Machine learning-based full-period safety management and control method and system for film bag sand bank slope

Abstract

The invention discloses a machine learning-based full-period safety management and control method and system for a film bag sand levee slope, and the method comprises the following steps: in a design stage, based on an initial design scheme, judging a damage mode through a classification model, selecting a corresponding regression prediction model to predict a limit height Hmax and a safety coefficient Fs, and comparing the limit height Hmax and the safety coefficient Fs with a safety set threshold value [Fs], judging whether the design scheme is feasible or not, and if not, generating an optimal scheme through the optimization model; in the construction stage, after displacement monitoring data passes through an inversion model to output new foundation soil parameters, original design parameters are replaced, an updating scheme and the monitoring data pass through a classification model, a time sequence model is used for predicting a safety coefficient Fsp, a [Fs] early warning threshold value is combined to judge a risk level, and if the safety range is exceeded, early warning is carried out, and an optimal adjustment scheme is generated through an optimization model. According to the method, a full-period closed loop of design pre-judgment, construction updating, real-time evaluation and optimization adjustment is constructed, monitoring data and a safety coefficient are organically combined, and dynamic full-period stability evaluation and adjustment are achieved.

Description

Technical Field

[0001] This invention relates to the field of intelligent engineering management and control technology, and in particular to a method and system for full-cycle safety management and control of membrane bag sand embankment slopes based on machine learning. Background Technology

[0002] Membrane bag sand embankments, with their advantages of high mechanization, fast construction speed, strong overall stability, and high adaptability, have become the core retaining structure for projects such as land reclamation, nearshore dikes, and waterway regulation. However, soft soil foundations are characterized by low strength and large deformation. Coupled with the pain points in embankment engineering, such as design relying on experience-based analogies, delayed construction monitoring, and untimely safety warnings, frequent cases of engineering instability occur. Under soft soil conditions, there may be "overall failure" or "partial failure" modes, such as insufficient foundation bearing capacity leading to embankment subsidence or combined sliding failure. In terms of design, conventional designs often use sandbags of uniform thickness and materials of uniform strength, without adequately reinforcing the bottom, which may affect the bearing capacity. Furthermore, the quality of membrane bag embankments is also closely dependent on the type, proportion, and filling process of the filling material.

[0003] Traditional methods for analyzing and controlling the stability of membrane bag sand embankments primarily rely on safety factor predictions during the design phase and real-time on-site monitoring during construction. However, as a flexible structure, membrane bag sand embankments experience continuous coordinated deformation and interaction between their structure and the foundation soil during construction and normal use. As the drainage consolidation process progresses, the stiffness of the cofferdam and the strength of the foundation soil change over time, resulting in dynamic adjustments to the overall stability of the embankment. Furthermore, the embankment's flexibility and the soft soil properties of the foundation contribute to significant and prolonged foundation deformation, making accurate stability assessment difficult solely based on deformation monitoring. Consequently, real-time monitoring data during construction is often difficult to link effectively with safety factors, hindering engineers from making precise judgments about the embankment's stability. Measures are only taken when significant anomalies appear in the monitoring data, without proactively anticipating and addressing trends based on these data.

[0004] To address the aforementioned issues, there is an urgent need to develop an intelligent stability management method and system that integrates multi-source information and is adapted to the entire lifecycle of membrane bag sand embankments. Machine learning technology, with its significant advantages in multimodal data fusion, dynamic prediction, and adaptive optimization, provides strong technical support for solving this engineering challenge. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a machine learning-based method and system for full-cycle safety management of membrane bag sand embankment slopes. It constructs a closed loop of design prediction, construction update, real-time evaluation, and optimization adjustment, covering the entire process from design to construction. It can provide early warnings and make timely response adjustments to continuously optimize the scheme.

[0006] This invention is achieved through the following technical solution: A machine learning-based method for full-cycle safety management of membrane bag sand embankment slopes includes the following steps: S1. During the design phase: S1-1. Based on the initial design scheme, the failure mode is determined by a classification model. The design scheme includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment cross section. S1-2. Based on the failure mode obtained in step S1-1, select the corresponding regression prediction model, and use the corresponding regression prediction model to predict the limit fill height H. max and safety factor F s ; S1-3, By predicting the safety factor F s With the safety factor set threshold [F] s In comparison, it is determined whether the design scheme is feasible. If it is feasible, the design scheme is the optimal scheme. If it is not feasible, the optimal scheme is generated through the optimization model. The optimal scheme includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment section. S2. During the construction phase: S2-1. Construction monitoring data is used to obtain new foundation soil parameters through an inversion model. The original foundation soil parameters are replaced with the new foundation soil parameters to update the optimal solution obtained in step S1. S2-2. The new optimal solution and new construction monitoring data are used to determine the failure mode through a classification model; S2-3. Select the corresponding time series model based on the failure mode, and predict the safety factor F using the corresponding time series model. sp ; S2-4, Based on the predicted safety factor F sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, no warning is needed and continuous monitoring is performed. If the safety risk level exceeds the set safety range, an warning is issued and an optimal solution is generated through an optimization model. The optimal solution includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment section. S2-5. During the construction process, following the methods in steps S2-2 to S2-4, conduct regular or irregular safety assessments and optimize the plan until the construction is completed.

[0007] Furthermore, the design parameters of the geomembrane bag sand embankment slope include the tensile stiffness J of the geomembrane bag and the undrained shear strength S of the soft soil. u The shear strength τ of the contact surface between the geomembrane bag and the foundation, the thickness t of the sand filling in the geomembrane bag, the internal friction angle φ of the filling sand, and the weight γ of the foundation soil. c'and membrane bag filling sand unit weight γ s The geometric parameters of the embankment slope section include the embankment slope width W, the embankment slope height H, and the embankment slope k; the parameters of the foundation soil include the undrained shear strength S of soft soil. u The construction monitoring data includes displacement data.

[0008] Furthermore, the classification model employs a random forest classifier, which balances the noise tolerance and classification accuracy of the engineering data. Its output destruction modes include overall destruction and partial destruction. The training method for the classification model is as follows: Data is collected from sources including actual engineering data (instability cases, including failure mode labels), simulation data, and indoor scaled physical model test data, including data under the condition of reconstituted sludge. The data was randomly divided into a training set and a test set in a 7:3 ratio. The hyperparameters were optimized using 5-fold cross-validation, with 100 decision trees and a maximum depth of 8 for each tree.

[0009] Furthermore, the regression prediction model adopts the XGBoost regression model structure, and the training method of the regression prediction model is as follows: The training set data sources include simulation data and indoor scaled physical model test data, which includes ultimate bearing capacity and safety factor data derived from the embankment load-settlement curve. The test set data includes actual engineering data; The hyperparameters were optimized using a grid search with a learning rate of 0.1, a tree depth of 6, and 200 iterations. Furthermore, the optimization model includes the NSGA-Ⅲ optimization algorithm, the classification model, and the regression prediction model. Its working process is as follows: under the condition of clearly defined optimization objective and constraints, candidate solutions are generated through the NSGA-Ⅲ optimization algorithm; Candidate solutions determine the failure mode using the classification model, select the corresponding regression prediction model based on the failure mode, and obtain the safety factor F through the corresponding regression prediction model. s ; Based on the safety coefficient Fs of each candidate scheme, feasible schemes are selected; Taking into account both economic and security factors, the optimal solution is selected from the feasible options.

[0010] Furthermore, the safety factor F of each candidate scheme... s The method for selecting feasible solutions is as follows: If it is in the design phase, the safety factor F of the candidate scheme will be... s With the safety factor set threshold [F]s If Fs ≥ [Fs], then the candidate solution is a feasible solution; otherwise, it is an infeasible solution. If it is during the construction phase, the safety factor F of the candidate scheme will be... sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, it is a feasible solution; otherwise, it is an infeasible solution. The constraints for the design phase are W∈30-120m, J∈140-8000kN / m, S u For parameters ≥5kPa, the corresponding constraint during the construction phase is that the adjustment range should be ≤20% of the original parameters to avoid significant changes to the construction plan.

[0011] Furthermore, the training method of the NSGA-Ⅲ optimization algorithm is as follows: using 2000 feasible schemes generated by simulation numerical simulation as the initial population, setting the crossover probability to 0.8, the mutation probability to 0.1, and the number of iterations to 50. Furthermore, the inversion model employs a Bayesian-optimized BP neural network (BO-BP) to balance inversion accuracy and computational efficiency. The input data is the monitored displacement data, and the output data is the undrained shear strength S of the soft soil. u ; The method for training the inversion model is as follows: The training set data sources include actual engineering data (construction-measured displacement and geotechnical test data) and simulation data. The simulation data includes data from S... u Simulated input parameters and displacement response data with values ​​ranging from 2.5 to 15 kPa, corresponding to different displacement outputs; The optimal initial weights and structure of a backpropagation (BP) neural network are searched using Bayesian optimization, including the number of hidden layers and the number of neurons per layer.

[0012] Furthermore, the time-series model employs an XGBoost-LSTM coupled model, comprising an XGBoost module, an LSTM module, and a feature fusion module. The XGBoost module extracts static nonlinear features from the static data in the optimal solution, the LSTM module extracts time-series features from the construction monitoring data, and the feature fusion module concatenates the static nonlinear feature vector output by the XGBoost module with the time-series feature vector output by the LSTM module, and completes feature fusion through a fully connected network to obtain a joint feature representation. This "feature-level fusion" achieves complementary advantages, addressing both the temporal continuity issue of monitoring data and adapting to the nonlinear prediction requirements after dynamic updates of engineering parameters, thus improving accuracy compared to a single LSTM.

[0013] The method for training the time series model is as follows: Data sources include construction monitoring time-series data from actual projects, with each data point containing a timestamp, horizontal displacement, vertical displacement, and cumulative settlement. Data preprocessing: A sliding window method (e.g., window size = 24) is used to construct time series samples, missing values ​​are filled by linear interpolation, and trend noise is eliminated by the difference method; Training and validation: The data is divided into training and testing sets in an 8:2 ratio, and hyperparameters are optimized, including hidden layer dimension, number of iterations, and batch size.

[0014] A machine learning-based full-cycle safety management system for membrane bag sand embankment slopes is used to implement the aforementioned machine learning-based full-cycle safety management method for membrane bag sand embankment slopes, and includes: The data collection module is used to collect actual engineering data, simulation data, and indoor scaled physical model test data for model training and verification, as well as scheme evaluation and optimization. The failure mode identification module is used to identify the failure mode of the design or optimization scheme through a classification model; The safety factor prediction module in the design phase is used to predict the maximum fill height Hmax and the safety factor F of the design scheme through a regression prediction model. s ; The construction phase safety factor prediction module is used to predict the safety factor F of the optimized scheme through a time series model. sp ; The foundation soil inversion module is used to obtain foundation soil parameters from construction monitoring data through an inversion model. The scheme optimization module is used to optimize the design scheme or optimization scheme through the optimization model when the safety factor of the scheme (design scheme or optimization scheme) exceeds the set value. It optimizes the initial design scheme in the design stage and continuously optimizes the optimized scheme (including the optimization scheme in the design stage and the construction stage) in the construction stage.

[0015] The main advantages of this invention compared to existing technologies are: This invention addresses the shortcomings of traditional methods that rely solely on static safety factors during the design phase and displacement monitoring data during the construction phase. These methods are insufficient to handle the continuous coordinated deformation of the embankment slope and soft soil foundation during construction, as well as the dynamic changes in cofferdam stiffness and foundation strength due to drainage consolidation. This results in distorted and delayed stability assessments during the construction phase. The invention constructs a closed-loop mechanism encompassing design prediction, construction updates, real-time assessment, and optimization adjustments. By organically combining displacement monitoring data with safety factors, it enables dynamic, full-cycle stability assessment and adjustment, thereby improving the systematic nature and reliability of safety management.

[0016] In the design phase, this invention uses a random forest classification model to distinguish between overall destruction and partial destruction modes, and combines this with an XGBoost regression model to predict the maximum fill height H. max With safety factor F s Risk points are identified in advance; during the construction phase, the undrained shear strength S of the foundation soft soil is dynamically updated based on monitoring data using a BO-BP inversion model. u The core parameters are then used to continuously predict the dynamic safety factor through the XGBoost-LSTM time series model, realizing the transformation from static prediction to full-cycle dynamic tracking.

[0017] This invention addresses the problems of instability in projects such as Tianjin Port and Shenjiamen Port in Zhoushan due to insufficient foundation bearing capacity and excessive settlement caused by traditional monitoring methods that are slow, untimely, and lack rapid response. It constructs an early warning mechanism based on an XGBoost-LSTM coupled model, dynamically predicting the safety factor. When the predicted safety level reaches a high-risk level, an early warning is triggered. The optimization model is then used to adjust and optimize the solution, enabling early identification of safety hazards and timely response. Simultaneously, it considers construction costs, such as reducing the amount of geosynthetic materials used, thus achieving a rapid response from early warning to optimization to execution.

[0018] This invention integrates data from multiple sources for prediction optimization and model training, achieving deep fusion of multimodal data. The data sources include actual design parameters (W, J, su, etc.), physical test (scaled physical model test) data, numerical simulation data, and construction monitoring data. This improves the accuracy of model judgment and prediction, making it closer to actual engineering conditions. It also overcomes the shortcomings of traditional design, such as the lack of consideration for the quantitative influence of core parameters like embankment bottom width, membrane bag stiffness, and foundation strength, which leads to insufficient design relevance. Furthermore, it addresses the problem of insufficient reliability in construction safety prediction caused by the failure to effectively link displacement monitoring data with safety factors during the conventional construction phase, thus ensuring the reliability of safety management. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the design phase of an embodiment of the present invention.

[0020] Figure 2 This is a flowchart illustrating the construction phase of an embodiment of the present invention.

[0021] Figure 3 This is a schematic diagram illustrating the working principle of the optimized model in this embodiment of the invention.

[0022] Figure 4 This is a schematic diagram illustrating the working principle of the timing model in an embodiment of the present invention.

[0023] Figure 5 This is a structural block diagram of the security control system in an embodiment of the present invention. Detailed Implementation

[0024] A machine learning-based method for full-cycle safety management of membrane bag sand embankment slopes includes the following steps: S1, such as Figure 1 As shown, during the design phase: S1-1. Based on the initial design scheme, the failure mode is determined by a classification model. The design scheme includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment cross section. S1-2. Based on the failure mode obtained in step S1-1, select the corresponding regression prediction model, and use the corresponding regression prediction model to predict the limit fill height H. max and safety factor F s ; S1-3, By predicting the safety factor F s With the safety factor set threshold [F] s In comparison, determine whether the design scheme is feasible. If feasible (F... s ≥[F s If the design scheme is not feasible (F), then the design scheme is the optimal solution; otherwise, it is not feasible (F). s <[F s The optimal solution is generated through an optimization model, which includes the design parameters of the membrane bag sand embankment slope and the geometric parameters of the embankment slope section. S2, such as Figure 2 As shown, during the construction phase: S2-1. Construction monitoring data is used to obtain new foundation soil parameters through an inversion model. The original foundation soil parameters are replaced with the new foundation soil parameters to update the optimal solution obtained in step S1. S2-2. The new optimal solution and new construction monitoring data are used to determine the failure mode through a classification model; S2-3. Select the corresponding time series model based on the failure mode, and predict the safety factor F using the corresponding time series model. sp ; S2-4, Based on the predicted safety factor F sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, no warning is needed and continuous monitoring is performed. If the safety risk level exceeds the set safety range (reaching the high-risk category), an warning is issued and the optimal solution is generated through the optimization model. The optimal solution includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment section. S2-5. During construction, following the methods described in steps S2-2 to S2-4, conduct regular or irregular safety assessments and optimize the plan until construction is completed. Regular assessments refer to risk assessments and plan optimizations performed at set time intervals, while irregular assessments refer to random sampling or assessments and optimizations performed when data is abnormal.

[0025] In this embodiment, the design parameters of the geomembrane bag sand embankment slope include the tensile stiffness J of the geomembrane bag and the undrained shear strength S of the soft soil. u The shear strength τ of the contact surface between the geomembrane bag and the foundation, the thickness t of the sand filling in the geomembrane bag, the internal friction angle φ of the filling sand, and the weight γ of the foundation soil. c 'and membrane bag filling sand unit weight γ s The geometric parameters of the embankment slope section include the embankment slope width W, the embankment slope height H, and the embankment slope k; the parameters of the foundation soil include the undrained shear strength S of soft soil. u The construction monitoring data includes displacement data, such as horizontal displacement, vertical displacement, and settlement value.

[0026] The core function of the classification model is to accurately identify the failure mode based on the input parameters, providing a basis for subsequent regression prediction. It needs to adapt to the high dimensionality and nonlinear characteristics of engineering data and have strong anti-interference capabilities. As one implementation method, the classification model uses a random forest classifier, balancing the noise tolerance and classification accuracy of the engineering data. Its output failure modes include overall failure and lateral failure. Overall failure is mainly caused by insufficient foundation bearing capacity, primarily manifested as overall subsidence of the embankment slope, with the failure surfaces in the foundation being basically symmetrical and intersecting below the center of the embankment slope. Lateral failure mainly manifests as localized foundation failure at the toe of the slope on both sides of the embankment slope.

[0027] For example, the training method for the classification model is as follows: (1) Collect data, including actual engineering data, simulation data and indoor scaled physical model test data, including data under the condition of remolding sludge.

[0028] The actual engineering data in this embodiment includes 12 sets of instability case data, with failure mode labels; the simulation data is obtained by numerical simulation calculation (finite element method), containing 2400 working conditions, including the bottom width W of the embankment slope, the tensile stiffness J of the membrane bag, the sand filling thickness t of the membrane bag, and the undrained shear strength S of the soft soil. u The parameters include the internal friction angle φ of the filling sand, the slope k of the embankment, and the shear strength τ of the contact surface between the geomembrane bag and the foundation; the indoor scaled physical model test data includes data from 30 sets of remolded silt conditions.

[0029] These data also need to be processed to filter out core features. Specifically, Z-score standardization can be used to eliminate the influence of dimensions, and redundant features can be eliminated through mutual information.

[0030] (2) The data is randomly divided into training set and test set in a ratio of 7:3.

[0031] (3) The hyperparameters were optimized using 5-fold cross-validation, with 100 decision trees and a maximum depth of 8 for each tree. Experiments showed that the obtained model could accurately distinguish between the two damage modes.

[0032] The main function of regression prediction models is to predict the limit height H. max and safety factor F s This meets engineering-grade accuracy requirements. It should be noted that although only the safety factor F is used in subsequent operations... s H was not used. max However, from the engineers' perspective, their main concern is the overall stability of the embankment and the safety factor F. s However, obtaining the safety factor requires knowing the ultimate fill height of the embankment slope. Therefore, predicting the ultimate fill height H... max Is it predicting F s This is a preliminary step, and it is also important for engineers to understand the maximum fill height of the embankment and determine whether it meets the engineering requirements.

[0033] This invention uses different regression prediction models for different failure modes. After the classification model determines whether the failure is overall or partial, the corresponding regression prediction model then predicts the prediction limit H based on the failure type. max and safety factor F s Different regression prediction models have the same model structure, but differ in the data types used for training, and their corresponding parameters may also differ. As one implementation method, in this embodiment, the regression prediction models all adopt the XGBoost regression model structure, which can effectively fit the complex relationship between "parameters and safety indicators," adapt to the nonlinear mapping relationship between parameters and target values, and support feature importance analysis.

[0034] The method for training a regression prediction model is as follows: (1) The data sources for the training set include simulation data and indoor scaled physical model test data. The indoor scaled physical model test data includes ultimate bearing capacity and safety factor data derived from the embankment load-settlement curve. In this embodiment, the simulation data is obtained through numerical simulation calculations, including but not limited to the finite element method, discrete element method, etc. The data specifically includes W, J, and S under different failure modes. u Parameters and H max F s The corresponding relationship data is mainly derived from the ultimate bearing capacity and safety factor data of the embankment load-settlement curve.

[0035] (2) The test set data includes actual engineering data.

[0036] In this embodiment, the actual engineering case data includes the actual data of the cofferdams in Tianjin Port and Shenjiamen Port, Zhoushan. The data for the geomembrane bag sand cofferdam for the immersed tunnel in Shenjiamen Port, Zhoushan are as follows: bottom width of the dike slope W=23.3m, height H=7m, slope k=1 / 2, geomembrane bag stiffness J=150kN / m, and filling sand unit weight is 20kN / m³. 3 Soil strength S u =25kPa, the buoyant unit weight of the soil is 9kN / m 3 .

[0037] The data for the geomembrane bag sand cofferdam for land reclamation in Tianjin Port are as follows: dam slope width W=60m, height H=6m, slope k=1 / 2, geomembrane bag stiffness J=140kN / m, and filling sand unit weight 18kN / m³. 3 Soil strength S u =5kPa, buoyant unit weight of soil 6.8kN / m 3 .

[0038] The number of parameters used in this verification case is less than the seven parameters input in the numerical simulation calculations mentioned above. This is because, according to the simulation results, some parameters have a relatively small impact on the ultimate fill height and stability of the embankment and can be ignored.

[0039] (3) Grid search was used to optimize the hyperparameters, with a learning rate of 0.1, a tree depth of 6, and 200 iterations. Verification showed that the obtained regression prediction model could achieve H... max With F s The predicted values ​​meet the requirements of engineering design and construction control. The function of the optimization model is to optimize the existing scheme, satisfy Fs≥[Fs] in the design stage, and be below the high-risk warning in the construction stage, generating the optimal parameter scheme. It needs to adapt to multiple constraints, such as material usage, construction difficulty, safety threshold, etc.

[0040] In one implementation, the optimization model in this embodiment is a composite model, which includes the NSGA-III optimization algorithm, the classification model, and the regression prediction model, as follows: Figure 3 As shown, its working process is as follows: a. Given a clear optimization objective and constraints, candidate solutions are generated using the NSGA-Ⅲ optimization algorithm.

[0041] For example, the constraints in the design phase are W∈30-120m, J∈140-8000kN / m, S u For parameters ≥5kPa, the corresponding constraint during the construction phase is that the adjustment range should be ≤20% of the original parameters to avoid significant changes to the construction plan.

[0042] The NSGA-Ⅲ optimization algorithm can simultaneously optimize three major objectives: maximizing the safety factor, minimizing the amount of geotechnical materials used, and minimizing the construction period.

[0043] b. The candidate solution determines the failure mode using the classification model, selects the corresponding regression prediction model based on the failure mode, and obtains the safety factor F through the corresponding regression prediction model. s .

[0044] c. Select feasible solutions based on the safety coefficient Fs of each candidate solution.

[0045] The specific method is as follows: During the design phase, the safety factor F of the candidate solutions is... s With the safety factor set threshold [F] s If Fs ≥ [Fs], then the candidate solution is a feasible solution; otherwise, it is an infeasible solution. During the construction phase, the safety factor F of the candidate scheme will be... sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, it is a feasible solution; otherwise, it is an infeasible solution.

[0046] d. Taking into account economic and safety factors, select the optimal solution from the feasible options, such as the most economical and safest option.

[0047] In this embodiment, the training method of the NSGA-Ⅲ optimization algorithm is as follows: using 2000 feasible schemes generated by simulation numerical modeling as the initial population, setting the crossover probability to 0.8, the mutation probability to 0.1, and the number of iterations to 50. Experiments show that the F-value of the optimized scheme is... s The compliance rate has been effectively improved, and the amount of geosynthetic materials used has been effectively reduced.

[0048] The core function of the inversion model is to invert key foundation parameters such as the undrained shear strength of soft soil based on construction monitoring displacement data, meeting the input requirements for dynamic prediction during construction. As one implementation method, in this embodiment, the inversion model employs a Bayesian optimized BP neural network (BO-BP), balancing inversion accuracy and computational efficiency. The input data is the monitored displacement data, and the output data is the undrained shear strength S of the soft soil. u .

[0049] For example, the method for training the inversion model is as follows: (1) The data sources for the training set include actual engineering data and simulation data.

[0050] In this embodiment, the actual engineering data includes 500 sets of measured displacement and geotechnical test data, and the simulation data specifically consists of 1000 sets of data obtained by simulating the relationship between "input parameters and displacement response". u The value ranges from 2.5 to 15 kPa, corresponding to different displacement outputs.

[0051] More specifically, the simulation method can be as follows: Using numerical simulation software, including but not limited to the finite element method and discrete element method, such as the commercial finite element software Abaqus, a finite element model is established based on the geometric dimensions and material parameters of the membrane bag sand embankment slope and foundation. This model simulates the actual construction process, applying loads until the embankment slope becomes unstable and fails, thereby obtaining the ultimate filling height of the embankment slope. Then, under the same geometric dimensions and material parameters, by setting different embankment slope filling heights, the safety factor corresponding to different filling heights is calculated using the strength reduction method. This establishes the relationship between the embankment slope geometric dimensions and material parameters and the ultimate filling height H. max Filling height H, safety factor F s The correspondence between them.

[0052] (2) The optimal initial weights and structure of the BP neural network were searched through Bayesian optimization, including the number of hidden layers and the number of neurons per layer. Finally, the number of hidden layers was 2 and the number of nodes was 32. The inversion parameters were highly accurate and the foundation parameters could be dynamically updated.

[0053] The main function of time series models is to dynamically predict safety factors and achieve risk warning by combining updated parameters with monitoring data. In this invention, different failure modes correspond to different time series models. After the classification model determines whether the failure is overall or partial, the corresponding time series model predicts the safety factor based on the failure type. Different regression time series models have the same model structure, but the data types trained on them differ, and their corresponding parameters may also differ.

[0054] As one implementation method, in this embodiment, the time series models all adopt the XGBoost-LSTM coupled model, such as... Figure 4 As shown, it includes an XGBoost module, an LSTM module, and a feature fusion module. The XGBoost module is used to extract static nonlinear features from the static data in the optimal solution. The LSTM module is used to extract time-series features from the construction monitoring data. The feature fusion module is used to concatenate the static nonlinear feature vector output by the XGBoost module and the time-series feature vector output by the LSTM module, and complete the feature fusion through one or two fully connected neural networks to obtain a joint feature representation for subsequent safety factor prediction.

[0055] The relevant formulas for the feature fusion module can be: , , .

[0056] This invention employs an XGBoost-LSTM coupled model to achieve complementary advantages through "feature-level fusion," which not only solves the problem of temporal continuity of monitoring data but also adapts to the nonlinear prediction requirements after dynamic updates of engineering parameters, thus improving accuracy compared to a single LSTM.

[0057] The static data of the XGBoost module comes from relevant data in the optimization scheme, such as W, H, J, and γ. c ′、γ s wait.

[0058] The training of the time series model involves independently training the XGBoost module to fully fit the nonlinear relationship between static design parameters and the safety factor. After training, it is used as a static feature extractor. The second stage, while keeping the XGBoost module parameters unchanged (frozen), utilizes its output static nonlinear features and the time series features extracted by LSTM from construction monitoring data to perform joint modeling in the feature-level fusion module. This is achieved by training the LSTM module and the feature fusion module end-to-end through backpropagation, thus predicting the dynamic safety factor. Specifically, the training method includes: (1) Data source: Construction monitoring time series data of actual projects, each data point includes timestamp, horizontal displacement, vertical displacement and cumulative settlement. In this embodiment, the construction monitoring time series data involves 3 actual projects, with a sampling frequency of 30 min / time, covering the entire stage of embankment filling, with a total of more than 100,000 time series data points.

[0059] (2) Data preprocessing: The time series samples were constructed using the sliding window method (window size = 24), missing values ​​were filled by linear interpolation, and trend noise was eliminated by the difference method.

[0060] (3) Training and Validation: The data was divided into training and testing sets in an 8:2 ratio, and hyperparameters were optimized, including the hidden layer dimension, number of iterations, and batch size. The final time series model had a hidden layer dimension of 64, 100 iterations, and a batch size of 32, and could effectively predict F. s It can issue early warnings of safety risks, and the warning values ​​can be flexibly adjusted according to actual project needs.

[0061] A machine learning-based full-cycle safety management system for membrane bag sand embankment slopes is used to implement the aforementioned machine learning-based full-cycle safety management method for membrane bag sand embankment slopes, such as... Figure 5As shown, the system includes a data collection module, a failure mode identification module, a safety factor prediction module for the design phase, a safety factor prediction module for the construction phase, a foundation soil inversion module, and a scheme optimization module. These modules can cover the entire lifecycle safety management of membrane bag sand embankments from the design phase to the construction phase, realizing closed-loop management from scheme design to construction control.

[0062] The data collection module is used to collect actual engineering data, simulation data, and indoor scaled physical model test data for model training and verification. At the same time, the data collection module is also used to collect data on the design scheme of the membrane bag sand embankment to be predicted and controlled for scheme evaluation and optimization.

[0063] The damage mode identification module is used to identify the damage modes of the design or optimized schemes using a classification model. During the design phase, damage mode identification is performed on the initial design scheme; during the construction phase, damage mode identification is performed on the optimized scheme. Damage modes include overall damage and partial damage, etc. In some embodiments, a random forest classifier is used as the classification model.

[0064] The safety factor prediction module in the design phase is used to predict the ultimate fill height H of the design scheme through a regression prediction model. max With safety factor F s Different failure modes are input into different regression prediction models. In some embodiments, the regression prediction models all adopt the XGBoost regression model structure.

[0065] The construction phase safety factor prediction module is used to predict the safety factor F of the optimized scheme through a time series model. sp Different failure modes are input into different time series models. In some embodiments, the time series models all adopt an XGBoost-LSTM coupled model structure.

[0066] The foundation soil inversion module is used to obtain foundation soil parameters from construction monitoring data through an inversion model. In some embodiments, the inversion model adopts a Bayesian optimized BP neural network (BO-BP) structure.

[0067] The scheme optimization module is used to optimize the design scheme or optimized scheme (design scheme or optimized scheme) through an optimization model when the safety factor of the scheme exceeds a set value. It optimizes the initial design scheme during the design phase and continuously optimizes the optimized scheme (including optimized schemes from the design and construction phases) during construction settlement. In some embodiments, the optimization model is a composite model, which includes the NSGA-Ⅲ optimization algorithm, the classification model, and the regression prediction model.

[0068] The above detailed description is a specific description of feasible embodiments of the present invention. These embodiments are not intended to limit the patent scope of the present invention. All equivalent implementations or modifications that do not depart from the present invention should be included in the patent scope of this case.

Claims

1. A machine learning-based method for full-cycle safety management of membrane bag sand embankment slopes, characterized in that, Includes the following steps: S1. During the design phase: S1-1. Based on the initial design scheme, the failure mode is determined by a classification model. The design scheme includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment cross section. S1-2. Based on the failure mode obtained in step S1-1, select the corresponding regression prediction model, and use the corresponding regression prediction model to predict the limit filling height H. max and safety factor F s ; S1-3, By predicting the safety factor F s With the safety factor set threshold [F] s In comparison, it is determined whether the design scheme is feasible. If it is feasible, the design scheme is the optimal scheme. If it is not feasible, the optimal scheme is generated through the optimization model. The optimal scheme includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment section. S2. During the construction phase: S2-1. Construction monitoring data is used to obtain new foundation soil parameters through an inversion model. The original foundation soil parameters are replaced with the new foundation soil parameters to update the optimal solution obtained in step S1. S2-2. The new optimal solution and new construction monitoring data are used to determine the failure mode through a classification model; S2-3. Select the corresponding time series model based on the failure mode, and predict the safety factor F using the corresponding time series model. sp ; S2-4, Based on the predicted safety factor F sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, no warning is needed and continuous monitoring is performed. If the safety risk level exceeds the set safety range, an warning is issued and an optimal solution is generated through an optimization model. The optimal solution includes the design parameters of the membrane bag sand embankment and the geometric parameters of the embankment section. S2-5. During the construction process, following the methods in steps S2-2 to S2-4, conduct regular or irregular safety assessments and optimize the plan until the construction is completed.

2. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The design parameters for the geomembrane bag sand embankment slope include the tensile stiffness J of the geomembrane bag and the undrained shear strength S of the soft soil. u The shear strength τ of the contact surface between the geomembrane bag and the foundation, the thickness t of the sand filling in the geomembrane bag, the internal friction angle φ of the filling sand, and the unit weight γ of the foundation soil. c 'and membrane bag filling sand density γ s The geometric parameters of the embankment slope section include the embankment slope width W, the embankment slope height H, and the embankment slope k; the parameters of the foundation soil include the undrained shear strength S of soft soil. u The construction monitoring data includes displacement data.

3. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The classification model uses a random forest classifier, and its output destruction modes include overall destruction and partial destruction. The training method for the classification model is as follows: Data is collected, and the data sources include actual engineering data, simulation data, and indoor scaled physical model test data, including data under the condition of remolded sludge. The data was randomly divided into a training set and a test set in a 7:3 ratio. The hyperparameters were optimized using 5-fold cross-validation, with 100 decision trees and a maximum depth of 8 for each tree.

4. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The regression prediction model adopts the XGBoost regression model structure, and the training method of the regression prediction model is as follows: The training set data sources include simulation data and indoor scaled physical model test data, which includes ultimate bearing capacity and safety factor data derived from the embankment load-settlement curve. The test set data includes actual engineering data; The hyperparameters were optimized using a grid search with a learning rate of 0.1, a tree depth of 6, and 200 iterations.

5. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The optimization model includes the NSGA-III optimization algorithm, the classification model, and the regression prediction model; the working process of the optimization model is as follows: Given a clear optimization objective and constraints, candidate solutions are generated using the NSGA-Ⅲ optimization algorithm. Candidate solutions determine the failure mode using the classification model, select the corresponding regression prediction model based on the failure mode, and obtain the safety factor F through the corresponding regression prediction model. s ; Based on the safety coefficient Fs of each candidate scheme, feasible schemes are selected; Taking into account both economic and security factors, the optimal solution is selected from the feasible options.

6. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 5, characterized in that, The safety factor F of each candidate scheme s The method for selecting feasible solutions is as follows: If it is in the design phase, the safety factor F of the candidate scheme will be... s With the safety factor set threshold [F] s If Fs ≥ [Fs], then the candidate solution is a feasible solution; otherwise, it is an infeasible solution. If it is during the construction phase, the safety factor F of the candidate scheme will be... sp With the safety factor set threshold [F] s In comparison, the safety risk level is determined. If the safety risk level is within the set safety range, it is a feasible solution; otherwise, it is an infeasible solution. The constraints for the design phase are W∈30-120m, J∈140-8000kN / m, S u For parameters ≥5kPa, the corresponding constraint during the construction phase is that the adjustment range is ≤20% of the original parameters.

7. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 5, characterized in that, The training method of the NSGA-Ⅲ optimization algorithm is as follows: 2000 feasible schemes generated by simulation numerical simulation are used as the initial population, and the crossover probability is set to 0.8, the mutation probability is set to 0.1, and the number of iterations is set to 50.

8. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The inversion model employs a Bayesian-optimized BP neural network. The input data is the monitored displacement data, and the output data is the undrained shear strength S of the soft soil. u ; The method for training the inversion model is as follows: The training set data sources include actual engineering data and simulation data, wherein the simulation data includes S u Simulated input parameter-displacement response data with values ​​ranging from 2.5 to 15 kPa; The optimal initial weights and structure of the BP neural network are searched using Bayesian optimization, including the number of hidden layers and the number of neurons per layer.

9. The method for full-cycle safety management of membrane bag sand embankment slopes based on machine learning according to claim 1, characterized in that, The time series model adopts an XGBoost-LSTM coupled model, which includes an XGBoost module, an LSTM module, and a feature fusion module. The XGBoost module is used to extract static nonlinear features from the static data in the optimal solution. The LSTM module is used to extract time series features from the construction monitoring data. The feature fusion module is used to concatenate the static nonlinear feature vector output by the XGBoost module and the time series feature vector output by the LSTM module, and complete the feature fusion through a fully connected network to obtain a joint feature representation. The training method for the time series model is as follows: Data sources include construction monitoring time-series data from actual projects, with each data point containing a timestamp, horizontal displacement, vertical displacement, and cumulative settlement. Data preprocessing: Time series samples were constructed using the sliding window method, missing values ​​were filled in using linear interpolation, and trend noise was eliminated using the difference method; Training and validation: The data is divided into training and testing sets in an 8:2 ratio, and hyperparameters are optimized, including hidden layer dimension, number of iterations, and batch size.

10. A machine learning-based full-cycle safety management system for membrane bag sand embankment slopes, used to implement the machine learning-based full-cycle safety management method for membrane bag sand embankment slopes as described in any one of claims 1 to 9, characterized in that, include: The data collection module is used to collect actual engineering data, simulation data, and indoor scaled physical model test data for model training and verification, as well as scheme evaluation and optimization. The failure mode identification module is used to identify the failure mode of the design or optimization scheme through a classification model; The safety factor prediction module in the design phase is used to predict the ultimate fill height H of the design scheme through a regression prediction model. max With safety factor F s ; The construction phase safety factor prediction module is used to predict the safety factor F of the optimized scheme through a time series model. sp ; The foundation soil inversion module is used to obtain foundation soil parameters based on construction monitoring data through an inversion model. The scheme optimization module is used to optimize the design scheme or optimization scheme through the optimization model when the safety factor of the scheme exceeds the set limit. It optimizes the initial design scheme in the design stage and continuously optimizes the optimized scheme in the construction stage.

Images