processing process is completed.

4. The intelligent mineral processing method based on digital brain according to claim 3 is characterized in that: The state space Including the content of useful minerals , impurity mineral content , Ore hardness , ball mill power , ball mill temperature , Flotation machine inflation volume ,granularity , Slurry concentration , Flotation reagent addition amount , Grinding product grade , Processing capacity , the current grinding control input , flotation control input at the current moment , the hierarchical control input at the current moment and the current dehydration control input Based on the state space Get the corresponding state vector .

5. The intelligent mineral processing method based on digital brain according to claim 3 is characterized in that: The action space Motion vectors included It is obtained according to the prediction time domain, control time domain and the corresponding adjustment coefficient of different ore dressing processes corresponding to the grinding process, flotation process, classification process and dehydration process.

6. The intelligent mineral processing method based on digital brain according to claim 3 is characterized in that: The state value function is updated according to the reward function value and the maximum state value under the next state vector after executing the action predicted by the value iteration algorithm.

7. The intelligent mineral processing method based on digital brain according to claim 6 is characterized in that: Get the reward function The methods include: Step 4.1, calculate the concentrate grade improvement reward based on the target concentrate grade and the current concentrate grade; Step 4.2: Calculate the recycling rate improvement reward based on the target recycling rate and the current recycling rate; Step 4.3, calculate the energy consumption reduction reward based on the initial energy consumption and the current energy consumption; Step 4.4, calculate the equipment wear reduction reward based on the initial wear degree and current wear degree of the equipment; Step 4.5: Calculate the reward function based on the sum of the concentrate grade improvement reward, recovery rate improvement reward, energy consumption reduction reward and equipment wear reduction reward .

8. The intelligent mineral processing method based on digital brain according to claim 7 is characterized in that: The method for predicting the maximum state value under the next state vector after executing the action by the value iteration algorithm includes: Future output of the grinding process predicted by the grinding process MPC model , the future output of the flotation process predicted by the MPC model of the flotation process , the future output of the grading process predicted by the grading process MPC model The future output of the dehydration process predicted by the dehydration process MPC model Get the next state vector, calculate the reward function value corresponding to the next state vector, calculate the corresponding state value function value for each action, and select the largest value among the state value function values ​​as the predicted maximum state value under the next state vector after executing the action.

9. The intelligent mineral processing method based on digital brain according to claim 1 is characterized in that: The method for verifying the collected data includes: Compare the collected data with the preset range threshold. If the collected data is within the preset range threshold, the data is valid. Otherwise, it is marked as abnormal data and collected again. Methods for assessing equipment operating temperature and issuing corresponding warnings include: Calculate the change rate of the equipment operating temperature based on the ratio of the temperature change value of the equipment at different time nodes to the time interval; The change rate of the equipment operating temperature is compared with a preset change rate threshold. When the change rate of the equipment operating temperature exceeds the preset change rate threshold, a corresponding equipment operating temperature abnormality warning is issued.

10. The intelligent mineral processing method based on digital brain according to claim 1 is characterized in that: The method for evaluating the vibration frequency and issuing a corresponding warning includes: Compare the device decibel with the preset device decibel threshold. When the device decibel is higher than the preset device decibel threshold, the device is judged to be operating abnormally and a corresponding decibel abnormality warning is issued; The vibration frequency is compared with a preset vibration frequency threshold. When the vibration frequency is higher than the preset vibration frequency threshold, the device is judged to be operating abnormally, and a corresponding vibration frequency abnormality warning is issued; Methods for evaluating equipment energy consumption and issuing corresponding warnings include: The actual energy consumption value is obtained by accumulating the real-time power of the device within the preset time, and the actual energy consumption value is compared with the preset energy consumption threshold. When the actual energy consumption value is higher than the preset energy consumption threshold, the device energy consumption is judged to be abnormal and an abnormal energy consumption warning is issued; otherwise, the device is judged to be working normally.

11. An intelligent mineral processing system based on a digital brain, implementing the intelligent mineral processing method based on a digital brain as described in any one of claims 1 to 10, characterized in that: include: Data acquisition module: collects raw ore property data, brain-like auditory data and brain-like sensory data; the raw ore property data includes the content of useful minerals in the raw ore, the content of impurity minerals and the hardness of the raw ore; the brain-like auditory data includes the decibel and vibration frequency of the equipment; the brain-like sensory data includes the equipment operating temperature, equipment energy consumption and process parameter data; Data evaluation module: cleans and verifies the collected data, and evaluates the brain-like auditory data, equipment operating temperature, and equipment energy consumption. When the evaluation is in an abnormal state, a corresponding warning is issued; Dynamic prediction module: When the evaluation is in a normal state, the grinding process MPC model, flotation process MPC model, classification process MPC model and dehydration process MPC model are built based on the MPC model to dynamically predict the mineral processing process; Intelligent mineral processing module: Based on the digital brain combined with the MPC model of the grinding process, the MPC model of the flotation process, the MPC model of the classification process and the MPC model of the dehydration process, the intelligent adjustment strategy of the mineral processing equipment is output, and the mineral processing equipment is intelligently adjusted according to the intelligent adjustment strategy through the brain-like motion equipment.

Images