Slope displacement prediction method based on IMVO-RELM model
The improved IMVO-RELM model solves the problems of low efficiency and insufficient accuracy of traditional slope displacement prediction methods, achieving high-precision and stable slope displacement prediction and improving the reliability of mine safety production.
Patent Information
- Application Number
- CN202511477523.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-16
- Publication Date
- 2026-01-09
AI Technical Summary
Traditional slope displacement prediction methods rely on human experience, which is inefficient and has limited accuracy, making it difficult to meet the needs of modern mine safety production.
A slope displacement prediction method based on the IMVO-RELM model is adopted. The search ability and convergence are improved by improving the multiverse optimization algorithm (MVO). The IMVO-RELM model is constructed by combining spiral update and adaptive compression factor. The RMSE, MAPE and R2 are used for evaluation.
It significantly improves the accuracy and stability of slope displacement prediction, achieving efficient and accurate slope displacement prediction and reducing risks to mine production and personnel safety.
Smart Images
Figure CN121301927A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to a slope displacement prediction method based on an IMVO-RELM model and belongs to the technical field of slope engineering. BACKGROUND
[0002] As the mining scale gradually expands, the height of the formed slope gradually increases, and gradually becomes a major hidden danger of mine safety production. Therefore, accurately predicting the slope displacement deformation is of great significance to the safety production and environmental protection of mines.
[0003] However, the traditional slope displacement prediction method often relies on artificial experience judgment, and these methods are not only inefficient, but also have limited accuracy, which is difficult to meet the needs of modern mine safety production. With the rapid development of deep learning technology, its strong ability in image recognition, pattern recognition and other fields provides a new solution for slope displacement prediction. The deep learning model can learn a large number of historical displacement data to find complex patterns and rules, thereby realizing high-precision prediction of slope displacement deformation. Based on deep learning, a high-precision prediction model is established to predict the displacement deformation of the open pit slope, thereby effectively reducing the risks of mine production and personnel safety. SUMMARY
[0004] In view of the problems of low prediction accuracy and difficulty in error control in the current slope displacement prediction research and practice, the application provides a slope displacement prediction method based on an IMVO-RELM model. The slope displacement data is collected, the training set and the test set are divided, and the data of the training set and the test set are normalized respectively. In order to improve the search ability and convergence ability of the MVO algorithm in the parameter optimization process of the RELM, the MVO algorithm is improved, the spiral update and the adaptive compression factor are introduced, the wormhole existence probability (WEP) and the travel distance rate (TDR) of the traditional multi-universe optimization algorithm are improved in the form of an exponential, and the IMVO algorithm is obtained. The parameters of the RELM are optimized by using the IMVO algorithm, and an IMVO-RELM slope displacement prediction model is constructed. The data of the training set are used to train the IMVO-RELM slope displacement prediction model, and the optimal training model is obtained. The data of the test set are input into the optimal training model for slope displacement prediction, and the RMSE, MAPE and R 2 The slope displacement prediction result is evaluated.
[0005] A slope displacement prediction method based on an IMVO-RELM model, and the specific steps are as follows: S1. Collect the slope displacement data, divide the training set and the test set, and normalize the data of the training set and the test set respectively; S2. To improve the search and convergence capabilities of the MVO algorithm in optimizing RELM parameters, the traditional MVO algorithm is improved to obtain the IMVO algorithm, which specifically includes: S21. Update the wormhole existence probability (WEP) and travel distance rate (TDR) in an exponential form; S22. Introduce a spiral update strategy to enhance the algorithm's global search capability; S23. Introduce an adaptive compression factor to adjust the position of the optimal universe; S3. To improve the stability of the ELM model optimization process, a regularization parameter is introduced. The RELM model is obtained by improving the ELM model; S4. The optimal weights and thresholds of the RELM model are obtained using the IMVO algorithm. The RELM model is then optimized using the optimal weights and thresholds to construct the IMVO-RELM slope displacement prediction model. The IMVO-RELM slope displacement prediction model is trained using the data from the training set to obtain the optimal training model. S5. Input the test set data into the optimal training model to predict slope displacement, using the root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient R0. 2 The slope displacement prediction results are evaluated.
[0006] The normalization formula for S1 is: ; in, For the data that currently needs to be normalized, The maximum value in the data that needs to be normalized. The minimum value of the data that needs to be normalized. It is the normalized data.
[0007] The update formulas for the S21. Wormhole Existence Probability (WEP) and Travel Distance Rate (TDR) are as follows: ; ; In the formula, The maximum number of iterations, This represents the current iteration number. This represents the probability of the maximum wormhole existing. Let K be the minimum probability of a wormhole existing, and K be a constant, where K = 5000. It is a constant. =500.
[0008] S22. A spiral update strategy is introduced to enhance the global search capability of the algorithm. The calculation formula is as follows: ; in: It is the current optimal universe's number One parameter; Indicates the first The first universe One parameter; The coefficients that describe the spiral shape, It is a random number between 0 and 1.
[0009] The formula for calculating the adaptive compression factor in S23 is as follows: ; in, Indicates the current iteration number. This represents the maximum number of iterations. Introducing an adaptive compression factor, the formula for updating an individual's cosmic position is:
[0010] In the formula, Indicates the lower limit. Indicates the upper limit. It is an adaptive compression factor. It is the current optimal universe's number One parameter, , , It is a random number between 0 and 1.
[0011] S3. Introduces regularization parameters. The update formula for the RELM model is: ; in: This is the output matrix of the neural network. It is the hidden layer output matrix The generalized inverse matrix, It is the identity matrix. It is the regularization coefficient.
[0012] The method for constructing the S4.IMVO-RELM slope displacement prediction model is as follows: S41. Calculate the current universe expansion rate by using the improved multiverse optimization algorithm. If the current universe expansion rate is better than the current universe expansion rate, update the current universe expansion rate; otherwise, retain the current universe expansion rate. S42. Using the weights and thresholds of the RELM model as the optimization objects of the IMVO algorithm, and the root mean square error between the RELM predicted value and the true value as the error fitness function, the optimal weights and thresholds that satisfy the iteration termination condition are the optimal smoothing factors of the IMVO-RELM model. S43. Determine if the termination condition is met. If it is, the iteration terminates and the dataset is input into the RELM model optimized by IMVO. If it is not met, the iteration count is incremented by 1, and the process returns to step S41. S44. Optimize the RELM model using the optimal weights and thresholds to obtain the IMVO-RELM slope displacement prediction model. Import the training set data into the IMVO-RELM slope displacement prediction model for slope displacement prediction training.
[0013] The formula for the root mean square error (RMSE) is: ; The formula for Mean Absolute Percentage Error (MAPE) is: ; Correlation coefficient R 2 The formula is: ; In the formula: For data length, For the first Predicted values of slope stability at any given time. For the first The true value of the slope stability state at any given moment. This represents the average value of the actual sample.
[0014] The beneficial effects of this invention are: (1) The slope displacement prediction method based on the IMVO-RELM model of this invention uses RELM as the main prediction model and updates its weights and bias parameters using an improved MVO algorithm. By changing the update method of wormhole existence probability (WEP) and travel distance rate (TDR), the algorithm converges at a faster speed, effectively improving the convergence speed and optimization efficiency of the algorithm. A spiral update method is introduced to enhance the global search capability of the algorithm, avoiding the algorithm from getting stuck in local optima and effectively improving the convergence accuracy. An adaptive compression factor is introduced so that the algorithm can dynamically adjust the position of the optimal universe during the iteration process to adapt to the optimization requirements of the algorithm at different times. The improved multi-universe optimization algorithm is used to optimize the regularized extreme learning machine model, achieving accurate prediction of higher displacement. (2) The present invention introduces regularization parameters to improve the ELM model, thereby improving the stability of the model, and uses RELM for displacement prediction. The IMVO-RELM model significantly improves the accuracy and stability of slope displacement prediction. Attached Figure Description
[0015] Figure 1 The prediction flowchart for the IMVO-IELM model; Figure 2 Conceptual diagram of the multiverse optimization algorithm; Figure 3 A comparison curve of the original and improved WEP in the MVO algorithm; Figure 4 A comparison graph of the original and improved TDR in the MVO algorithm; Figure 5 Here is a diagram of the ELM model structure; Figure 6 The curve shown is the prediction result curve in the example. Detailed Implementation
[0016] The present invention will be further described in detail below with reference to specific embodiments, but the scope of protection of the present invention is not limited to the content described. Invention Overview A slope displacement prediction method based on the IMVO-RELM model (see...) Figure 1 and 2 The specific steps are as follows: S1. Collect slope displacement data, divide the data into training and test sets, and normalize the data in both sets. The normalization formula is: (1); in, For the data that currently needs to be normalized, The maximum value in the data that needs to be normalized. The minimum value of the data that needs to be normalized. It is normalized data; S2. To improve the search and convergence capabilities of the MVO algorithm in optimizing RELM parameters, the traditional MVO algorithm is improved to obtain the IMVO algorithm. Specifically, the MVO algorithm is an intelligent optimization algorithm based on the multiverse theory of physics. It uses the target optimization value as the cosmic expansion rate and establishes a mathematical model by simulating the motion behavior of a multidimensional cosmic population under the combined influence of white holes, black holes, and wormholes. The optimal solution is obtained through the evolution of the cosmic population, demonstrating significant advantages in solving low-dimensional, small-scale optimization problems. The cosmic population U is denoted as: (2) in, Indicates the number of cosmic groups; Indicates the number of variables; The normalized expansion rates of each universe are arranged, and a roulette wheel strategy is used to select white holes and black holes for matter exchange, realizing the rapid evolution of individual universes into "high-quality universes." The mathematical model is as follows: (3) in, Represents the universe The normalized inflation rate, A random number between 0 and 1 This refers to the white hole selected based on the roulette strategy. The Dimensional components; Individuals in the universe choose to travel through wormholes to reach the vicinity of the optimal universe, increasing their expansion rate to ensure population diversity; the iterative optimization mathematical model is as follows: (4) In the formula, Indicates the adaptive compression factor. This represents the current optimal universe. One object, As the lower limit, The upper limit, , , A random number between 0 and 1; TDR and WEP are the cosmic travel distance rate and the probability of wormhole existence, respectively. They are important adjustment parameters in the MVO algorithm and are updated using equations (5) and (6). (5) (6) In the formula, This represents the current iteration number. The maximum number of iterations, This represents the minimum probability of a wormhole existing (an empirical value of 0.2 can be used). This represents the maximum probability of a wormhole existing (an empirical value of 1 can be used). It is a parameter that characterizes the accuracy of algorithm development; S21. The wormhole existence probability (WEP) and travel distance rate (TDR) are updated using an exponential form. The update formulas for the wormhole existence probability (WEP) and travel distance rate (TDR) are as follows: (7) (8) In the formula, The maximum number of iterations, This represents the current iteration number. This represents the maximum probability of a wormhole existing (an empirical value of 1 can be used). Let K be the minimum probability of a wormhole existing (an empirical value of 0.2 can be taken), and K be a constant, K=5000. It is a constant. =500; S22. A spiral update strategy is introduced to enhance the algorithm's global search capability. The calculation formula is as follows: (9) in: Indicates the first The first universe One parameter; It is the current optimal universe's number One parameter; The coefficients that describe the spiral shape, It is a random number between 0 and 1; S23. An adaptive compression factor is introduced to adjust the position of the optimal universe. The calculation formula is as follows: (10) in, Indicates the current iteration number. This represents the maximum number of iterations. After adding the adaptive compression factor, the formula for updating the individual's cosmic position is as follows: (11) Introducing an improvement strategy, the iteratively optimized mathematical model is updated as follows: (12) In the formula, Indicates the first The first universe One parameter; It is an adaptive compression factor. It is the current optimal universe's number One parameter, Indicates the lower limit. Indicates the upper limit. , , It is a random number between 0 and 1; S3. To improve the stability of the ELM model optimization process, a regularization parameter is introduced. The RELM model is obtained by improving the ELM model; specifically, Traditional ELM algorithm outputs weight matrix for (13) in, Represents the hidden layer output matrix The generalized inverse matrix can be obtained analytically through orthogonal projection or singular decomposition. When there are many outliers in the training samples, the weight matrices of the hidden layer and the output layer... The uncertainty may affect the robustness and generalization of the model. Therefore, equation (13) is decomposed and transformed using the reflexivity of the generalized inverse matrix to obtain equation (14). (14) The regularization parameter is introduced into equation (14) to improve the model stability, thus obtaining the RELM model algorithm; (15) in: This is the output matrix of the neural network. It is the hidden layer output matrix The generalized inverse matrix, It is the identity matrix. It is the regularization coefficient; S4. The optimal weights and bias parameters of the RELM model are obtained using the IMVO algorithm. The RELM model is then optimized using the optimal weights and bias parameters to construct the IMVO-RELM slope displacement prediction model. The IMVO-RELM slope displacement prediction model is trained using the training set data to obtain the optimal training model. Specifically, the construction method of the IMVO-RELM slope displacement prediction model is as follows: S41. Calculate the current universe expansion rate by using the improved multiverse optimization algorithm. If the current universe expansion rate is better than the current universe expansion rate, update the current universe expansion rate; otherwise, retain the current universe expansion rate. S42. The weights and thresholds of the RELM model are used as the optimization objects of the IMVO algorithm. The root mean square error between the RELM predicted value and the true value is used as the error fitness function. The optimal weights and thresholds that satisfy the iteration termination condition are the optimal weights and thresholds of the IMVO-RELM model. S43. Determine whether the termination condition (reaching the optimal cosmic position) is met. If it is met, the iteration terminates, and the dataset is input into the RELM model optimized by IMVO. If it is not met, the iteration count is incremented by 1, and the process returns to step S41. S44. Optimize the RELM model using the optimal weights and thresholds to obtain the IMVO-RELM slope displacement prediction model. Import the training set data into the IMVO-RELM slope displacement prediction model for slope displacement prediction training. S5. Input the test set data into the optimal training model to predict slope displacement, using the root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient R0.2 The slope displacement prediction results are evaluated; specifically, the formula for the root mean square error (RMSE) is: ; The formula for Mean Absolute Percentage Error (MAPE) is: ; Correlation coefficient R 2 The formula is: ; In the formula: For data length, For the first Predicted values of slope stability at any given time. For the first The true value of the slope stability state at any given moment. This represents the average value of the actual sample.
[0018] Example 1: Using actual monitored displacement data from a mine in Yunnan Province from July 1, 2023 to May 29, 2024 as an example dataset, the average value of the monitored coordinates each day represents the valid coordinates for that day, with one valid data point taken each day, for a total of 333 data points; the dataset is divided into a training set and a test set, with the period from July 1, 2023 to March 22, 2024 used as modeling data to predict the displacement values from March 23, 2024 to May 29, 2024; the ELM, RELM, and MVO-RELM methods are compared with the IMVO-RELM method proposed in this invention; A slope displacement prediction method based on the IMVO-RELM model, the specific steps of which are as follows: S1. Collect slope displacement data, divide the data into training and test sets at a 3:1 ratio, and normalize the data in both sets. The normalization formula is: ; in, For the data that currently needs to be normalized, The maximum value in the data that needs to be normalized. The minimum value of the data that needs to be normalized. It is normalized data; S2. To improve the search and convergence capabilities of the MVO algorithm in optimizing RELM parameters, the traditional MVO algorithm is improved to obtain the IMVO algorithm. Specifically, the mathematical model for iterative optimization is as follows: ; In the formula, Indicates the adaptive compression factor. This represents the current optimal universe. One object, As the lower limit, The upper limit, , , A random number between 0 and 1; TDR and WEP are the cosmic travel distance rate and the probability of wormhole existence, respectively, and the update method is as follows: ; ; In the formula, This represents the current iteration number. The maximum number of iterations, This represents the minimum probability of a wormhole existing (an empirical value of 0.2 can be used). This represents the maximum probability of a wormhole existing (an empirical value of 1 can be used). It is a parameter that characterizes the accuracy of algorithm development; S21. The wormhole existence probability (WEP) and travel distance rate (TDR) are updated using an exponential form. The update formulas for the wormhole existence probability (WEP) and travel distance rate (TDR) are as follows: ; ; In the formula, The maximum number of iterations, This represents the current iteration number. This represents the maximum probability of a wormhole existing (an empirical value of 1 can be used). Let K be the minimum probability of a wormhole existing (an empirical value of 0.2 can be taken), and K be a constant, K=5000. It is a constant. =500; S22. A spiral update strategy is introduced to enhance the algorithm's global search capability. The calculation formula is as follows: ; in: Indicates the first The first universe One parameter; It is the current optimal universe's number One parameter; The coefficients that describe the spiral shape, It is a random number between 0 and 1; S23. An adaptive compression factor is introduced to adjust the position of the optimal universe. The calculation formula is as follows: ; in, Indicates the current iteration number. This represents the maximum number of iterations. After adding the adaptive compression factor, the formula for updating the individual's cosmic position is as follows: ; By introducing an improved strategy, the mathematical model for iterative optimization can be modified as follows: ; In the formula, Indicates the adaptive compression factor. This represents the current optimal universe. One object, As the lower limit, The upper limit, , , S3. To improve the stability of the ELM model optimization process, a regularization parameter is introduced. The RELM model is obtained by improving the ELM model; ; in: This is the output matrix of the neural network. It is the hidden layer output matrix The generalized inverse matrix, It is the identity matrix. It is the regularization coefficient; S4. The optimal weights and thresholds of the RELM model are obtained using the IMVO algorithm. The RELM model is then optimized using the optimal weights and thresholds to construct the IMVO-RELM slope displacement prediction model. The IMVO-RELM slope displacement prediction model is trained using the training set data to obtain the optimal training model. Specifically, the construction method of the IMVO-RELM slope displacement prediction model is as follows: S41. Calculate the current universe expansion rate by using the improved multiverse optimization algorithm. If the current universe expansion rate is better than the current universe expansion rate, update the current universe expansion rate; otherwise, retain the current universe expansion rate. S42. Using the weights and thresholds of the RELM model as the optimization objects of the IMVO algorithm, and the root mean square error between the RELM predicted value and the true value as the error fitness function, the optimal weights and thresholds that satisfy the iteration termination condition are the optimal weights and thresholds of the IMVO-RELM model. S43. Determine whether the termination condition (reaching the optimal cosmic position) is met. If it is met, the iteration terminates, and the dataset is input into the RELM model optimized by IMVO. If it is not met, the iteration count is incremented by 1, and the process returns to step S41. S44. Optimize the RELM model using the obtained optimal weights and thresholds to obtain the IMVO-RELM slope displacement prediction model. Import the training set data into the IMVO-RELM slope displacement prediction model for slope displacement prediction training. S5. Input the test set data into the optimal training model to predict slope displacement, using the root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient R0. 2 The slope displacement prediction results are evaluated; specifically, the formula for the root mean square error (RMSE) is: ; The formula for Mean Absolute Percentage Error (MAPE) is: ; Correlation coefficient R 2 The formula is: ; In the formula: For data length, For the first Predicted values of slope stability at any given time. For the first The true value of the slope stability state at any given moment. This represents the average value of the actual sample. In this embodiment, the dataset samples are input into three models: ELM, RELM, and MVO-RELM, and then into the IMVO-RELM method of this embodiment for 10 training and testing iterations. The model output results are recorded for each iteration. After the output is completed, the RMSE, MAPE, and R of the 10 test set samples are calculated. 2 The optimal value, mean, and standard deviation are used as evaluation parameters for the slope displacement prediction model. This reduces the randomness of the prediction results and comprehensively demonstrates the stability and generalization of the model performance. The evaluation indicators for the prediction results of each model are shown in Table 1, and the optimal prediction curves for each model are shown in Table 2. Figure 6 ; Table 1 Evaluation Indicators for Prediction Results of Each Model ; Table 1 shows that the standard deviations of the RMSE prediction results of the ELM model, RELM model, MVO-RELM model, and IMVO-RELM model are 1.256, 1.216, 1.480, and 1.268, respectively. The MVO-RELM model has poor prediction stability. The standard deviations of the RMSE, MPAE, and R of the ELM model are also relatively low. 2 The optimal and average values were the worst among the four models. RELM and MVO-RELM showed a certain degree of improvement over ELM, indicating that the prediction performance of the MVO-RELM and RELM models was significantly better than that of the ELM model. RMSE, MPAE, and R of the IMVO-RELM model 2 The optimal values were 4.456 mm, 0.26%, and 0.962, respectively, and the average values were 6.201 mm, 0.39%, and 0.924, respectively. Compared with the ELM model, RELM model, and MVO-RELM model, the IMVO-RELM model achieved better prediction performance. The RELM model improved prediction accuracy, and the MVO-RELM model improved prediction precision, but the prediction results had slightly higher dispersion. The IMVO-RELM model combines all the advantages of the RELM model and the MVO-RELM model in optimizing and improving the ELM model, achieving higher prediction accuracy and prediction stability.
[0019] The specific embodiments of the present invention have been described in detail above. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.
Claims
1. A slope displacement prediction method based on the IMVO-RELM model, characterized in that, The specific steps are as follows: S1. Collect slope displacement data, divide it into training set and test set, and normalize the data in the training set and test set respectively; S2. The traditional MVO algorithm is improved to obtain the IMVO algorithm, which specifically includes: S21. Update the wormhole existence probability (WEP) and travel distance rate (TDR) in an exponential form; S22. Introduce a spiral update strategy to enhance the algorithm's global search capability; S23. Introduce an adaptive compression factor to adjust the position of the optimal universe; S3. Introduce regularization parameters The RELM model is obtained by improving the ELM model; S4. The optimal weights and thresholds of the RELM model are obtained using the IMVO algorithm. The RELM model is then optimized using the optimal weights and thresholds to construct the IMVO-RELM slope displacement prediction model. The IMVO-RELM slope displacement prediction model is trained using the training set data to obtain the optimal training model. S5. Input the test set data into the optimal training model to predict slope displacement, using the root mean square error (RMSE), mean absolute percentage error (MAPE), and correlation coefficient R0. 2 The slope displacement prediction results are evaluated.
2. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: S1. The normalization formula is: ； in, For the data that currently needs to be normalized, The maximum value in the data that needs to be normalized. The minimum value of the data that needs to be normalized. It is the normalized data.
3. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: S21. The update formulas for the wormhole existence probability (WEP) and travel distance rate (TDR) are as follows: ； ； In the formula, The maximum number of iterations, This represents the current iteration number. This represents the probability of the maximum wormhole existing. This represents the probability of the minimum wormhole existing. K It is a constant. K =5000, It is a constant. =500.
4. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: S22. A spiral update strategy is introduced to enhance the algorithm's global search capability. The calculation formula is as follows: ； in: It is the current optimal universe's number One parameter; Indicates the first The first universe One parameter; The coefficients that describe the spiral shape, It is a random number between 0 and 1.
5. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: S23. The formula for calculating the adaptive compression factor is: ； in, Indicates the current iteration number. This represents the maximum number of iterations. Introducing an adaptive compression factor, the formula for updating an individual's cosmic position is: ； In the formula, Indicates the lower limit. Indicates the upper limit. It is an adaptive compression factor. It is the current optimal universe's number One object, , It is a random number between 0 and 1.
6. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: S3. Introduce regularization parameters The update formula for the RELM model is: ； in: This is the output matrix of the neural network. It is the hidden layer output matrix The generalized inverse matrix, It is the identity matrix. It is the regularization coefficient.
7. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: The construction method of the S4.IMVO-RELM slope displacement prediction model is as follows: S41. Calculate the current universe expansion rate by using the improved multiverse optimization algorithm. If the current universe expansion rate is better than the current universe expansion rate, update the current universe expansion rate; otherwise, retain the current universe expansion rate. S42. The weights and biases of the RELM model are used as the optimization objects of the IMVO algorithm. The root mean square error between the RELM prediction and the true value is used as the error fitness function. The optimal weights and biases that satisfy the iteration termination condition are the optimal smoothing factors of the IMVO-RELM model. S43. Determine if the termination condition is met. If it is, the iteration terminates and the dataset is input into the RELM model optimized by IMVO. If it is not met, the iteration count is incremented by 1, and the process returns to step S41. S44. Optimize the RELM model using the optimal weights and bias parameters to obtain the IMVO-RELM slope displacement prediction model. Import the training set data into the IMVO-RELM slope displacement prediction model for slope displacement prediction training.
8. The slope displacement prediction method based on the IMVO-RELM model according to claim 1, characterized in that: The formula for the root mean square error (RMSE) is: ； The formula for Mean Absolute Percentage Error (MAPE) is: ； Correlation coefficient R 2 The formula is: ； In the formula: For data length, For the first Predicted values of slope stability at any given time. For the first The true value of the slope stability state at any given moment. This represents the average value of the actual sample.