Intelligent stability collaborative control system for complex stratum roadbed

Abstract

The application discloses a complex stratum roadbed intelligent stability collaborative control system, and belongs to the technical field of geological engineering. The system comprises a multi-source sensing network building unit, which is used for arranging sensing nodes in a roadbed cross section, longitudinal section and depth layer position and collecting roadbed state data and historical disturbance event data; a response feature extraction unit, which is used for carrying out hierarchical, zonal and time period processing on the roadbed state data to form a multi-dimensional structure response feature set; and a response fingerprint construction unit, which is used for decomposing the multi-dimensional structure response feature set into response components corresponding to different instability mechanisms. The complex stratum roadbed intelligent stability collaborative control system carries out multi-dimensional structure feature extraction on roadbed state data, decomposes the multi-dimensional structure feature into response components of different instability mechanisms, constructs response decoupling fingerprints, and provides important support for in-depth analysis and accurate prediction of instability mechanisms.

Description

Technical Field

[0001] This invention relates to the field of geological engineering, and more specifically, to an intelligent stability and collaborative control system for roadbeds in complex strata. Background Technology

[0002] Roadbeds are a crucial foundational structure in transportation engineering, and their stability directly impacts traffic safety and the lifespan of the project. In complex geological environments, roadbeds may experience various types of instability phenomena, such as settlement, landslides, and cracking, due to factors such as soil physical properties, groundwater variations, and climate change. Therefore, real-time monitoring of roadbed condition changes and the identification and control of instability mechanisms through intelligent methods have become a research hotspot in modern civil engineering.

[0003] Currently, roadbed stability assessment typically relies on traditional manual monitoring and simple physical models. These methods suffer from incomplete monitoring, poor real-time performance, and insufficient emergency response capabilities. With the rapid development of technologies such as the Internet of Things, artificial intelligence, and big data, sensor network-based data acquisition and intelligent analysis methods are gradually being applied to roadbed monitoring. However, most existing systems are limited to monitoring and analysis of single data sources, lacking multi-source data fusion and intelligent collaborative control, making it difficult to effectively cope with complex geological disturbances and changing environmental conditions. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent stability collaborative control system for roadbeds in complex strata. This system solves the problem that the stability assessment of existing roadbeds usually relies on traditional manual monitoring and simple physical models, which are mostly limited to monitoring and analysis of a single data source. They lack multi-source data fusion and intelligent collaborative control, making it difficult to effectively cope with complex strata disturbances and changing environmental conditions, and thus failing to meet the application requirements.

[0005] This invention achieves the above objectives through the following technical solution: an intelligent stability and collaborative control system for roadbeds in complex geological formations, the system comprising: Multi-source sensing network building unit, response feature extraction unit, response fingerprint construction unit, perturbation memory field construction unit, coupling analysis and judgment unit, cooperative strategy generation unit, and dynamic update unit; Multi-source sensing network building unit is used to deploy sensing nodes in the cross section, longitudinal segment and depth layer of the roadbed and collect roadbed status data and historical disturbance event data. The response feature extraction unit is used to perform hierarchical, partitioned, and time-segmented processing on the roadbed state data to form a multi-dimensional structural response feature set. The response fingerprint construction unit is used to decompose the multidimensional structural response feature set into response components corresponding to different instability mechanisms, and assign each response component a response intensity value, influence range value, development rate value and dominant probability value to form a stratigraphic structure response decoupling fingerprint. The perturbation memory field construction unit is used to classify and encode the historical perturbation event data and generate memory parameters that characterize the residual effects to form a perturbation memory field. The coupling analysis and determination unit is used to couple the decoupled fingerprint of the geological structure response with the disturbance memory field and output the dominant instability source type, instability evolution stage and main control risk area. The collaborative strategy generation unit is used to generate partitioned, hierarchical, and phased collaborative control strategies based on the coupling analysis results. The dynamic update unit is used to introduce newly acquired data after control execution to update the formation structure response decoupling fingerprint and the disturbance memory field, and to adjust the cooperative control strategy according to the update results.

[0006] Furthermore, the multi-source sensing network building unit includes: Hierarchical deployment sub-unit, partition mapping sub-unit, and database creation sub-unit; The layered deployment subunit is used to deploy settlement, displacement, pore pressure, water migration and load response sensing nodes around the surface layer, deep layer and interlayer interface; The partition mapping subunit is used to establish spatial mapping relationships according to cross-sectional functional areas and longitudinal construction areas; The data database building subunit is used to perform standardization, noise reduction, anomaly removal, missing data completion, and classification storage on the collected data to form a current state database and a historical event database. It also supports fast retrieval of time series and regional indexes as well as data calls for subsequent model training.

[0007] Furthermore, the response feature extraction unit includes the following steps: Vertical compression response features, lateral displacement expansion features, water content change sensitivity features, pore pressure release or accumulation features, local stiffness difference features, deformation diffusion path features, and hysteresis recovery features after load application are extracted from the roadbed state data. A one-to-one correspondence is established according to the roadbed spatial layer, structural zoning, and monitoring time scale to form a multidimensional structural response feature set characterizing the current overall state of the roadbed. The monitoring timescale includes at least hourly, daily, and weekly levels, and different timescales correspond to different response evolution interpretation windows.

[0008] Furthermore, the current state database and the historical event database are jointly connected to the model training and optimization unit. The model training and optimization unit is used to train, verify, test and optimize the roadbed instability mechanism identification and coupling analysis model based on historical monitoring data, disturbance event data and instability judgment labels, and send the optimized model parameters to the response feature extraction unit, response fingerprint construction unit, disturbance memory field construction unit and coupling analysis judgment unit to support online inference. The instability determination label includes at least the dominant instability source, instability stage, and main control risk area, and is used to constrain the consistency between the model output and the actual operating conditions, and to support the feedback correction of subsequent control results.

[0009] Furthermore, the response fingerprint construction unit includes the following steps: The multidimensional structural response feature set is decomposed into compaction-dominant response components, shear slip-dominant response components, moisture-induced softening-dominant response components, and load disturbance amplification response components. The response intensity value, influence range value, development rate value, and dominant probability value of each response component are calculated respectively. Then, a stratigraphic structural response decoupling fingerprint composed of response components and their corresponding parameters is constructed to characterize the mechanism composition and evolution trend of the current anomalous state. Each response component and its corresponding parameter are stored in association with the region location and depth layer, and are used to distinguish between the dominant anomaly region and the secondary anomaly region.

[0010] Furthermore, the perturbation memory field construction unit includes the following steps: The disturbances caused by construction filling, temporary excavation, rainfall infiltration, heavy traffic, changes in drainage conditions, preloading or unloading, and reinforcement are classified and coded. For each disturbance event, memory parameters are generated corresponding to the occurrence time, area of ​​effect, depth of influence, residual influence intensity, decay rate, spatial propagation direction, and superposition relationship, thereby forming a regionalized disturbance memory field that evolves over time. The classification coding consists of perturbation category coding and perturbation detail coding, and is used to achieve a unified representation of heterogeneous perturbation events.

[0011] Furthermore, the perturbation memory field construction unit also includes the following steps: The attenuation rate is corrected according to the disturbance type, subgrade soil type and disturbance depth, and the current residual influence intensity is calculated based on the corrected attenuation rate and the initial influence intensity of the disturbance. The roadbed soil types include at least cohesive soil, sandy soil, and gravelly soil. The current residual influence intensity is used as the historical residual effect input when the coupled analysis judgment unit performs coupled calculations. The correction process simultaneously considers the proportional relationship between the disturbance effect depth and the critical depth, as well as the difference in residual effect decay rate between different soil types.

[0012] Furthermore, the coupling analysis and determination unit is used to first establish a matching pair between response components and disturbance events based on the correspondence between instability mechanisms and disturbance types, and then perform weighted matching calculations on the dominant probability value of each response component and the current residual influence intensity of the corresponding disturbance event to obtain risk values ​​for different instability types. The coupling analysis and determination unit is also used to determine the dominant instability source type according to the magnitude of the risk value, and to determine the initial instability stage, evolution amplification stage and sensitive recovery stage according to the preset threshold range. At the same time, it identifies whether the anomaly is triggered by the release of the residual effect of historical disturbance, locates the spatial range of the main control risk area, and outputs the risk ranking result.

[0013] Furthermore, the collaborative strategy generation unit includes the following steps: Based on the dominant instability source type, instability evolution stage, main control risk area, and disturbance residual characteristics, a collaborative control strategy is generated, which includes drainage intensity adjustment, drainage path switching, local unloading or load limiting, grouting area and grouting sequence adjustment, reinforcement and strengthening area redivision, lateral support intensity adjustment, preloading rhythm adjustment, construction machinery operation area and timing rearrangement, traffic diversion or load restriction, and risk area isolation control. The strategy also outputs the combination of control actions to be executed with the highest priority and the types of control actions that should not be applied. The execution intensity of each control action is matched according to the risk level, and the implementation order is allocated according to the main control risk area and the non-main control risk area.

[0014] Furthermore, the dynamic update unit is used to continuously receive new monitoring data after the collaborative control strategy is executed, update the dominant probability value, influence range value and development rate value in the stratigraphic structure response decoupling fingerprint, and simultaneously update the residual influence intensity and decay rate in the disturbance memory field. When the updated risk value decreases and the dominant response component weakens, maintain the current control strategy; When the updated risk value undergoes spatial shift or a new dominant instability mechanism is formed, the collaborative control boundary is redefined and the collaborative strategy generation unit is triggered to output a new partitioned control strategy. At the same time, the control data of this round is written into the historical database to perform incremental optimization, thereby forming a closed-loop adaptive control mechanism for roadbeds with complex strata and improving the consistency of subsequent coupling judgment and strategy generation.

[0015] The beneficial effects of this invention are as follows: 1. By utilizing multi-source sensing technology to deploy sensor nodes at different depths and locations, comprehensive roadbed status data and historical disturbance event data are collected, providing more accurate and comprehensive roadbed status information than traditional monitoring methods.

[0016] 2. By extracting multidimensional structural features from the roadbed state data and decomposing it into response components of different instability mechanisms, a response decoupling fingerprint is constructed, providing important support for in-depth analysis and accurate prediction of instability mechanisms.

[0017] 3. By constructing a disturbance memory field, the residual effects of historical disturbance events can be dynamically tracked. Combined with formation response decoupling fingerprint for coupling analysis, the dominant instability source and evolution stage can be identified in real time, providing a more accurate prediction of instability risk.

[0018] 4. Based on the coupling analysis results, the system can generate coordinated control strategies for different instability risks at different zones, levels, and stages, and adjust control measures in real time to ensure the stability of the roadbed. The dynamic update unit continuously receives new data after control is executed to ensure the effectiveness and timeliness of the control strategy.

[0019] 5. By continuously updating the formation structure response decoupling fingerprint and disturbance memory field, the system can automatically adjust the control strategy with the support of new monitoring data, realize closed-loop adaptive control, and effectively cope with complex formations and changing environmental conditions. Attached Figure Description

[0020] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a flowchart illustrating the overall system flow of the present invention. Figure 2 This is an internal flowchart of the multi-source sensing network building unit of the present invention; Figure 3 This is an internal flowchart of the coupling analysis and determination unit of the present invention. Detailed Implementation

[0021] The present application will now be described in further detail with reference to the accompanying drawings. It should be noted that the following specific embodiments are only used to further illustrate the present application and should not be construed as limiting the scope of protection of the present application. Those skilled in the art can make some non-essential improvements and adjustments to the present application based on the above application content.

[0022] Example 1: Please see Figure 1-3 This invention provides a technical solution: an intelligent stability and collaborative control system for roadbeds in complex geological formations, the system comprising: The multi-source sensing network building unit is used to establish a roadbed multi-source sensing network. Multi-source sensing nodes are deployed in the roadbed cross section, longitudinal segment and depth layer to collect roadbed status data and historical event data, and to establish a current status database and a historical event database. Among them, multi-source sensing nodes are equipment points deployed at different locations on the roadbed, including cross-sections, longitudinal segments, and depth layers, used to collect data. They can collect various types of data to provide basic information for subsequent analysis; roadbed status data reflects the current physical and mechanical characteristics of the roadbed, such as settlement, displacement, and stress data; historical event data contains information related to various events that have occurred on the roadbed in the past, which may include records of events that have affected the roadbed, such as construction events and natural disasters; the current status database stores the current roadbed status data collected by the multi-source sensing nodes, used to understand the current condition of the roadbed in real time; and the historical event database stores historical event data, providing a basis for analyzing the historical conditions of the roadbed and predicting future trends. The model training and optimization unit is used to build and train a roadbed instability mechanism identification and coupling analysis model. Based on historical monitoring data and disturbance event data, it completes the training, verification and optimization of the model and determines the optimal parameters of the model. Among them, the subgrade instability mechanism identification and coupling analysis model is a model constructed through mathematical algorithms and logical relationships, aiming to identify the intrinsic mechanism of subgrade instability and analyze the coupling relationship between multiple factors; historical monitoring data, data obtained from past monitoring of the subgrade, is used for model training to enable it to learn the normal and abnormal state characteristics of the subgrade; disturbance event data, data related to events that cause disturbances to the subgrade, such as earthquakes, sudden changes in vehicle load, etc., are used for model training to enhance its ability to identify and analyze disturbance factors; optimal model parameters, the parameter combination determined after continuous adjustment and optimization during model training to achieve the best model performance, ensuring that the model can accurately identify the subgrade instability mechanism and perform coupling analysis; The response feature extraction unit is used to extract multi-scale structural response features. The current monitoring data is input into the trained model, which processes the data by layer, partition and time period to extract the multi-dimensional structural response features of the roadbed and map them to the spatial layer and structural partition to form a multi-dimensional feature set. Among them, multi-scale structural response characteristics refer to the response characteristics of the subgrade structure under the influence of load, hydrological changes, construction disturbances, and stratum differences at the overall structural scale, local structural scale, and microstructural scale. These characteristics include displacement, strain, pore water pressure changes, water content migration changes, and vibration response characteristics. Layered, zoned, and time-series processing are used to process the input current monitoring data according to the depth layer, spatial region, and time series of the subgrade, so as to analyze the subgrade response characteristics in more detail. The multi-dimensional feature set is a collection formed by mapping the extracted multi-dimensional structural response characteristics of the subgrade according to spatial layer and structural zone, which comprehensively reflects the response of the subgrade in different dimensions. The response fingerprint construction unit is used to construct the response decoupling fingerprint of the geological structure. The model uses a multi-dimensional feature set to perform mechanism decomposition on the current roadbed anomaly state, obtains response components containing independent mechanism attributes, and assigns feature parameters to each response component to form the response decoupling fingerprint of the geological structure. Among them, the decoupled fingerprint of the geological structure response is obtained by mechanistic decomposition of the multidimensional feature set to obtain response components containing independent mechanistic attributes, and the fingerprint is formed by assigning characteristic parameters to each response component. It is used to accurately describe the response characteristics of the geological structure under different mechanistic effects. Mechanism decomposition decomposes the complex subgrade response into multiple response components with independent mechanistic attributes, so as to analyze the influence of different factors on the subgrade response separately. Characteristic parameters are parameters used to describe the characteristics of each response component, such as the amplitude, frequency, and phase of the response, so that the response components can be accurately quantified and distinguished. The perturbation memory field construction unit is used to construct the perturbation memory field. The model classifies and encodes various perturbation events based on the historical event database, extracts the memory parameters of each perturbation event, and establishes perturbation memory fields that evolve over time for different areas of the roadbed. The disturbance memory field, based on a historical event database, classifies and encodes various disturbance events. After extracting the memory parameters of each disturbance event, it establishes a field that evolves over time for different areas of the roadbed. This field is used to record and reflect the degree and historical status of the impact of different disturbance events on different areas of the roadbed. The classification and coding process categorizes various disturbance events according to certain rules and assigns them unique codes to facilitate systematic management and analysis of disturbance events. The memory parameters are used to describe the impact of disturbance events on the roadbed, such as the intensity, duration, and range of the disturbance event. These parameters are extracted to construct the disturbance memory field. The coupling analysis and judgment unit is used to perform coupling analysis of decoupled fingerprint and disturbance memory field. The model couples the current geological structure response decoupled fingerprint with the disturbance memory field of the corresponding area to obtain the subgrade instability judgment result. Among them, the coupling calculation performs mathematical calculations and analyses on the decoupled fingerprint of the current geological structure response and the disturbance memory field of the corresponding region to determine their interrelationship and degree of influence; the roadbed instability judgment result, based on the results of the coupling calculation, determines whether the roadbed is in an unstable state and the degree of instability, providing a basis for the generation of subsequent control strategies; The collaborative strategy generation unit is used to generate collaborative control strategies. Based on the coupling analysis results, it generates collaborative control strategies in the form of partitions, levels, and stages. Among them, the coordinated control strategy is a zoned, graded, and phased control measure formulated for roadbed instability. It ensures the stability of the roadbed through the coordinated action of multiple aspects. Zoned means to formulate different strategies according to the characteristics of different areas of the roadbed; graded means to formulate control measures of different intensities according to the degree of instability; phased means to implement the control strategy step by step in chronological order. The dynamic update unit is used to execute control and dynamically update. After executing the collaborative control strategy, it continuously collects new monitoring data, inputs the new data into the model to complete the update of the fingerprint and memory field, analyzes the response changes after control and adjusts the control strategy to form a closed-loop collaborative control. The process includes: executing a collaborative control strategy, which involves performing actual control operations on the roadbed according to the generated strategy, such as reinforcement and drainage; new data, which is continuously collected by multi-source sensing nodes after the control strategy is executed, reflecting the new state and response of the roadbed; updating the fingerprint and memory field, which involves inputting the newly collected data into the model to update the decoupling fingerprint and disturbance memory field of the geological structure response, so that it can reflect the latest state and historical conditions of the roadbed; and closed-loop collaborative control, which involves continuously collecting new data, updating the fingerprint and memory field, analyzing the response changes after control, and adjusting the control strategy to form a continuously cyclical and self-optimizing control process to ensure that the roadbed is always in a stable state.

[0023] It should be noted that during use, the multi-source sensing network building unit comprehensively collects data, providing a rich information foundation for the system and facilitating accurate understanding of the roadbed condition. The model training and optimization unit ensures that the model can accurately identify the roadbed instability mechanism and analyze coupling relationships, improving the accuracy of judgment. The response feature extraction and fingerprint construction unit extracts features from multiple scales and constructs decoupled fingerprints, which can clearly analyze the internal mechanism of abnormal roadbed states. The disturbance memory field construction unit records historical disturbances, providing historical evidence for analysis. The coupling analysis and judgment unit comprehensively considers multiple factors to arrive at an instability judgment result that is scientific and reasonable. The collaborative strategy generation unit formulates zoning, hierarchical, and phased strategies to enhance the targeting and effectiveness of control. The dynamic update unit forms a closed-loop control, continuously optimizing strategies so that the system can adapt to roadbed changes in real time, ensuring the long-term stability of roadbeds in complex geological formations and reducing instability risks and maintenance costs.

[0024] In one embodiment, the subgrade status data consists of real-time dynamic monitoring data of the subgrade at different spatial locations and depths, including at least the surface and deep settlement of the subgrade, horizontal and lateral displacement gradients, pore water pressure changes, soil water migration information, layered compression deformation information, interlayer faulting information, and load response information of different areas of the subgrade. The historical event data consists of records of various disturbances generated during the subgrade construction and operation periods, including at least historical construction disturbance records, rainfall records, traffic load records, and reinforcement records. After standardization, noise reduction, and completion preprocessing, the two types of data are respectively entered into the current status database and the historical event database to achieve classified storage and rapid retrieval of data. The monitoring thresholds for physical quantities such as settlement and displacement are the safety limits specified in the subgrade design documents, which are comprehensively determined by the subgrade soil and rock parameters, design bearing capacity, and engineering usage requirements. The outlier judgment threshold for data preprocessing is three times the standard deviation of the mean, and data exceeding the threshold are completed using interpolation.

[0025] This design specifies and preprocesses roadbed condition and historical event data in detail, then categorizes and stores them. Comprehensive roadbed condition data covers multi-dimensional information and can accurately reflect the real-time condition of the roadbed. Historical event data provides historical basis for analysis. Standardization, noise reduction, and completion preprocessing can improve data quality and ensure data accuracy and completeness. Categorized storage and rapid retrieval facilitate subsequent model training and analysis, enabling the system to quickly obtain roadbed information based on reliable data. This provides a solid foundation for accurately identifying instability mechanisms, constructing decoupled fingerprints and perturbation memory fields, and improving the accuracy and timeliness of the entire system's judgment of roadbed condition.

[0026] In one embodiment, the roadbed instability mechanism identification and coupling analysis model is a deep learning model incorporating an attention mechanism. The model's basic architecture adopts a combination of Transformer and fully connected layers. The input includes a structural response feature branch and a disturbance event feature branch, and the output includes three sub-modules: decoupled fingerprint generation, disturbance memory field construction, and coupling risk calculation. The specific steps for model training are as follows: Dataset construction and partitioning: Structural response feature data, disturbance event data and corresponding roadbed instability judgment labels were extracted from historical databases and divided into training set, validation set and test set in a 7:2:1 ratio. The labels were marked by engineering experts based on actual instability cases and included information such as the dominant instability source, instability stage and main risk area. For model initialization, the Transformer encoder is set to have 8 multi-head attention heads, the hidden layer dimension is 512, the number of hidden nodes in the fully connected layer is 256, 128, and 64 respectively, the activation function is ReLU, the output layer activation function is Sigmoid, the weights are initialized using Xavier normal distribution, and the bias term is initialized to 0. Model training involves inputting the training set data into the model to initialize it, using the mean squared error. Let be the loss function, and its expression is:

[0027] in, The actual label value. These are the model's predicted values. The sample size is 0.001. The Adam optimizer is used to update the parameters. The initial learning rate is set to 0.001, the weight decay coefficient is set to 0.0001, the batch size is set to 32, and the number of training epochs is set to 200. Model validation and early stopping: Every 5 training rounds, the validation set data is input into the model, and the loss value and accuracy of the validation set are calculated. If the loss value of the validation set does not decrease for 20 consecutive rounds, the early stopping mechanism is triggered to terminate the training and avoid model overfitting. Model testing and optimization involves inputting test set data into the trained model and calculating its precision, recall, and accuracy. Value, accuracy Recall rate F1 value ,in, True positive False positive It is a false negative; If the test set value If necessary, adjust the number of attention heads or the dimension of hidden layers and retrain until... value Determine and save the optimal parameters of the model; Model deployment involves deploying the trained, optimal model to the roadbed intelligent control terminal, supporting real-time online inference and computation of data, and controlling the inference latency threshold. .

[0028] This design employs a deep learning model with an integrated attention mechanism and specifies detailed training steps. The integrated attention mechanism can focus on key information and improve the model's ability to process complex data. The detailed training steps, from dataset partitioning to model deployment, ensure that the model training is scientific and reasonable. Through rigorous verification, testing, and optimization, an accurate and reliable model can be obtained. When facing the problem of roadbed instability in complex strata, it can accurately identify the mechanism, construct decoupled fingerprints and perturbation memory fields, and provide strong support for subsequent coupling analysis and control strategy generation. This improves the system's accuracy in judging roadbed instability and its ability to respond.

[0029] In one embodiment, the model performs layered processing of the current monitoring data according to the layered structure of the roadbed soil and rock, zoning processing according to the functional zoning of the roadbed cross section and longitudinal segment, and time-segmented processing according to the monitoring time scale of hours / days / weeks. The extracted multidimensional structural response features are the core features that can characterize the changes in the mechanical and hydrological properties of the roadbed structure, including at least vertical compression response features, lateral displacement expansion features, water content change sensitivity features, pore pressure release or accumulation features, local stiffness difference features, deformation diffusion path features, and hysteresis recovery features after load application. Each feature is mapped one-to-one with the corresponding roadbed spatial layer and structural partition, forming a multi-dimensional feature set that can comprehensively characterize the current overall state of the roadbed. The anomaly thresholds for each feature were calibrated using indoor geotechnical tests, in-situ test data, and the "Highway Subgrade Design Specifications." The time scale thresholds for feature extraction were set based on the subgrade instability evolution law: hourly thresholds were used to capture sudden disturbance responses, while daily / weekly thresholds were used to analyze long-term evolution trends. The feature importance of the extracted features was determined using the random forest method, and the feature importance thresholds were set accordingly. The features are incorporated into the multidimensional feature set.

[0030] This design specifies the model's stratification, zoning, and time-segmentation processing of current monitoring data, as well as the feature extraction method. Stratification is based on soil and rock structure, zoning is based on functional zoning, and time-segmentation is based on monitoring scale. This allows for a comprehensive and detailed capture of changes in roadbed condition. The extracted multidimensional structural response features can accurately characterize changes in roadbed mechanical and hydrological properties. Multidimensional feature sets are formed by spatial and zonal mapping to comprehensively reflect the overall condition of the roadbed. Reasonable setting of anomaly judgment and time scale thresholds can promptly capture sudden and long-term changes. The random forest method is used to determine the importance of features, ensuring the effectiveness of the included features and improving the accuracy and reliability of model analysis.

[0031] In one embodiment, the response components output by the model, which contain independent mechanistic attributes, are characteristic components that can reflect different disaster-causing mechanisms of subgrade instability. These include at least a compaction-dominant component, a shear slip-dominant component, a moisture-induced softening-dominant response component, and a load disturbance amplification response component. Each component is independent and can completely cover the main instability types of subgrades in complex strata. The characteristic parameters assigned to each response component are core parameters for quantifying the component's state, including response intensity values. Scope of influence Development rate value and dominant probability value Each feature parameter is calculated by the model through a fully connected layer, and the calculation method is as follows: Response intensity value:

[0032] in, The measured monitoring value of the response component, This serves as the benchmark reference value for the response component under normal and stable roadbed conditions, determined through statistical analysis of measured data from similar projects and numerical simulation. The engineering safety threshold for this response component is determined by adjusting it according to the "Railway Subgrade Design Code" and the actual working conditions of the project. The range of values ​​is The larger the value, the higher the degree of abnormality of the component; Among them, the dominant response component of moisture-induced softening refers to the response component corresponding to the decrease in strength, decrease in stiffness, and increase in deformation sensitivity of the subgrade soil caused by the increase in water content, changes in pore water pressure, rainfall infiltration, or changes in groundwater conditions.

[0033] Impact range value:

[0034] in, The area / volume of the roadbed region exhibiting an anomalous state in this response component. To correspond to the overall monitoring area / volume of the roadbed, The range of values ​​is The larger the value, the wider the scope of the abnormal influence of that component. The anomaly detection threshold is... That is, the scope of the abnormal impact exceeds the monitoring area. When the condition is met, it is determined to be a regional anomaly; Development rate value:

[0035] in, For this response component in time Changes in monitored values ​​within the area The positive and negative values ​​represent the development trend of the component; the larger the absolute value, the faster the component evolves. The danger threshold is set according to the roadbed instability warning level, with the first-level warning threshold being... Exceeding the normal evolution rate by 3 times or more; Dominant probability value:

[0036] in, For the first The response intensity value of each response component. The total number of response components, and , The range of values ​​is The larger the value, the higher the dominance of the component in the current subgrade instability state. The threshold for determining the dominant component is... That is, the dominant probability of a certain component exceeds When this occurs, it is determined to be the dominant mechanism component of the current instability; Error thresholds calculated by the model for each parameter The accuracy level of the monitoring data and the model fit are determined by the accuracy level of the monitoring data. When the error threshold is exceeded, the model automatically triggers a data re-acquisition mechanism.

[0037] This design specifies the response components and assigned characteristic parameters of the model output. The response components are independent of each other and cover the main instability types, accurately reflecting different disaster-causing mechanisms. The assigned characteristic parameters quantify the component states and are obtained through specific calculation methods with a small calculation error threshold, ensuring parameter accuracy. These parameters provide key data for subsequent construction of decoupled fingerprints and analysis of instability mechanisms, enabling the system to accurately quantify the roadbed instability state, clarify the contribution of each mechanism, provide a scientific basis for formulating targeted control strategies, and improve the ability to cope with roadbed instability.

[0038] In one embodiment, the decoupled fingerprint of the geological structure response is a set of features output by the model that uniquely characterizes the current subgrade instability mechanism, consisting of each response component and its corresponding response intensity value. Scope of influence Development rate value Dominant probability value It is constructed in a one-to-one correspondence, and its expression is:

[0039] in, For the first Each response component The fingerprint, represented by a positive integer, is output by the decoupled fingerprint generation submodule of the model. It enables the mechanistic decomposition and quantitative characterization of the current abnormal state of the roadbed, providing data support for subsequent instability source identification. The fingerprint validity threshold is the calculation error of each feature parameter. Furthermore, the fingerprint features matched the historical instability cases. The matching degree is calculated from the cosine similarity.

[0040] This design defines the composition and validity determination of the decoupled fingerprint of the geological structure response, which consists of response components and corresponding characteristic parameters. It can uniquely characterize the current subgrade instability mechanism, realize the mechanism decomposition and quantitative characterization of abnormal states, and set a validity determination threshold to ensure that the fingerprint is accurate and reliable, with a high degree of matching with historical instability cases. It can provide accurate data support for subsequent instability source identification. This helps the system to clearly understand the internal mechanism of subgrade instability, accurately determine the instability type, provide key basis for formulating effective control strategies, and improve the system's ability to identify and respond to subgrade instability.

[0041] In one embodiment, the disturbance events generated by the model based on the historical event database are various events that may affect the stability of the roadbed during the roadbed construction and operation periods. These include at least construction filling disturbances, temporary excavation disturbances, heavy rainfall infiltration disturbances, heavy-load traffic disturbances, drainage condition change disturbances, preloading or unloading disturbances, and historical grouting reinforcement disturbances. The model classifies and encodes various disturbance events according to their risk-causing mechanisms and modes of action. The encoding uses an 8-digit code, with the first 2 digits representing the disturbance category code and the last 6 digits representing the disturbance details code, thereby achieving standardized identification of disturbance types. The memory parameters are the core parameters for the model to quantify the residual impact of disturbance events. They include at least the disturbance type, occurrence time, area of ​​effect, depth of influence, residual impact intensity, decay rate, spatial propagation direction, and superposition relationship with other disturbances. Each parameter is obtained by the model's disturbance memory field construction submodule based on the original records in the historical event database through statistical analysis and quantitative calculation. The impact depth threshold of disturbance events is determined based on the effective stress layer depth of the roadbed. Disturbances exceeding the effective stress layer depth are classified as low-impact disturbances and are not included in the calculation of core memory parameters. The model's classification and coding accuracy threshold for disturbance events is also included. .

[0042] This design specifies the disturbance events and memory parameters generated by the model, classifies and encodes various disturbance events to achieve standardized identification, facilitates model processing and analysis, and quantifies the residual impact of disturbance events through detailed calculations. It also sets thresholds for impact depth and classification coding accuracy to ensure data accuracy and validity. These designs enable the model to accurately record and analyze the impact of historical disturbances on the roadbed, providing reliable data for constructing a disturbance memory field. This allows for accurate consideration of historical disturbance factors in coupled analysis, improving the accuracy and comprehensiveness of the system's judgment on roadbed instability.

[0043] In one embodiment, residual influence intensity The intensity of the impact of a disturbance event on the roadbed from the time the event occurred until the current moment, calculated for the model, is as follows:

[0044] in, The initial impact intensity at the beginning of a disturbance event is determined by the model through engineering quantitative analysis based on the scale, intensity, and mode of action of the disturbance event. The range of values ​​is The value is based on the reduction ratio of the roadbed soil strength caused by the disturbance. When the soil strength is reduced... hour, When reduced hour, ; This is the attenuation coefficient for this type of disturbance, characterizing how quickly the disturbance's influence decays over time. It is dynamically calculated by the model in conjunction with soil properties. The time interval from the occurrence of the disturbance event to the current moment; Attenuation coefficient The model undergoes dual correction based on both the disturbance type and the characteristics of the subgrade soil to better reflect the actual engineering properties of subgrades in complex geological formations. The correction formula is as follows:

[0045] in, This is the foundation attenuation coefficient for this type of disturbance under standard soil conditions, obtained by the model based on engineering tests and historical data statistics, such as heavy rainfall infiltration disturbance. Construction excavation disturbance ; This is a correction factor for the subgrade soil type. The model automatically assigns values ​​based on the lithology, density, and water content of the subgrade soil. (For cohesive soil...) , sandy soil gravel soil The values ​​are determined by the model in conjunction with the soil permeability coefficient and shear strength parameters from indoor geotechnical tests; The disturbance depth correction factor is automatically determined by the model based on the ratio of the disturbance depth to the critical depth of the roadbed. hour, , hour, ,ratio hour, ; and The values ​​are all calibrated by the model in combination with engineering experience and measured data, and then corrected. Error in value ; The perturbation memory field construction submodule of the model supports real-time updates of memory parameters, with the update frequency consistent with the monitoring data acquisition frequency, which is adjustable at the minute / hour level.

[0046] This design specifies the calculation and correction methods for residual influence intensity. By calculating the initial influence intensity, attenuation coefficient, and time interval, it can accurately reflect the influence intensity on the roadbed from the occurrence of the disturbance event to the present moment. The attenuation coefficient is double-corrected, taking into account the disturbance type and the characteristics of the roadbed soil, making it more consistent with the actual engineering characteristics. The corrected value has a small error and supports real-time updates of memory parameters to ensure the timeliness of data. This helps the model accurately quantify the residual influence of historical disturbances, more accurately consider historical factors in coupled analysis, and improve the scientificity and accuracy of the system's judgment on roadbed instability.

[0047] In one embodiment, the coupling calculation is performed by the model's coupling risk calculation submodule. This is a quantitative analysis process that deeply integrates the mechanistic characteristics of the decoupling fingerprint of the geological structure response with the historical residual characteristics of the disturbance memory field. Specifically, the method is as follows: The model first establishes a matching relationship between response components and disturbance events based on the correspondence between instability mechanisms and disturbance types. Then, it performs weighted matching calculations between the characteristic parameters of each response component in the decoupled fingerprint of the formation structure response and the memory parameters of the corresponding region and type of disturbance in the disturbance memory field to obtain the risk value of each instability type. The calculation formula is:

[0048] in, For the first The dominant probability value of each response component represents the degree of mechanistic dominance of that component; For the first The matching weights of each response component and its corresponding disturbance are calculated by the model using the Analytic Hierarchy Process (AHP), and their values ​​range from [value range missing]. The basis for determination is the contribution rate of the disturbance to the instability mechanism. hour, Contribution rate hour, Contribution rate hour, Contribution rate hour, Furthermore, the sum of the weights of all matching pairs satisfies the normalization requirement, and the model's consistency ratio threshold for weight calculation is met. ; To characterize the residual effect of historical disturbances in order to reflect the current residual impact strength of the corresponding disturbances; The number of matched logarithms between the response components and the disturbance involved in the coupling; The model calculates the risk values ​​for each type of instability. The instability types are sorted in descending order of their risk values, and the type with the highest risk value is the dominant instability source type. Furthermore, the risk values ​​are determined according to engineering safety standards. The threshold range is divided into three stages of instability evolution: the initial instability stage (…). ), evolutionary amplification stage ( Sensitive recovery phase ( ),in, , The risk threshold set by the model based on engineering experience, roadbed design parameters, and field measurement data is the default value. Take 0.3, Take 0.7, and The threshold correction is based on the safety level of the roadbed. The model supports automatic threshold adjustment according to the roadbed safety level, for a first-level safety level roadbed. Reduced to 0.2 Reduced to 0.6, Level III safety grade roadbed Increased to 0.4 Increased to 0.8.

[0049] This design, which specifies the coupling calculation method and risk value calculation formula, establishes a matching relationship between response components and disturbance events, and performs weighted matching calculation to obtain the risk value. It can deeply integrate the mechanism characteristics of the decoupling fingerprint of the geological structure response with the historical residual characteristics of the disturbance memory field, reasonably set the risk threshold range to divide the instability evolution stage, and support automatic adjustment of the threshold according to the roadbed safety level. This helps the system to accurately determine the current dominant instability source type and instability evolution stage, providing a key basis for generating targeted control strategies and improving the system's ability to cope with and control roadbed instability.

[0050] In one embodiment, the subgrade instability determination result is a comprehensive determination information output after model coupling analysis that can guide the generation of subsequent control strategies. It includes at least the current dominant instability source type, instability evolution stage, whether the anomaly is caused by the release of historical disturbance residual effects, the main control risk area, the type of control action least suitable to be applied, and the combination of control actions to be executed with the highest priority. Among them, the main control risk zone is determined by the model based on the risk value of each region. The judgment threshold is determined by combining the characteristic parameters of the response components with the regional risk value. And the development rate of the dominant component If the risk threshold is exceeded, it is identified as a core control risk area; The least suitable and highest priority control actions are determined by the model based on the characteristics of the instability mechanism and the influence law of disturbance residuals, and the weights are matched accordingly. The control action corresponding to the disturbance is determined as the highest priority action, and the control action that acts in the opposite direction to the instability mechanism is determined as the least appropriate action. The instability determination accuracy threshold of the model is set. .

[0051] This design specifies the content and method of determining roadbed instability. The comprehensive determination information fully covers the key information of the current roadbed instability, providing clear basis for determining the main control risk zone, control action type, etc. The reasonable determination threshold is set to ensure that the determination results are accurate and reliable. This helps the system to accurately understand the roadbed instability status, provides strong support for generating scientific and reasonable collaborative control strategies, makes the control strategies more targeted and effective, and improves the system's control capability and safety for roadbed instability.

[0052] In one embodiment, the collaborative control strategy is a multi-means, linkage control scheme generated by the model based on the coupling analysis results, targeting the current instability characteristics of the roadbed. It includes at least the following: drainage intensity adjustment, drainage path switching, local unloading or load limiting, grouting area and grouting sequence adjustment, reinforcement and strengthening area redivision, lateral support intensity adjustment, preloading rhythm adjustment, construction machinery operation area and timing rearrangement, traffic diversion or load restriction, and risk area isolation control. The control actions are not executed independently, but rather by the model in a targeted, coordinated manner based on the causative mechanism of the dominant instability source and the influence of historical disturbance residuals. The weight allocation for this coordinated matching is as described above. Maintain consistency to achieve a precise match between control measures and instability mechanisms; The execution intensity threshold for each control action is determined by the model based on the risk value. Automatically determined, the basic strength is applied during the initial instability phase, with a threshold value equal to the design strength. During the evolution and amplification phase, the enhancement intensity is increased, with the threshold being the design intensity. During the sensitive recovery phase, the maximum strength is applied, with the threshold being the design strength. The model supports dynamically adjusting the execution intensity based on the control effect, thereby improving the effectiveness and relevance of the control strategy.

[0053] This design specifies the content and execution methods of the collaborative control strategy. The multi-means, linkage-based control scheme covers a variety of control actions, which can comprehensively address the problem of roadbed instability. Based on the disaster-causing mechanism of the dominant instability source and the influence relationship of the residual effects of historical disturbances, targeted linkage matching execution is carried out to achieve precise matching between control means and instability mechanisms. The execution intensity threshold is reasonably set and supports dynamic adjustment, which improves the effectiveness and pertinence of the control strategy. This helps the system to formulate a scientific and reasonable control scheme, effectively address roadbed instability, and ensure the stability and safety of the roadbed.

[0054] In one embodiment, dynamic updating is a process of online adaptive updating of the model and adjustment of the control strategy based on control effect feedback, aiming to ensure that the control strategy continuously adapts to the real-time state changes of the roadbed. Specifically, the method is as follows: After data re-acquisition and model input, and the execution of the control strategy, new roadbed monitoring data are continuously collected at a preset frequency. After preprocessing, the data is input into the trained roadbed instability mechanism identification and coupling analysis model. Model parameters and features are updated. The model automatically updates the feature parameters of the stratigraphic response decoupling fingerprint and the memory parameters of the perturbation memory field based on the new data. The updated parameters serve as the basis for the next coupling analysis. The control effect is determined by comparing the dominant probability values ​​before and after the update. Risk Value residual influence intensity To determine the control effect, if the dominant probability value of the dominant instability component is... It shows a downward trend, and the residual influence intensity of the corresponding perturbation in the perturbation memory field is... Entering the continuous decay phase, the decay rate If the attenuation is effective, it indicates that the current control strategy is effective, so the current control strategy is maintained. If the dominant probability value of a certain type of response component There was no downward trend, or even an upward trend, or the original response component transformed into other unstable components, i.e., the new component dominated the probability. If the current control strategy does not match the instability mechanism, the model will re-identify the new dominant instability mechanism and generate a new adaptive cooperative control strategy. If the instability risk value of a local roadbed area Significant decline, good recovery, decline range and The risk value of adjacent areas It shows an upward trend, new anomalies have emerged, and the magnitude of the increase... and If this indicates that the risk of roadbed instability has shifted spatially, the model will redefine the roadbed collaborative control boundary and generate targeted zonal control strategies for the new control boundary. The model undergoes iterative optimization by adding monitoring data, judgment results, control strategies, and execution effects from each control operation to the historical database as new samples. Every 100 new samples accumulated, the model automatically triggers incremental training to achieve iterative optimization. The test set of the incrementally trained model is then used. The value needs to be maintained. ; The thresholds for determining the control effect are all set by the model according to the instability warning and treatment specifications for roadbed engineering, and dynamically corrected in combination with the real-time changes in on-site monitoring data.

[0055] This design specifies the dynamic update method and process. Through steps such as data re-acquisition, model parameter and feature update, and control effect judgment, it realizes online adaptive updating of the model and adjustment of control strategy based on control effect feedback. Iterative optimization of the model can continuously improve performance and ensure that the system continuously adapts to the real-time changes in the roadbed state. This helps the system maintain efficient operation in complex and ever-changing environments, adjust control strategies in a timely manner, improve the control capability and adaptability to roadbed instability, and ensure the long-term stability of the roadbed.

[0056] Those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0057] The above embodiments provide a detailed description of the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A smart stability and collaborative control system for roadbeds in complex geological formations, characterized in that, The system includes: Multi-source sensing network building unit, response feature extraction unit, response fingerprint construction unit, perturbation memory field construction unit, coupling analysis and judgment unit, cooperative strategy generation unit, and dynamic update unit; Multi-source sensing network building unit is used to deploy sensing nodes in the cross section, longitudinal segment and depth layer of the roadbed and collect roadbed status data and historical disturbance event data. The response feature extraction unit is used to perform hierarchical, partitioned, and time-segmented processing on the roadbed state data to form a multi-dimensional structural response feature set. The response fingerprint construction unit is used to decompose the multidimensional structural response feature set into response components corresponding to different instability mechanisms, and assign each response component a response intensity value, influence range value, development rate value and dominant probability value to form a stratigraphic structure response decoupling fingerprint. The perturbation memory field construction unit is used to classify and encode the historical perturbation event data and generate memory parameters that characterize the residual effects to form a perturbation memory field. The coupling analysis and determination unit is used to couple the decoupled fingerprint of the geological structure response with the disturbance memory field and output the dominant instability source type, instability evolution stage and main control risk area. The collaborative strategy generation unit is used to generate partitioned, hierarchical, and phased collaborative control strategies based on the coupling analysis results. The dynamic update unit is used to introduce newly acquired data after control execution to update the formation structure response decoupling fingerprint and the disturbance memory field, and to adjust the cooperative control strategy according to the update results.

2. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 1, characterized in that, The multi-source sensing network building unit includes: Hierarchical deployment sub-unit, partition mapping sub-unit, and database creation sub-unit; The layered deployment subunit is used to deploy settlement, displacement, pore pressure, water migration and load response sensing nodes around the surface layer, deep layer and interlayer interface; The partition mapping subunit is used to establish spatial mapping relationships according to cross-sectional functional areas and longitudinal construction areas; The data database building subunit is used to perform standardization, noise reduction, anomaly removal, missing data completion, and classification storage on the collected data to form a current state database and a historical event database. It also supports fast retrieval of time series and regional indexes as well as data calls for subsequent model training.

3. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 1, characterized in that, The response feature extraction unit includes the following steps: Vertical compression response features, lateral displacement expansion features, water content change sensitivity features, pore pressure release or accumulation features, local stiffness difference features, deformation diffusion path features, and hysteresis recovery features after load application are extracted from the roadbed state data. A one-to-one correspondence is established according to the roadbed spatial layer, structural zoning, and monitoring time scale to form a multidimensional structural response feature set characterizing the current overall state of the roadbed. The monitoring timescale includes at least hourly, daily, and weekly levels, and different timescales correspond to different response evolution interpretation windows.

4. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 2, characterized in that: The current state database and the historical event database are jointly connected to the model training and optimization unit. The model training and optimization unit is used to train, verify, test and optimize the roadbed instability mechanism identification and coupling analysis model based on historical monitoring data, disturbance event data and instability judgment labels. The optimized model parameters are sent to the response feature extraction unit, response fingerprint construction unit, disturbance memory field construction unit and coupling analysis judgment unit to support online inference. The instability determination label includes at least the dominant instability source, instability stage, and main control risk area, and is used to constrain the consistency between the model output and the actual operating conditions, and to support the feedback correction of subsequent control results.

5. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 3, characterized in that, The response fingerprint construction unit includes the following steps: The multidimensional structural response feature set is decomposed into compaction-dominant response components, shear slip-dominant response components, moisture-induced softening-dominant response components, and load disturbance amplification response components. The response intensity value, influence range value, development rate value, and dominant probability value of each response component are calculated respectively. Then, a stratigraphic structural response decoupling fingerprint composed of response components and their corresponding parameters is constructed to characterize the mechanism composition and evolution trend of the current anomalous state. Each response component and its corresponding parameter are stored in association with the region location and depth layer, and are used to distinguish between the dominant anomaly region and the secondary anomaly region.

6. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 1, characterized in that, The perturbation memory field construction unit includes the following steps: The disturbances caused by construction filling, temporary excavation, rainfall infiltration, heavy traffic, changes in drainage conditions, preloading or unloading, and reinforcement are classified and coded. For each disturbance event, memory parameters are generated corresponding to the occurrence time, area of ​​effect, depth of influence, residual influence intensity, decay rate, spatial propagation direction, and superposition relationship, thereby forming a regionalized disturbance memory field that evolves over time. The classification coding consists of perturbation category coding and perturbation detail coding, and is used to achieve a unified representation of heterogeneous perturbation events.

7. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 6, characterized in that, The perturbation memory field construction unit further includes the following steps: The attenuation rate is corrected according to the disturbance type, subgrade soil type and disturbance depth, and the current residual influence intensity is calculated based on the corrected attenuation rate and the initial influence intensity of the disturbance. The roadbed soil types include at least cohesive soil, sandy soil, and gravelly soil. The current residual influence intensity is used as the historical residual effect input when the coupled analysis judgment unit performs coupled calculations. The correction process simultaneously considers the proportional relationship between the disturbance effect depth and the critical depth, as well as the difference in residual effect decay rate between different soil types.

8. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 5, characterized in that: The coupling analysis and determination unit is used to first establish a matching pair between response components and disturbance events based on the correspondence between instability mechanism and disturbance type, and then perform weighted matching calculation between the dominant probability value of each response component and the current residual influence intensity of the corresponding disturbance event to obtain the risk value of different instability types. The coupling analysis and determination unit is also used to determine the dominant instability source type according to the magnitude of the risk value, and to determine the initial instability stage, evolution amplification stage and sensitive recovery stage according to the preset threshold range. At the same time, it identifies whether the anomaly is triggered by the release of the residual effect of historical disturbance, locates the spatial range of the main control risk area, and outputs the risk ranking result.

9. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 7, characterized in that, The collaborative strategy generation unit includes the following steps: Based on the dominant instability source type, instability evolution stage, main control risk area, and disturbance residual characteristics, a collaborative control strategy is generated, which includes drainage intensity adjustment, drainage path switching, local unloading or load limiting, grouting area and grouting sequence adjustment, reinforcement and strengthening area redivision, lateral support intensity adjustment, preloading rhythm adjustment, construction machinery operation area and timing rearrangement, traffic diversion or load restriction, and risk area isolation control. The strategy also outputs the combination of control actions to be executed with the highest priority and the types of control actions that should not be applied. The execution intensity of each control action is matched according to the risk level, and the implementation order is allocated according to the main control risk area and the non-main control risk area.

10. The intelligent stability and collaborative control system for roadbeds in complex strata according to claim 8, characterized in that: The dynamic update unit is used to continuously receive new monitoring data after the collaborative control strategy is executed, update the dominant probability value, influence range value and development rate value in the stratigraphic structure response decoupling fingerprint, and synchronously update the residual influence intensity and decay rate in the disturbance memory field. When the updated risk value decreases and the dominant response component weakens, maintain the current control strategy; When the updated risk value undergoes spatial shift or a new dominant instability mechanism is formed, the collaborative control boundary is redefined and the collaborative strategy generation unit is triggered to output a new partitioned control strategy. At the same time, the control data of this round is written into the historical database to perform incremental optimization, thereby forming a closed-loop adaptive control mechanism for roadbeds with complex strata and improving the consistency of subsequent coupling judgment and strategy generation.

Images