Oxygen-enriched bottom blowing copper matte grade prediction method based on GA-BP neural network

Abstract

The invention discloses an oxygen-enriched bottom blowing copper matte grade prediction method based on a GA-BP neural network, which improves optimization efficiency and model prediction precision by improving a core mechanism of a genetic algorithm. The method comprises the following steps: collecting actual production data in a stable production period of a copper smelting plant; an existing GA-BP neural network is improved by adding a self-adaptive crossover probability adjustment mechanism and a self-adaptive mutation probability adjustment mechanism in a genetic algorithm, and the self-adaptive crossover probability adjustment mechanism not only depends on individual fitness, but also is combined with global and local information to be used for enhancing global search capability and local optimization capability. The adaptive mutation probability adjustment mechanism not only combines individual fitness, but also introduces global entropy and dynamic noise influence, and meanwhile, allows the mutation probability to be increased randomly in the later stage of evolution so as to jump out of local optimum; and by utilizing the improved GA-BP neural network, taking actual production data as input, predicting the oxygen-enriched bottom blowing copper matte grade and outputting a copper matte grade value.

Description

Technical Field

[0001] The embodiments of the present application relate to the technical field of oxygen-enriched bottom-blown copper matte grade prediction, and in particular to a method for oxygen-enriched bottom-blown copper matte grade prediction based on a GA-BP neural network. Background Art

[0002] Oxygen-enriched bottom-blowing copper smelting technology is a new type of copper smelting method with the advantages of high efficiency, energy saving and environmental protection. Its main technical principles include oxygen bottom-blowing smelting, bottom-blowing slag beneficiation and oxygen bottom-blowing, which can improve the copper recovery rate, reduce energy consumption and reduce environmental pollution. In general, oxygen-enriched bottom-blowing copper smelting technology is a copper smelting method with broad application prospects, which can bring significant economic and social benefits to copper smelting enterprises.

[0003] The copper matte grade refers to the proportion of copper in the copper matte. It is an important indicator to measure the quality of copper matte and the effect of related production links such as copper smelting. It is necessary to comprehensively consider various factors such as raw materials and processes to make optimization and adjustments to ensure efficient and high-quality production of qualified copper products.

[0004] With the rapid development of artificial intelligence and optimization algorithms, the combination of genetic algorithm (GA) and BP neural network (BP Neural Network) has made the traditional BP neural network widely used in the copper matte grade prediction analysis in oxygen-enriched bottom blowing copper smelting technology with its powerful learning ability. However, due to the shortcomings of traditional BP neural network such as easy to fall into local optimality, slow training speed, and sensitivity to initial weights and thresholds, it is often difficult to achieve ideal results when dealing with complex optimization problems. Summary of the invention

[0005] In order to overcome the above technical defects and solve the problems of insufficient population diversity, weak global search ability, weak local optimization ability and easy loss of elite individuals caused by the fixed crossover probability and mutation probability of the traditional GA-BP neural network during the optimization process, the present application embodiment provides a method for predicting the grade of oxygen-enriched bottom-blown copper matte based on GA-BP neural network, and its main technical solutions are as follows:

[0006] S1. Collect the actual production data of the copper smelter during the stable production cycle, wherein the actual production data includes oxygen consumption x1, solvent rate x2, oxygen enrichment concentration x3, copper content x4, sulfur content x5, iron content x6, silicon dioxide content x7 and magnesium oxide content x8;

[0007] S2, improving the existing GA-BP neural network by adding adaptive crossover probability adjustment mechanism and adaptive mutation probability adjustment mechanism to the genetic algorithm;

[0008] The adaptive crossover probability adjustment mechanism not only relies on individual fitness, but also combines global and local information to enhance global search capabilities and local optimization capabilities; the adaptive mutation probability adjustment mechanism not only combines individual fitness, but also introduces global entropy and dynamic noise, while allowing the mutation probability to increase randomly in the late stage of evolution to jump out of the local optimum. The global entropy value is used to measure population diversity;

[0009] S3. Using the improved GA-BP neural network, the actual production data is used as input to predict the grade of oxygen-enriched bottom-blown copper matte and output the copper matte grade value.

[0010] Specifically, the adaptive crossover probability adjustment mechanism introduces fitness influencing factors, evolutionary dynamics influencing factors, fitness ranking factors and dynamic weights; the adaptive mutation probability adjustment mechanism introduces population entropy factors, fitness adjustment factors and dynamic noise factors.

[0011] The embodiments of the present application adopt the above technical solution to achieve the following technical effects:

[0012] By improving the core mechanism of the genetic algorithm, the crossover probability and mutation probability adjustment mechanism in the traditional genetic algorithm are improved. Among them, the adaptive crossover probability adjustment mechanism introduces evolutionary dynamic influencing factors, fitness ranking factors, and dynamic weights. The high crossover probability in the early stage of evolution is conducive to exploring the solution space, and it gradually decreases in the later stage to protect the better gene combination and improve the algorithm optimization efficiency. The crossover probability of individuals with high fitness is low, which can protect excellent genes, ensure the algorithm to evolve to the best, and improve the optimization efficiency. The dynamic weight makes the crossover probability transition smoothly, ensures that the crossover operation at each stage of evolution is reasonable, and thus improves the optimization efficiency of the genetic algorithm and the prediction accuracy of the BP neural network. The adaptive mutation probability adjustment mechanism introduces population entropy factor, fitness adjustment factor, and dynamic noise factor. When the population diversity is low, this factor will increase the mutation probability, introduce new genes, avoid local convergence, and improve the optimization efficiency. Due to this factor, individuals with low fitness obtain a higher mutation probability, increase population diversity, help to jump out of the local optimum, and improve the algorithm efficiency. The dynamic noise factor introduces randomness to prevent the population from converging prematurely, prompting the algorithm to continuously explore the new solution space, improve the optimization efficiency of the genetic algorithm, and thus improve the prediction accuracy of the BP neural network. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings exemplarily illustrate the embodiments and constitute a part of the specification, and together with the text description of the specification, are used to explain the exemplary implementation of the embodiments. The embodiments shown are for illustrative purposes only and do not limit the scope of the claims. In all drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0014] Figure 1It is a flow chart of the method for predicting the grade of oxygen-enriched bottom-blown copper matte based on GA-BP neural network in this application. DETAILED DESCRIPTION

[0015] In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work are within the scope of protection of the present application.

[0016] It should be noted that the descriptions involving "first", "second", etc. in the embodiments of the present application are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of technical features indicated. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such combination of technical solutions does not exist and is not within the scope of protection required by this application.

[0017] In the description of the present application, it should be understood that the numerical labels before the steps do not indicate the order in which the steps are executed, but are only used to facilitate the description of the present application and to distinguish each step, and therefore should not be understood as a limitation on the present application.

[0018] This application aims to solve the following key problems of the traditional GA-BP neural network in the optimization process: insufficient population diversity caused by fixed crossover probability and mutation probability, weak global search ability, weak local optimization ability, and easy loss of elite individuals. By improving the core mechanism of the genetic algorithm, a GA-BP neural network optimization method based on a super-complex adaptive crossover probability and mutation probability adjustment mechanism is proposed to improve the optimization efficiency and model prediction accuracy.

[0019] To facilitate those skilled in the art to understand the technical solutions provided in the embodiments of the present application, the following describes the related technologies. Figure 1 In the embodiment of the present application, the oxygen-enriched bottom-blown copper matte grade prediction method based on GA-BP neural network includes:

[0020] S1. Collect actual production data of the copper smelter during the stable production cycle, where the actual production data includes oxygen consumption x1, solvent rate x2, oxygen enrichment concentration x3, copper content x4, sulfur content x5, iron content x6, silicon dioxide content x7 and magnesium oxide content x8.

[0021] Specifically, the copper smelter uses oxygen-enriched bottom-blown copper smelting technology for copper smelting. After collecting the data production data, the data is normalized and divided into a training set, a test set, and a validation set, and the division ratio can be: 8:2, 7:3 or 9:1.

[0022] S2. The existing GA-BP neural network is improved by adding adaptive crossover probability adjustment mechanism and adaptive mutation probability adjustment mechanism into the genetic algorithm.

[0023] The adaptive crossover probability adjustment mechanism not only relies on individual fitness, but also combines global and local information to enhance global search capabilities and local optimization capabilities; specifically, global and local information includes: population fitness distribution, evolutionary historical trends, population diversity dynamics, and individual fitness rankings. The adaptive crossover probability adjustment mechanism introduces fitness influencing factors, evolutionary dynamic influencing factors, fitness ranking factors, and dynamic weights. The fitness influencing factors are used to balance the crossover probabilities of individuals with high fitness and low fitness, and individuals with low fitness are more inclined to have higher crossover probabilities. The evolutionary dynamic influencing factors are used to dynamically adjust the crossover probability based on the relationship between the current number of iterations and the total number of iterations, so that the crossover probability decreases over time, thereby gradually changing the search strategy during the evolution process, such as encouraging more exploration in the early stage and tending to use existing better solutions in the later stage. The fitness ranking factor determines the crossover probability of an individual by calculating its fitness ranking in the population. High-ranking individuals (higher fitness) obtain lower crossover probabilities, which can protect excellent genes and avoid excessive damage during the crossover process. The dynamic weight makes the crossover probability gradually and smoothly transition to the later stage as the number of iterations increases, controls the changing trend of the crossover probability during the evolution process, and ensures that the crossover operation can be performed reasonably at different stages.

[0024] The adaptive mutation probability adjustment mechanism not only combines individual fitness, but also introduces global entropy and dynamic noise effects, while allowing the mutation probability to increase randomly in the late stage of evolution to jump out of the local optimum. The global entropy is used to measure population diversity. The population entropy factor, fitness adjustment factor and dynamic noise factor are introduced into the adaptive mutation probability adjustment mechanism. The population entropy factor adjusts the mutation probability based on the entropy of the current population. When the population diversity is low, the mutation probability is appropriately increased to introduce new genes to avoid falling into the local optimum. The fitness adjustment factor compares the current individual fitness with the maximum fitness in the population, so that individuals with low fitness have a higher mutation probability, thereby increasing the diversity of individuals in the population and increasing the possibility of searching for the global optimal solution. The dynamic noise factor introduces a small amount of randomness with time and fitness changes to avoid premature convergence of the population, so that the algorithm can jump out of the local optimal solution during the search process and continue to explore a wider solution space.

[0025] Further specifically and optionally, the adaptive crossover probability adjustment mechanism expression is:

[0026]

[0027] Where P c (i) is the crossover probability of the i-th individual, P cmin , P cmax are the upper and lower bounds of the crossover probability, is the fitness influencing factor, where f i is the current individual fitness, f avg is the average fitness of the current population, W1(t) is the evolutionary dynamic influencing factor, R(i, t) is the fitness ranking factor, and W2(t) is the dynamic weight.

[0028] More specifically, the expression of W1(t) is: Where α is a control parameter that determines the rate of decrease over time, t is the current iteration number, and T is the total number of iterations. The expression of R(i, t) is: Where rank(f i ) is the fitness ranking of the i-th individual, N is the population size, and high-ranking individuals have lower crossover probability, protecting excellent genes. The expression of W2(t) is: Where β is a parameter that controls the steepness of the curve, usually ranging from 5 to 10, t is the current number of iterations, and T is the total number of iterations. As the number of iterations increases, the crossover probability gradually transitions to a stable state in the later stage.

[0029] In a further specific embodiment, the expression of the adaptive mutation probability adjustment mechanism is:

[0030]

[0031] Where P m (i) is the mutation probability of the i-th individual, P mmax , P mmin are the upper and lower limits of the mutation probability, H(t) is the population entropy factor, N(i, t) is the dynamic noise factor, is the fitness adjustment factor, where f i is the current individual fitness, f max It is the maximum fitness in the population. The fitness adjustment factor can make individuals with low fitness obtain a higher mutation probability.

[0032] To be more specific, the entropy value H(t) of the current population represents the diversity of the population, and its expression is: is the frequency of the jth genotype in the population. N(i, t) introduces a small amount of randomness over time and fitness changes to avoid premature convergence of the population. Its expression is: γ is the dynamic noise amplitude control parameter, and δ is the dynamic coefficient associated with individual fitness.

[0033] S3. Using the improved GA-BP neural network, the actual production data is used as input to predict the grade of oxygen-enriched bottom-blown copper matte and output the copper matte grade value.

[0034] That is, the mineral unit oxygen consumption X1, solvent rate X2, oxygen enrichment concentration X3, copper content X4, sulfur content X5, iron content X6, silicon dioxide content X7 and magnesium oxide content X8 are used as input data of the improved GA-BP neural network.

[0035] In the technical solution of the present application, by improving the core mechanism of the genetic algorithm, the crossover probability and mutation probability adjustment mechanism in the traditional genetic algorithm is improved, and an ultra-complex adaptive adjustment model is introduced to enable it to dynamically adapt to the population status and evolution stage, thereby improving the optimization efficiency of the genetic algorithm and the prediction accuracy of the BP neural network.

[0036] Based on the improvement in S2 above, the specific steps of the modified genetic algorithm include:

[0037] S2.1, population initialization, and set population generation g = 1;

[0038] S2.2. Calculate the fitness of the g-th generation population. The calculation expression is as follows:

[0039] Among them, f itk represents the fitness function, y k is the true value, y′ k is the predicted value, w1 and w2 are weight coefficients used to adjust the impact of the two error terms, and ∈ is a very small value to prevent division by zero.

[0040] S2.3, select the best individual of the current iteration and its fitness function value, and determine whether the preset stop condition is met or the set maximum number of iterations is reached. If the stop condition is met or the maximum number of iterations is reached, the iteration is stopped, and the one with the smallest fitness function value is output as the best individual. If the stop condition is not met or the maximum number of iterations is not reached, execute step S2.4;

[0041] S2.4, select operator;

[0042] S2.5, crossover operator through adaptive crossover probability adjustment mechanism;

[0043] S2.6, mutation operator through adaptive mutation probability adjustment mechanism;

[0044] S2.7. Retain the best individual of the current generation through the elite individual retention strategy, generate a new population of (g+1) generations, and jump to step S2.2, using g+1 as the initial value of step S2.2 for iteration.

[0045] The technical solution of the present application also includes verifying the prediction accuracy of the improved GA-BP neural network model, which can be verified using the root mean square error RMSE and the mean absolute error MAE. The calculation formulas of the two evaluation indicators are as follows:

[0046] Among them, y represents the actual value, y represents the predicted value of the model, and |*| represents the absolute value. Indicates summing the internal data.

[0047] Obviously, those skilled in the art should understand that the modules or steps of the above-mentioned embodiments of the present application can be implemented by general-purpose computer devices, they can be concentrated on a single computer device, or distributed on a network composed of multiple computer devices, optionally, they can be implemented by executable program codes of computer devices, so that they can be stored in a storage device and executed by the computer device, and in some cases, the steps shown or described can be executed in a different order from that herein, or they can be made into individual integrated circuit modules, or multiple modules or steps therein can be made into a single integrated circuit module for implementation. In this way, the embodiments of the present application are not limited to any specific combination of hardware and software.

[0048] It should be noted that the above are only preferred embodiments of the present application, and the patent protection scope of the present application is not limited thereto. Any equivalent structure or equivalent process transformation made using the contents of the specification and drawings of the present application, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for predicting the grade of oxygen-enriched bottom-blown copper matte based on GA-BP neural network, characterized in that: include: S1. Collect the actual production data of the copper smelter during the stable production cycle, wherein the actual production data includes oxygen consumption x1, solvent rate x2, oxygen enrichment concentration x3, copper content x4, sulfur content x5, iron content x6, silicon dioxide content x7 and magnesium oxide content x8; S2, improving the existing GA-BP neural network by adding adaptive crossover probability adjustment mechanism and adaptive mutation probability adjustment mechanism to the genetic algorithm; The adaptive crossover probability adjustment mechanism not only relies on individual fitness, but also combines global and local information to enhance global search capabilities and local optimization capabilities; the adaptive mutation probability adjustment mechanism not only combines individual fitness, but also introduces global entropy and dynamic noise effects, while allowing the mutation probability to increase randomly in the late stage of evolution to jump out of the local optimum, and the global entropy is used to measure population diversity; S3. Using the improved GA-BP neural network, the actual production data is used as input to predict the grade of oxygen-enriched bottom-blown copper matte and output the copper matte grade value.

2. The method for predicting the grade of oxygen-enriched bottom-blown copper matte according to claim 1, characterized in that: The adaptive crossover probability adjustment mechanism introduces fitness influencing factors, evolutionary dynamics influencing factors, fitness ranking factors and dynamic weights; the adaptive mutation probability adjustment mechanism introduces population entropy factors, fitness adjustment factors and dynamic noise factors.

3. The method for predicting the grade of oxygen-enriched bottom-blown copper matte according to claim 2, characterized in that: The expression of the adaptive crossover probability adjustment mechanism is: Where P c (i) is the crossover probability of the i-th individual, P cmin , P cmax are the upper and lower bounds of the crossover probability, is the fitness influencing factor, where f i is the current individual fitness, f avg is the average fitness of the current population, W1(t) is the evolutionary dynamic influencing factor, R(i, t) is the fitness ranking factor, and W2(t) is the dynamic weight.

4. The method for predicting the grade of oxygen-enriched bottom-blown copper matte according to claim 2 or 3, characterized in that: The expression of the adaptive mutation probability adjustment mechanism is: Where P m (i) is the mutation probability of the i-th individual, P m max , P m min are the upper and lower limits of the mutation probability, H(t) is the population entropy factor, N(i, t) is the dynamic noise factor, is the fitness adjustment factor, where f i is the current individual fitness, f max It is the maximum fitness in the population. The fitness adjustment factor can make individuals with low fitness obtain a higher mutation probability.

5. The method for predicting the grade of oxygen-enriched bottom-blown copper matte according to claim 1, characterized in that: Based on the improvement in S2 above, the specific steps of the modified genetic algorithm include: S2.1, population initialization, and set population generation g = 1; S2.

2. Calculate the fitness of the g-th generation population. The calculation expression is as follows: Among them, f itk represents the fitness function, y k is the true value, y′ k is the predicted value, w1 and w2 are weight coefficients used to adjust the influence of the two error terms, and ∈ is a very small value to prevent division by zero; S2.3, select the best individual of the current iteration and its fitness function value, and determine whether the preset stop condition is met or the set maximum number of iterations is reached. If the stop condition is met or the maximum number of iterations is reached, the iteration is stopped, and the one with the smallest fitness function value is output as the best individual. If the stop condition is not met or the maximum number of iterations is not reached, execute step S2.4; S2.4, select operator; S2.5, crossover operator through adaptive crossover probability adjustment mechanism; S2.6, mutation operator through adaptive mutation probability adjustment mechanism; S2.

7. Retain the best individual of the current generation through the elite individual retention strategy, generate a new population of (g+1) generations, and jump to step S2.2, using g+1 as the initial value of step S2.2 for iteration.

Images