Multi-ore joint production ore blending method and system based on multi-target ore grade constraint

Abstract

The invention discloses a multi-ore joint production ore blending method based on multi-target ore grade constraints. The method comprises the steps that S1, N ore spots are selected, and the total ore quantity T containing M ore grades is set as an ore blending target; s2, constructing an annular neighborhood topological structure by a multi-target ore blending optimization model, and performing iteration and position updating processing by adopting an improved annular neighborhood topological hybrid particle swarm optimization algorithm to obtain an optimal solution of a target function; selecting the historical optimal solution of the individual particles as the historical optimal solution of the annular neighborhood, and taking the optimal solutions of the objective functions in all the annular neighborhoods as the global historical optimal solution; and S3, stopping iteration of the multi-target ore blending optimization model, and outputting optimal ore blending amounts corresponding to the N ore deposits respectively. According to the method, the multi-ore-ore joint optimization ore blending scheme output under the constraint of the multi-target ore grade is realized, the element errors are balanced through the adaptive target function, the element grade requirements of the multi-target ore can be met, and the product quality stability is improved.

Description

Technical Field

[0001] This invention relates to the field of multi-objective intelligent ore blending, and in particular to a multi-ore joint production blending method and system based on multi-objective ore grade constraints. Background Technology

[0002] With declining mineral resource grades, stricter environmental standards, and increasing market demand for consistent product quality, a key aspect of consistent product quality is ensuring that the ore grade at the mining site meets target requirements. Ore grade is a core indicator of ore quality, referring to the content of useful components (such as metallic elements and valuable minerals) in the ore. As product quality requirements (especially for high-end products) rise, and the large-scale mining of mineral resources leads to a decline in high-grade ore resources, it is difficult to meet the ore grade requirements for producing high-end and high-quality products unless relying solely on a single high-grade mineral resource. However, high-grade mineral resources are relatively limited and face a supply shortage. Therefore, it is necessary to rationally and effectively blend minerals from multiple grade mineral resource sites to meet the target ore grade required for production. Multi-grade ore blending is a technical means in mining production that scientifically adjusts the proportion of ores of different grades (high, medium, and low) to stabilize the grade of the blended ore within a target range, so as to meet the requirements of subsequent beneficiation and smelting processes or the efficient utilization of resources. Its core is to balance the fluctuation of ore grade and achieve the unity of maximizing resource utilization and production stability.

[0003] Multi-grade ore blending offers the following advantages: 1. It avoids fluctuations in beneficiation or smelting process parameters (such as reagent consumption, energy consumption, and product recovery rate) caused by excessively high or low grades in a single batch of ore, ensuring continuous and stable production. 2. By blending low-grade ore (which has low economic value when mined alone) with high and medium-grade ore, it improves resource utilization and avoids waste caused by "abandoning poor-grade ore for rich-grade ore." 3. Rational ore blending can reduce the mining and processing costs per unit product, while extending the mine's service life by balancing the consumption of ores of different grades. Multi-grade ore blending is a core element of "refined management" in mines, especially for mines with large grade fluctuations and a high proportion of low-grade resources. It can resolve the contradiction of "wasting high-grade ore and leaving low-grade ore idle," while ensuring stable downstream production, making it a key technology for achieving sustainable resource utilization. Currently, multi-grade ore blending mainly relies on manual experience calculations (such as simple linear programming calculations). These calculations are highly random and usually pre-set the blending quantities for certain mining sites based on production capacity or reserves. Achieving effective blending often requires very complex calculations. Fluctuations in mining site grades also necessitate recalculation. Traditional methods make it difficult to achieve refined management of multi-grade ore blending and cannot quickly obtain high-precision, executable blending schemes. Summary of the Invention

[0004] The purpose of this invention is to overcome the technical problems of existing multi-grade ore blending and to provide a multi-mine joint production blending method based on multi-objective ore grade constraints. The multi-objective blending optimization model is constructed with an objective function and a ring-shaped neighborhood topology. An improved ring-shaped neighborhood topology hybrid particle swarm algorithm is used for iteration and position update processing to obtain the optimal solution of the objective function. This realizes the joint optimization of blending schemes for multiple mining sites under multi-objective ore grade constraints. By balancing the errors of each element through an adaptive objective function, it can meet the element grade requirements of multi-objective ores, providing an automated and high-precision multi-mine joint production blending scheme and improving product quality stability.

[0005] The objective of this invention is achieved through the following technical solution: A multi-mine joint production and blending method based on multi-objective ore grade constraints, the method comprising: S1. Select N mining sites and set the total ore quantity T containing the grades of M kinds of ore elements as the ore blending target, and construct a multi-objective ore blending optimization model containing the objective function. S2. The multi-objective ore blending optimization model constructs a ring-shaped neighborhood topology structure. An improved ring-shaped neighborhood topology hybrid particle swarm algorithm is used for iterative and position update processing to obtain the optimal solution of the objective function. During position update, the historical optimal solution of individual particles is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function in all ring-shaped neighborhoods are taken as the global historical optimal solution. S3. The multi-objective ore blending optimization model iteratively terminates and outputs the optimal ore blending amount corresponding to N mining points.

[0006] To better implement this invention, in method S1, the objective function in the multi-objective ore blending optimization model is... The expression is as follows: ,in Let n be the amount of ore to be allocated at mineral deposit n. Let m be the grade of ore element m in mineral deposit n. The weighting coefficient for ore element m is... To determine the target grade for ore element m. This is the penalty coefficient for the total amount constraint.

[0007] Preferably, the multi-objective ore blending optimization model further includes the following constraint function: ; ; ;in The maximum mining volume set for mining site n. The grade deviation threshold set for ore element m.

[0008] Preferably, the improved ring-neighborhood topology hybrid particle swarm optimization algorithm at iteration time t+1... The update expression for the factor learning mechanism is as follows: ;in For individual learning factors, As a global social learning factor, As a local social learning factor, These are random numbers in the interval [0,1]. For particles Individual historical optimal solution. This is the globally historical optimal solution. For particles The historical best solution within the annular neighborhood For inertial weights, For particles The velocity at iteration time t+1 For particles The velocity at iteration time t For particles The position at iteration time t.

[0009] Preferably, after the location is updated, the multi-objective ore blending optimization model performs constraint condition function constraint satisfaction processing through the projection operator.

[0010] Preferably, the improved ring-neighborhood topology hybrid particle swarm optimization algorithm includes the following methods during position update processing: Update individual optimality: if particle If the fitness value of the current position is better than its historical best, then update the current particle. The position of the particle Individual historical optimal solution; Update the optimal ring neighborhood: After updating the individual optimal, select the historical optimal solution of the particle with the best objective function in the ring neighborhood as the historical optimal solution of the ring neighborhood; Update the global optimum: After updating the historical optimum of the annular neighborhood, select the historical optimum of the annular neighborhood of the particle with the best objective function in all annular neighborhoods as the global historical optimum.

[0011] Preferably, the improved ring-neighborhood topology hybrid particle swarm optimization algorithm introduces a periodic crossover and mutation operation, the method of which includes: Randomly select particle pairs New particle pairs are generated by performing crossover operations as follows. : , where M is the mask matrix, determined by the Bernoulli distribution; particle The perturbation is performed according to the following expression: ,in The position after the disturbance. For Gaussian perturbation, To limit the magnitude of variation; For inertia weight and parameters for limiting the magnitude of variation The following parameters are dynamically adjusted: Inertia weight adjustment formula: ,in This is the inertia weight attenuation coefficient; Formula for adjusting the parameter to limit the amplitude of variation: ,in This represents the current iteration number. This represents the maximum number of iterations.

[0012] Preferably, the improved iterative method for the ring-neighborhood topology hybrid particle swarm optimization algorithm is as follows: Within a constrained solution space, a particle swarm is randomly initialized. For each particle, an objective function is calculated as its fitness value, and the individual optimum, neighborhood optimum, and global optimum are calculated sequentially. For the first round of initialization, its individual optimum is set to the current initial value. The velocity and position of the particles in the new round are updated according to the velocity and position update formulas, and the individual optimum, neighborhood optimum, and global optimum of the new round of particle swarm are calculated. A crossover and mutation operation is performed on the original particles to generate new particles. The position and velocity of the new particles generated by crossover and mutation are recalculated. If the fitness of the new particles is better, the original particles are replaced; otherwise, the inertia weight and mutation operation magnitude are dynamically adjusted.

[0013] A multi-objective ore grade-constrained blending system based on polymetallic mine joint production includes a multi-objective blending optimization model and a blending parameter input module. The blending parameter input module is used to input mine point data and blending target data. The mine point data includes N mine points and their parameters, including ore and ore grade. The blending target data includes the total quantity T of M ore element grades, where the total quantity T includes M types of ore, and each ore has a target grade. The multi-objective blending optimization model includes an objective function. The multi-objective blending optimization model constructs a ring-shaped neighborhood topology and uses an improved ring-shaped neighborhood topology hybrid particle swarm algorithm for iteration and position update processing to obtain the optimal solution of the objective function. During position update, the historical optimal solution of individual particles is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function in all ring-shaped neighborhoods are taken as the global historical optimal solution. The multi-objective blending optimization model terminates its iteration and outputs the optimal blending quantity corresponding to each of the N mine points.

[0014] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) The multi-objective ore blending optimization model of the present invention has an objective function and a ring neighborhood topology. The optimal solution of the objective function is obtained by iterative and position update processing using the improved ring neighborhood topology hybrid particle swarm algorithm. This realizes the output of the joint optimization ore blending scheme of multiple mining sites under the constraint of multi-objective ore grade. By balancing the errors of each element through the adaptive objective function, it can meet the element grade requirements of multi-objective ore and automatically obtain a high-precision multi-mine joint production ore blending scheme, thereby improving the stability of product quality.

[0015] (2) The present invention uses the projection operator to satisfy the constraint condition function. The improved position update process of the ring neighborhood topology hybrid particle swarm algorithm includes updating the individual optimum, updating the ring neighborhood optimum and updating the ring neighborhood optimum, which prevents getting trapped in local optima and also improves the convergence speed.

[0016] (3) The improved ring neighborhood topology hybrid particle swarm algorithm of the present invention achieves a balance between exploration and development through neighborhood information sharing and cross-mutation mechanism. It generates new particles through cross-operation to update particles, adds Gaussian perturbation to prevent local optima and limit the mutation amplitude, and dynamically adjusts the inertia weight and parameters through dynamic parameter adjustment. This allows the improved ring neighborhood topology hybrid particle swarm algorithm to quickly find the optimal solution and improves the search efficiency. Attached Figure Description

[0017] Figure 1 This is a flowchart of the multi-mine joint production and ore blending method of the present invention; Figure 2 This is a schematic diagram illustrating the principle of the multi-mine joint production and ore blending method in the embodiment. Figure 3 The diagram shows a comparison of the convergence of the improved ring-neighborhood topology hybrid particle swarm algorithm with the standard particle swarm algorithm and the standard genetic algorithm applied to the objective function in this embodiment. Detailed Implementation

[0018] The present invention will be further described in detail below with reference to embodiments: Example

[0019] like Figure 1 As shown, a multi-ore co-production blending method based on multi-objective ore grade constraints includes the following steps: S1. Select N mining sites and set the total ore quantity T, containing M types of ore element grades, as the ore blending objective. Construct a multi-objective ore blending optimization model with an objective function. The objective function of the multi-objective ore blending optimization model is... The expression is as follows: ,in Let n be the amount of ore to be allocated at mineral deposit n. This represents the grade of element m in the ore at mineral deposit n (i.e., the grade of the element in the ore at mineral deposit n). The weighting coefficient of ore element m (multi-objective balance is achieved through the weighting coefficient). To determine the target grade for ore element m. This represents the penalty coefficient for total quantity constraints. The objective function constructed in this invention satisfies the requirement of selecting N different grade ores from various mining sites. Through the objective function of the multi-objective ore blending optimization model, the optimized ore grade meets the optimal feed grade, improving the overall mining utilization rate of the deposit and effectively limiting grade fluctuations.

[0020] In some embodiments, the multi-objective ore blending optimization model further includes the following constraint function: ; ; ;in The maximum mining volume set for mining site n (the ore blending volume of each mining site cannot exceed the maximum ore production volume of the mining site at the blasting point). The grade deviation threshold set for ore element m.

[0021] S2. A multi-objective ore blending optimization model is constructed using a ring-shaped neighborhood topology. An improved ring-shaped neighborhood topology hybrid particle swarm optimization algorithm is employed for iterative and position update processing to obtain the optimal solution of the objective function. The advantage of the ring-shaped neighborhood topology lies in balancing exploration and development capabilities: each particle interacts only with a fixed number of neighboring particles, forming multiple local search regions. This avoids premature convergence caused by global topology while preventing the inefficiency of purely local searches. The ring-shaped neighborhood topology is particularly suitable for the characteristics of ore blending problems because different ore combinations may form multiple relatively optimal solution regions, which can be explored in parallel. This hybrid strategy significantly improves the probability of finding the global optimum while maintaining a fast convergence speed. During position update, the historical optimal solution of an individual particle is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function within all ring-shaped neighborhoods are taken as the global historical optimal solution. Figure 2 As shown, the steps of the improved ring-neighborhood topology hybrid particle swarm optimization algorithm are as follows: Initialize all particles by randomly assigning their positions and velocities within a defined range. Each particle's position represents a solution to the equation, and its velocity represents the direction of change of that solution. Let D be the number of solutions to the equation, and the particle swarm contain n particles. Assign each particle's position... ,in Substitute the solution into the objective function to obtain their respective initializations. The current group's gbest is The solution corresponding to the optimal value. This invention introduces a ring-shaped neighborhood topology to additionally calculate the neighborhood optimal value for each particle. , The optimal point is selected from the k nearest points around the particle. This term is introduced mainly to prevent the method from getting stuck in local optima.

[0022] After obtaining the global optimum gbest at time t in the previous round, the entire particle swarm needs to move towards the position corresponding to gbest at time t+1. That is, the solution of the equation needs to move closer to the objective function in a more ideal state. This update process first requires determining how to move, that is, determining the speed of each particle.

[0023] The velocity update formula of the ring neighborhood topology hybrid particle swarm optimization algorithm consists of three terms: inertia term, individual influence factor, and social influence factor. To further optimize the problem of getting trapped in local optima, the social factor in the iteration velocity influence factor of the ring neighborhood topology hybrid particle swarm optimization algorithm is split into local social factor and global social factor, forcing the calculation of local optima in the iteration and slowing down the convergence speed. The local optima are determined by the constraints of the ring neighborhood topology.

[0024] By comparing the sum and corresponding adaptiveness of each round, the values ​​of each round are updated. The difference is that it does not need to be compared with the value of the previous round. It is the position of the particle with the highest adaptiveness among all particles in the sphere with the particle as the center and R as the radius. It depends on the position of the particle in each iteration and will also change with the change of particle position distribution.

[0025] In some embodiments, the improved ring-neighborhood topology hybrid particle swarm optimization algorithm at iteration time t+1... The update expression for the factor learning mechanism is as follows: ;in This serves as an individual learning factor (preserving the particle's own search characteristics). It serves as a global social learning factor (guiding convergence towards the global optimum). It serves as a local social learning factor (maintaining neighborhood diversity and preventing getting trapped in local optima). These are random numbers in the interval [0,1] (for the introduced random perturbation). For particles Individual historical optimal solution. This is the globally historical optimal solution. For particles The historical best solution within the annular neighborhood The inertial weight (this invention employs dynamic parameter adjustment; preferably, it uses an inertial weight attenuation coefficient for dynamic adjustment, the formula of which is...) In this embodiment, the initial inertia weight is 0.8, and the inertia weight decay coefficient is selected as 0.99. This balances the historical velocity and the current gradient. In the early exploration, a larger inertia weight is needed to quickly search for the optimal location. In the later stage, a smaller inertia weight is needed to avoid failing to converge to the accurate value. Therefore, this process is a slow, exponentially decreasing decay process. For particles The velocity at iteration time t+1 For particles The velocity at iteration time t For particles The position at iteration time t.

[0026] The inertial term preserves the original motion trend of the particle, and its weight decreases with iteration to balance exploration and development.

[0027] As an individual cognitive term, it guides particles to learn towards their own historical best position. As a global social term, it drives particles to move closer to the global optimal solution of the population. For the local social term: local optimal information is obtained through the ring-shaped neighborhood topology to avoid premature convergence. In the process of searching for the optimal solution within a certain solution space, the particle movement speed also needs to be constrained. Therefore, if the speed in the iteration exceeds the change, the boundary value within the constraint condition is taken.

[0028] The multi-objective ore blending optimization model processes the constraint condition function through a projection operator after the position update. The original position at time t is known. Movement speed at time t+1 The new position at time t+1 can be obtained using the following formula. By superimposing the next time step's movement vector onto the original position, a new position is obtained. Since the position represents the solution, and the solution only exists in a finite solution space, the position needs to be updated by using the projection operator to satisfy constraints. The projection operator ensures that the solution after each update exists in a space that satisfies all constraints.

[0029] The projection operator guarantees that: Boundary constraints: Total quantity constraints: ; The projection steps of the projection operator are as follows: 1. Temporary location: ; 2. Boundary truncation: ; 3. Total amount normalization: = ; in, This represents the amount of ore mixed at mineral deposit j for the t-th generation particle i. This represents the velocity of particle i in dimension j of generation t; T represents the maximum exploitable quantity of mineral deposit j; T is the target total mineral quantity.

[0030] In some embodiments, such as Figure 2 As shown, the improved ring-neighborhood topological hybrid particle swarm optimization algorithm obtains a new set of positions, i.e., a new set of solutions, after one update of velocity and position during the position update process. For this new set of solutions, the position of this round needs to be recalculated. , and This includes the following methods: Update individual optimality: if particle If the fitness value of the current position is better than its historical best, then update the current particle. The position of the particle The individual's historical best solution. If the particle's current fitness value is better than its historical best value, then update. This represents the current position of the particle; each particle independently maintains its own position. And retain the optimal solution found by the particle itself, avoiding the loss of local potential advantages, as expressed below: .

[0031] Update the circumferential neighborhood optimum: After updating the individual optimum, the historical optimum of the particle with the best objective function in the circumferential neighborhood is selected as the historical optimum of the circumferential neighborhood. For each particle i, the expression is as follows: The domain is defined as follows: K is the neighborhood radius. Introducing the neighborhood optimum term can limit the range of information propagation through a circular topology, maintain population diversity, and avoid premature convergence caused by global topology.

[0032] Update the global optimum: After updating the historical optimum of the annular neighborhood, select the historical optimum of the annular neighborhood of all particles with the best objective function within the annular neighborhood as the global historical optimum. Then update the annular topology corresponding to each particle. Then, select the particle with the optimal objective function within all annular neighborhoods. As . It is the sole global guide shared by the entire population, but can be used independently. It may cause premature convergence, therefore, it needs to be coordinated. The mechanism is used, and its expression is as follows: , i=1,2,…,N; where N is the population size.

[0033] To further improve the premature convergence problem of the standard particle swarm optimization (PSO) algorithm, this invention not only introduces a ring-shaped neighborhood topology and local social factors to achieve a balance between local fine-grained search and global exploration, but also combines global search capabilities by introducing periodic crossover and mutation operations after velocity and position updates to increase population diversity and avoid getting trapped in local optima. The improved ring-shaped neighborhood topology hybrid PSO algorithm of this invention introduces periodic crossover and mutation operations, such as... Figure 2 As shown, the method includes: Randomly select particle pairs New particle pairs are generated by performing crossover operations as follows. : , where M is the mask matrix, determined by the Bernoulli distribution; the purpose is to enhance population diversity through inter-particle information recombination.

[0034] particle Perturbate using the following expression (to help escape local optima): ,in The position after the disturbance. For Gaussian perturbation, Parameters to limit the magnitude of variation.

[0035] For inertia weight and parameters for limiting the magnitude of variation Perform the following parameter dynamic adjustments (also known as adaptive parameter updates): Inertia weight adjustment formula: ,in This is the inertia weight decay coefficient; the inertia weight decay coefficient is used to improve the local search in the later stages.

[0036] Formula for adjusting the parameter to limit the amplitude of variation: ,in This represents the current iteration number. The parameter that limits the mutation magnitude is the maximum number of iterations. The number of iterations gradually decreases.

[0037] In some embodiments, the improved iterative method of the ring-neighborhood topology hybrid particle swarm optimization algorithm is as follows: Within a constrained solution space, a particle swarm is randomly initialized. For each particle, an objective function is calculated as its fitness value, and the individual optimum, neighborhood optimum, and global optimum are calculated sequentially. For the first round of initialization, its individual optimum is set to the current initial value. The velocity and position of the particles in the new round are updated according to the velocity and position update formulas, and the individual optimum, neighborhood optimum, and global optimum of the new round of particle swarm are calculated. A crossover and mutation operation is performed on the original particles to generate new particles. The position and velocity of the new particles generated by crossover and mutation are recalculated. If the fitness of the new particles is better, the original particles are replaced; otherwise, the inertia weight and mutation operation magnitude are dynamically adjusted.

[0038] S3. The multi-objective ore blending optimization model terminates iteratively and outputs the optimal ore blending quantities for each of the N ore deposits. After the multi-objective ore blending optimization model reaches the iteration termination condition, the optimal solution or optimal value is output when the objective function converges. middle This represents the ore allocation amount for mining site n, which means obtaining the optimal ore allocation amount for each of the N mining sites.

[0039] In this embodiment, the objective function is iteratively converged using an improved cyclic neighborhood topology hybrid particle swarm optimization algorithm. This embodiment uses three algorithms—the improved cyclic neighborhood topology hybrid particle swarm optimization algorithm (the method used in the multi-objective ore blending optimization model of this invention, abbreviated as Optimized_PSO), the standard particle swarm optimization algorithm (Standard_PSO), and the standard genetic algorithm (Standard_GA)—on data with 100 ore points and 10 ore elements to evaluate their convergence speed. Figure 3 As shown, compared with the standard particle swarm optimization (PSO) algorithm and the standard genetic algorithm (GA), the improved ring neighborhood topology hybrid particle swarm optimization algorithm of this invention significantly improves the convergence speed. The improved ring neighborhood topology hybrid particle swarm optimization algorithm also achieves optimal fitness upon final convergence (in...). Figure 3 In the figure, the horizontal axis represents the number of iterations, and the vertical axis represents the global best fitness.

[0040] A multi-objective ore grade-constrained blending system based on polymetallic mine joint production includes a multi-objective blending optimization model and a blending parameter input module. The blending parameter input module is used to input mine point data and blending target data. The mine point data includes N mine points and their parameters, including ore and ore element grades. The blending target data includes the total ore quantity T of M ore element grades, where the total ore quantity T includes M types of ore, each with a target grade. The multi-objective blending optimization model includes an objective function. The multi-objective blending optimization model constructs a ring-shaped neighborhood topology and uses an improved ring-shaped neighborhood topology hybrid particle swarm algorithm for iteration and position update processing to obtain the optimal solution of the objective function. During position update, the historical optimal solution of individual particles is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function in all ring-shaped neighborhoods are taken as the global historical optimal solution. The multi-objective blending optimization model terminates its iteration and outputs the optimal blending quantity corresponding to each of the N mine points.

[0041] This embodiment takes a ore blending target of 23, 22, and 21 for ore elements A, B, and C, and a total ore quantity T of 10,000 t (tons) as an example. Ore elements A, B, and C each correspond to three metal elements (such as lead, zinc, and silver, or copper, molybdenum, and cobalt). Ore blending is constrained based on a multi-objective ore blending optimization model from four ore deposits (i.e., ore sources). The data for the four ore deposits are shown in Table 1 below: Table 1. Parameter table of four mining sites as examples Mining site number Maximum mining capacity (t) of the mining site Grade (%) of element A in ore deposit Grade (%) of element B in ore deposit C grade (%) of ore in mineral deposits loc_1 18000 15 19 27 loc_2 16000 28 25 15 loc_3 20000 19 11 26 loc_4 17000 27 23 21

[0042] According to the multi-objective ore grade-constrained blending system or method of the present invention, the optimal blending quantities for the four mining sites are obtained as shown in Table 2 below: Table 2 Optimal Ore Blending for Four Mining Sites Mining site number Maximum mining capacity (t) of the mining site Optimize ore blending (t) Normalized ore blending ratio (%) Grade (%) of element A in ore deposit Grade (%) of element B in ore deposit C grade (%) of ore in mineral deposits loc_1 18000 3439.597 34.39597 15 19 27 loc_2 16000 3691.275 36.91275 28 25 15 loc_3 20000 302.0135 3.020135 19 11 26 loc_4 17000 2567.114 25.67114 27 23 21

[0043] As shown in the table above, to achieve the set target of 10,000 tons of ore blending, the objective function of the multi-objective ore blending optimization model is optimized using the improved cyclic neighborhood topology hybrid particle swarm algorithm to output the optimal solution. The ore blending amounts for the four mining sites are as follows: the ore blending amount for mining site loc_1 is approximately 3,440 tons, the ore blending amount for mining site loc_2 is approximately 3,691 tons, the ore blending amount for mining site loc_3 is approximately 302 tons, and the ore blending amount for mining site loc_4 is approximately 3,567 tons. The grade fluctuations of ore elements A, B, and C are -3.3e-08, -6.3e-08, and -6.0e-08, respectively.

[0044] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multi-mine joint production and blending method based on multi-objective ore grade constraints, characterized in that: The methods include: S1. Select N mining sites and set the total ore quantity T containing the grades of M kinds of ore elements as the ore blending target, and construct a multi-objective ore blending optimization model containing the objective function. S2. The multi-objective ore blending optimization model constructs a ring-shaped neighborhood topology structure. An improved ring-shaped neighborhood topology hybrid particle swarm algorithm is used for iterative and position update processing to obtain the optimal solution of the objective function. During position update, the historical optimal solution of individual particles is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function in all ring-shaped neighborhoods are taken as the global historical optimal solution. S3. The multi-objective ore blending optimization model iteratively terminates and outputs the optimal ore blending amount corresponding to N mining points.

2. The multi-ore joint production and blending method based on multi-objective ore grade constraints according to claim 1, characterized in that: In method S1, the objective function The expression is as follows: ,in Let n be the amount of ore to be allocated at mineral deposit n. Let m be the grade of ore element m in mineral deposit n. The weighting coefficient for ore element m is... To determine the target grade for ore element m. This is the penalty coefficient for the total amount constraint.

3. The multi-mine joint production and blending method based on multi-objective ore grade constraints according to claim 2, characterized in that: The multi-objective ore blending optimization model also includes the following constraint functions: ; ; ;in The maximum mining volume set for mining site n. The grade deviation threshold set for ore element m.

4. The multi-mine joint production and blending method based on multi-objective ore grade constraints according to claim 1, characterized in that: The improved ring-neighborhood topology hybrid particle swarm optimization algorithm at iteration time t+1... The update expression for the factor learning mechanism is as follows: ;in For individual learning factors, As a global social learning factor, As a local social learning factor, These are random numbers in the interval [0,1]. For particles Individual historical optimal solution. This is the globally historical optimal solution. For particles The historical best solution within the annular neighborhood For inertial weights, For particles The velocity at iteration time t+1 For particles The velocity at iteration time t For particles The position at iteration time t.

5. The multi-mine joint production and blending method based on multi-objective ore grade constraints according to claim 3, characterized in that: After the location is updated, the multi-objective ore blending optimization model uses the projection operator to handle the constraint satisfaction of the constraint function.

6. The multi-ore joint production and blending method based on multi-objective ore grade constraints according to claim 1 or 4, characterized in that: The improved ring-neighborhood topology hybrid particle swarm optimization algorithm includes the following methods for position update processing: Update individual optimality: if particle If the fitness value of the current position is better than its historical best, then update the current particle. The position of the particle Individual historical optimal solution; Update the optimal ring neighborhood: After updating the individual optimal, select the historical optimal solution of the particle with the best objective function in the ring neighborhood as the historical optimal solution of the ring neighborhood; Update the global optimum: After updating the historical optimum of the annular neighborhood, select the historical optimum of the annular neighborhood of the particle with the best objective function in all annular neighborhoods as the global historical optimum.

7. The multi-mine joint production and blending method based on multi-objective ore grade constraints according to claim 4, characterized in that: The improved ring-neighborhood topology hybrid particle swarm optimization algorithm introduces periodic crossover and mutation operations, including: Randomly select particle pairs New particle pairs are generated by performing crossover operations as follows. : , where M is the mask matrix, determined by the Bernoulli distribution; particle The perturbation is performed according to the following expression: ,in The position after the disturbance. For Gaussian perturbation, To limit the magnitude of variation; For inertia weight and parameters for limiting the magnitude of variation The following parameters are dynamically adjusted: Inertia weight adjustment formula: ,in This is the inertia weight decay coefficient; Formula for adjusting the parameter to limit the amplitude of variation: ,in This represents the current iteration number. This represents the maximum number of iterations.

8. The multi-mine joint production and blending method based on multi-objective ore grade constraints according to claim 1, characterized in that: The improved iterative method for the ring-neighborhood topology hybrid particle swarm optimization algorithm is as follows: Within a constrained solution space, a particle swarm is randomly initialized. For each particle, an objective function is calculated as its fitness value, and the individual optimum, neighborhood optimum, and global optimum are calculated sequentially. For the first round of initialization, its individual optimum is set to the current initial value. The velocity and position of the particles in the new round are updated according to the velocity and position update formulas, and the individual optimum, neighborhood optimum, and global optimum of the new round of particle swarm are calculated. A crossover and mutation operation is performed on the original particles to generate new particles. The position and velocity of the new particles generated by crossover and mutation are recalculated. If the fitness of the new particles is better, the original particles are replaced; otherwise, the inertia weight and mutation operation magnitude are dynamically adjusted.

9. A multi-mine joint production and blending system based on multi-objective ore grade constraints, characterized in that: The system includes a multi-objective ore blending optimization model and an ore blending parameter input module. The ore blending parameter input module is used to input ore point data and ore blending target data. The ore point data includes N ore points and their parameters, including ore and ore element grades. The ore blending target data includes the total ore quantity T of M types of ore element grades, where the total ore quantity T includes M types of ore, each with a target grade. The multi-objective ore blending optimization model includes an objective function. The multi-objective ore blending optimization model constructs a ring-shaped neighborhood topology and uses an improved ring-shaped neighborhood topology hybrid particle swarm algorithm for iteration and position update processing to obtain the optimal solution of the objective function. During position update, the historical optimal solution of individual particles is selected as the historical optimal solution of the ring-shaped neighborhood, and the optimal solutions of the objective function in all ring-shaped neighborhoods are taken as the global historical optimal solution. The multi-objective ore blending optimization model terminates its iteration and outputs the optimal ore blending quantity corresponding to each of the N ore points.

Images