Enhanced slope stability classification method and system based on LSTM and Bayesian optimization

Abstract

The invention relates to an enhanced slope stability classification method and system based on LSTM and Bayesian optimization, and the method comprises the following steps: collecting original data of a plurality of slopes, correspondingly extracting feature data, and adding a stability label; preprocessing the extracted feature data, and constructing a slope stability training sample set; constructing a slope stability classification model based on the LSTM network, inputting a slope stability training sample set to train the slope stability classification model, and performing iterative tuning on hyper-parameters of the LSTM network by adopting a Bayesian optimization technology to obtain a trained slope stability classification model; and monitoring the stability of the target slope through the trained slope stability classification model.

Description

TECHNICAL FIELD

[0001] The application relates to an enhanced slope stability classification method and system based on LSTM and Bayesian optimization, and belongs to the technical field of geological disaster monitoring. BACKGROUND

[0002] A slope area is a key area of interaction between geological disasters and human activities, and is also a sensitive and fragile part of the geological environment. With the intensification of climate change and human engineering activities, landslide disasters caused by slope instability have become one of the main geological disasters faced by global society, posing a serious threat to infrastructure safety and human life and restricting regional economic development. Therefore, it is crucial to build a high-precision slope stability classification model to identify and quantify the key driving factors leading to instability disasters for disaster prevention and mitigation. Although a variety of methods have been applied to slope stability analysis and prediction, the existing slope stability evaluation models still have uncertainties, especially in capturing the non-linear relationship between driving factors and instability and determining the weight contribution of driving factors. Although the traditional limit equilibrium method provides a certain theoretical basis, its calculation method based on simplified assumptions often cannot accurately reflect complex geological conditions and dynamic changes.

[0003] Machine learning models, as data-driven techniques, can more effectively perform non-linear modeling and reveal the complex dependence between disaster occurrence and hazard factors compared to traditional limit equilibrium methods. Machine learning models have shown significant potential in slope stability prediction and risk management due to their ability to handle large-scale data and mine hidden patterns. However, traditional machine learning methods such as random forests and support vector machines have the "black box" characteristic, i.e., the opacity of the internal decision-making process of the model, making it difficult to explain the prediction results and limiting their application in actual engineering decisions. In addition, the trial-and-error process of hyperparameter adjustment is often time-consuming and inefficient, making it difficult to achieve optimal model performance. SUMMARY

[0004] To solve the problems existing in the prior art, the application provides an enhanced slope stability classification method and system based on LSTM and Bayesian optimization, which combines long short-term memory networks and Bayesian optimization deep learning methods to build an enhanced slope stability classification model, uses Bayesian optimization to improve model performance, and enhances prediction ability through multi-model fusion. By introducing explainability analysis and real-time monitoring technology, the limitations of traditional methods are solved, accurate evaluation and dynamic management of slope instability are achieved, and scientific basis is provided for infrastructure safety.

[0005] The technical solutions of the application are as follows:

[0006] In one aspect, the application provides an enhanced slope stability classification method based on LSTM and Bayesian optimization, comprising the following steps:

[0007] Collecting original data of a plurality of slopes, corresponding to extracting feature data and adding stability labels;

[0008] Pretreating the extracted feature data to construct a slope stability training sample set;

[0009] Constructing a slope stability classification model based on an LSTM network, inputting the slope stability training sample set to train the slope stability classification model, and simultaneously using Bayesian optimization technology to iteratively optimize the hyperparameters of the LSTM network to obtain a trained slope stability classification model;

[0010] Monitoring the stability of a target slope by using the trained slope stability classification model.

[0011] As a preferred embodiment, the feature data includes:

[0012] Rock-soil unit weight, rock-soil cohesion, rock-soil internal friction angle, rock-soil pore water pressure ratio, slope height and slope angle;

[0013] The step of pretreating the extracted feature data includes standardizing each feature data.

[0014] As a preferred embodiment, the step of using Bayesian optimization technology to iteratively optimize the hyperparameters of the LSTM network is specifically:

[0015] Randomly initializing a plurality of groups of hyperparameters of the LSTM network as a hyperparameter space, selecting a plurality of sampling points in the hyperparameter space, inputting the slope stability training sample set to train the LSTM network, and recording the model performance corresponding to each sampling point;

[0016] Through each sampling point and the corresponding model performance, the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point is calculated by a Gaussian process;

[0017] According to the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point, the next sampling point is selected by an expected improvement acquisition function, and the LSTM network is trained by using the sampling point to record the corresponding model performance;

[0018] Iteratively execute the Gaussian process and select the next sampling point by the expected improvement acquisition function until the iteration end condition is reached;

[0019] Outputting the sampling point corresponding to the optimal model performance as the combination of hyperparameters of the LSTM network.

[0020] As a preferred embodiment, the step of monitoring the stability of the target slope by using the trained slope stability classification model comprises:

[0021] The original data of the target slope in a period of time is obtained, the feature data of each time step is extracted and preprocessed, and then input into the trained slope stability classification model to obtain the slope instability probability of the target slope at the corresponding time step;

[0022] A time window is selected, and the dynamic risk score of the target slope is calculated by using the sliding window method, which includes:

[0023] In the selected time window, the slope instability probability of the target slope output by the slope stability classification model at each time step is obtained, and the average value is calculated as the slope instability risk value of the target slope in the time window;

[0024] When the slope instability risk value exceeds the preset threshold, a warning is given.

[0025] On the other hand, the present application also proposes an enhanced slope stability classification system based on LSTM and Bayesian optimization, which comprises:

[0026] A data acquisition module is used to acquire the original data of a plurality of slopes, extract the feature data, and add stability labels;

[0027] A data set construction module is used to preprocess the extracted feature data and construct a slope stability training sample set;

[0028] A model training module is used to construct a slope stability classification model based on an LSTM network, input the slope stability training sample set to train the slope stability classification model, and simultaneously use the Bayesian optimization technology to iteratively optimize the hyperparameters of the LSTM network to obtain a trained slope stability classification model;

[0029] A slope monitoring module is used to monitor the stability of the target slope by using the trained slope stability classification model.

[0030] As a preferred embodiment, the feature data includes:

[0031] The unit weight of rock and soil, the cohesion of rock and soil, the internal friction angle of rock and soil, the pore water pressure ratio of rock and soil, the slope height and the slope angle;

[0032] The step of preprocessing the extracted feature data includes standardizing each feature data.

[0033] As a preferred embodiment, the step of using the Bayesian optimization technology to iteratively optimize the hyperparameters of the LSTM network specifically comprises:

[0034] Randomly initialize a plurality of sets of hyperparameters of a plurality of sets of LSTM networks as a hyperparameter space, select a plurality of sampling points in the hyperparameter space, input a slope stability training sample set to train the LSTM network, and record the model performance corresponding to each sampling point;

[0035] Through each sampling point and the corresponding model performance, the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point is calculated by a Gaussian process;

[0036] According to the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point, the next sampling point is selected by an expected improvement acquisition function, and the LSTM network is trained through the sampling point to record the corresponding model performance;

[0037] Iteratively execute the above Gaussian process and select the next sampling point by the expected improvement acquisition function until the iteration end condition is reached.

[0038] The sampling point corresponding to the optimal model performance is output as the hyperparameter combination of the LSTM network.

[0039] As a preferred embodiment, the step of monitoring the stability of the target slope by the trained slope stability classification model comprises:

[0040] Obtain the original data of the target slope within a period of time, extract the feature data of each time step and perform preprocessing, and input it into the trained slope stability classification model to obtain the slope instability probability of the target slope at the corresponding time step;

[0041] Select a time window, and calculate the dynamic risk score of the target slope by a sliding window method, comprising:

[0042] In the selected time window, obtain the slope instability probability of the target slope output by the slope stability classification model at each time step, and calculate the mean value as the slope instability risk value of the target slope within the time window;

[0043] When the slope instability risk value exceeds a preset threshold, a warning is given.

[0044] In another aspect, the present application also proposes an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to realize the enhanced slope stability classification method based on LSTM and Bayesian optimization as described in any embodiment of the present application.

[0045] In another aspect, the present application also proposes a computer readable storage medium having a computer program stored thereon, wherein the program is executed by a processor to realize the enhanced slope stability classification method based on LSTM and Bayesian optimization as described in any embodiment of the present application.

[0046] The beneficial effects of the present application are that:

[0047] 1. By combining the LSTM deep learning model and the Bayesian optimization technology, the accuracy of slope stability classification is significantly improved, the nonlinear relationship between geological parameters can be better captured, and the risk of misjudgment is reduced.

[0048] 2. The Gaussian process and the expected improvement acquisition function are used for Bayesian optimization, the calculation cost and time of manual parameter adjustment are reduced, the best hyperparameters are automatically searched, the model generalization ability is improved, and the method is suitable for large-scale real slope data.

[0049] Additional aspects and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the application. Furthermore, various aspects and advantages of the present application can be realized and attained by means of the instrumentalities and combinations particularly pointed out in the appended claims. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The method flowchart of the embodiment one of the present application. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] It should be understood that the step numbers used herein are only for the convenience of description, and are not limited to the execution sequence of the steps.

[0053] It should be understood that the terms used in the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the present application specification and the appended claims, unless otherwise clear from the context, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0054] The terms "include" and "contain" indicate the presence of the described features, whole, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, whole, steps, operations, elements, components and / or sets thereof.

[0055] The term "and / or" means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0056] Embodiment one:

[0057] Referring to Figure 1 , the embodiment proposes an enhanced slope stability classification method based on LSTM and Bayesian optimization, including the following steps:

[0058] S100, collecting original data of a plurality of slopes, corresponding to extracting feature data and adding stability labels (representing the instability probability of the slope);

[0059] S200, preprocessing the extracted feature data, and constructing a slope stability training sample set;

[0060] S300, constructing a slope stability classification model based on a long short-term memory network (LSTM) network, inputting the slope stability training sample set to train the slope stability classification model, and simultaneously using Bayesian optimization technology to iteratively optimize the hyperparameters of the LSTM network, to obtain a trained slope stability classification model;

[0061] S400, performing stability monitoring of the target slope by using the trained slope stability classification model.

[0062] As a preferred embodiment of the embodiment, the feature data includes:

[0063] Four geotechnical parameters (unit weight kN / m 3 , cohesion kPa, internal friction angle, and pore water pressure ratio) and two slope features (slope height m and slope angle degree), as well as binary stability labels (stable or unstable).

[0064] The step of preprocessing the extracted feature data includes preprocessing the unit weight (kN / m 3 ), cohesion (kPa), internal friction angle (degree), pore water pressure ratio, slope height (m), and slope angle (degree) and other features using a standardization method; the Min-Max standardization method is used to linearly transform the data to the [0, 1] interval, and the formula is:

[0065]

[0066] In the formula, X normalized is the standardized value, X is the original data value, X min is the minimum value of the feature, and X max is the maximum value of the feature; subsequently, the training set and the test set are proportionally stratified, to ensure balanced data distribution, eliminate the influence of dimensions, and eliminate differences in variable variation sizes.

[0067] As a preferred embodiment of the present embodiment, in step S300, the time sequence dependence of the slope feature data is processed by using an LSTM network structure, and a slope stability binary classification model (stable or unstable) is constructed; the LSTM unit includes a forgetting gate, an input gate, a candidate cell state, a cell state update, an output gate and a hidden state update, and the long-term dependence is captured through these gate mechanisms to avoid the problem of gradient disappearance or explosion.

[0068] The forgetting gate of the LSTM is used to decide to retain or discard the cell state information of the previous time, and its calculation formula is:

[0069] f t =σ(W f ·[h t-1 ,x t ]+b f );

[0070] In the formula, f t is the forgetting gate output vector, sigma is the sigmoid activation function, W f is the weight matrix of the forgetting gate, b f is the bias vector, h t-1 is the hidden state of the previous time, x t is the input vector of the current time, and [] represents vector splicing.

[0071] The input gate of the LSTM is used to decide which new information is stored in the cell state, and its calculation formula is:

[0072] i t =σ(W i ·[h t-1 ,x t ]+b i );

[0073] In the formula, i t is the input gate output vector, W i is the weight matrix of the input gate, and b i is the bias vector.

[0074] The candidate cell state is used to create a new candidate value, and its calculation formula is:

[0075] In the formula, is the candidate cell state, tanh is the hyperbolic tangent activation function, W C is the weight matrix of the candidate cell state, and b C is the bias vector.

[0076] The cell state update combines the forgetting gate and the input gate, and its calculation formula is:

[0077]

[0078] where C t is the current cell state, C t-1 is the previous cell state, and ⊙ denotes element-wise multiplication.

[0079] The output gate of LSTM is used to determine which information to output, and its calculation formula is:

[0080] h t = o t ⊙ tanh(C t );

[0081] where htis the current hidden state, which is used to pass to the next time or as the model output.

[0082] In step S300, the step of iteratively optimizing the hyperparameters of the LSTM network using the Bayesian optimization technique is specifically:

[0083] Randomly initialize several groups of hyperparameters of LSTM network as hyperparameter space, including the number of neurons (16-128), dropout rate (0.1-0.3), learning rate (0.001-0.1), number of layers (1-8), and activation function (relu, tanh, sigmoid). Select several sampling points in the hyperparameter space, input the slope stability training sample set to train the LSTM network, and record the model performance corresponding to each sampling point; for example: select 5 sampling points, each sampling point is a vector of a group of hyperparameter combinations, such as: x1: (16, 0.1, 0.001, 1,'relu'); through each sampling point x i Train the LSTM network to obtain the corresponding model f(x i ).

[0084] where the five-fold stratified cross-validation is used to verify the optimized LSTM model, and based on the test data set, the accuracy, precision, recall, specificity, F1 score and AUC are used to evaluate the model performance, and the calculation formula of each index is:

[0085]

[0086] In the formula, TP represents a true case (correctly predicted as unstable); FP represents a false positive (incorrectly predicted as unstable); FN represents a false negative (incorrectly predicted as stable); and TN represents a true negative (correctly predicted as stable). The accuracy, precision, recall, specificity, and F1-score have a value range of [0, 1]. The accuracy is used to measure the percentage of the number of correct predictions in the total number of predictions. The closer the accuracy of the model is to 1, the higher the overall precision of the model. The closer the precision of the model is to 1, the smaller the probability of misjudgment of the model in predicting unstable results. The closer the recall of the model is to 1, the stronger the prediction ability of the model for unstable slopes. The closer the specificity is to 1, the stronger the prediction ability of the model for stable slopes. The F1-score combines precision and recall, and the larger the value, the better the prediction performance of the model.

[0087] Based on the test data set, the model performance is evaluated using the receiver operating characteristic curve (ROC). The receiver operating characteristic curve is a tool commonly used to evaluate the performance of a binary classification model. It is a curve plotted with the true positive rate (recall) as the vertical coordinate and the false positive rate (1-specificity) as the horizontal coordinate. The area under the ROC curve is the AUC value, and the larger the AUC value, the better the model performance.

[0088] Through each sampling point x i and the corresponding model performance f(x i ), the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point is calculated by the Gaussian process, and the specific formula is:

[0089]

[0090] where μ(x) is the posterior mean, k(x, X) is the covariance vector of the new sampling point x and the hyperparameter space X, K is the covariance matrix between sampling points, is the noise variance, I is the identity matrix, and y is the model performance value vector.

[0091] According to the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point, the next sampling point is selected by the expected improvement acquisition function, and the LSTM network is trained through the sampling point to record the corresponding model performance. Specifically, the calculation formula of the expected improvement acquisition function is:

[0092] EI(x)=(μ(x)-f(x + ))Φ(Z)+σ(x)φ(Z);

[0093] EI(x) is the expected improvement value, f(x +) is the model performance value corresponding to the current best sampling point, Φ(Z) is the standard normal cumulative distribution function, and φ(Z) is the standard normal probability density function,

[0094] The candidate sampling point with the largest expected improvement value is selected as the next sampling point, and the candidate sampling point is selected from the hyperparameter space.

[0095] The above Gaussian process and the next sampling point selection by the expected improvement acquisition function are iteratively performed until an iteration end condition is reached, and the iteration end condition can be the number of iterations, for example, 20 times.

[0096] The sampling point corresponding to the optimal model performance is output as the hyperparameter combination of the LSTM network.

[0097] As a preferred embodiment of the present embodiment, the step of monitoring the stability of the target slope by using the trained slope stability classification model comprises:

[0098] Obtain the original data of the target slope in a period of time, extract the feature data of each time step and pre-process, and input into the trained slope stability classification model to obtain the slope instability probability of the corresponding time step;

[0099] Select a time window, and calculate the dynamic risk score of the target slope by the sliding window method, including:

[0100] In the selected time window, obtain the slope instability probability of the target slope output by the slope stability classification model at each time step, calculate the mean value as the slope instability risk value of the target slope in the time window, and the specific formula is:

[0101]

[0102] Where, R t is the dynamic risk score at time t, w is the size of the sliding window, P(unstable|X i ) is the slope instability probability output by the model at the i-th time step;

[0103] When R t exceeds a preset threshold (for example, 0.7), an alarm is triggered.

[0104] Embodiment two:

[0105] The present embodiment proposes an enhanced slope stability classification system based on LSTM and Bayesian optimization, comprising:

[0106] The data acquisition module is configured to acquire original data of a plurality of slopes, extract feature data, and add stability labels, and is configured to implement the function of step S100 in Embodiment 1 and will not be described herein again.

[0107] The data set construction module is configured to pre-process the extracted feature data and construct a slope stability training sample set, and is configured to implement the function of step S200 in Embodiment 1 and will not be described herein again.

[0108] The model training module is configured to construct a slope stability classification model based on an LSTM network, input the slope stability training sample set to train the slope stability classification model, and use a Bayesian optimization technique to iteratively optimize hyperparameters of the LSTM network to obtain a trained slope stability classification model, and is configured to implement the function of step S300 in Embodiment 1 and will not be described herein again.

[0109] The slope monitoring module is configured to monitor the stability of a target slope by using the trained slope stability classification model, and is configured to implement the function of step S400 in Embodiment 1 and will not be described herein again.

[0110] As a preferred embodiment of the present embodiment, the feature data includes:

[0111] unit weight of rock and soil, cohesion of rock and soil, internal friction angle of rock and soil, pore water pressure ratio of rock and soil, slope height, and slope angle.

[0112] The step of pre-processing the extracted feature data includes standardizing each feature data.

[0113] As a preferred embodiment of the present embodiment, the step of using the Bayesian optimization technique to iteratively optimize the hyperparameters of the LSTM network includes:

[0114] randomly initializing a plurality of groups of hyperparameters of the LSTM network as a hyperparameter space, selecting a plurality of sampling points in the hyperparameter space, inputting the slope stability training sample set to train the LSTM network, and recording the model performance corresponding to each sampling point.

[0115] calculating, by using a Gaussian process, posterior probability distribution information of the hyperparameter space corresponding to each sampling point according to each sampling point and the corresponding model performance;

[0116] selecting a next sampling point by using an expected improvement acquisition function according to the posterior probability distribution information of the hyperparameter space corresponding to the current sampling point, training the LSTM network by using the sampling point, and recording the corresponding model performance;

[0117] The above Gaussian process is iteratively executed, and the next sampling point is selected by improving the acquisition function through expectation until an iteration end condition is reached.

[0118] The sampling point corresponding to the optimal model performance is output as the hyperparameter combination of the LSTM network.

[0119] As a preferred embodiment of the present embodiment, the step of monitoring the stability of the target slope by using the trained slope stability classification model comprises:

[0120] The original data of the target slope in a period of time is obtained, the feature data of each time step is extracted and preprocessed, and then input into the trained slope stability classification model to obtain the slope instability probability of the target slope at the corresponding time step;

[0121] A time window is selected, and the dynamic risk score of the target slope is calculated by using the sliding window method, including:

[0122] In the selected time window, the slope instability probability of the target slope output by the slope stability classification model at each time step is obtained, and the average value is calculated as the slope instability risk value of the target slope in the time window;

[0123] When the slope instability risk value exceeds a preset threshold, a warning is given.

[0124] Embodiment three:

[0125] The present embodiment provides an electronic device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the method according to any one of the embodiments of the present application when executing the program.

[0126] Embodiment four:

[0127] The present embodiment provides a computer readable storage medium, which stores a computer program executable by a processor to implement the method according to any one of the embodiments of the present application.

[0128] In the embodiments of the present application, "at least one" means one or more, and "multiple" means two or more. The "and / or" describes the association relationship between the associated objects, which means that there can be three kinds of relationships, for example, A and / or B, which means that A exists alone, A and B exist together, and B exists alone. Wherein A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects before and after it. "At least one of the following" and the like means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b and c can mean: a, b, c, a and b, a and c, b and c, or a and b and c, wherein a, b, and c can be single or multiple.

[0129] Those skilled in the art can clearly understand that the units and algorithm steps described in the embodiments disclosed herein can be realized by electronic hardware, computer software and combination of the electronic hardware and the computer software. Whether the functions are realized in the hardware or the software mode depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but the realization should not be considered beyond the scope of the present application.

[0130] Those skilled in the art can clearly understand that, for the convenience and brevity of the description, the specific working processes of the above-described system, device and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0131] In several embodiments provided in the present application, any function realized in the form of a software function unit and sold or used as an independent product can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts of the technical solutions that make contributions to the prior art or the parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the embodiments of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0132] The above only describes the embodiments of the present application, and does not limit the patent scope of the present application. Any equivalent structure or equivalent process transformation based on the content of the present application specification and drawings, or direct or indirect application in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An enhanced slope stability classification method based on LSTM and Bayesian optimization, characterized in that, Includes the following steps: Collect raw data of several slopes, extract corresponding feature data and add stability labels; The extracted feature data are preprocessed to construct a slope stability training sample set; A slope stability classification model based on an LSTM network is constructed. The slope stability training sample set is input to train the slope stability classification model. At the same time, Bayesian optimization technology is used to iteratively tune the hyperparameters of the LSTM network to obtain a well-trained slope stability classification model. The stability of the target slope is monitored by using a trained slope stability classification model.

2. The enhanced slope stability classification method based on LSTM and Bayesian optimization according to claim 1, characterized in that, The feature data includes: Unit weight of soil and rock, cohesion of soil and rock, internal friction angle of soil and rock, pore water pressure ratio of soil and rock, slope height and slope angle; The step of preprocessing the extracted feature data includes standardizing each feature data.

3. The enhanced slope stability classification method based on LSTM and Bayesian optimization according to claim 1, characterized in that, The specific steps for iteratively tuning the hyperparameters of the LSTM network using Bayesian optimization techniques are as follows: Several sets of hyperparameters of LSTM networks are randomly initialized as hyperparameter space. Several sampling points in the hyperparameter space are selected and input into the slope stability training sample set to train the LSTM network. The model performance corresponding to each sampling point is recorded. Based on each sampling point and the corresponding model performance, the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point is calculated using a Gaussian process. Based on the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point, the next sampling point is selected by improving the sampling function as expected, and the LSTM network is trained through the sampling point, and the corresponding model performance is recorded. The above Gaussian process is executed iteratively, and the next sampling point is selected by the expected improved acquisition function, until the iteration termination condition is met. The sampling points corresponding to the optimal output model performance are used as the hyperparameter combination of the LSTM network.

4. The enhanced slope stability classification method based on LSTM and Bayesian optimization according to claim 1, characterized in that, The steps for monitoring the stability of a target slope using a trained slope stability classification model include: Obtain the raw data of the target slope over a period of time, extract the feature data of each time step and preprocess it, and input it into the trained slope stability classification model to obtain the slope instability probability of the corresponding time step. Select a time window and calculate the dynamic risk score of the target slope using a sliding window method, including: Within the selected time window, obtain the slope instability probability of the target slope output by the slope stability classification model at each time step, and calculate the mean as the slope instability risk value of the target slope within that time window. An early warning will be issued when the slope instability risk value exceeds a preset threshold.

5. An enhanced slope stability classification system based on LSTM and Bayesian optimization, characterized in that, include: The data acquisition module is used to collect raw data from several slopes, extract corresponding feature data, and add stability labels. The dataset construction module is used to preprocess the extracted feature data and construct a training sample set for slope stability. The model training module is used to build a slope stability classification model based on an LSTM network. The slope stability training sample set is input to train the slope stability classification model. At the same time, Bayesian optimization technology is used to iteratively tune the hyperparameters of the LSTM network to obtain a trained slope stability classification model. The slope monitoring module is used to monitor the stability of a target slope using a trained slope stability classification model.

6. The enhanced slope stability classification system based on LSTM and Bayesian optimization according to claim 5, characterized in that, The feature data includes: Unit weight of soil and rock, cohesion of soil and rock, internal friction angle of soil and rock, pore water pressure ratio of soil and rock, slope height and slope angle; The step of preprocessing the extracted feature data includes standardizing each feature data.

7. The enhanced slope stability classification system based on LSTM and Bayesian optimization according to claim 5, characterized in that, The specific steps for iteratively tuning the hyperparameters of the LSTM network using Bayesian optimization techniques are as follows: Several sets of hyperparameters of LSTM networks are randomly initialized as hyperparameter space. Several sampling points in the hyperparameter space are selected and input into the slope stability training sample set to train the LSTM network. The model performance corresponding to each sampling point is recorded. Based on each sampling point and the corresponding model performance, the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point is calculated using a Gaussian process. Based on the posterior probability distribution information of the hyperparameter space corresponding to the currently selected sampling point, the next sampling point is selected by improving the sampling function as expected, and the LSTM network is trained through the sampling point, and the corresponding model performance is recorded. The above Gaussian process is executed iteratively, and the next sampling point is selected by the expected improved acquisition function, until the iteration termination condition is met. The sampling points corresponding to the optimal output model performance are used as the hyperparameter combination of the LSTM network.

8. The enhanced slope stability classification system based on LSTM and Bayesian optimization according to claim 5, characterized in that, The steps for monitoring the stability of a target slope using a trained slope stability classification model include: Obtain the raw data of the target slope over a period of time, extract the feature data of each time step and preprocess it, and input it into the trained slope stability classification model to obtain the slope instability probability of the corresponding time step. Select a time window and calculate the dynamic risk score of the target slope using a sliding window method, including: Within the selected time window, obtain the slope instability probability of the target slope output by the slope stability classification model at each time step, and calculate the mean as the slope instability risk value of the target slope within that time window. An early warning will be issued when the slope instability risk value exceeds a preset threshold.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the enhanced slope stability classification method based on LSTM and Bayesian optimization as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the enhanced slope stability classification method based on LSTM and Bayesian optimization as described in any one of claims 1 to 4.

Images