Photoelectric separation parameter optimization method and system for copper-containing waste rock

Abstract

The invention relates to the technical field of mineral production, in particular to a photoelectric separation parameter optimization method and system for copper-containing waste rocks, and the method comprises the following steps: obtaining physical and chemical characteristics of a copper-containing waste rock sample by utilizing a photoelectric sensor, and obtaining separation parameters and separation indexes by utilizing a multi-component separation test of a dry separator on the copper-containing waste rock sample; sorting an effect prediction model by using a neural network model; converting the sorting effect prediction model into an optimization target, solving the optimization target by using an optimization algorithm to obtain an optimal sorting parameter, and performing reinforcement learning on the sorting effect prediction model by using the optimal sorting parameter; and establishing a sorting parameter optimization model by using the neural network model. According to the method, the neural network model and the multi-target optimization algorithm are combined, the optimal sorting parameters of the copper-containing waste rocks in different grade intervals can be objectively determined, accurate optimization of the sorting parameters is achieved, efficient adaptation of raw material properties and sorting equipment parameters is achieved, and optimization of the sorting effect is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mineral production, in particular to a photoelectric sorting parameter optimization method and system for copper-containing waste rock. BACKGROUND

[0002] Copper-containing waste rock is an important by-product in the process of mining production, and its efficient utilization has become a research hotspot in the field of mining. Photoelectric sorting technology, as an efficient sorting technology based on the physical properties of minerals (such as color, shape, reflectivity, etc.), has the advantages of high sorting efficiency, good environmental performance, high intelligence, and great potential in copper-containing waste rock resource recovery.

[0003] However, the existing photoelectric sorting technology still has some technical bottlenecks when dealing with copper-containing waste rock. On the one hand, the setting of sorting parameters has certain human subjectivity and is not fine enough, resulting in insufficient stability and reliability of sorting effect, which makes it difficult to meet the needs of industrial production; on the other hand, the sorting parameters for different grade intervals of copper-containing waste rock lack targeted optimization, and the matching relationship between raw material properties, equipment performance and sorting process is not fully considered, affecting the overall sorting efficiency and economy. SUMMARY

[0004] The purpose of the present application is to provide a photoelectric sorting parameter optimization method for copper-containing waste rock to solve the technical problems of insufficient setting precision of sorting parameters and lack of targeted optimization in the prior art.

[0005] To solve the above technical problems, the present application specifically provides the following technical solutions: A photoelectric sorting parameter optimization method for copper-containing waste rock, comprising the following steps: Using a photoelectric sensor to obtain the physicochemical properties of a copper-containing waste rock sample, and using a dry separator to obtain sorting parameters and sorting indexes through multiple sorting tests on the copper-containing waste rock sample; Using a neural network model to establish a mapping relationship between the physicochemical properties, sorting parameters and sorting indexes, and obtaining a sorting effect prediction model for predicting the sorting indexes according to the physicochemical properties and sorting parameters; Converting the sorting effect prediction model into an optimization target, using an optimization algorithm to solve the optimization target to obtain optimal sorting parameters for achieving optimal sorting indexes of the physicochemical properties, and using the optimal sorting parameters to enhance the learning of the sorting effect prediction model; Using a neural network model to establish a mapping relationship between the physicochemical properties, sorting indexes and optimal sorting parameters, and obtaining a sorting parameter optimization model for predicting the optimal sorting parameters according to the physicochemical properties and sorting indexes; The optimal sorting parameters of the dry separator are determined in real time by using a sorting parameter optimization model based on the physicochemical properties of the copper-containing waste rock to be sorted in real time and the sorting index that matches the preset sorting requirements.

[0006] As a preferred embodiment of the present invention, the physical and chemical properties include color, shape, reflectivity, texture, and volume; the sorting parameters include density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening time; and the sorting indicators include concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate. Among them, the quality of Cu in copper-containing waste rock samples before sorting was determined. Weight of copper-containing waste rock samples before sorting The Cu content of copper-containing waste rock samples that have been sorted and entered the concentrate category is used as the Cu content of the concentrate. The weight of copper-containing waste rock samples that entered the concentrate category after sorting was determined. ; pass Calculate the Cu recovery rate; pass Calculate the enrichment ratio; pass Calculate the waste rock disposal rate.

[0007] As a preferred embodiment of the present invention, the method for constructing a sorting effect prediction model includes: Using the physicochemical properties and sorting parameters of copper-containing waste rock samples as input terms to a neural network model, and the sorting indices of the copper-containing waste rock samples as output terms, a multilayer perceptron is trained to obtain a sorting performance prediction model. In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron. The loss function of the sorting effect prediction model In the formula, For the true value of the sorting index, These are the true values ​​for Cu grade in concentrate, Cu recovery rate, enrichment ratio, and waste rock disposal rate, respectively. This is the cross-entropy operation formula.

[0008] As a preferred embodiment of the present invention, the method for solving the optimal sorting parameters includes: Sorting effect prediction model Convert to a multi-objective optimization function In the formula, max is the maximization identifier; The physicochemical properties of each copper-containing waste rock sample were fixed one by one. Using the threshold range of the dry separator's sorting parameters as the search space, a genetic algorithm or particle swarm optimization algorithm was employed. Assuming the fitness function, the multi-objective optimization function is solved to obtain the optimal sorting parameters for each copper-containing waste rock sample. In the formula , for The weight, for The normalized result.

[0009] As a preferred embodiment of the present invention, the method for performing reinforcement learning on a sorting effect prediction model using optimal sorting parameters includes: The physicochemical properties s and optimal sorting parameters of copper-containing waste rock samples were analyzed. Separation indexes from copper-containing waste rock Construct an augmented dataset; Using augmented datasets to predict sorting performance The model was retrained to obtain an enhanced prediction model for sorting performance.

[0010] As a preferred embodiment of the present invention, the method for constructing the sorting parameter optimization model includes: To enhance the centralized physical properties of the data and sorting indicators As input, with optimal sorting parameters For output, a multilayer perceptron is trained to obtain the sorting parameter optimization model. In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters output by the model are optimized to improve the sorting parameters. MLP stands for Multilayer Perceptron. The loss function of the sorting parameter optimization model ; In the formula, To enhance the optimal sorting parameters in the dataset, For the sorting effect prediction model based on Output sorting metrics for The fitness value of the corresponding sorting index, for The fitness value, where min is the minimum identifier. This is the cross-entropy operation formula.

[0011] As a preferred embodiment of the present invention, the present invention provides a photoelectric sorting parameter optimization system for copper-containing waste rock, which is applied to a photoelectric sorting parameter optimization method for copper-containing waste rock. The system includes: Photoelectric sensors are used to acquire the physicochemical properties of copper-containing waste rock samples; The dry separator module is used for multi-component sorting tests on copper-containing waste rock samples to obtain sorting parameters and sorting indices. A computer processor is used to establish a mapping relationship between the physical and chemical properties, sorting parameters, and sorting indicators using a neural network model, thereby obtaining a sorting effect prediction model that predicts the sorting indicators based on the physical and chemical properties and sorting parameters; the sorting effect prediction model is converted into an optimization objective, and an optimization algorithm is used to solve the optimization objective to obtain the optimal sorting parameters for achieving the optimal sorting indicators based on the physical and chemical properties, and the optimal sorting parameters are used to perform reinforcement learning on the sorting effect prediction model; a sorting parameter optimization model is used to establish a mapping relationship between the physical and chemical properties, sorting indicators, and optimal sorting parameters using a neural network model, thereby obtaining a sorting parameter optimization model that predicts the optimal sorting parameters based on the physical and chemical properties of the copper-containing waste rock to be sorted in real time and the sorting indicators that match the preset sorting requirements, and the optimal sorting parameters of the dry separator are determined in real time using the sorting parameter optimization model.

[0012] As a preferred embodiment of the present invention, the sorting effect prediction model In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron.

[0013] As a preferred embodiment of the present invention, the sorting parameter optimization model In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters are output by the model to optimize the sorting parameters. MLP stands for Multilayer Perceptron.

[0014] As a preferred embodiment of the present invention, the Cu recovery rate The calculation formula is: ; The enrichment ratio ER is calculated using the following formula: ; The formula for calculating the waste rock disposal rate P is as follows: ; In the formula, To determine the Cu content in copper-containing waste rock samples before sorting, This represents the weight of the copper-containing waste rock sample before sorting. For Cu concentrate, This represents the weight of the copper-containing waste rock sample that has been sorted and is now classified as a concentrate.

[0015] Compared with the prior art, the present invention has the following advantages: This invention utilizes a combination of neural network models and multi-objective optimization algorithms to objectively determine the optimal sorting parameters for copper-containing waste rock of different grade ranges, achieving precise optimization of sorting parameters, efficient adaptation of raw material properties and sorting equipment parameters, and ensuring optimal sorting results. Attached Figure Description

[0016] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely exemplary, and those skilled in the art can derive other embodiments based on the provided drawings without creative effort.

[0017] Figure 1 This is a flowchart of the photoelectric sorting parameter optimization method provided in an embodiment of the present invention; Figure 2 This is a block diagram of the photoelectric sorting parameter optimization system provided in an embodiment of the present invention; Figure 3 The image shows the photoelectric sorting results provided in an embodiment of the present invention.

[0018] The labels in the diagram represent the following: 1. Buffer silo; 2. Vibrating discharge hopper; 3. Vibrating distributor; 4. Feeding belt; 5. Computer processor; 6. Photoelectric sensor; 7. Dust collector; 8. Air compressor; 9. Air tank; 10. Refrigerated dryer; 11. Diverter filter; 12. Nozzle; 13. Tailings belt; 14. Concentrate belt; 15. Belt scale on tailings belt; 16. Belt scale on concentrate belt. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] like Figure 1As shown, this invention provides a method for optimizing photoelectric sorting parameters for copper-containing waste rock, comprising the following steps: The physicochemical properties of copper-containing waste rock samples (with different quality ranges) were obtained using photoelectric sensors, and the sorting parameters and sorting indices were obtained by multi-component sorting tests of the copper-containing waste rock samples using a dry separator. A neural network model is used to establish the mapping relationship between physical properties, sorting parameters and sorting indicators, and a sorting effect prediction model is obtained to predict the sorting indicators based on physical properties and sorting parameters. The sorting effect prediction model is transformed into an optimization objective. The optimization objective is solved using an optimization algorithm to obtain the optimal sorting parameters that enable the physical and chemical properties to reach the optimal sorting index. The optimal sorting parameters are then used to perform reinforcement learning on the sorting effect prediction model. By using a neural network model, a mapping relationship is established between physical properties, sorting indices and optimal sorting parameters, resulting in a sorting parameter optimization model that predicts the optimal sorting parameters based on physical properties and sorting indices. The optimal sorting parameters for the dry separator are determined in real time using a sorting parameter optimization model based on the physicochemical properties of the copper-containing waste rock to be sorted and the sorting indices that match the preset sorting requirements. The sorting results obtained based on the optimal sorting parameters are as follows: Figure 3 As shown in the red box, the results indicate that this invention can effectively select concentrate products.

[0021] The physical and chemical properties include color, shape, reflectivity, texture, and volume. The sorting parameters include density threshold, vibrating feeder frequency (corresponding to vibrating distributor 3), belt speed (corresponding to feed belt 4, tailings belt 13, and concentrate belt 14), high-pressure air injection delay time (corresponding to air compressor 8), and spray valve opening time (corresponding to nozzle 12, or pneumatic plate). The sorting indicators include concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate. The function of the tailings conveyor belt is to allow the waste rock products, which account for a large proportion after sorting, to fall freely into the tailings conveyor belt. The function of the concentrate conveyor belt is to allow the concentrate product, which accounts for a smaller proportion after sorting, to fall freely into the sorting product conveyor belt after being separated by air jets from the nozzles.

[0022] Among them, the quality of Cu in copper-containing waste rock samples before sorting was determined. Weight of copper-containing waste rock samples before sorting The Cu content of copper-containing waste rock samples that have been sorted and entered the concentrate category is used as the Cu content of the concentrate. The weight of copper-containing waste rock samples that entered the concentrate category after sorting was determined by belt scale. ; pass Calculate the Cu recovery rate; pass Calculate the enrichment ratio; pass Calculate the waste rock disposal rate.

[0023] This invention involves selecting copper-bearing ore samples from different grade ranges for sorting tests. The sorting parameters of a dry separator are randomly set to sort the copper-bearing ore samples, and the sorting results are recorded, such as the weight of the ore sample sorted onto the concentrate conveyor belt and the Cu grade of the concentrate is determined. Based on the total ore sample weight and total ore sample Cu grade recorded before sorting, sorting indicators are calculated, and the sorting effect corresponding to each sorting parameter is intuitively quantified.

[0024] This invention uses data accumulated from sorting experiments to train a neural network, constructing a neural network model that predicts the corresponding sorting indicators (sorting effect) based on physical and chemical properties and sorting parameters, i.e., a sorting effect prediction model.

[0025] This invention merely randomly combines sorting parameters and physicochemical properties in the sorting sample to obtain the corresponding sorting index (sorting effect). It cannot determine the optimal sorting parameters and the corresponding optimal sorting index (sorting effect) under a specific physicochemical property. Therefore, this invention utilizes a multi-objective optimization algorithm to solidify the physicochemical properties, thus modifying the original sorting effect prediction model. The physical and chemical properties s in the middle are weakened and transformed into form.

[0026] When optimizing the sorting parameters, the target to be optimized corresponds to each index in the sorting index, namely, maximizing the Cu grade of the concentrate. Maximize Cu recovery rate Maximize enrichment ratio And to maximize the waste rock disposal rate Thus, a search and solution can be performed within the parameter allowable range of the dry separator. ,get The optimal solution is obtained, thus yielding the most suitable sorting parameters for each physical and chemical property.

[0027] After obtaining the optimal sorting parameters, this invention establishes a new matching relationship between physical properties and sorting parameters (an enhanced dataset). This matching relationship represents a high-quality, high-value data point located in the globally optimal performance region. Using this data to retrain the original sorting effect prediction model enhances the model's prediction accuracy in and around the optimal region. In the next multi-objective optimization, the search based on this more accurate model will find a solution closer to the true optimum, creating a positive feedback loop. Simultaneously, the model continuously absorbs new data, expanding its knowledge boundaries to better adapt to changes in raw materials. It is no longer a static, one-off model, but a dynamic model that evolves alongside the production process, with its generalization ability and robustness continuously increasing over time.

[0028] After training the sorting effect prediction model, this invention yields a correlation formula between sorting parameters and sorting indicators (sorting effect), which is then applied to the multi-objective optimization of sorting parameters. Methods for constructing sorting performance prediction models include: Using the physicochemical properties and sorting parameters of copper-containing waste rock samples as input terms to a neural network model, and the sorting indices of the copper-containing waste rock samples as output terms, a multilayer perceptron is trained to obtain a sorting performance prediction model. In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron. Loss function of sorting effect prediction model In the formula, For the true value of the sorting index, These are the true values ​​for Cu grade in concentrate, Cu recovery rate, enrichment ratio, and waste rock disposal rate, respectively. This is the cross-entropy operation formula.

[0029] Methods for solving the optimal sorting parameters include: Sorting effect prediction model Convert to a multi-objective optimization function In the formula, max is the maximization identifier; The physicochemical properties of each copper-containing waste rock sample were fixed one by one. Using the threshold range of the dry separator's sorting parameters as the search space, a genetic algorithm or particle swarm optimization algorithm was employed. Assuming a fitness function, a multi-objective optimization function is solved to obtain the optimal sorting parameters for each copper-containing waste rock sample. In the formula , for The weight (set as needed). for The normalization result (eliminating dimensional errors).

[0030] Taking genetic algorithms as an example: Step 1, Initialization: Randomly generate a population of N individuals. Each individual represents a possible combination of parameters, for example, x_j=[ [] (Represents a set of parameter settings). Each individual is represented by a "chromosome", which is usually encoded in binary or real numbers.

[0031] The second step is to invoke the neural network. Evaluate: After decoding each individual (parameter combination) in the population, the data is input into the trained neural network model. In the middle, calculate the fitness F of each individual. Fitness is an indicator that measures whether an individual is good or bad.

[0032] Third step, select: Objective: To select superior individuals from the current population to serve as parents for the next generation. Individuals with higher fitness have a greater probability of being selected.

[0033] Methods: Roulette wheel selection or tournament selection are used. In roulette wheel selection, the probability of each individual being selected is proportional to its fitness. In tournament selection, several individuals are randomly selected, and the one with the highest fitness wins.

[0034] Step 4, Cross: Objective: To simulate biological hybridization and create new offspring from two parent individuals, aiming to combine the advantages of both parents.

[0035] Method: Single-point intersection, for example, parent A:[ Parent generation B:[ Randomly select a crossover point, exchange some genes, and you may get offspring:[ ].

[0036] Step 5, Mutation: Objective: To introduce new genetic material by randomly altering certain genes in individuals with a small probability, thereby maintaining population diversity and avoiding local optima.

[0037] Methods: bit flipping or Gaussian perturbation, for example, child [ The second gene may mutate and become [ ].

[0038] Step 6: Formation of a new population and iteration: The new individuals produced through selection, crossover, and mutation form a new population, replacing the old population.

[0039] Repeat steps two through five (evaluation → selection → crossover → mutation) to evolve generation after generation.

[0040] Termination condition: The algorithm stops when the maximum number of iterations is reached, or when the optimal fitness no longer significantly improves after several generations. The individual with the highest fitness across all generations is output as the recommended optimal sorting parameter.

[0041] Methods for using optimal sorting parameters to perform reinforcement learning on sorting performance prediction models include: The physicochemical properties s and optimal sorting parameters of copper-containing waste rock samples were analyzed. Separation indexes from copper-containing waste rock Construct an augmented dataset; Using augmented datasets to predict sorting performance The model was retrained to obtain an enhanced prediction model for sorting performance.

[0042] Methods for constructing sorting parameter optimization models include: To enhance the centralized physical properties of the data and sorting indicators As input, with optimal sorting parameters For the output, a multilayer perceptron is trained to obtain a sorting parameter optimization model. In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters output by the model are optimized to improve the sorting parameters. MLP stands for Multilayer Perceptron. Loss function of sorting parameter optimization model ; In the formula, To enhance the optimal sorting parameters in the dataset, For the sorting effect prediction model based on Output sorting metrics for The fitness value of the corresponding sorting index, for The fitness value, where min is the minimum identifier. This is the cross-entropy operation formula.

[0043] Furthermore, based on the augmented dataset, this invention establishes a sorting parameter optimization model that predicts the optimal sorting parameters based on physicochemical properties and sorting indices. This model is the opposite of the prediction logic of the sorting effect prediction model. When the model is put into practical use after training, the non-sorting indices usually refer to the sorting requirements or the desired sorting effect. This model can determine the optimal sorting parameters based on sorting requirements (e.g., Cu grade of concentrate > 15%, recovery rate > 65%, etc.) under a specific physicochemical property.

[0044] Therefore, constructing a sorting parameter optimization model can directly determine the optimal sorting parameters without the need for sorting experiments or multi-objective optimization solutions, thus improving efficiency.

[0045] This invention employs a three-part loss term when training the sorting parameter optimization model. The first part is... It is the expected model prediction result. With truth value (that is, augmenting the dataset) If the loss is consistent with the model's loss, training with this loss ensures that the prediction loss of the sorting parameter optimization model is minimized and the prediction accuracy is maximized.

[0046] The second part is ,in It is the prediction result of the sorting parameter optimization model. Substitute these into the sorting effect prediction model to obtain the sorting index (sorting effect) predicted by the sorting effect prediction model. It is the basis of expectations Obtain the sorting indicators (sorting effect). The sorting index (sorting requirements) input to the sorting parameter optimization model. Consistent, that is, able to pass get It can also be done through Restore By utilizing a two-way correspondence, this further ensures The accuracy of the prediction.

[0047] The third part is ,in It is the multi-objective optimization function in multi-objective optimization algorithms. The calculated sorting index (sorting effect) should correspond to the maximum fitness, while the sorting index (sorting requirements) should correspond to the maximum fitness. This should also correspond to the maximum fitness value. It is an expectation The obtained fitness value and The obtained fitness values ​​are consistent, and using these as the loss for training can ensure that the fitness value loss of the sorting parameter optimization model is minimized and the prediction accuracy is maximized.

[0048] Therefore, based on the combination of the above three parts, the optimal prediction performance of the sorting parameter optimization model is obtained, and the accurate prediction of sorting parameters is achieved.

[0049] like Figure 2 As shown, this invention provides a photoelectric sorting parameter optimization system for copper-containing waste rock, applied to a photoelectric sorting parameter optimization method for copper-containing waste rock. The system includes: Photoelectric sensors are used to acquire the physicochemical properties of copper-containing waste rock samples; The dry separator module is used for multi-component sorting tests on copper-containing waste rock samples to obtain sorting parameters and sorting indices. A computer processor is used to establish a mapping relationship between physical and chemical properties, sorting parameters, and sorting indicators using a neural network model, thereby obtaining a sorting effect prediction model that predicts the sorting indicators based on physical and chemical properties and sorting parameters. This model is then converted into an optimization objective, and an optimization algorithm is used to solve the objective to obtain the optimal sorting parameters that enable the physical and chemical properties to achieve the optimal sorting indicators. The optimal sorting parameters are then used to perform reinforcement learning on the sorting effect prediction model. Furthermore, a sorting parameter optimization model is established using a neural network model to establish a mapping relationship between physical and chemical properties, sorting indicators, and optimal sorting parameters, thereby obtaining a sorting parameter optimization model that predicts the optimal sorting parameters based on physical and chemical properties and sorting indicators. Finally, the optimal sorting parameters for the dry separator are determined in real-time based on the physical and chemical properties of the copper-containing waste rock to be sorted and the sorting indicators that match the preset sorting requirements. The computer processor also sets the parameters of the dry separator based on the optimal sorting parameters in order to control the dry separator.

[0050] Sorting effect prediction model In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron.

[0051] Sorting parameter optimization model In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters are output by the model to optimize the sorting parameters. MLP stands for Multilayer Perceptron.

[0052] Cu recovery rate The calculation formula is: ; The formula for calculating the enrichment ratio ER is as follows: ; The formula for calculating the waste rock disposal rate P is: ; In the formula, To determine the Cu content in copper-containing waste rock samples before sorting, This represents the weight of the copper-containing waste rock sample before sorting. For Cu concentrate, This represents the weight of the copper-containing waste rock sample that has been sorted and is now classified as a concentrate.

[0053] This invention utilizes a combination of neural network models and multi-objective optimization algorithms to objectively determine the optimal sorting parameters for copper-containing waste rock of different grade ranges, achieving precise optimization of sorting parameters, efficient adaptation of raw material properties and sorting equipment parameters, and ensuring optimal sorting results.

[0054] The above embodiments are merely exemplary embodiments of this application and are not intended to limit this application. The scope of protection of this application is defined by the claims. Those skilled in the art can make various modifications or equivalent substitutions to this application within its substance and scope of protection, and such modifications or equivalent substitutions should also be considered to fall within the scope of protection of this application.

Claims

1. A method for optimizing photoelectric sorting parameters for copper-containing waste rock, characterized in that, Includes the following steps: The physicochemical properties of copper-containing waste rock samples were obtained using photoelectric sensors, and the sorting parameters and sorting indices were obtained by multi-component sorting tests on the copper-containing waste rock samples using a dry separator. A neural network model is used to establish the mapping relationship between the physical properties, sorting parameters and sorting indicators, so as to obtain a sorting effect prediction model that predicts the sorting indicators based on the physical properties and sorting parameters. The sorting effect prediction model is transformed into an optimization objective. The optimization objective is solved using an optimization algorithm to obtain the optimal sorting parameters that enable the physical and chemical properties to reach the optimal sorting index. The optimal sorting parameters are then used to perform reinforcement learning on the sorting effect prediction model. A neural network model is used to establish the mapping relationship between the physical properties, sorting indicators and optimal sorting parameters, so as to obtain a sorting parameter optimization model that predicts the optimal sorting parameters based on the physical properties and sorting indicators. The optimal sorting parameters of the dry separator are determined in real time by using a sorting parameter optimization model based on the physicochemical properties of the copper-containing waste rock to be sorted in real time and the sorting index that matches the preset sorting requirements.

2. The method for optimizing photoelectric sorting parameters for copper-containing waste rock according to claim 1, characterized in that: The physical and chemical properties include color, shape, reflectivity, texture, and volume. The sorting parameters include density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening time. The sorting indicators include concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate. Among them, the quality of Cu in copper-containing waste rock samples before sorting was determined. Weight of copper-containing waste rock samples before sorting The Cu content of copper-containing waste rock samples that have been sorted and entered the concentrate category is used as the Cu content of the concentrate. The weight of copper-containing waste rock samples that entered the concentrate category after sorting was determined. ; pass Calculate the Cu recovery rate; pass Calculate the enrichment ratio; pass Calculate the waste rock disposal rate.

3. The method for optimizing photoelectric sorting parameters for copper-containing waste rock according to claim 2, characterized in that: Methods for constructing sorting performance prediction models include: Using the physicochemical properties and sorting parameters of copper-containing waste rock samples as input terms to a neural network model, and the sorting indices of the copper-containing waste rock samples as output terms, a multilayer perceptron is trained to obtain a sorting performance prediction model. In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron. The loss function of the sorting effect prediction model In the formula, For the true value of the sorting index, These are the true values ​​for Cu grade in concentrate, Cu recovery rate, enrichment ratio, and waste rock disposal rate, respectively. This is the cross-entropy operation formula.

4. The method for optimizing photoelectric sorting parameters for copper-containing waste rock according to claim 3, characterized in that: Methods for solving the optimal sorting parameters include: Sorting effect prediction model Convert to a multi-objective optimization function In the formula, max is the maximization identifier; The physicochemical properties of each copper-containing waste rock sample were fixed one by one. Using the threshold range of the dry separator's sorting parameters as the search space, a genetic algorithm or particle swarm optimization algorithm was employed. Assuming the fitness function, the multi-objective optimization function is solved to obtain the optimal sorting parameters for each copper-containing waste rock sample. In the formula , for The weight, for The normalized result.

5. The method for optimizing photoelectric sorting parameters for copper-containing waste rock according to claim 4, characterized in that: Methods for using optimal sorting parameters to perform reinforcement learning on sorting performance prediction models include: The physicochemical properties s and optimal sorting parameters of copper-containing waste rock samples were analyzed. Separation indexes from copper-containing waste rock Construct an augmented dataset; Using augmented datasets to predict sorting performance The model was retrained to obtain an enhanced prediction model for sorting performance.

6. The method for optimizing photoelectric sorting parameters for copper-containing waste rock according to claim 5, characterized in that: Methods for constructing sorting parameter optimization models include: To enhance the centralized physical properties of the data and sorting indicators As input, with optimal sorting parameters For output, a multilayer perceptron is trained to obtain the sorting parameter optimization model. In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters output by the model are optimized to improve the sorting parameters. MLP stands for Multilayer Perceptron. The loss function of the sorting parameter optimization model ; In the formula, To enhance the optimal sorting parameters in the dataset, For the sorting effect prediction model based on Output sorting metrics for The fitness value of the corresponding sorting index, for The fitness value, where min is the minimum identifier. This is the cross-entropy operation formula.

7. A photoelectric sorting parameter optimization system for copper-containing waste rock, characterized in that, The system, which is applied to the photoelectric sorting parameter optimization method for copper-containing waste rock according to any one of claims 1-6, comprises: Photoelectric sensors are used to acquire the physicochemical properties of copper-containing waste rock samples; The dry separator module is used for multi-component sorting tests on copper-containing waste rock samples to obtain sorting parameters and sorting indices. A computer processor is used to establish a mapping relationship between the physical and chemical properties, sorting parameters, and sorting indicators using a neural network model, thereby obtaining a sorting effect prediction model that predicts the sorting indicators based on the physical and chemical properties and sorting parameters; the sorting effect prediction model is converted into an optimization objective, and an optimization algorithm is used to solve the optimization objective to obtain the optimal sorting parameters for achieving the optimal sorting indicators based on the physical and chemical properties, and the optimal sorting parameters are used to perform reinforcement learning on the sorting effect prediction model; a sorting parameter optimization model is used to establish a mapping relationship between the physical and chemical properties, sorting indicators, and optimal sorting parameters using a neural network model, thereby obtaining a sorting parameter optimization model that predicts the optimal sorting parameters based on the physical and chemical properties of the copper-containing waste rock to be sorted in real time and the sorting indicators that match the preset sorting requirements, and the optimal sorting parameters of the dry separator are determined in real time using the sorting parameter optimization model.

8. The photoelectric sorting parameter optimization system for copper-containing waste rock according to claim 7, characterized in that: The sorting effect prediction model In the formula, The sorting index is the output of the sorting effect prediction model. These represent the concentrate Cu grade, Cu recovery rate, enrichment ratio, and waste rock disposal rate output by the sorting effect prediction model, respectively, where s is the physicochemical property. For sorting parameters, These are density threshold, vibrating feeder frequency, belt speed, high-pressure air injection delay time, and spray valve opening duration, respectively. MLP stands for Multilayer Perceptron.

9. The photoelectric sorting parameter optimization system for copper-containing waste rock according to claim 8, characterized in that: The sorting parameter optimization model In the formula, The sorting index is the input to the sorting parameter optimization model, where s is the physical and chemical property. The optimal sorting parameters are output by the model to optimize the sorting parameters. MLP stands for Multilayer Perceptron.

10. The photoelectric sorting parameter optimization system for copper-containing waste rock according to claim 7, characterized in that: Cu recovery rate The calculation formula is: ; The enrichment ratio ER is calculated using the following formula: ; The formula for calculating the waste rock disposal rate P is as follows: ; In the formula, To determine the Cu content in copper-containing waste rock samples before sorting, This represents the weight of the copper-containing waste rock sample before sorting. For Cu concentrate, This represents the weight of the copper-containing waste rock sample that has been sorted and is now classified as a concentrate.

Images