Ore body volume prediction model training method, prediction method, equipment and medium

Abstract

The invention relates to an ore body volume prediction model training method, a prediction method, equipment and a medium, an ore body volume prediction model adopting a BP neural network is constructed, and based on a mapping relation between a BP neural network weight and an individual position in an individual position optimization algorithm, an individual position optimization algorithm is utilized to optimize the volume of an ore body. According to the method, the initial weight of the BP neural network is optimized, and the trained BP neural network can output the corresponding ore body volume predicted value result after inputting the normalized geological attribute strongly related to the ore body volume, so that the situation that the ore body volume data is obtained through large-scale geological exploration is avoided, manpower, material resources and time cost are saved, and the working efficiency is improved. The efficiency is greatly improved.

Description

Technical Field

[0001] The present application relates to the technical field of mineral resource exploration and development, and in particular to a training method, prediction method, equipment and medium for an ore body volume prediction model. Background Art

[0002] In the field of mineral resource exploration and development, mineral volume prediction plays an extremely critical role and has far-reaching significance for the scientific development, efficient operation and sustainable advancement of the entire industry. Mineral volume prediction can significantly improve the accuracy and stability of mineral resource estimation. By predicting the volume of ore bodies, the scale of mineral resources can be clearly grasped, which not only helps mining companies accurately assess the economic value of mines, rationally plan investment budgets and development progress, but also provides a scientific basis for countries and regions to formulate macro-resource strategies, ensure the rational allocation and efficient use of resources, and avoid decision-making errors and resource waste caused by resource estimation errors.

[0003] Traditional mineral exploration is often like looking for a needle in a haystack, requiring large-scale geological exploration work in vast areas, which results in a huge amount of manpower, material resources and time costs in mineral volume prediction. Summary of the invention

[0004] The present application proposes a training method, prediction method, equipment and medium for an ore body volume prediction model, which can solve the problem of low efficiency of traditional mineral volume prediction.

[0005] In order to achieve the above objectives, this application adopts the following technical solutions:

[0006] In a first aspect, a training method for an ore body volume prediction model is provided, wherein the prediction model adopts a back propagation BP neural network, and the training set includes: normalized geological attributes strongly related to the ore body volume, the true value of the ore body volume, and the predicted value of the ore body volume, and the training method includes:

[0007] Establishing a mapping relationship between the BP neural network weights and the individual positions in the individual position optimization algorithm, wherein the individual position optimization algorithm adopts an optimal ore body volume prediction objective function;

[0008] The BP neural network is trained using the training set, and at the same time, the initialization weights of the BP neural network are optimized using the individual position optimization algorithm.

[0009] Based on the above technical solution, a ore body volume prediction model using a BP neural network was constructed. Based on the mapping relationship between the BP neural network weights and the individual positions in the individual position optimization algorithm, the individual position optimization algorithm was used to optimize the initial weights of the BP neural network. The trained BP neural network can output the corresponding ore body volume prediction value results after inputting normalized geological attributes that are strongly correlated with the ore body volume. In this way, large-scale geological exploration to obtain ore body volume data is avoided, manpower, material and time costs are saved, and efficiency is greatly improved.

[0010] In a possible design of the first aspect, the initialization weights of the BP neural network are optimized by using the individual position optimization algorithm, specifically including:

[0011] Initializing the individual positions to obtain an initial optimal individual position set that meets the requirements of the optimal ore body volume prediction objective function;

[0012] Before the maximum number of training iterations is reached, when the current Sine chaotic map value meets the random number requirement, the individual positions in the initial optimal individual position set are sorted and grouped according to their fitness, and the escape optimization algorithm is used to update the individual positions in the initial optimal individual position set. Otherwise, the individual positions in the initial optimal individual position set are updated by the wavelet mutation strategy that obeys the t-distribution probability.

[0013] When the maximum number of training iterations is reached, the optimal individual position is output as the optimal initialization weight of the BP neural network.

[0014] Based on the above technical solution, the wavelet mutation strategy that obeys the probability density of t distribution is used to operate the optimal solution, which changes the development direction of the fleeing crowd and makes them enter other areas for development. It improves the global search capability of the algorithm, accelerates the convergence speed, and effectively improves the accuracy and speed of ore volume prediction.

[0015] In a possible design of the first aspect, the training method further includes: determining whether the current Sine chaotic map value meets the random number requirement, specifically: determining whether the current Sine chaotic map value γ is greater than the random number rand, wherein:

[0016]

[0017] FE represents the current iteration number, γ FE represents the new Sine chaotic map at the FE iteration, γ FE+1It represents the new Sine chaotic map at the time of iteration FE+1, which controls the algorithm to update the position. Cos represents the cosine value operation. Sin represents the sine value operation. Pi represents the pi. A represents a random number between (0, 4). The initial value of γ is a random number between 0 and 1.

[0018] Based on the above technical solution, the chaotic characteristics of Sine chaotic mapping are used to generate unpredictable dynamic behaviors, so that the algorithm can flexibly switch between global search and local search, and can better jump out of the local optimal solution, further improving the accuracy and stability of ore volume prediction.

[0019] In a possible design manner of the first aspect, the escape optimization algorithm is used to update the individual positions in the initial optimal individual position set, specifically including:

[0020] The individual positions in the initial optimal individual position set are divided into a calm group q, a follow-the-flow group f and a panic group l;

[0021] A nonlinear panic factor is used to update the individual positions in the calm group, the follow-the-crowd group, and the panic group respectively.

[0022] Based on the above technical solution, by introducing a nonlinear panic factor, the escape optimization algorithm can perform more refined local optimization in the later stage, improve the optimization accuracy and convergence speed, and further improve the accuracy and stability of ore volume prediction.

[0023] In a possible design of the first aspect, the following formula is used to update the individual positions in the quiet group:

[0024]

[0025] Among them, ES i,j represents the position of the i-th individual in the j-th dimension, represents the updated individual position, Q j is the average of all individual positions in the calm group in the jth dimension, τ1 takes the value 0 or 1 with equal probability for the calm group, P(FE) is the nonlinear panic factor, v q,j represents the moving speed of the calm group, ω1 is the adaptive Levy weight,

[0026]

[0027] μ represents a random number in the interval [0,1], δ represents a constant value of 20, FE max represents the maximum number of iterations,

[0028]

[0029] vq,j represents the moving speed of the calm group, H q,j is a randomly generated position within the range of the calm group, represents small adjustments in individual movements,

[0030]

[0031] represents the minimum and maximum values ​​of the jth dimension of all individual positions in the calm group, represents the individual position in the calm group,

[0032]

[0033] r j represents a random variable ranging from 0 to 1 and follows a standard normal distribution.

[0034]

[0035] v and x follow a normal distribution, v j ~N(0,1), is the standard normal distribution, x j ~N(0,σ 2 ), n takes values ​​of 1, 2,

[0036]

[0037] represents the parameters that are dynamically adjusted by the algorithm, Γ represents the gamma function,

[0038]

[0039] yes The initial value of is 1.5, π represents pi;

[0040] The following formula is used to update the individual positions in the group following the general flow:

[0041]

[0042] ES l,j is the position of an individual randomly selected from the panic group, indicating the potential direction of panic-driven movement, ω1 and ω2 are adaptive Levy weights, τ2 takes values ​​of 0 or 1 with equal probability as obtained by the mass flow group, and v f,j It indicates the moving speed of the group following the large flow.

[0043]

[0044] v f,j Indicates the moving speed of the group following the large flow, H f,jIt indicates that the position is randomly generated within the large flow group.

[0045]

[0046] represents the minimum and maximum values ​​of the jth dimension of all individual positions in the large flow group, represents the individual position in the group following the general flow;

[0047] The following formula is used to update the individual positions in the panic group:

[0048]

[0049] D j represents the optimal solution in the initial optimal individual position set, ES rand,j represents the individual position randomly selected from the initial optimal individual position set, ES l,j is the position of an individual randomly selected from the panic group, indicating the potential direction of panic-driven movement, v l,j Indicates the movement speed of the panic group.

[0050]

[0051] v l,j represents the moving speed of the panic group, H l,j It represents the randomly generated individual position in the panic group.

[0052]

[0053] H l,j It represents the randomly generated individual position in the panic group. represents the minimum and maximum values ​​of the jth dimension of all individual positions in the panic group, Indicates the individual position in the panic group.

[0054] In a possible design manner of the first aspect, the individual positions in the initial optimal individual position set are updated by a wavelet mutation strategy that obeys the t-distribution probability, specifically by using the following formula to update the individual positions in the initial optimal individual position set:

[0055]

[0056] t k represents the wavelet mutation probability that follows the variation law of the t-distribution probability density function, a1 is a random number in [0,1]. <t k When , all contemporary individual positions are fine-tuned by wavelet. represents the mutation value obtained by wavelet mutation,

[0057]

[0058] v represents the degree of freedom, and the best effect is when it is 3. Γ represents the gamma function.

[0059]

[0060] is the variation value of the optimal solution obtained through wavelet mutation, Y j,max and Y j,min The optimal solutions D j The search upper and lower boundaries of the j-th dimension, φ is the value of the wavelet function,

[0061]

[0062] z is randomly generated from [-2.5n, 2.5n], e represents the exponential function, and n represents the fine-tuning parameter to meet the purpose of wavelet mutation fine-tuning.

[0063]

[0064] ln represents the logarithmic function, x is the upper limit of n, and its value is 10000, and ζ represents the shape parameter, and its value is 5.

[0065] Based on the above technical solution, by introducing a probability density function that obeys the t distribution, the probability of the algorithm mutating in the early and late stages of the iteration is relatively small, which helps to speed up the convergence of the algorithm. In the middle of the iteration, the algorithm will have a greater probability of disturbance, which can effectively help the algorithm escape from the local optimal solution and effectively improve the accuracy and stability of the ore volume prediction.

[0066] In a possible design manner of the first aspect, the optimal ore body volume prediction objective function is:

[0067]

[0068] V Fitness represents the objective function, M represents the total number of samples, V actual,m represents the true value of the ore volume of the mth ore body, V predicted,m It represents the predicted value of the ore body volume of the mth ore body predicted by the BP neural network.

[0069] In a second aspect, a method for predicting the volume of an ore body is provided, wherein the method is based on the BP neural network trained as described above.

[0070] In a third aspect, an electronic device is provided, comprising: a processor, and a memory coupled to the processor, the memory being used to store a computer program; the processor being used to execute the computer program stored in the memory, so that the electronic device performs a training method as any possible implementation of the first aspect, or performs a prediction method as any possible implementation of the second aspect.

[0071] In a fourth aspect, a computer-readable storage medium is provided, comprising a computer program or instructions, which, when executed on a computer, enables the computer to execute a training method as in any possible implementation of the first aspect, or to execute a prediction method as in any possible implementation of the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the embodiments or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0073] Figure 1 It is a flow chart of optimizing the initialization weights of the BP neural network based on the improved escape optimization algorithm in the embodiment of the present application;

[0074] Figure 2 It is a convergence comparison diagram of the improved escape optimization algorithm of the embodiment of the present application for optimizing the initialization weights of the BP neural network;

[0075] Figure 3 It is a comparison chart of the mineral volume value predicted by the BP neural network based on the improved escape optimization algorithm in the embodiment of the present application, the volume value predicted by the BP neural network based on the escape optimization algorithm and the actual value. DETAILED DESCRIPTION

[0076] In order to make the purpose, technical solution 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 used to limit the present application.

[0077] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0078] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0079] An embodiment of the present application provides an efficient method for predicting the volume of an ore body. An improved escape optimization algorithm is used to optimize the initialization weights of a BP neural network to obtain the optimal solution of the initialization weights. On this basis, a highly efficient BP neural network is obtained. The BP neural network can be used to predict the volume of an ore body, thereby improving the efficiency, accuracy and stability of mineral volume prediction. The detailed steps are as follows:

[0080] S1, establish a mapping relationship between the initial weights of the BP neural network and the individual positions of the improved escape optimization algorithm;

[0081] S2. Use the improved escape optimization algorithm to optimize the initialization weights of the BP neural network, and then obtain the optimal solution of the initialization weights;

[0082] S3: Input the optimal solution of the initialization weights obtained in stage S2 into the BP neural network, and then construct a highly efficient BP neural network.

[0083] In this way, the constructed BP neural network is used to predict the volume of the geological model of laterite bauxite, and finally outputs the corresponding ore body volume prediction value. Of course, in other embodiments, the volume of other types of ore bodies can also be predicted.

[0084] The improved escape optimization algorithm is used to optimize the initial weights of the BP neural network, and the initial weights of the optimal geological model volume prediction are optimized through the objective function. Among them, the objective function V Fitness It is expressed by formula (1);

[0085]

[0086] In formula (1), V Fitness represents the objective function, M represents the total number of samples, V actual,m Represents the true volume value of the mth ore body, V predicted,m Represents the volume of the mth ore body predicted by the BP neural network.

[0087] Shallow wells or drilling projects are carried out at key locations in the laterite bauxite area, and cores are extracted for testing to record geological parameters related to the ore body volume. Then, the collected data is cleaned and the cleaned data set is divided into three parts: training set, validation set and test set. Using the drilling data and shallow well engineering data obtained from the exploration work, the variables that affect the volume of laterite bauxite are collected: the top plate of the ore body block, the bottom plate of the ore body block, the terrain undulation, thickness, and the density of the ore body block.

[0088] The Kendall rank correlation coefficient was used to analyze the main factors affecting the volume of laterite bauxite, and the influencing factors with strong correlation were extracted and used as the input data of the BP neural network. The correlation was calculated using formula (2);

[0089]

[0090] In formula (2), τ b represents the Kendall rank correlation coefficient, represents the number of concordances between two variables, β represents the number of discordant pairs between two variables, and p represents the number of samples of laterite bauxite volume.

[0091] The factors with strong correlation are taken as the input items of BP neural network. At the same time, the collected data set is normalized by mean, and the normalized data set is divided into training set, validation set and test set according to a certain ratio, where the training set accounts for 70%, the validation set accounts for 15%, and the test set accounts for 15%. The calculation of mean normalization is expressed as formula (3);

[0092]

[0093] In formula (3), h norm represents the value after data mean normalization, h m represents the mth value in the collected ore volume data set, h mean represents the average value of the collected ore volume data set, h max represents the maximum value in the collected ore volume data set, h min represents the minimum value in the collected ore volume data set, and m represents the mth data point in the data set.

[0094] The training set is input into the BP neural network constructed in S2 for training. Then, the performance of the trained BP neural network is evaluated using the test set and validation set. After the evaluation, the actual data is input into the BP neural network. After the output result is obtained, it is denormalized to obtain the predicted value of the ore volume.

[0095] Different from other prediction models, BP neural network is a learning model constructed by multi-layer neural network. The input data will be processed by hidden layer and output layer in turn to obtain the prediction result of ore volume, which is expressed as formula (4);

[0096] g=A3(W3A2(W2A1(TW1+b1)+b2)+b3) (4)

[0098] In formula (4), T represents the input value, W1 represents the weight value from the input layer to the first hidden layer, W2 represents the weight value from the first hidden layer to the second hidden layer, W3 represents the weight value from the second hidden layer to the output layer, b1 represents the bias term of the first hidden layer, b2 represents the bias term of the second hidden layer, b3 represents the bias term of the output layer, A1 represents the GELU activation function from the input layer to the first hidden layer, which is formula (5), A2 represents the tanh activation function from the first hidden layer to the second hidden layer, which is formula (6), and A3 represents the Softplus activation function from the second hidden layer to the output layer, which is formula (7).

[0099]

[0100] In formula (5), A1 represents the GELU activation function from the input layer to the hidden layer, T represents the input data, and tanh() represents the hyperbolic tangent function;

[0101]

[0102] In formula (6), A2 represents the tanh activation function from the first hidden layer to the second hidden layer, e represents the exponential constant, and T represents the input data;

[0103] A3(T)=log(1+e T ) (7)

[0104] In formula (7), A3 represents the Softplus activation function from the second hidden layer to the output layer, log represents the logarithmic function, e represents the logarithmic constant, and T represents the input data;

[0105] The ore volume is predicted using the BP neural network optimized by the improved escape optimization algorithm. The input of the BP neural network is the variables affecting the ore volume obtained from the exploration project, and the output result is the predicted ore volume value.

[0106] Step S2 uses the improved escape optimization algorithm to optimize the initialization weights of the BP neural network and obtains the optimal solution of the initialization weights as follows: Figure 1 As shown, the specific steps are as follows:

[0107] S21, use formula (8) to initialize the population,

[0108] ES i =LB+(UB-LB)·rand(0,1) (8)

[0109] In formula (8), ES i represents the i-th candidate solution, LB represents the lower boundary of the search space, UB represents the upper boundary of the search space, and rand represents a random number in the interval [0,1].

[0110] S22. Use formula (1) to calculate the objective function values ​​of all individuals in the population, and select the best individuals from them and store them in the elite pool G, which is expressed as formula (9).

[0111] G={ES1,ES2,…,ES exist} (9)

[0113] In formula (9), G represents the number of potential solutions discovered by the group, {ES1,ES2,…,ES exist} represents the best potential individual.

[0114] S23, judging the current number of iterations FE and the maximum number of iterations FE max If FE is greater than FE max , jump to execute S28.

[0115] S24, perform algorithm judgment through the new Sine chaotic map γ. If γ is greater than rand, execute step S25 to enter the evolution of crowd behavior during the evacuation process for algorithm exploration and update. Otherwise, execute step S26 to enter the algorithm development stage for individual position update, where rand represents a random number in the interval [0,1], and the new Sine chaotic map γ is expressed as formula (10),

[0116]

[0117] In formula (10), FE represents the current iteration number, γ FE represents the new Sine chaotic map at the FE iteration, γ FE+1 It represents the new Sine chaotic map at the time of iteration FE+1, which controls the algorithm to update the position. Cos represents the cosine value operation. Sin represents the sine value operation. Pi represents the pi. A represents a random number between (0, 4). The initial value of γ is a random number between 0 and 1.

[0118] A new chaotic feature of the Sine chaotic map γ is proposed to generate unpredictable dynamic behavior, which enables the algorithm to flexibly switch between global search and local search, and can better jump out of the local optimal solution, effectively improving the accuracy and stability of the algorithm in geological model volume prediction in laterite bauxite.

[0119] S25. Use formula (1) to calculate the fitness of the objective function values ​​of all individuals in the elite pool G (the fitness of the objective function is usually calculated by converting the objective function into a fitness function). According to the individual fitness, the greedy strategy is prioritized to sort the individuals, thereby dividing the individuals into a calm group, a follow-the-crowd group, and a panic group. Through the escape optimization algorithm, the evolution of crowd behavior during the evacuation process is entered to update the individual positions.

[0120] S251. Escape optimization algorithm divides individuals into three groups according to their different behavioral responses during the evacuation process, q = 15%, f = 35%, l = 50%, which are the calm group q, the follow-the-flow group f and the panic group l. The subsequent steps are as follows:

[0121] S252, the panic factor P(FE) at the beginning of each iteration of FE is represented by a nonlinear factor, which is expressed as formula (11),

[0122]

[0123] In formula (11), μ represents a random number in the interval [0,1], cos represents the operation of finding the cosine value, δ represents a constant value of 20, FE represents the current number of iterations, and FE max Indicates the maximum number of iterations.

[0124] By introducing new nonlinear factors, the escape optimization algorithm can perform more refined local optimization in the later stage, improve the optimization accuracy and convergence speed, and significantly improve the accuracy and stability of the algorithm in the volume prediction of geological models in laterite bauxite.

[0125] S253, use the calm group to update the individual position, towards the center position Q of the group decision j Move, expressed as formula (12):

[0126]

[0127] In formula (12), represents the newly updated individual value, Q j is the average of all individuals in the calm group in the jth dimension, τ1 takes the value of 0 or 1 with equal probability in the calm group, P(FE) is the panic factor, which is formula (11), v q,jrepresents the moving speed of the calm group, which is formula (13), and ω1 is an adaptive Levy weight, which is formula (16).

[0128]

[0129] In formula (13), v q,j represents the moving speed of the calm group, H q,j is a randomly generated position within the range of the calm group, which is formula (14). represents a small adjustment of the individual motion, which is formula (15).

[0130]

[0131] In formula (14), H q,j is a randomly generated position within the range of the calm group. represents the minimum and maximum values ​​of the jth dimension of all individuals in the calm group, represents individuals in the calm group.

[0132]

[0133] In formula (15), represents a small adjustment of the individual motion, r j represents a random variable ranging from 0 to 1 and follows a standard normal distribution.

[0134]

[0135] In formula (16), v and x obey normal distribution, v j ~N(0,1), is the standard normal distribution, x j ~N(0,σ 2 ),

[0136]

[0137] In formula (17), represents the parameter that is dynamically adjusted by the algorithm, which is formula (18), and Γ represents the gamma function.

[0138]

[0139] In formula (18), yes The initial value is 1.5, sin represents the sine function, FE represents the current number of iterations, FE max represents the maximum number of iterations, and π represents the ratio of pi.

[0140] S254. Use the following flow group to update the individual bit, which is expressed as formula (19):

[0141]

[0142] In formula (19), Q j is the average of all individuals in the calm group in the jth dimension, ES l,j is a randomly selected individual from the panic group, indicating the potential direction of panic-driven movement, ω1 and ω2 are adaptive Levy weights, expressed as (16), τ1 takes values ​​of 0 or 1 with the same probability as in the calm group, P(FE) is the panic factor, τ2 takes values ​​of 0 or 1 with the same probability as in the follow-the-flow group, v f,j It represents the moving speed of the large flow group, which is formula (20);

[0143]

[0144] In formula (20), v f,j Indicates the moving speed of the group following the large flow, H f,j It represents the position randomly generated within the large flow group, which is formula (21). represents a small adjustment of the individual motion, which is formula (15).

[0145]

[0146] H f,j It indicates that the position is randomly generated within the large flow group. represents the minimum and maximum values ​​of the j-th dimension of all individuals in the large flow group, Represents individuals in the group that follows the crowd.

[0147] S255, use the panic group to update the individual position to the center position D of the group decision j Move, expressed as formula (22):

[0149]

[0150] In formula (22), represents the newly updated individual value, D j represents the best solution in the population, and a panic-driven individual might go to this exit. ES rand,j represents an individual randomly selected from the population, ES l,j is a randomly selected individual from the panic group, indicating the potential direction of panic-driven movement, ω1 and ω2 are adaptive Levy weights, expressed as (16), τ1 takes values ​​of 0 or 1 with the same probability as in the calm group, τ2 takes values ​​of 0 or 1 with the same probability as in the follow-the-flow group, P(FE) is the panic factor, v l,j It represents the moving speed of the panic group, which is formula (23);

[0151]

[0152] In formula (23), v l,j represents the moving speed of the panic group, H l,j It represents the randomly generated position in the panic group, which is formula (24). represents a small adjustment of the individual motion, which is formula (15).

[0153]

[0154] In formula (24), H l,j It indicates a randomly generated position within the panic group. represents the minimum and maximum values ​​of the jth dimension of all individuals in the panic group, represents individuals in the panic group.

[0155] S26. Enter the algorithm development phase to update individual positions.

[0156] S261, update the individual position by the wavelet mutation strategy that obeys the t-distribution probability, expressed as formula (25);

[0157]

[0158] In formula (25), t k represents the wavelet mutation probability, that is, the probability obeys the variation law of the probability density function of the t distribution, and the formula is expressed as (26), a1 is a random number in [0,1]. When a1 <t k When all individuals of the current era are fine-tuned by wavelet, ES i,j represents the position of the i-th individual in the j-th dimension, D j represents the best solution in the population, represents the variation value obtained by wavelet mutation, which is formula (27), ES rand,j is the position of an individual randomly selected from the population, ω1 and ω2 are adaptive Levy weights, expressed as formula (16), τ1 takes values ​​of 0 or 1 with equal probability as in the calm group, and τ2 takes values ​​of 0 or 1 with equal probability as in the following group.

[0159]

[0160] In formula (26), t k represents the probability of wavelet mutation, v represents the degree of freedom, and the best effect is when the value is 3, FE represents the current number of iterations, FE max represents the maximum number of iterations, and Γ represents the gamma function.

[0161] By introducing a probability density function that obeys the t distribution, the probability of the algorithm mutating in the early and late stages of the iteration is relatively small, which helps to speed up the convergence of the algorithm. In the middle of the iteration, the algorithm will have a greater probability of being disturbed, helping the algorithm escape from the local optimal solution.

[0162]

[0163] In formula (26), is the mutation value of the optimal solution obtained through wavelet mutation, D j represents the best solution in the population, Y j,max and Y j,min The best solution D j The search upper and lower boundaries of the j-th dimension, φ is the wavelet function value, which is formula (27),

[0164]

[0165] In formula (27), φ represents the wavelet function value, z is randomly generated from [-2.5n, 2.5n], e represents the exponential function, and n represents the fine-tuning parameter to meet the purpose of wavelet mutation fine-tuning, which is formula (28).

[0166]

[0167] In formula (28), FE represents the current iteration number, FE max represents the maximum number of iterations, ln represents the logarithmic function, e represents the exponential function, x is the upper limit of n, and its value is 10000, and ζ represents the shape parameter, and its value is 5.

[0168] S262. Furthermore, the wavelet mutation strategy that obeys the probability density of t distribution is used to operate the optimal solution, which changes the development direction of the fleeing people and makes them enter other areas for development, improves the global search capability of the algorithm, accelerates the convergence speed, and thus improves the accuracy and speed of the algorithm in the volume prediction of geological models in laterite bauxite.

[0169] S27. Use formula (1) to calculate the objective function values ​​of all individuals in the population, then retain the individual with the smallest objective function value as the individual in the next generation population, and then jump to execute step S23.

[0170] S28. Output the optimal solution for initializing weights.

[0171] In the specific implementation process, the data set is first screened. In this data set, the factors that affect the volume of the ore body include: the top plate of the ore body block, the bottom plate of the ore body block, the terrain undulation, the thickness, the density of the ore body block, etc. After the data preprocessing is completed, it is used for the training of the BP neural network. After that, the experiment is carried out in the MATLAB environment. The experimental conditions set are: the maximum number of iterations FE max Set to 200, the population size N pop is set to 50, the problem dimension Dim is set to 8, and the upper and lower boundaries are set to UB = [1,1,1,1,1,1,1,1] and LB = [-1,-1,-1,-1,-1,-1,-1,-1] respectively.

[0172] Figure 2 The presented graph is a comparison of the convergence of the improved escape optimization algorithm (CNWESCC) when optimizing the initial weights of the BP neural network. The methods involved in the comparison include the escape optimization algorithm (ESC), the termite life cycle optimization algorithm (TLCO) and the arithmetic optimization algorithm (AOA). Figure 2 It can be found that the ore volume prediction method based on the improved escape optimization algorithm has a faster convergence speed and higher convergence accuracy.

[0173] Figure 3 The comparison chart between the predicted value and the actual value of the ore volume obtained by the BP neural network prediction based on the improved escape optimization algorithm (CNWESC-BP) is shown. The method to be compared is the BP neural network prediction based on the escape optimization algorithm (ESC-BP). Figure 3 It can be seen that the prediction error produced by the ore volume prediction method based on the improved escape optimization algorithm is relatively small.

[0174] The above-mentioned efficient method for predicting the ore body volume includes both the training method of the ore body volume prediction model of the embodiment of the present application and the prediction method of the ore body volume based on the trained model.

[0175] An embodiment of the present invention further provides an electronic device, comprising: a processor, and a memory coupled to the processor, the memory being used to store a computer program; the processor being used to execute the computer program stored in the memory, so that the electronic device executes a training method as in any possible implementation manner described above, or executes a prediction method as in any possible implementation manner described above. .

[0176] The electronic device may be a computing device such as a desktop computer, a notebook, a palmtop computer, a cloud server, etc. The electronic device may include, but is not limited to, a processor and a memory.

[0177] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the electronic device, and uses various interfaces and lines to connect various parts of the entire device.

[0178] The memory may be used to store the computer program, and the processor implements various functions of the electronic device by running or executing the computer program stored in the memory and calling the data stored in the memory.

[0179] The memory may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, at least one application required for a function, etc.; the data storage area may store data created according to the use of the mobile phone, etc. In addition, the memory may include a high-speed random access memory, and may also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0180] The embodiment of the present invention further provides a storage medium, the storage medium is a computer-readable storage medium, the computer program is stored in the computer-readable storage medium, and the computer program, when executed by the processor, can implement the steps of the above-mentioned various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0181] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A training method for an ore body volume prediction model, characterized in that: The prediction model adopts a back propagation BP neural network, and the training set includes: normalized geological attributes strongly related to the ore body volume, the actual value of the ore body volume, and the predicted value of the ore body volume. The training method includes: Establishing a mapping relationship between the BP neural network weights and the individual positions in the individual position optimization algorithm, wherein the individual position optimization algorithm adopts an optimal ore body volume prediction objective function; The BP neural network is trained using the training set, and at the same time, the initialization weights of the BP neural network are optimized using the individual position optimization algorithm.

2. The training method according to claim 1, characterized in that: The individual position optimization algorithm is used to optimize the initialization weights of the BP neural network, specifically including: Initializing the individual positions to obtain an initial optimal individual position set that meets the requirements of the optimal ore body volume prediction objective function; Before the maximum number of training iterations is reached, when the current Sine chaotic map value meets the random number requirement, the individual positions in the initial optimal individual position set are sorted and grouped according to fitness, and the escape optimization algorithm is used to update the individual positions in the initial optimal individual position set. Otherwise, the individual positions in the initial optimal individual position set are updated through the wavelet mutation strategy that obeys the t-distribution probability. When the maximum number of training iterations is reached, the optimal individual position is output as the optimal initialization weight of the BP neural network.

3. The training method according to claim 2, characterized in that: The training method further includes: determining whether the current Sine chaotic mapping value meets the random number requirement, specifically: determining whether the current Sine chaotic mapping value γ is greater than the random number rand, wherein: FE represents the current iteration number, γ FE represents the new Sine chaotic map at the FE iteration, γ FE+1 It represents the new Sine chaotic map at the time of iteration FE+1, which controls the algorithm to update the position. Cos represents the cosine value operation. Sin represents the sine value operation. Pi represents the pi. A represents a random number between (0, 4). The initial value of γ is a random number between 0 and 1.

4. The training method according to claim 2, characterized in that: The escape optimization algorithm is used to update the individual positions in the initial optimal individual position set, specifically including: The individual positions in the initial optimal individual position set are divided into a calm group q, a follow-the-flow group f and a panic group l; A nonlinear panic factor is used to update the individual positions in the calm group, the follow-the-crowd group, and the panic group respectively.

5. The training method according to claim 4, characterized in that: The following formula is used to update the individual positions in the calm group: Among them, ES i,j represents the position of the i-th individual in the j-th dimension, represents the updated individual position, Q j is the average of all individual positions in the calm group in the jth dimension, τ1 takes the value 0 or 1 with equal probability for the calm group, P(FE) is the nonlinear panic factor, v q,j represents the moving speed of the calm group, ω1 is the adaptive Levy weight, μ represents a random number in the interval [0,1], δ represents a constant value of 20, FE max represents the maximum number of iterations, v q,j represents the moving speed of the calm group, H q,j is a randomly generated position within the range of the calm group, represents small adjustments in individual movements, represents the minimum and maximum values ​​of the jth dimension of all individual positions in the calm group, represents the individual position in the calm group, r j represents a random variable ranging from 0 to 1 and follows a standard normal distribution. v and x follow a normal distribution, v j ~N(0,1), is the standard normal distribution, x j ~N(0,σ 2 ), n takes values ​​of 1, 2, represents the parameters that are dynamically adjusted by the algorithm, Γ represents the gamma function, yes The initial value of is 1.5, π represents pi; The following formula is used to update the individual positions in the group following the general flow: ES l,j is the position of an individual randomly selected from the panic group, indicating the potential direction of panic-driven movement, ω1 and ω2 are adaptive Levy weights, τ2 takes values ​​of 0 or 1 with equal probability as obtained by the mass flow group, and v f,j It indicates the moving speed of the group following the large flow. v f,j Indicates the moving speed of the group following the large flow, H f,j It indicates that the position is randomly generated within the large flow group. represents the minimum and maximum values ​​of the jth dimension of all individual positions in the large flow group, represents the individual position in the group following the general flow; The following formula is used to update the individual positions in the panic group: D j represents the optimal solution in the initial optimal individual position set, ES rand,j represents the individual position randomly selected from the initial optimal individual position set, ES l,j is the position of an individual randomly selected from the panic group, indicating the potential direction of panic-driven movement, v l,j Indicates the movement speed of the panic group. v l,j represents the moving speed of the panic group, H l,j It represents the randomly generated individual position in the panic group. H l,j It represents the randomly generated individual position in the panic group. represents the minimum and maximum values ​​of the jth dimension of all individual positions in the panic group, Indicates the individual position in the panic group.

6. The training method according to claim 2, characterized in that: The individual positions in the initial optimal individual position set are updated by a wavelet mutation strategy that obeys the t-distribution probability. Specifically, the individual positions in the initial optimal individual position set are updated by using the following formula: t k represents the wavelet mutation probability that follows the variation law of the t-distribution probability density function, a1 is a random number in [0,1]. <t k When , all contemporary individual positions are fine-tuned by wavelet. represents the mutation value obtained by wavelet mutation, v represents the degree of freedom, and the best effect is when it is 3. Γ represents the gamma function. is the variation value of the optimal solution obtained through wavelet mutation, Y j,max and Y j,min The optimal solutions D j The search upper and lower boundaries of the j-th dimension, φ is the value of the wavelet function, z is randomly generated from [-2.5n, 2.5n], e represents the exponential function, and n represents the fine-tuning parameter to meet the purpose of wavelet mutation fine-tuning. ln represents the logarithmic function, x is the upper limit of n, and its value is 10000, and ζ represents the shape parameter, and its value is 5.

7. The training method according to claim 1, characterized in that: The optimal ore body volume prediction objective function is: V Fitness represents the objective function, M represents the total number of samples, V actual,m represents the true value of the ore volume of the mth ore body, V predicted,m It represents the predicted value of the ore body volume of the mth ore body predicted by the BP neural network.

8. A method for predicting ore body volume, characterized in that: The prediction method is based on the BP neural network trained according to any one of claims 1-7.

9. An electronic device, characterized in that: The electronic device comprises: a processor, and a memory coupled to the processor, The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the electronic device performs the training method as described in any one of claims 1 to 7, or performs the prediction method as described in claim 8.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a computer program or instructions. When the computer program or instructions are executed on a computer, the computer executes the training method according to any one of claims 1 to 7, or executes the prediction method according to claim 8.

Images