Tailing dam risk assessment method and system based on LSTM safety coefficient prediction

Abstract

The invention discloses a tailing dam risk assessment method and system based on LSTM safety coefficient prediction, and relates to the field of rock and soil safety monitoring, and the method comprises the steps: constructing a historical dam body environment sequence and a safety coefficient sequence; derivative features are extracted through feature engineering, and a composite feature sequence including static features and dynamic time features is constructed in combination with normalized time and attenuation coefficients. Then, constructing an LSTM model, and performing model training by taking the historical dam body environment composite feature sequence as input data and taking the historical safety coefficient sequence as output data; and inputting the current dam body environment composite feature sequence into the trained model, predicting a safety coefficient sequence, calculating a rainfall cumulative effect according to rainfall intensity and rainfall duration, and performing comprehensive risk assessment on the tailing dam. According to the method, the representativeness and accuracy of safety coefficient prediction of the tailing dam are enhanced, risk judgment is performed by using the rainfall cumulative effect value, and the accuracy of risk assessment of the tailing dam is further improved.

Description

Technical Field

[0001] This application relates to the field of geotechnical safety monitoring, and in particular to a tailings dam risk assessment method and system based on LSTM safety factor prediction. Background Technology

[0002] Tailings dams are the core structures for storing tailings during mining operations. Their stability is directly related to the safety of life and property of miners, the safety of the ecological environment, and the continuity of mining production. The safety factor is the core quantitative indicator for assessing the stability of tailings dams, and accurate and real-time prediction of the safety factor is the key to tailings dam safety management.

[0003] Currently, tailings dam safety factor prediction mainly relies on two traditional methods, both of which have significant limitations: First, numerical simulation methods based on physical mechanisms (such as the finite element method and the limit equilibrium method) require precise geotechnical constitutive models and parameter settings, resulting in complex and time-consuming calculations that are difficult to meet real-time prediction needs. More importantly, these methods often employ static assumptions when dealing with time-varying loads such as rainfall, failing to quantify the impact of accumulated rainfall over time and dynamic changes in pore water pressure on the safety factor, thus limiting prediction accuracy. Second, statistical empirical formula methods, while computationally simple, suffer from low accuracy and poor generalization ability. Their formulas are often derived from data fitting under specific operating conditions, failing to capture the complex nonlinear interactions between multiple factors such as dam parameters and hydrological conditions. Furthermore, they can only output the safety factor at a single time point, unable to achieve sequential prediction of the safety factor's evolution over time. In some cases, machine learning methods are gradually being applied to this field, but existing applications mostly employ static neural networks or support vector machines, only enabling single-point prediction of the safety factor under a single operating condition, failing to fully utilize the temporal evolution characteristics of the safety factor.

[0004] In summary, existing tailings dam risk assessments suffer from poor representativeness and low accuracy in safety factor predictions, and fail to fully consider the dynamic impact of cumulative rainfall on dam stability, leading to inaccurate assessment results. Summary of the Invention

[0005] The purpose of this application is to provide a tailings dam risk assessment method and system based on LSTM safety factor prediction, which can enhance the representativeness and accuracy of tailings dam safety factor prediction and improve the accuracy of tailings dam risk assessment.

[0006] To achieve the above objectives, this application provides the following solution: Firstly, this application provides a tailings dam risk assessment method based on LSTM safety factor prediction. The method includes: acquiring historical dam environmental data and corresponding historical safety factor data for continuous time points of the target tailings dam, and obtaining historical dam environmental sequences and historical safety factor sequences by constructing time series data; the historical dam environmental data includes: dam height, dam slope ratio, cohesion, internal friction angle, permeability coefficient, reservoir water level, rainfall intensity, and rainfall duration; using feature engineering to extract derived features from samples in the historical dam environmental sequences one by one, obtaining historical derived features, normalized time, and attenuation coefficients of the samples; merging the historical derived features with the features in the historical dam environmental sequences as static features, and using the normalized time and attenuation coefficients as dynamic time features to construct a composite feature sequence of the historical dam environment; The historical features include: dam height-to-slope ratio, shear strength parameter, logarithm of permeability, rainfall-to-dam height ratio, reservoir water level-to-dam height ratio, and dam morphology. An LSTM model is constructed, using the historical dam environment composite feature sequence as input data and the historical safety factor sequence as output data. The LSTM model is trained with the minimum loss function as the optimization objective. The current dam environment composite feature sequence of the target tailings dam is obtained and input into the trained LSTM model to obtain the predicted safety factor sequence. The cumulative rainfall effect value is calculated using the rainfall intensity and duration at preset time points. Based on the cumulative rainfall effect value and the predicted safety factor sequence, a risk assessment of the target tailings dam is conducted, resulting in a risk assessment report.

[0007] Secondly, this application also provides a computer system, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the tailings dam risk assessment method based on LSTM safety factor prediction as described in the first aspect.

[0008] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application constructs historical dam environment sequences and safety factor sequences by acquiring historical dam environmental data and corresponding historical safety factor data at continuous time points for the target tailings dam. The process of acquiring multi-source data enhances the data's representational capabilities. Feature engineering is used to extract derived features, and combined with normalized time and decay coefficients to construct a composite feature sequence including static and dynamic time features, further improving the representativeness of the LSTM model input data. Then, an LSTM model is constructed, using the historical dam environment composite feature sequence as input data and the historical safety factor sequence as output data for model training. The current dam environment composite feature sequence is input into the trained model to predict the safety factor sequence, and the cumulative rainfall effect is calculated based on rainfall intensity and duration to complete a comprehensive risk assessment of the tailings dam. In summary, this application enhances the representativeness and accuracy of tailings dam safety factor prediction and further improves the accuracy of tailings dam risk assessment by utilizing the cumulative rainfall effect value for risk determination. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a flowchart illustrating a tailings dam risk assessment method based on LSTM safety factor prediction, provided as an embodiment of this application.

[0011] Figure 2 This is an internal structure diagram of a computer system provided in an embodiment of this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0013] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0014] Example 1, such as Figure 1 As shown in the figure, this embodiment provides a tailings dam risk assessment method based on LSTM safety factor prediction. The method includes: S1. Obtain historical dam environmental data and corresponding historical safety factor data for continuous time points of the target tailings dam, and obtain historical dam environmental sequence and historical safety factor sequence by constructing time series respectively; historical dam environmental data includes: dam height, dam slope ratio, cohesion, internal friction angle, permeability coefficient, reservoir water level, rainfall intensity and rainfall duration.

[0015] Optionally, historical dam environmental data can be obtained by collecting historical engineering case data and dam environmental simulation data obtained by Founder.

[0016] Furthermore, after step S1, the method further includes: preprocessing the historical dam body environmental data and historical safety factor data respectively, and updating the historical dam body environmental data and historical safety factor data; the preprocessing includes at least: cleaning, outlier removal and normalization.

[0017] S2. Using feature engineering, derivative features are extracted from samples in the historical dam environment sequence one by one to obtain the historical derivative features, normalized time, and decay coefficient of the samples. The historical derivative features are then combined with the features in the historical dam environment sequence as static features, and the normalized time and decay coefficient are used as dynamic time features to construct a composite feature sequence of historical dam environment. The historical derivative features include: dam height-to-slope ratio (HNR), shear strength factor (SSF), log permeability feature (LPF), rainfall-to-height ratio (RHR), water-level-to-height ratio (WHR), and dam morphology feature (GMF).

[0018] Furthermore, the formula for calculating the dam height-to-slope ratio characteristic is as follows: .

[0019] The formula for calculating the characteristic shear strength parameters is as follows: .

[0020] The formula for calculating the logarithmic characteristic of osmosis is as follows: .

[0021] The formula for calculating the height-to-depth ratio of rainfall dams is as follows: .

[0022] The formula for calculating the reservoir water level to dam height ratio is as follows: .

[0023] The formula for calculating the dam's morphological characteristics is as follows: .

[0024] In the formula, Characteristics of dam height and slope ratio; Characteristic parameters of shear strength; The characteristic of permeability is logarithmic. The characteristics of the dam height ratio during rainfall; The ratio of reservoir water level to dam height is a characteristic feature; Features of the dam's morphology; The dam is high; The dam slope ratio; It is cohesive force; It is the internal friction angle; Permeability coefficient; This refers to the reservoir water level. This refers to the intensity of rainfall.

[0025] In practical applications, the characteristics of dam height-slope ratio Comprehensive geometric features used to characterize the geometry and overall stability of a dam.

[0026] Shear strength parameter characteristics Core mechanical parameters used to characterize the shear strength of soil and rock masses.

[0027] Permeation logarithmic characteristics Logarithmically dimensionless features used to characterize the seepage properties of dams.

[0028] Rainfall Dam Height Ratio Characteristics Used to characterize the relative influence of rainfall intensity on dam height.

[0029] Reservoir water level to dam height ratio characteristics Used to characterize the relative impact of reservoir water level on dam stability.

[0030] Dam morphological characteristics Used to characterize the geometric regularity of the dam body and the morphological characteristics of the slope.

[0031] In practical applications, for each sample, the combined feature vector of the historical dam environment composite feature sequence is extended along the time axis to form an input feature sequence X_seq with a dimension of [total number of features × time step]. The corresponding target output is a historical safety coefficient sequence Y_seq with a dimension of [1 × time step].

[0032] Optionally, the formula for calculating normalized time is: .

[0033] In the formula, T_max represents the maximum time span of the historical dam environment sequence.

[0034] The formula for calculating decay time is: .

[0035] In the formula, For the preset attenuation coefficient, .

[0036] S3. Construct an LSTM model, and use the historical dam environment composite feature sequence as input data and the historical safety coefficient sequence as output data. Use the minimum loss function as the optimization objective to train the LSTM model and obtain the trained LSTM model.

[0037] Furthermore, step S3 specifically includes: S31. Construct the LSTM model.

[0038] S32. Use the historical dam environment composite feature sequence as input data and the historical safety coefficient sequence as output data, and divide the training set and validation set according to a preset ratio.

[0039] Optional, the default ratio is 8:2.

[0040] S33. Set the initial learning rate, learning rate decay coefficient, maximum number of training epochs, and early stopping patience value for the Adam optimizer in the LSTM model, and then train the LSTM model.

[0041] S34. Use the validation set to select the optimal parameters of the LSTM model, and use the model with the minimum loss function corresponding to the safety factor prediction error as the trained LSTM model.

[0042] Furthermore, the network architecture of the LSTM model includes: a sequence input layer, a first LSTM layer, a first Dropout layer, a second LSTM layer, a second Dropout layer, a fully connected layer, and a regression output layer, connected in sequence. The output mode of the first LSTM layer is set to sequence mode, which is used to output sequence features to the second LSTM layer. The Dropout probability of the first Dropout layer is set to 0.2~0.4; the Dropout probability of the second Dropout layer is set to 0.1~0.3. The Dropout probability is used to suppress model overfitting.

[0043] S4. Obtain the current composite feature sequence of the target tailings dam's environment and input it into the trained LSTM model to obtain the predicted safety factor sequence.

[0044] Optionally, the maximum time span for predicting the safety factor sequence is 48 hours.

[0045] In practical applications, the actual prediction process of the trained LSTM model also utilizes a query optimization module. This module pre-stores a lookup table generated from the training data. During the prediction phase, the query optimization module first performs a matching query on the new input parameter set. If the input parameter completely matches an entry in the lookup table, the corresponding safety coefficient sequence is read from the lookup table first, without needing to call the LSTM time series modeling and training unit. If no completely matching entry is found, the LSTM time series modeling and training unit is triggered to load the optimal model and execute the prediction process, balancing prediction efficiency and accuracy.

[0046] S5. Calculate the cumulative rainfall effect value using the rainfall intensity and duration at preset time points, and conduct a risk assessment of the target tailings dam based on the cumulative rainfall effect value and the predicted safety factor sequence, obtain the risk assessment result of the target tailings dam, and generate the corresponding risk assessment report.

[0047] Furthermore, step S5 specifically includes: S51. Calculate the cumulative effect value of rainfall using the rainfall intensity and duration at preset time points.

[0048] S52. Predict the safety factor at the corresponding preset time point in the safety factor sequence based on the cumulative effect value index of rainfall.

[0049] S53. The prediction uncertainty is calculated based on the cumulative effect value of rainfall, and the safety factor at the corresponding preset time point is weighted according to the prediction uncertainty to obtain the weighted safety factor.

[0050] S54. Determine the risk assessment results based on the weighted safety coefficient; the risk assessment results include: high risk, medium risk, and low risk.

[0051] Furthermore, step S54 specifically includes: S541. When the weighted safety factor is less than the first safety factor threshold, the risk assessment result is determined to be high risk.

[0052] S542. When the pre-weighted safety factor is greater than or equal to the first safety factor threshold and less than the second safety factor threshold, the risk assessment result is determined to be medium risk; the first safety factor threshold is less than the second safety factor threshold.

[0053] S543. When the weighted safety factor is greater than or equal to the second safety factor threshold, the risk assessment result is determined to be low risk.

[0054] The first safety threshold is 1.2; the second safety threshold is 1.5.

[0055] Furthermore, the formula for calculating the uncertainty in prediction is as follows: .

[0056] In the formula, To predict uncertainties; The base is a preset uncertainty value, base∈[0.02,0.05]; This represents the cumulative effect of rainfall.

[0057] The formula for calculating the weighted safety factor is as follows: .

[0058] In the formula, The weighted safety factor; For safety factor; This represents the cumulative effect of rainfall.

[0059] Furthermore, the formula for calculating the cumulative effect of rainfall is as follows: .

[0060] In the formula, This represents the cumulative effect of rainfall. Rainfall intensity; This refers to the duration of rainfall.

[0061] S55. Generate a risk assessment report corresponding to the risk assessment results.

[0062] Optionally, the risk assessment report shall include at least: the predicted safety factor sequence, confidence interval, risk warnings and early warning recommendations.

[0063] As an optional implementation method, this application takes the upstream tailings dam in the Poyang Lake area as an example to illustrate the specific implementation steps of this application in detail: I. Model Training Phase The first step involves establishing orthogonal experimental numerical simulation data based on actual engineering parameters and historical monitoring data of the upstream tailings dams in the Poyang Lake area. Stability calculations are then performed, and the results are compiled into two files: train_data.xlsx and test_data.xlsx. Train_data.xlsx contains 1000 data points, and test_data.xlsx contains 100 data points. Each sample includes: dam height h (m), dam slope ratio n (e.g., 1:3), cohesion c (kPa), and internal friction angle. (Unit: degree), permeability coefficient k (unit: m / s), reservoir water level w (unit: m, negative distance from dam crest), rainfall intensity q (unit: mm / d), rainfall duration t (unit: h), and safety factor sequence Fs corresponding to 0, 3, 6, ..., 48 hours (a total of 17 time points).

[0064] The second step is to run the main function `tailings_dam_safety_prediction_system_v9()`. The system will enter the command line main menu, displaying five function options: "1. Train a new model, 2. Interactive prediction, 3. Batch prediction, 4. Model validation, 5. Exit".

[0065] The third step is to select a training model and call the `train_new_model_LSTM_optimized_fixed()` function, which in turn calls the `execute_training_LSTM_fixed()` function to execute the core training process.

[0066] The fourth step involves calling the `load_and_preprocess_data` function via the `execute_training_LSTM_fixed()` function to read the `train_data.xlsx` file, removing outlier samples, filling in missing values, and normalizing all parameters and safety factors using Z-score standardization. Subsequently, the `extract_features` function is called to calculate 14-dimensional static features (h, n, c, ...) based on the original parameters. ,k,w,q,t,h / n,c×tan( The dynamic time features t / 48 (normalized time, T_max=48h) and exp(-0.05×t) (attenuation coefficient α=0.05) are calculated for each of the 17 time points. Finally, each sample is converted into a 16×17 input feature sequence X_cell{i}, and the corresponding target sequence Y_cell{i} is a 1×17 safety coefficient sequence.

[0067] Step 5: Data partitioning and model building: Divide the 1000 samples into a training set (800 samples) and a validation set (200 samples) in an 8:2 ratio; Calculate the mean (x_mean, y_mean) and standard deviation (x_std, y_std) of the features and the target based on the training set data, and save them to the normalized parameter file; Build an LSTM network.

[0068] Step 6, Model Training: Set the initial learning rate of the Adam optimizer to 0.001, the learning rate decay coefficient to 0.95, the maximum number of training epochs to 100, and the early stopping patience value to 10 (stop training if the loss on the validation set does not decrease after 10 consecutive epochs); use the mean squared error (MSE) as the loss function, train the model on the training set, monitor the model performance on the validation set, and finally stop training when the optimal performance is reached.

[0069] Step 7, Model Validation and Saving: After training is complete, the system calls the `validate_model_LSTM` function, loads the `test_data.xlsx` test set, and calculates the model's prediction accuracy: R 2 =0.962, mean absolute error =0.043, root mean square error =0.058, meeting the preset accuracy requirements; save the trained LSTM network (lstm_net), normalized parameters (x_mean, x_std, y_mean, y_std), hyperparameter configuration and other key data to the lstm_optimized_model.mat file to complete model training.

[0070] II. Online Prediction Phase Step 8: Run the main function tailings_dam_safety_prediction_system_v9(), enter the main menu, input "2", and select "Interactive Prediction"; the system will automatically check the lstm_optimized_model.mat file and load the LSTM optimal model, normalization parameters, and lookup table from it.

[0071] Step 9: The system prompts the user to input various parameters in sequence. The user inputs the following parameters according to the actual working conditions: dam height h=120m, dam slope ratio n=1:3, cohesion c=8kPa, internal friction angle =32°, permeability coefficient k=5.0e-5m / s, reservoir water level w=-7m (7m from the top of the dam), rainfall intensity q=400mm / d, rainfall duration t=24h.

[0072] Step 10: Call the query optimization module to match the input parameters with the lookup table. If there is no complete match, call the predict_safety_factor_LSTM_optimized(120,'1:3',8,32,5.0e-5,-7,400,24) function to execute the normal prediction process.

[0073] Step 11: Call the extract_features function to calculate the 14-dimensional static features corresponding to the input parameters, and then calculate the dynamic time features for 17 time points to form a 16×17 feature sequence. Use the normalization parameters saved during the training phase to preprocess (standardize) the feature sequence and input it into the loaded LSTM model. The model outputs a 1×17 normalized safety coefficient sequence. After inverse normalization, the predicted safety coefficient values ​​for 0~48 hours (17 time points) are obtained, for example: [1.85,1.82,1.79,1.76,1.73,1.70,1.68,1.66,1.65,1.64,1.63,1.62,1.61,1.60,1.59,1.58,1.65].

[0074] Step 12: Call the `calculate_rainfall_effect` function to calculate the cumulative rainfall effect: `rainfall_effect=min(0.1×400×24 / 1000,0.5)=min(0.96,0.5)=0.5`; Set the base uncertainty value `base=0.03` and calculate the prediction uncertainty: `uncertainty=0.03+0.5×0.08=0.07`; Calculate the adjusted safety factor corresponding to 48 hours: `adjusted_factor=1.65×(1-0.5×0.3)≈1.40`; Based on the preset risk level standards (`adjusted_factor≥1.5` is low risk, `1.2≤adjusted_factor<1.5` is medium risk, and `adjusted_factor<1.2` is high risk), the prediction result is determined to be medium risk, corresponding to the risk warning "Note: The impact of rainfall is significant, the safety margin is reduced, it is recommended to strengthen the monitoring frequency of the dam body and prepare for early warning".

[0075] Step 13: The system displays the predicted safety factor, adjusted safety factor, prediction uncertainty, and risk level for 17 time points in text format. It also generates a line graph of the safety factor changing over time (including a 95% confidence interval). The prediction results (parameters, prediction sequence, and risk assessment) are saved to the 202XXXXX_pred.xlsx file in the prediction_results folder for users to query and export later.

[0076] The technical effects of this application are as follows: 1) This application utilizes the time series modeling capabilities of LSTM networks to capture the dynamic evolution of the safety factor under complex dynamic effects such as rainfall infiltration and reservoir water level changes; through multi-dimensional composite features, it characterizes the dam body characteristics, hydrological conditions and time effects, thereby enhancing the representativeness and accuracy of tailings dam safety factor prediction.

[0077] 2) In addition to risk assessment based on safety factors, this application also integrates rainfall cumulative effect model to quantify prediction uncertainty, assess risk level, and output risk assessment report, thereby improving the accuracy of tailings dam risk assessment and meeting the decision-making needs of mine safety management.

[0078] 3) Improved the efficiency of risk assessment and provided technical support for rapid early warning under extreme conditions such as sudden rainfall.

[0079] Example 2: This example provides a computer system, which can be a server or a terminal, and its internal structure diagram can be as follows. Figure 2As shown, the computer system includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores forced oscillation samples and sub / supersynchronous oscillation samples. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the aforementioned method for rapid prediction and identification of the dominant frequency of sub / supersynchronous oscillations in new energy power systems based on transfer learning.

[0080] Those skilled in the art will understand that Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer system to which the present application is applied. A specific computer system may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0081] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).

[0082] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0083] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0084] This document uses specific examples to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A tailings dam risk assessment method based on LSTM safety factor prediction, characterized in that, The method includes: Historical environmental data of the target tailings dam and corresponding historical safety factor data at continuous time points are obtained, and historical environmental sequences and historical safety factor sequences are obtained by constructing time series. The historical environmental data of the dam includes: dam height, dam slope ratio, cohesion, internal friction angle, permeability coefficient, reservoir water level, rainfall intensity and rainfall duration. Feature engineering is used to extract derived features from samples in the historical dam environment sequence one by one, obtaining the historical derived features, normalized time, and decay coefficient of the samples. The historical derived features are then combined with the features in the historical dam environment sequence as static features, and the normalized time and decay coefficient are used as dynamic time features to construct a composite feature sequence of historical dam environment. The historical derived features include: dam height-to-slope ratio, shear strength parameter, logarithm of seepage, rainfall-to-dam height ratio, reservoir water level-to-dam height ratio, and dam morphology. An LSTM model is constructed, and the historical dam environment composite feature sequence is used as input data, and the historical safety coefficient sequence is used as output data. The LSTM model is trained with the minimum loss function as the optimization objective, and the trained LSTM model is obtained. Obtain the current composite feature sequence of the target tailings dam's environment and input it into the trained LSTM model to obtain the predicted safety factor sequence; The cumulative effect value of rainfall is calculated using the rainfall intensity and duration at preset time points. Based on the cumulative effect value of rainfall and the predicted safety factor sequence, a risk assessment of the target tailings dam is conducted to obtain the risk assessment result of the target tailings dam and generate the corresponding risk assessment report.

2. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, After acquiring historical environmental data and corresponding historical safety factor data of the target tailings dam at continuous time points, and obtaining the historical dam environmental sequence and historical safety factor sequence by constructing time series, the process also includes: The historical dam environment data and historical safety factor data are preprocessed and updated respectively; the preprocessing includes at least: cleaning, outlier removal and normalization.

3. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, The formula for calculating the dam height-to-slope ratio characteristic is as follows: ; The calculation formula for the shear strength parameter characteristics is as follows: ; The formula for calculating the logarithmic characteristic of permeability is as follows: ; The formula for calculating the height-to-depth ratio of the rainfall dam is as follows: ; The formula for calculating the reservoir water level to dam height ratio is as follows: ; The calculation formula for the dam's morphological characteristics is as follows: ; In the formula, Characteristics of dam height and slope ratio; Characteristic parameters of shear strength; The characteristic of permeability is logarithmic. The characteristics of the dam height ratio during rainfall; The ratio of reservoir water level to dam height is a characteristic feature; Features of the dam's morphology; The dam is high; The dam slope ratio; It is cohesive force; It is the internal friction angle; Permeability coefficient; This refers to the reservoir water level. This refers to the intensity of rainfall.

4. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, An LSTM model is constructed, using the historical dam environment composite feature sequence as input data and the historical safety coefficient sequence as output data. The LSTM model is trained with the minimum loss function as the optimization objective, resulting in a trained LSTM model. Specifically, this includes: Construct an LSTM model; The historical dam environment composite feature sequence is used as input data, and the historical safety coefficient sequence is used as output data. The training set and the validation set are divided according to a preset ratio. Set the initial learning rate, learning rate decay coefficient, maximum number of training epochs, and early stopping patience value for the Adam optimizer in the LSTM model, and then train the LSTM model. The optimal parameters of the LSTM model are selected using a validation set, and the model with the minimum loss function corresponding to the safety factor prediction error is used as the trained LSTM model.

5. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, The network architecture of the LSTM model includes: a sequence input layer, a first LSTM layer, a first Dropout layer, a second LSTM layer, a second Dropout layer, a fully connected layer, and a regression output layer, connected in sequence. The output mode of the first LSTM layer is set to sequence mode, which is used to output sequence features to the second LSTM layer. The Dropout probability of the first Dropout layer is set to 0.2~0.4; the Dropout probability of the second Dropout layer is set to 0.1~0.3; the Dropout probability is used to suppress model overfitting.

6. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, The cumulative rainfall effect is calculated using the rainfall intensity and duration at preset time points. Based on the cumulative rainfall effect and the predicted safety factor sequence, a risk assessment is conducted on the target tailings dam, yielding the risk assessment results and generating a corresponding risk assessment report, which includes: Calculate the cumulative effect of rainfall using the rainfall intensity and duration at preset time points; Predict the safety factor at a preset time point in the safety factor sequence based on the cumulative rainfall effect index; The prediction uncertainty is calculated based on the cumulative effect of rainfall, and the safety factor at the corresponding preset time point is weighted according to the prediction uncertainty to obtain the weighted safety factor. The risk assessment results are determined based on a weighted safety coefficient; the risk assessment results include: high risk, medium risk, and low risk. Generate a risk assessment report corresponding to the risk assessment results.

7. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 6, characterized in that, The risk assessment results are determined based on a weighted safety factor, specifically including: When the weighted safety factor is less than the first safety factor threshold, the risk assessment result is determined to be high risk; When the pre-weighted safety factor is greater than or equal to the first safety factor threshold and less than the second safety factor threshold, the risk assessment result is determined to be medium risk; the first safety factor threshold is less than the second safety factor threshold. When the weighted safety factor is greater than or equal to the second safety factor threshold, the risk assessment result is determined to be low risk.

8. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 6, characterized in that, The formula for calculating the prediction uncertainty is as follows: ； In the formula, To predict uncertainties; The base is a preset uncertainty value, base∈[0.02,0.05]; This represents the cumulative effect of rainfall. The formula for calculating the weighted safety factor is as follows: ； In the formula, The weighted safety factor; For safety factor; This represents the cumulative effect of rainfall.

9. The tailings dam risk assessment method based on LSTM safety factor prediction according to claim 1, characterized in that, The formula for calculating the cumulative effect of rainfall is as follows: ； In the formula, This represents the cumulative effect of rainfall. Rainfall intensity; This refers to the duration of rainfall.

10. A computer system, comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that the processor executes the computer program to implement the tailings dam risk assessment method based on LSTM safety factor prediction as described in any one of claims 1-9.

Images