VMD-LSTM-based mountain road slope displacement prediction method and device and storage medium

Abstract

The invention provides a VMD-LSTM mountain road slope displacement prediction method based on a dung beetle optimization algorithm, and the method comprises the following steps: collecting daily displacement data of a to-be-measured slope, carrying out the data preprocessing and normalization processing of the collected displacement data, and dividing the data into a training set and a test set; constructing a VMD-LSTM mountain road slope displacement prediction model, and pre-training the prediction model by using the training set and the test set; optimizing hyper-parameters in the prediction model by adopting a dung beetle optimization algorithm; and carrying out mountain road slope displacement prediction based on the optimized prediction model. According to the mountain road slope displacement prediction method, the accuracy and reliability of mountain road slope displacement prediction can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of highway slope displacement prediction, and particularly relates to a method for predicting the displacement of mountain highway slopes based on the dung beetle optimization algorithm and VMD-LSTM. Background Art

[0002] As an important channel connecting remote areas with the outside world, mountain highways not only promote local economic development but also play an irreplaceable role in improving residents' living conditions and promoting cultural exchanges. However, due to the complex mountain terrain and variable geological conditions, the stability problem of highway slopes is particularly prominent, and landslides have become one of the main factors threatening the safe operation of mountain highways. Accurately predicting slope displacement is crucial for preventing disasters and ensuring traffic safety. Predicting the displacement of mountain highway slopes can not only help relevant departments take timely measures to reduce losses caused by natural disasters but also provide a scientific basis for highway maintenance and management.

[0003] In existing research on slope displacement prediction, due to the limitations of traditional prediction methods such as empirical formulas and statistical models in dealing with complex nonlinear problems, in recent years, prediction models based on machine learning and deep learning have gradually received attention. Variational mode decomposition (VMD), as an effective signal processing technology, has been widely applied to landslide displacement prediction. VMD can decompose complex nonlinear sequences into relatively stable subsequences, thus simplifying the subsequent analysis and prediction process. Long short-term memory network (LSTM) has been widely used in landslide displacement prediction due to its advantages in processing time series data. LSTM can effectively capture long-term dependencies in time series, thereby improving the accuracy of prediction.

[0004] In summary, although multiple models have been proposed and applied in practice, in a complex mountain environment, existing slope monitoring often has difficulty accurately capturing small but critical displacement changes, and the prediction accuracy is difficult to reach an ideal level. Summary of the Invention

[0005] In view of the above background, the present invention proposes a method for predicting the displacement of mountain highway slopes based on the dung beetle optimization algorithm and VMD-LSTM to improve the accuracy and reliability of mountain highway slope displacement prediction.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The first aspect of the present invention provides a method for predicting the displacement of mountain highway slopes based on the dung beetle optimization algorithm and VMD-LSTM, including the following steps:

[0008] Collect the daily displacement data of the slope to be measured, and perform data preprocessing and normalization on the collected displacement data, and divide it into a training set and a test set;

[0009] Construct a VMD-LSTM mountain highway slope displacement prediction model, and use the training set and the test set to pre-train the prediction model;

[0010] Use the dung beetle optimization algorithm to optimize the hyperparameters in the prediction model;

[0011] Based on the optimized prediction model, predict the displacement of the mountain highway slope.

[0012] Furthermore, collecting the daily displacement data of the slope to be measured, and performing data preprocessing and normalization on the collected displacement data includes:

[0013] Collect the slope displacement data in the (x, y, z) directions at the preset monitoring points of the slope to be measured based on the surface displacement sensor;

[0014] Adopt the Euclidean norm Fuse the slope displacement data in the (x, y, z) directions of each monitoring point to obtain the cumulative slope displacement;

[0015] Use the min-max normalization method to process the cumulative slope displacement data, that is

[0016]

[0017] where x is the original cumulative slope displacement data; x ′ is the normalized cumulative slope displacement data; max(x) is the maximum value in the sample; min(x) is the minimum value in the sample.

[0018] Furthermore, constructing a VMD-LSTM mountain highway slope displacement prediction model includes:

[0019] Construct a VMD-LSTM model, where the VMD model is used to decompose the normalized cumulative slope displacement data to obtain a series of sub-modal components {u 1 , u 2 , …, u k}, and use them as the input of the LSTM model. The LSTM model outputs the predicted values Y 1 , Y 2 , …, Y k ;

[0020] Sum up and superimpose the predicted values to obtain the final predicted value of the mountain highway slope displacement.

[0021] Further, pre-training the prediction model using the training set and the test set includes:

[0022] Establish an evaluation function, which includes the mean absolute error MAE, the root mean square error RMSE, and the mean absolute percentage error MAPE. Their expressions are respectively:

[0023]

[0024] where N is the number of samples in the test set; and Y i are respectively the predicted displacement value and the actual displacement value of the i-th sample in the test set;

[0025] Use the slope cumulative displacement data in the training set as the input of the VMD-LSTM model to obtain the corresponding predicted displacement value of the mountain highway slope. Combine the actual displacement value and use the evaluation function to evaluate the model performance to obtain the initial evaluation result.

[0026] Further, optimizing the hyperparameters in the prediction model using the dung beetle optimization algorithm includes:

[0027] Initialize the algorithm parameters and define the search space. The algorithm parameters include the population size and the number of iterations. Defining the search space includes setting reasonable value ranges for the number of decomposition modes k, the balance parameter α in the VMD model, and the hyperparameters of the LSTM model. The hyperparameters at least include the number of hidden layers, the number of neurons, the learning rate, and the batch size;

[0028] Encode all the hyperparameters to be optimized to form a multi-dimensional vector to represent the position of each dung beetle, and randomly initialize multiple hyperparameter combinations as the initial population;

[0029] Based on the hyperparameter configuration represented by each individual in the initial population, train the VMD-LSTM model and use the test set to perform fitness evaluation based on the evaluation function;

[0030] Simulate the behavior of dung beetles rolling dung balls, consider the balance between local search and global search, and update the position of each dung beetle;

[0031] Check whether the preset stopping criterion is met after each iteration, including reaching the maximum number of iterations or the best fitness not changing significantly in several consecutive iterations;

[0032] When the preset stopping criterion is met, select a set of hyperparameter configurations with the lowest fitness value from all historical records as the final optimization result.

[0033] Further, simulating the behavior of dung beetles rolling dung balls, considering the balance between local search and global search, and updating the position of each dung beetle includes:

[0034] At initialization, a position vector x i (t) is defined for each dung beetle, where i = 1, 2, ..., N represents the i-th dung beetle, N is the population size, and t represents the current iteration number; each position vector corresponds to a set of hyperparameter configurations, and a velocity vector v i (t) is used to guide the position update; let x best (t) be the best position found in the population so far up to the current iteration;

[0035] During each iteration, update the velocity and position of each dung beetle according to the following rules:

[0036] Update the velocity v i (t + 1) = w·v i (t) + c 1 r 1 (x best (t) - x i (t)) + c 2 r 2 (x pbest,i - x i (t)), where w is the inertia weight, c 1 and c 2 are the cognitive and social factors respectively, r 1 and r 2 are two random numbers independently distributed between [0, 1], and x p best,i is the best position found by the i-th dung beetle;

[0037] Update the position: x i (t + 1) = x i (t) + v i (t + 1);

[0038] Balance the local search and global search in the dung beetle optimization algorithm by adjusting the inertia weight w, and w gradually decreases from a larger value;

[0039] Meanwhile, set a small probability P rand , when the randomly generated probability is less than P rand , let x i (t + 1) be a random position within the search space;

[0040] After each iteration, calculate the fitness value according to the evaluation function and check whether the preset stopping criterion is satisfied. If the stopping condition is met, select a set of hyperparameter configurations with the lowest fitness value from all historical records as the final optimization result; otherwise, continue the next iteration.

[0041] In a second aspect of the present invention, a computer device is provided, including a memory and a processor. Computer instructions are stored in the memory, and the processor executes the mountain highway slope displacement prediction method as described in the first aspect above by executing the computer instructions.

[0042] In a third aspect of the present invention, a computer-readable storage medium is provided. The storage medium stores computer instructions, and when the computer instructions are run by the computer, they execute the mountain highway slope displacement prediction method as described in the first aspect above.

[0043] The beneficial technical effects of the present invention are as follows:

[0044] The mountain highway slope displacement prediction method of the present invention simultaneously introduces variational mode decomposition (VMD) and long short-term memory network (LSTM) to predict the slope displacement. By introducing the VMD model to preprocess the original slope displacement data, the complex non-linear time series is decomposed into multiple relatively stable sub-modal components, effectively reducing the influence of data noise. Then, combined with the LSTM model, it can more accurately capture the time-dependent relationships of each sub-modal component, thereby improving the overall accuracy of the mountain highway slope displacement prediction. At the same time, for the first time in the field of slope displacement, the dung beetle optimization algorithm is applied to the VMD-LSTM slope displacement prediction model. The dung beetle optimization algorithm is used to intelligently optimize the key hyperparameters in the VMD-LSTM model. By simulating the behavior pattern of dung beetles rolling dung balls, the local and global search capabilities are dynamically balanced in the search space, and a certain probability of random exploration mechanism is retained, so as to ensure finding the optimal or near-optimal hyperparameter configuration, further enhancing the generalization ability and prediction performance of the VMD-LSTM model. Description of the Drawings

[0045] Figure 1 It is a schematic flow chart of an embodiment of the mountain highway slope displacement prediction method of the present invention.

[0046] Figure 2 It is a schematic flow principle diagram of an embodiment of the mountain highway slope displacement prediction method of the present invention.

[0047] Figure 3 It is a schematic diagram of the unit structure of LSTM.

[0048] Figure 4 It is a slope displacement diagram after data preprocessing in an embodiment of the present invention.

[0049] Figure 5 It is a training loss function diagram in an embodiment of the present invention. Detailed Embodiments

[0050] Embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present invention. It should be understood that the drawings and embodiments of the present invention are only for exemplary purposes and are not used to limit the protection scope of the present invention.

[0051] See Figure 1 and Figure 2 , embodiments of the present invention provide a method for predicting the displacement of mountain highway slopes based on the dung beetle optimization algorithm, including the following steps:

[0052] S1. Collect the daily displacement data of the slope to be measured, and perform data preprocessing and normalization on the collected displacement data, and divide it into a training set and a test set.

[0053] As a preferred implementation, in this embodiment, the slope displacement data in the (x, y, z) directions at the preset monitoring points of the slope to be measured is collected based on the surface displacement sensor;

[0054] Then, use the Euclidean norm to fuse the slope displacement data in the (x, y, z) directions of each monitoring point to obtain the cumulative slope displacement;

[0055] Finally, use the maximum-minimum normalization method to process the cumulative slope displacement data, that is

[0056]

[0057] where x is the original cumulative slope displacement data; x ′ is the normalized cumulative slope displacement data; max(x) is the maximum value in the sample; min(x) is the minimum value in the sample.

[0058] S2. Construct a VMD-LSTM mountain highway slope displacement prediction model, and use the training set and the test set to pre-train the prediction model.

[0059] As a preferred implementation, in this embodiment, first construct a VMD-LSTM model.

[0060] Among them, the VMD model is used to decompose the normalized cumulative slope displacement data to obtain a series of sub-modal components {u 1 , u 2 , …, u kWhen decomposing the cumulative displacement data of the slope, the number of decomposed sub-modal components k is determined by observing the stability of the central frequency. When the central frequency of the last layer is relatively stable, k is selected as the current value.

[0061] Then, each sub-modal component obtained by VMD decomposition is used as the input of the LSTM model, and single-step prediction is performed to obtain the predicted values of each sub-modal component, that is, Y 1 ,Y 2 ,…,Y k 。

[0062] Finally, the predicted values of each item are summarized and superimposed to obtain the final predicted value of the displacement of the mountain highway slope.

[0063] For the convenience of those skilled in the art to understand and master, the VMD model and the LSTM model are further described below.

[0064] Variational Mode Decomposition (VMD) is a new signal processing method used to decompose complex non-stationary signals into several modal components with different central frequencies. Compared with the traditional Empirical Mode Decomposition (EMD), VMD has better stability and noise resistance, and can extract the multi-scale features of signals more accurately. The goal of VMD is to decompose the input signal x(t) into K modal components u k (t), and each modal component has a central frequency ω k 。VMD realizes signal decomposition by solving the following optimization problem:

[0065]

[0066] Among them: represents the Hilbert transform, represents the time derivative, ∥·∥ 2 represents L 2 norm.

[0067] To ensure the uniqueness and stability of the decomposition, VMD also introduces a constraint condition, that is, the sum of all modal components should be equal to the original signal:

[0068]

[0069] The algorithm steps of VMD are as follows:

[0070] ①Initialization: Set the number of modal components K and the initial central frequency ω k ;

[0071] ②Alternating Direction Method of Multipliers (ADMM): Solve the above optimization problem by the Alternating Direction Method of Multipliers (ADMM) to update the modal component uk (t) and the central frequency ω k ;

[0072] ③ Convergence judgment: When the convergence condition is met (for example, the change in the modal component is less than a certain threshold), stop the iteration;

[0073] ④ Output result: Output the decomposed modal component u k (t) and the central frequency ω k .

[0074] Long Short-Term Memory (LSTM) is a special type of Recurrent Neural Network (RNN) specifically designed to handle long-term dependencies in sequential data. By introducing a gating mechanism, LSTM effectively solves the problems of vanishing gradients and exploding gradients in traditional RNNs and can maintain information transmission in long time series. In the present invention, LSTM is used to model slope displacement data to capture its non-linear and time-varying characteristics.

[0075] The basic unit of LSTM contains three gating mechanisms: Input Gate, Forget Gate, and Output Gate, and the unit structure is as Figure 3 shown. These gating mechanisms control the inflow, retention, and outflow of information, thereby achieving effective management of long-term and short-term information.

[0076] Input Gate: Determines which information will be added to the cell state;

[0077] i t = σ(W i [h t-1 , x t +b i )

[0078] Forget Gate: Determines which information will be deleted from the cell state;

[0079] f t = σ(W f [h t-1 , x t +b f )

[0080] Cell state update: Calculate the new candidate cell state;

[0081]

[0082] Cell state: Update the cell state;

[0083]

[0084] Output gate: determines which information will be output;

[0085] o t = σ(W o [h t-1 , x t +b o )

[0086] Hidden state: calculates the hidden state at the current moment;

[0087] h t = o t ⊙ tanh(c t )

[0088] Where: σ represents the sigmoid activation function, tanh represents the hyperbolic tangent activation function, and ⊙ represents element-wise multiplication.

[0089] As a preferred implementation, in this embodiment, the prediction model is pre-trained using the training set and the test set in the following manner:

[0090] First, establish an evaluation function, including the mean absolute error MAE, the root mean square error RMSE, and the mean absolute percentage error MAPE, and their expressions are respectively:

[0091]

[0092] Where N is the number of test set samples; and Y i are respectively the predicted displacement value and the actual displacement value of the i-th sample in the test set;

[0093] Then, use the slope cumulative displacement data in the training set as the input of the VMD-LSTM model, obtain the corresponding mountain highway slope displacement prediction value, and combine the actual displacement value to evaluate the model performance using the above evaluation function to obtain the initial evaluation result.

[0094] S3. Use the dung beetle optimization algorithm to optimize the hyperparameters in the prediction model to improve the model prediction performance.

[0095] As a preferred implementation, in this embodiment, the hyperparameters in the prediction model are optimized in the following manner:

[0096] Initialize the algorithm parameters and define the search space.

[0097] Among them, the algorithm parameters include the population size and the number of iterations. The population size determines the number of "dung beetles" participating in the optimization process. This number needs to be large enough to ensure sufficient exploration of the solution space, but also take into account the limitations of computing resources. Set the maximum number of loops for the algorithm to run to ensure that the algorithm can converge to a better solution within a limited time.

[0098] Define the search space, including setting reasonable value ranges for the number of decomposition modes k, the balance parameter α in the VMD model, and the hyperparameters of the LSTM model. Among them, the hyperparameters at least include the number of hidden layers, the number of neurons, the learning rate, and the batch size.

[0099] Encode all the hyperparameters to be optimized to form a multi-dimensional vector representing the position of each dung beetle. For example, a possible position vector can be {K, alpha, neurons, learning_rate, batch_size}. Randomly generate the initial population, that is, randomly initialize multiple combinations of hyperparameters as starting points.

[0100] Based on the hyperparameter configuration represented by each individual in the initial population, train the VMD-LSTM model, and use the test set to perform fitness evaluation based on the evaluation function;

[0101] Simulate the behavior of dung beetles rolling dung balls, consider the balance between local search (following the current optimal solution) and global search (exploring new potential solutions), imitate the behavior of the best individuals in the group, and at the same time retain a certain probability of random exploration ability to update the position of each dung beetle.

[0102] Specifically, at initialization, define a position vector x i (t) for each dung beetle, where i = 1, 2,..., N represents the i-th dung beetle, N is the population size, and t represents the current number of iterations; each position vector corresponds to a set of hyperparameter configurations, and initialize a velocity vector v i (t) to guide the position update; let x best (t) be the best position found in the group so far in the current iteration;

[0103] In each iteration process, update the velocity and position of each dung beetle according to the following rules:

[0104] Update the velocity v i (t + 1) = w·v i (t) + c 1 r 1 (x best (t) - x i (t)) + c 2 r 2 (x pbest,i - x i(t)), where w is the inertia weight, c 1 and c 2 are the cognitive and social factors respectively, r 1 and r 2 are two random numbers independently distributed between [0, 1], x p best,i is the best position found by the i-th dung beetle;

[0105] Update the position: x i (t + 1) = x i (t) + v i (t + 1);

[0106] The balance between local search and global search in the dung beetle optimization algorithm is achieved by adjusting the inertia weight w. Usually, w gradually decreases from a larger value to ensure a larger global search space in the initial stage and enhance the local search ability in the later stage;

[0107] Meanwhile, to maintain a certain random exploration ability, a small probability P rand is set, so that the dung beetle has a certain chance to ignore the above update rule and instead randomly select a new position in the search space as the position for the next iteration, that is, when the randomly generated probability is less than P rand , let x i (t + 1) be a random position in the search space;

[0108] After each iteration, calculate the fitness value according to the evaluation function and check whether the preset stop criterion is met, for example, reaching the maximum number of iterations or the best fitness not changing significantly in several consecutive iterations. If the stop condition is met, select the set of hyperparameter configurations with the lowest fitness value from all historical records as the final optimization result; otherwise, continue the next round of iteration.

[0109] S4. Conduct the displacement prediction of the mountain highway slope based on the optimized prediction model.

[0110] The following uses a specific example to illustrate the progress of the mountain highway slope displacement prediction method of the present invention compared with the prior art.

[0111] Select a certain slope of the national highway in Xi'an, Shaanxi Province as the experimental prediction object of the example of the present invention, and use the surface displacement sensor to monitor the slope displacement. In this example, the data measured by the surface displacement sensor at the monitoring point G108 from November 28, 2023 to October 15, 2024 for a total of 324 days are selected as the sample set for verification. The surface displacement sensor can collect the slope displacement data in the X, Y, and Z directions, and the Euclidean norm is used to fuse the data in the (x, y, z) directions of each monitoring point to obtain the cumulative slope displacement. Part of the monitoring data is shown in Table 1.

[0112] Table 1

[0113]

[0114]

[0115] To eliminate the problem of increased training time caused by singular samples, the min-max normalization method is used to preprocess the data, that is

[0116]

[0117] where x is the original data; x ′ is the normalized data; max(x) is the maximum value in the sample; min(x) is the minimum value in the sample. The processed slope displacement is as Figure 4 shown

[0118] Next, the VMD model is applied to decompose the slope displacement time series to obtain several sub-modal components with clear physical meanings. Appropriate values of k (the number of decomposition modes) and α (the balance parameter) are selected to ensure that each decomposed modal component can effectively represent different frequency components of the original signal. The decomposition parameter settings in this example are shown in Table 2

[0119] Table 2

[0120] Parameter Description Set Value K Number of Decomposition Modes 5 α Balance Parameter 2000 τ Noise Tolerance 0 DC Whether to Include DC Component False init Initial Center Frequency 'random' tol Convergence Threshold 1e-7

[0121] All algorithms and models in this example are compiled based on the Python 3.7 and PyTorch 1.6.0 deep learning frameworks in the Windows 10 environment. The computer configuration is: the CPU processor is Intel(R) Core(TM) i7-10700 CPU @ 2.90GHz, the graphics card configuration is NVDIA GeForce RTX 3080, the video memory is 16GB, and the memory is 32GB

[0122] A total of 324 pieces of data are collected in this example. The modal components of the first 270 pieces of data are selected as training samples, and the last 54 pieces of data are used as test samples for comparative analysis. The input dimension is set to 5. The initial hyperparameters of the LSTM model are set as follows: the hidden layer is set to 2, the number of neurons is 50, the initial learning rate is 0.001, and the number of times is set to 1000 times. From Figure 5 the shown training loss function graph, it can be seen that after training for 600 times, the loss gradually approaches 0 and stabilizes near 0, indicating that the model has converged

[0123] To further improve the performance of the VMD-LSTM model, the dung beetle optimization algorithm is introduced to automatically tune the key hyperparameters. By adjusting parameters such as the K value and α value in the VMD model, and the number of hidden layer neurons and learning rate in the LSTM model, lower prediction errors and faster convergence speeds are achieved. The comparison of hyperparameters before and after optimization is shown in Table 3, and the comparison of the performance of the prediction models before and after optimization is shown in Table 4.

[0124] Table 3

[0125] Parameter Unoptimized After DBO Optimization VMD-K 5 7 VMD-α 2000 2500 Number of Hidden Layers in LSTM 2 3 Number of Neurons in Each Layer of LSTM 50 64 Learning Rate of LSTM 0.001 0.0008 Batch Size of LSTM 32 64

[0126] Table 4

[0127] Metric Unoptimized Model Model after DBO Optimization MAE 0.021 0.017 RMSE 0.034 0.028 MAPE 1.52％ 1.23％

[0128] Furthermore, to verify the effectiveness and superiority of the VMD-LSTM model optimized by the dung beetle optimization algorithm in this example, it is compared with other traditional methods. Specifically, ARIMA, BP neural network, and the unoptimized VMD-LSTM model are selected as the control groups, and performance evaluations are carried out from two aspects: short-term and long-term predictions. The performance evaluation results are shown in Table 5.

[0129] Table 5

[0130] Model MAE RMSE MAPE ARIMA 0.042 0.067 3.14％ BP Neural Network 0.038 0.059 2.87％ VMD-LSTM (Unoptimized) 0.021 0.034 1.52％ VMD-LSTM (Optimized) 0.017 0.028 1.23％

[0131] As can be seen from Table 5, compared with the traditional ARIMA and BP neural networks, even the unoptimized VMD-LSTM model shows better prediction accuracy. And the VMD-LSTM model with hyperparameters tuned by introducing the dung beetle optimization algorithm further improves the prediction effect, with significant improvements in all three evaluation indicators of MAE, RMSE, and MAPE.

[0132] Another embodiment of the present invention also provides a computer device, including a memory and a processor. Computer instructions are stored in the memory, and the processor executes the computer instructions to execute the mountain highway slope displacement prediction method disclosed in the foregoing embodiments.

[0133] Another embodiment of the present invention also provides a computer-readable storage medium. The storage medium stores computer instructions, and when the computer instructions are run by the computer, the mountain highway slope displacement prediction method disclosed in the foregoing embodiments is executed.

[0134] It should be noted that the method according to the embodiments of the present invention can be executed by a single device, such as a computer or a server. The method of this embodiment can also be applied to a distributed scenario and completed by the cooperation of multiple devices. In such a distributed scenario, one of the multiple devices can only execute one or more steps of the method according to the embodiments of the present invention, and these multiple devices will interact with each other to complete the described method.

[0135] It should be noted that some embodiments of the present invention have been described above. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be executed in a different order from that in the above embodiments and still achieve the desired result. Additionally, the processes depicted in the drawings do not necessarily require the specific order or sequential order shown to achieve the desired result. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0136] The embodiments of the present invention are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the embodiments of the present invention shall be included within the protection scope of the present invention.

Claims

1. A VMD-LSTM mountain highway slope displacement prediction method based on dung beetle optimization algorithm, characterized in that: The steps include: Collect the daily displacement data of the slope to be tested, perform data preprocessing and normalization on the collected displacement data, and divide them into training set and test set; Constructing a VMD-LSTM mountain highway slope displacement prediction model, and pre-training the prediction model using the training set and the test set; Using a dung beetle optimization algorithm to optimize the hyperparameters in the prediction model; The slope displacement prediction of mountain highway is carried out based on the optimized prediction model.

2. The VMD-LSTM mountain highway slope displacement prediction method based on the dung beetle optimization algorithm as claimed in claim 1, characterized in that: Collect the daily displacement data of the slope to be measured, and perform data preprocessing and normalization on the collected displacement data, including: The slope displacement data in the (x, y, z) direction at the preset monitoring point of the slope to be measured is collected based on the surface displacement sensor; Using Euclidean norm The slope displacement data in the (x, y, z) direction of each monitoring point are fused to obtain the accumulated slope displacement; The maximum and minimum normalization method is used to process the slope cumulative displacement data, that is, Where x is the original slope cumulative displacement data; x′ is the normalized slope cumulative displacement data; max(x) is the maximum value in the sample; min(x) is the minimum value in the sample.

3. The VMD-LSTM mountain highway slope displacement prediction method based on the dung beetle optimization algorithm as claimed in claim 2 is characterized in that: The construction of VMD-LSTM mountain highway slope displacement prediction model includes: A VMD-LSTM model is constructed, in which the VMD model is used to decompose the normalized slope cumulative displacement data to obtain a series of sub-modal components {u1,u2,…,u k } and used as the input of the LSTM model, which outputs the predicted values ​​of each sub-modal component Y1, Y2, …, Y k ; The various predicted values ​​are summarized and superimposed to obtain the final predicted value of slope displacement of mountain highway.

4. The VMD-LSTM mountain highway slope displacement prediction method based on the dung beetle optimization algorithm as claimed in claim 3 is characterized in that: Pre-training the prediction model using the training set and the test set includes: An evaluation function is established, which includes mean absolute error MAE, root mean square error RMSE and mean absolute percentage error MAPE, and their expressions are: Where N is the number of test set samples; and Y i are the predicted displacement value and actual displacement value of the i-th sample in the test set, respectively; The accumulated displacement data of the slope in the training set is used as the input of the VMD-LSTM model to obtain the corresponding predicted displacement value of the slope of the mountain highway, and the evaluation function is used to evaluate the model performance in combination with the actual displacement value to obtain an initial evaluation result.

5. The VMD-LSTM mountain highway slope displacement prediction method based on dung beetle optimization algorithm as claimed in claim 4 is characterized in that: Using the dung beetle optimization algorithm to optimize the hyperparameters in the prediction model includes: Initializing algorithm parameters and defining a search space, wherein the algorithm parameters include a population size and a number of iterations, wherein defining the search space includes setting a reasonable value range for the number of decomposition modes k, the balance parameter α, and the hyperparameters of the LSTM model in the VMD model, wherein the hyperparameters include at least the number of hidden layers, the number of neurons, the learning rate, and the batch size; Encode all the hyperparameters to be optimized to form a multi-dimensional vector to represent the position of each dung beetle, and randomly initialize multiple hyperparameter combinations as the initial population; Based on the hyperparameter configuration represented by each individual in the initial population, the VMD-LSTM model is trained, and fitness evaluation is performed based on the evaluation function using a test set; Simulate the behavior of dung beetles rolling dung balls, consider the balance between local search and global search, and update the position of each dung beetle; After each iteration, check whether the preset stopping criteria are met, including reaching the maximum number of iterations or the best fitness has not changed significantly in several consecutive iterations; When the preset stopping criterion is met, a set of hyperparameter configurations with the lowest fitness value is selected from all historical records as the final optimization result.

6. The VMD-LSTM mountain highway slope displacement prediction method based on dung beetle optimization algorithm as claimed in claim 5, characterized in that: Simulating the behavior of dung beetles rolling dung balls, considering the balance between local search and global search, updating the position of each dung beetle includes: During initialization, a position vector x is defined for each dung beetle i (t), where i = 1, 2, ..., N represents the i-th dung beetle, N is the population size, and t represents the current iteration number; each position vector corresponds to a set of hyperparameter configurations, and a velocity vector v is initialized i (t) is used to guide the position update; let x best (t) is the best position found in the population up to the current iteration; During each iteration, the velocity and position of each dung beetle are updated according to the following rules: Update speed i (t+1)=w·v i (t)+c1r1(x best (t)-x i (t))+c2r2(x pbest,i -x i (t)), where w is the inertia weight, c1 and c2 are cognitive and social factors, r1 and r2 are two random numbers independently distributed between [0,1], and x p best,i is the best position found by the i-th dung beetle; Update position: x i (t+1)=x i (t)+v i (t+1); The balance between local search and global search in the dung beetle optimization algorithm is achieved by adjusting the inertia weight w, which gradually decreases from a larger value; At the same time, set a small probability P rand , when the probability of random generation is less than P rand When x i (t+1) is a random position in the search space; After each iteration, the fitness value is calculated according to the evaluation function, and it is checked whether the preset stopping criteria are met. If the stopping criteria are met, a set of hyperparameter configurations with the lowest fitness value is selected from all historical records as the final optimization result; otherwise, continue to the next round of iteration.

7. A computer device, characterized in that: The method comprises a memory and a processor, wherein the memory stores computer instructions, and the processor executes the method for predicting the displacement of a mountain highway slope as claimed in any one of claims 1 to 6 by executing the computer instructions.

8. A computer-readable storage medium, characterized in that: The storage medium stores computer instructions, and when the computer instructions are executed by a computer, the method for predicting the displacement of a mountain highway slope as claimed in any one of claims 1 to 6 is executed.

Images