Strip mine slope deformation prediction method and system based on SA-PSO-MLP fusion algorithm

Abstract

The invention discloses a strip mine slope deformation prediction method based on an SA-PSO-MLP fusion algorithm. The strip mine slope deformation prediction method comprises the following steps: step 1, data preprocessing; 2, performing global search on an initial weight and a threshold value of a multi-layer perceptron (MLP) model by using a particle swarm optimization (PSO) algorithm; and step 3, refining a global optimal solution obtained by searching of a particle swarm optimization (PSO) algorithm by using a simulated annealing algorithm (SA) to obtain a weight and a threshold value of an optimal multi-layer perceptron (MLP) network. The invention also provides a strip mine slope deformation prediction system based on the SA-PSO-MLP fusion algorithm. According to the method, the simulated annealing algorithm and the particle swarm optimization algorithm are combined, the global optimization ability of PSO and the local fine search ability of SA are fully played, the problems that a traditional model is prone to falling into local optimum and high in parameter sensitivity are effectively solved, and the method has the advantages of being high in prediction precision, high in convergence speed and high in robustness; the method has a good application effect on deformation prediction of strip mine slopes and similar geologic bodies.

Description

Technical Field

[0001] This invention belongs to the field of mine safety monitoring and geological disaster prediction technology, and specifically relates to a method and system for predicting slope deformation in open-pit mines based on the SA-PSO-MLP fusion algorithm. Background Technology

[0002] Slope instability in open-pit mines is one of the main hazards to mine safety. Existing slope deformation prediction technologies mainly include finite element methods based on mechanical models, statistical regression models, and prediction algorithms based on machine learning. Traditional mechanical models rely on complex geological parameters and are difficult to reflect the dynamic changes in surface deformation in real time; statistical methods such as the grey prediction model GM(1,1) have low accuracy when there are insufficient samples or high noise.

[0003] In recent years, neural networks and intelligent optimization algorithms have been widely applied in the field of slope monitoring. For example, methods such as GA-BP, PSO-MLP, and SSA-BP improve model accuracy by optimizing the initial weights of the network, but they generally suffer from problems such as local optima and high parameter sensitivity. Especially under the complex geological conditions of open-pit mines, the predictive stability and generalization performance of these algorithms are still insufficient.

[0004] Therefore, how to construct a slope deformation prediction model that combines global optimization capability with local convergence accuracy to improve prediction accuracy and model robustness has become an important research direction in this field. Summary of the Invention

[0005] To address the aforementioned technical problems, the present invention aims to provide a method and system for predicting slope deformation in open-pit mines based on the SA-PSO-MLP fusion algorithm, wherein... SA The algorithm used is Simulated Annealing (PSO), Particle Swarm Optimization (PSO), and Multilayer Perceptron (MLP). This invention utilizes the Simulated Annealing algorithm to improve the search capability of PSO and combines it with the Multilayer Perceptron to achieve high-precision prediction of open-pit mine slope displacement.

[0006] To achieve the above objectives, this invention proposes a method for predicting the deformation of open-pit mine slopes based on SA-PSO-MLP, comprising the following steps: Step 1, Data Preprocessing: The raw displacement data from GNSS monitoring in open-pit mines are processed, including outlier detection, missing value imputation, and normalization, to obtain preprocessed data. The preprocessed data is then divided into training and testing sets, which are used as inputs to the MLP model. Step 2 involves using the Particle Swarm Optimization (PSO) algorithm to perform a global search for the initial weights and thresholds of the Multilayer Perceptron (MLP) model. This includes the following steps: Step 201, Set the particle swarm size Maximum number of iterations and inertial weight ; Step 202: Use the initial weights and thresholds of the MLP model as particles in the particle swarm, and initialize the position and velocity of the particles as random sequences respectively; Step 203: For each particle, calculate its current fitness value, evaluate each particle using the fitness value, and obtain the global optimal position. If the fitness value is better than the particle's historical best fitness value, then the current fitness value is taken as the new optimal position for the individual. p i and the current position x i Updated to the individual's optimal position; If the fitness value is better than the current global optimum position g If the fitness value is used as the new global optimal solution, the current position of the particle is recorded as the new global optimal position; and then proceed to step 301, where the global optimal solution is used as the initial solution. Otherwise, proceed to step 204; Step 204: Update the velocity and position of each particle; Step 3: Refine the global optimal solution obtained by the Particle Swarm Optimization (PSO) algorithm using the Simulated Annealing (SA) algorithm to obtain the weights and thresholds of the optimal Multilayer Perceptron (MLP) network. This specifically includes the following steps: Step 301, given the initial temperature And input the initial solution obtained in step 204; Step 302, based on the current temperature Generate a new solution and compute the new solution. The change in fitness value Δ of the current solution E : ; Where, Δ E This represents the change in fitness values ​​between the new solution and the current solution; E new This represents the fitness value of the new solution. E old This represents the fitness value of the current solution. Step 303, determine Δ E Is it less than 0? If Δ E If the value is less than 0, then accept the new solution and proceed to step 306; If Δ E If ≥0, proceed to step 304; Step 304, determine the probability Is it greater than rand , randThis indicates that a random number is generated from the interval [0,1]. Among them, probability P , The current temperature; like P>rand If so, accept the new solution and proceed to step 306; like P≤rand Then return to step 302; Step 306: Determine if the termination condition is met. The termination condition is that the temperature drops to a certain minimum value. Or the preset maximum number of iterations has been reached. T If the conditions are met, the optimal solution is output, i.e., the multilayer perceptron. MLP Network weights and thresholds; If the condition is not met, temperature decay is performed, and the process returns to step 302 to continue iterating.

[0007] Furthermore, in step 1, outlier detection employs... Z-score The specific calculation formula is as follows: ; In the formula: Z For the standardized data values, if | Z If |>3, then the point is considered an outlier; X For a certain data point; μ The mean of the data; σ The standard deviation is denoted as .

[0008] Furthermore, the missing values ​​in step 1 are filled in using cubic spline interpolation.

[0009] Furthermore, the inertial weight w Adaptive update strategy adopted: ; in, w max , w min These are the maximum inertia weight and the minimum inertia weight, respectively. t This represents the current iteration number. T This represents the maximum number of iterations.

[0010] Furthermore, the MLP model includes an input layer, two hidden layers, and an output layer. The activation function of the hidden layers is ReLU, and the output layer is a linear function.

[0011] Furthermore, in step 1, the data preprocessing uses a time window method to generate samples with a window length of 7 days, and uses the data from the first 7 days to predict the displacement value on the 8th day.

[0012] Furthermore, the fitness value is determined by the fitness function. F The fitness function was calculated. F The weighted combination of mean squared error and coefficient of determination is used, and the calculation formula is as follows: ; in, The weighting coefficient has a value range of [0,1]. MSE Mean square error, ,in, y m For the first m The true value of each sample; For the first m The model prediction value for each sample; n The number of samples; R 2 As the coefficient of determination, ; in, The average of the true values. .

[0013] Furthermore, in step 204, the update formulas for velocity and position are: ; ; in, t For the number of iterations, For the first i Individual particles t The speed of generation; For the first i Individual particles t The position of the generation; p i The optimal position for an individual; g The globally optimal position; c 1. c 2 is the learning factor; r1 and r2 are random numbers in the interval [0,1].

[0014] Furthermore, the temperature decay in step 306 employs an exponential cooling strategy: ; in, The cooling coefficient has a range of values. ; This is the current temperature.

[0015] This invention also provides an open-pit mine slope deformation prediction system based on the SA-PSO-MLP fusion algorithm, used to execute the open-pit mine slope deformation prediction method based on the SA-PSO-MLP fusion algorithm.

[0016] Compared with the prior art, the present invention has the following beneficial effects: The prediction method of this invention first preprocesses the raw displacement data monitored by GNSS radar on the slope, including outlier detection, missing value imputation, normalization, and dataset partitioning, to eliminate monitoring errors and obtain smooth time-series data. Then, in the parameter optimization stage, the initial weights and thresholds of the multilayer perceptron (MLP) model are globally searched using the particle swarm optimization (PSO) algorithm, and the search process of PSO is refined by the simulated annealing (SA) algorithm. Temperature decay and adaptive weight adjustment are introduced during model training, and the inertial weights and learning factors are dynamically adjusted under the cooling mechanism to improve the globality and stability of the search. Finally, the MLP model is trained using the double-optimized parameters, and the prediction is performed with the mean square error and coefficient of determination as the target, which can take into account both error control and goodness of fit, and output the displacement change results of the open-pit mine slope.

[0017] The SA-PSO-MLP model and prediction method of this invention predict the mean square error in complex geological environments of open-pit mines. MSE The coefficient of determination is 0.5099. R The accuracy reaches 0.9229, significantly improving the prediction accuracy of the SA-PSO-MLP model of this invention compared to the traditional MLP model. This fully verifies the applicability and reliability of the method of this invention in complex geological environments.

[0018] This invention combines simulated annealing (SSO) with particle swarm optimization (PSO), fully leveraging the global optimization capability of PSO and the local fine-grained search capability of SA. This effectively overcomes the problems of traditional models easily getting trapped in local optima and being highly sensitive to parameters. The improved SA-PSO-MLP model possesses advantages such as high prediction accuracy, fast convergence speed, and strong robustness, demonstrating excellent application results in predicting deformation of open-pit mine slopes and similar geological bodies. Furthermore, the model structure of this invention is highly scalable and can be extended to scenarios such as reservoir slopes, spoil heaps, and landslide deformation prediction. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1This is a schematic diagram of the principle of an open-pit mine slope deformation prediction method based on SA-PSO-MLP according to the present invention. Figure 2 This is a schematic diagram of the multilayer perceptron (MLP) network structure of the present invention; Figure 3 This is a flowchart illustrating the fusion of particle swarm optimization and simulated annealing algorithms of the present invention. Figure 4 The cubic spline interpolation of this invention after processing missing values x Directional deformation displacement polygonal graph; Figure 5 This is a comparison chart of the prediction results of the SA-PSO-MLP algorithm model of this invention with the actual values ​​of five other models.

[0021] Figure 6 The graph shows the absolute error of the SA-PSO-MLP algorithm model of this invention compared with five other models. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0023] like Figure 1 As shown, a method for predicting the deformation of open-pit mine slopes based on SA-PSO-MLP includes the following steps: Step 1, Data Preprocessing: The raw displacement data from GNSS monitoring in open-pit mines are processed, including outlier detection, missing value imputation, and normalization, to obtain preprocessed data. The preprocessed data is then divided into training and testing sets, which are used as inputs to the MLP model. This embodiment of the experiment selected a GNSS equipment observation point at the Dongbang branch of the Heishantou open-pit mine in Xinjiang. x The directional deformation displacement data was used as the research object. The data collection period was from March to August 2024. The GNSS monitoring equipment collected data at 08:00 am every day. A total of 182 sets of raw monitoring data were obtained, which were numbered as 1#, 2#...182#, as shown in Table 1.

[0024] Table 1 GNSS monitoring displacement data

[0025] Due to changes in the slope monitoring environment or factors such as the base station and monitoring points, the data acquired through real-time monitoring may contain some errors. Therefore, the monitoring data needs to be preprocessed before being used for model training and prediction. Errors generated during real-time slope monitoring can be classified into three categories: gross errors, systematic errors, and random errors. The data preprocessing section mainly deals with random errors and gross errors.

[0026] Outliers are data points that are significantly different from other data points due to measurement errors, data recording errors, or anomalies. Methods for handling outliers include: deleting outliers, replacing outliers, standardizing outliers, and using robust regression models. Because the data distribution is not asymmetrical, [the following is used]... Z-score The method for detecting whether data exceeds the standard deviation range considers points greater than 3 or less than -3 as outliers. The calculation formula is as follows: ; In the formula: Z For the standardized data values, if | Z If |>3, then the point is considered an outlier; X For a certain data point; μ The mean of the data; σ Standard deviation; The test revealed that the data were all within a reasonable range and there were no outliers. Meanwhile, it was discovered that data for March 30, 2024, was missing. Therefore, interpolation was needed to handle the missing values. Due to the complex data trends, cubic spline interpolation was chosen for better curve fitting and to preserve the smoothness of the data. The processed x-direction deformation displacement polygon is shown below. Figure 4 As shown.

[0027] After handling outliers and missing values, normalization was performed to form a training sample dataset of 183 groups. The first 80% was used as the training set for model training, and the last 20% was used as the validation set to test the model's fit. like Figure 1 and Figure 3 As shown, step 2 involves using the Particle Swarm Optimization (PSO) algorithm to perform a global search for the initial weights and thresholds of the Multilayer Perceptron (MLP) model. This includes the following steps: Step 201, Set the particle swarm size Maximum number of iterations and inertial weight ; In this embodiment, the particle swarm size is set. N =50, maximum number of iterations T =1000, inertia weight w According to the formula Adaptive adjustment, learning factorc 1= c 2 = 2.0; in, w max , w min These are the maximum inertia weight and the minimum inertia weight, respectively. t This represents the current iteration number. T This represents the maximum number of iterations. Step 202: Use the initial weights and thresholds of the MLP model as particles in the particle swarm, and initialize the position and velocity of the particles as random sequences respectively; Step 203: For each particle, calculate its current fitness value, evaluate each particle using the fitness value, and obtain the global optimal position. If the fitness value is better than the particle's historical best fitness value, then the current fitness value is taken as the new optimal solution for the individual. p i and the current position x i Updated to the individual's optimal position; If the fitness value is better than the current global optimum position g If the fitness value is used as the new global optimal solution, the current position of the particle is recorded as the new global optimal position; and then proceed to step 301, where the global optimal solution is used as the initial solution. Otherwise, proceed to step 204; The fitness value is obtained through the fitness function. F The fitness function was calculated. F The weighted combination of mean squared error and coefficient of determination is used, and the calculation formula is as follows: ; in, The weighting coefficient has a value range of [0,1]. MSE Mean square error, ,in, y m For the first m The true value of each sample; For the first m The model prediction value for each sample; n The number of samples; R 2 As the coefficient of determination, ;in, The average of the true values. ; Step 204, update the velocity and position of each particle, using the following formula: ; ; in, t For the number of iterations, For the first i Individual particles t The speed of generation; For the first i Individual particles t The position of the generation; p i The optimal position for an individual; g The globally optimal position; c 1. c 2 is the learning factor; r1 and r2 are random numbers in the interval [0,1]. Step 3, using the simulated annealing algorithm SA The global optimal solution obtained by the Particle Swarm Optimization (PSO) algorithm is refined to obtain the weights and thresholds of the optimal Multilayer Perceptron (MLP) network. This process includes the following steps: Step 301, given the initial temperature In this embodiment, the initial temperature =100, and input the initial solution obtained in step 204; Step 302, based on the current temperature Generate a new solution And calculate the new solution. The change in fitness value Δ of the current solution E : ; Where, Δ E This represents the change in fitness values ​​between the new solution and the current solution; E new This represents the fitness value of the new solution. E old This represents the fitness value of the current solution. Step 303, determine Δ E Is it less than 0? If Δ E If the value is less than 0, then accept the new solution and proceed to step 306; If Δ E If ≥0, proceed to step 304; Step 304, determine the probability Is it greater than rand , rand This indicates that a random number is generated from the interval [0,1]. Among them, probability P , The current temperature; like P>randIf so, accept the new solution and proceed to step 306; like P≤rand Then return to step 302; Step 306: Determine if the termination condition is met. The termination condition is that the temperature drops to a certain minimum value. Or the preset maximum number of iterations has been reached. T If the conditions are met, the optimal solution is output, which is the weights and thresholds of the multilayer perceptron (MLP) network. If the condition is not met, temperature decay is performed, and the process returns to step 302 to continue iterating.

[0028] Temperature decay employs an exponential cooling strategy: ; in, The cooling coefficient has a range of values. ; This is the current temperature.

[0029] In this embodiment, the cooling coefficient α =0.95.

[0030] like Figure 2 As shown, in this embodiment, the MLP model includes an input layer (7 nodes), two hidden layers (the first and second hidden layers), and an output layer (1 node). The hidden layer activation function is ReLU, the output layer is a linear function, and the optimizer... Adam The hyperparameters of the SA-PSO-MLP algorithm model of this invention are optimized using a Bayesian optimization model (existing technology). The parameter optimization space includes the number of neurons in the hidden layer, the learning rate, and the regularization coefficient. The optimal hyperparameters after optimization are: 101 neurons in the first hidden layer, 89 neurons in the second hidden layer, a learning rate of 0.0071, and a regularization coefficient of 0.07440.

[0031] For MLP models, the time window is set to 7, meaning the first 7 days are used. d Displacement data to predict the 8th d The displacement value; The formula for calculating the first hidden layer is: ; in, , is the output of the first hidden layer; RELU For activation functions; , which is the bias term for the first hidden layer; , which are the input features, i.e., the top 7 d Displacement value data; , where is the weight matrix of the first hidden layer; The formula for calculating the second hidden layer is: ; in, , is the output of the second hidden layer; RELU For activation functions; , which is the bias term of the second hidden layer; , where is the weight matrix of the second hidden layer; The formula for calculating the output layer is: ; in, , is the weight matrix of the output layer; , which is the bias term for the output layer; , which is the output feature of the output layer, i.e., the 8th d The displacement value; Experimental verification: Under the same experimental conditions, the SA-PSO-MLP algorithm model of this invention was compared with the genetic algorithm-improved BP neural network (GA-BP) and support vector machine regression (SVR). GM (1,1) Prediction accuracy of five models—grey prediction, MLP, and Particle Swarm Optimization algorithm-improved multilayer perceptron (PSO-MLP)—on monitoring data from the Heishantou open-pit mine in Xinjiang (evaluation metrics include mean squared error). MSE and coefficient of determination R 2 , MSE The smaller the value, the better. R 2 This is used to measure the goodness of fit of the model; the closer the value is to 1, the better the prediction effect. A comparison is shown in Table 2.

[0032] Table 2 Comparison of the accuracy of various prediction models

[0033] As can be seen from the data in Table 2, the SA-PSO-MLP algorithm model of this invention has a smaller mean square error overall compared to the other five models. MSE and a larger coefficient of determination R 2 This indicates that the prediction performance of SA-PSO-MLP is better than that of the comparison model. It also shows that the prediction results of SA-PSO-MLP are closer to the actual slope displacement monitoring values, thus verifying the feasibility of SA-PSO-MLP in slope deformation prediction.

[0034] In addition, the displacement prediction results of the SA-PSO-MLP algorithm model of this invention and five other models for data 148#-183# are compared, such as... Figure 5 As shown, the prediction curve of the SA-PSO-MLP algorithm model of this invention is in high agreement with the actual monitoring curve, indicating that the model of this invention has high accuracy and robustness in time series prediction.

[0035] The predicted values ​​and errors of each model were compared, as shown in Table 3. All errors in Table 3 are absolute errors, and both the error and predicted value are in units of 1 / 2 ppm. mm The absolute prediction errors of the six algorithms and such Figure 6 As shown.

[0036] For the m One sample, error Defined as: ; in, y m This is the actual measured value. mm ; For predicted values, mm ; The formula for calculating the sum of absolute errors is:

[0037] in, n The total number of samples; Table 3. Comparison of predicted values ​​and errors between the present invention and five other models.

[0038] As can be seen from Table 3, the SA-PSO-MLP model has the smallest absolute error.

[0039] Therefore, in summary, this embodiment demonstrates that the observation point of a GNSS device at the eastern side of the Heishantou open-pit mine in Xinjiang... x In predicting directional deformation displacement, the SA-PSO-MLP model of this invention not only has higher prediction accuracy, but also exhibits stronger robustness.

[0040] In summary, the SA-PSO-MLP of this invention has good predictive performance and practical feasibility in slope deformation prediction.

[0041] Although the present invention has been described in detail above with general descriptions and specific embodiments, modifications or improvements can be made to it, which will be obvious to those skilled in the art. Therefore, all such modifications or improvements made without departing from the spirit of the present invention fall within the scope of protection claimed by the present invention.

Claims

1. A method for predicting slope deformation in open-pit mines based on SA-PSO-MLP, characterized in that, Includes the following steps: Step 1, Data Preprocessing: The raw displacement data from GNSS monitoring in open-pit mines are processed, including outlier detection, missing value imputation, and normalization, to obtain preprocessed data. The preprocessed data is then divided into training and testing sets, which are used as inputs to the MLP model. Step 2 involves using the Particle Swarm Optimization (PSO) algorithm to perform a global search for the initial weights and thresholds of the Multilayer Perceptron (MLP) model. This includes the following steps: Step 201, Set the particle swarm size Maximum number of iterations and inertial weight ; Step 202: Use the initial weights and thresholds of the MLP model as particles in the particle swarm, and initialize the position and velocity of the particles as random sequences respectively; Step 203: For each particle, calculate its current fitness value, evaluate each particle using the fitness value, and obtain the global optimal position. If the fitness value is better than the particle's historical best fitness value, then the current fitness value is taken as the new optimal position for the individual. p i and the current position x i Updated to the individual's optimal position; If the fitness value is better than the current global best position g If the fitness value is used as the new global optimal solution, the current position of the particle is recorded as the new global optimal position; and then proceed to step 301, where the global optimal solution is used as the initial solution. Otherwise, proceed to step 204; Step 204: Update the velocity and position of each particle; Step 3: Refine the global optimal solution obtained by the Particle Swarm Optimization (PSO) algorithm using the Simulated Annealing (SA) algorithm to obtain the weights and thresholds of the optimal Multilayer Perceptron (MLP) network. This specifically includes the following steps: Step 301, given the initial temperature And input the initial solution obtained in step 204; Step 302, based on the current temperature Generate a new solution and compute the new solution. The change in fitness value Δ of the current solution E : ； Where, Δ E This represents the change in fitness values ​​between the new solution and the current solution; E new This represents the fitness value of the new solution. E old This represents the fitness value of the current solution. Step 303, determine Δ E Is it less than 0? If Δ E If the value is less than 0, then accept the new solution and proceed to step 306; If Δ E If ≥0, proceed to step 304; Step 304, determine the probability Is it greater than rand , rand This indicates that a random number is generated from the interval [0,1]. Among them, probability P , The current temperature; like P > rand If so, accept the new solution and proceed to step 306; like P ≤ rand Then return to step 302; Step 306: Determine if the termination condition is met. The termination condition is that the temperature drops to a certain minimum value. Or the preset maximum number of iterations has been reached. T If the conditions are met, the optimal solution is output, which is the weights and thresholds of the multilayer perceptron (MLP) network. If the condition is not met, temperature decay is performed, and the process returns to step 302 to continue iterating.

2. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP as described in claim 1, characterized in that, In step 1, outlier detection is performed using... Z-score The specific calculation formula is as follows: ； In the formula: Z is the standardized data value. If |Z|>3, the point is considered an outlier. X For a certain data point; μ The mean of the data; σ The standard deviation is denoted as .

3. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP as described in claim 1, characterized in that, The missing values ​​in step 1 are filled in by cubic spline interpolation.

4. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP as described in claim 1, characterized in that, The inertial weight w Adaptive update strategy adopted: ； in, w max , w min These are the maximum inertia weight and the minimum inertia weight, respectively. t This represents the current iteration number. T This represents the maximum number of iterations.

5. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP according to claim 1, characterized in that, The MLP model includes an input layer, two hidden layers, and an output layer. The activation function of the hidden layers is ReLU, and the output layer is a linear function.

6. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP according to claim 1, characterized in that, The data preprocessing in step 1 uses a time window method to generate samples with a window length of 7 days, and uses the data from the first 7 days to predict the displacement value on the 8th day.

7. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP according to claim 1, characterized in that, The fitness value is obtained through the fitness function. F The fitness function was calculated. F The weighted combination of mean squared error and coefficient of determination is used, and the calculation formula is as follows: ； in, The weighting coefficient has a value range of [0,1]. MSE Mean square error, ,in, y m For the first m The true value of each sample; For the first m The model prediction value for each sample; n The number of samples; R 2 As the coefficient of determination, ; in, The average of the true values. .

8. The method for predicting slope deformation in open-pit mines based on SA-PSO-MLP according to claim 1, characterized in that, In step 204, the update formulas for velocity and position are as follows: ； ； in, t For the number of iterations, For the first i Individual particles t The speed of generation; For the first i Individual particles t The position of the generation; p i The optimal position for an individual; g The globally optimal position; c 1. c 2 is the learning factor; r1 and r2 are random numbers in the interval [0,1].

9. The method for predicting open-pit mine slope deformation based on SA-PSO-MLP according to claim 1, characterized in that, The temperature decay in step 306 employs an exponential cooling strategy: ； in, The cooling coefficient has a range of values. ; This is the current temperature.

10. A method based on SA-PSO-ML P The open-pit mine slope deformation prediction system based on fusion algorithms is characterized by, The method for predicting the deformation of open-pit mine slopes based on the SA-PSO-MLP fusion algorithm as described in any one of claims 1 to 9.

Images