An adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system and method

Abstract

The application discloses a kind of self-adapting multi-source fusion roadbed intelligent monitoring and dynamic regulation system and method, and relates to civil engineering monitoring technical field.The system includes differentiating sensing unit, data transmission and processing unit, mixed model prediction unit and dynamic regulation unit;Differentiating sensing unit acquires multidimensional data by composite anchoring structure, magnetic attraction follow-up module and fiber-optic monitoring network;Data transmission and processing unit transmit data using LoRaWAN protocol and pre-process;Mixed model prediction unit predicts settlement trend and risk level based on STGCN-MHSA hybrid model and MOPSO algorithm;Dynamic regulation unit makes corresponding regulation according to prediction result.The self-adapting multi-source fusion roadbed intelligent monitoring and dynamic regulation system and method proposed by the application can realize high-precision monitoring and prediction, have active regulation capability, and are suitable for roadbed health management in complex scenarios.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of civil engineering monitoring, in particular to a self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system and method. BACKGROUND

[0002] As the core bearing structure in road engineering, the roadbed will be disturbed by various natural environments in daily use, and its stability directly determines the safety and service life of the road. Especially in special environments such as highways and high fill roadbeds, due to the changeable environmental load and complex geological conditions, the roadbed is prone to uneven settlement, leading to serious problems such as road surface cracking and even structural instability. Therefore, it is of great significance to prevent engineering disasters and reduce maintenance costs to monitor and predict the settlement of the roadbed in real time with high precision and timely regulation.

[0003] At present, the roadbed settlement monitoring technology mainly includes physical sensing devices and data analysis models. Traditional physical monitoring devices mostly rely on pre-embedded cement blocks or rigid structures (such as cone pipes and spiral blades), which have a complicated construction process and are prone to displacement of the reference point due to precipitation or freezing-thawing in soft soil or frozen soil, affecting the monitoring accuracy. At the same time, the existing follow-up monitoring structure lacks adaptability to multi-dimensional settlement and cannot accurately capture the inclination or local deformation. For monitoring data, it is mostly limited to reporting and does not integrate an active regulation module. In the face of sudden settlement or extreme environments (such as freezing-thawing cycles), it cannot delay the settlement process through dynamic adjustment of load distribution or temperature control, resulting in high passive maintenance costs. From the perspective of data analysis models, although the CNN-BiLSTM architecture used in current mainstream prediction models can extract time sequence features, it does not fully utilize the topological relationship of the roadbed sensor network and lacks sufficient spatio-temporal feature fusion capability. In addition, the hyperparameter optimization used relies on the genetic algorithm (GA), which is prone to local optimization and slow convergence, affecting the model deployment efficiency. SUMMARY

[0004] The purpose of the present application is to provide a self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system and method, which significantly improves the problems of traditional rigid sensing devices that cannot adapt to multi-dimensional deformation monitoring in complex geological environments, traditional roadbed settlement prediction models that have low hyperparameter optimization efficiency and insufficient spatio-temporal feature fusion, and traditional roadbed settlement monitoring systems that lack regulation functions and cannot respond to settlement risks in real time.

[0005] To achieve the above purpose, the present application provides a self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system, which comprises:

[0006] The differentiated sensing unit includes a composite anchoring structure, a magnetic servo module, and a fiber optic monitoring network. The composite anchoring structure comprises a tapered tube with staggered helical blades on its outer wall, and miniature pressure sensors integrated at the blade edges. The magnetic servo module consists of a magnetic sleeve and an elastic plate. The sleeve contains a Hall sensor that dynamically attracts a signal receiver at the end of the flexible pipe using magnetic attraction. The elastic plate contains a shape memory alloy (SMA). The fiber optic monitoring network is equipped with distributed fiber optic grating (FBG) sensors layered along the roadbed to collect strain and temperature data in real time.

[0007] The data transmission and processing unit uses the LoRaWAN wireless transmission protocol to upload multi-source data to the edge computing node. The data fusion layer in the edge computing node performs spatiotemporal alignment and outlier filtering on the multi-source data to generate a standardized feature matrix.

[0008] The hybrid model prediction unit constructs an STGCN-MHSA hybrid model based on the improved spatiotemporal graph convolutional network STGCN and the multi-head attention mechanism MHSA. The hyperparameters of the STGCN-MHSA hybrid model are optimized by the multi-objective particle swarm optimization algorithm MOPSO. The normalized feature matrix is ​​input to predict the settlement trend and risk level.

[0009] The dynamic control unit includes a hydraulic support system, a temperature control module, and a blockchain storage module. The hydraulic support system is used to adjust the local load distribution of the roadbed; the temperature control module is used to initiate geothermal circulation in permafrost areas; and the blockchain storage module is used to upload all operation records to the blockchain for storage.

[0010] Preferably, in the differentiated sensing unit, the deformation of the shape memory alloy (SMA) is... The calculation formula is as follows:

[0011] ;

[0012] in, α The coefficient of thermal expansion is β The stress response coefficient is... σ For soil stress, The soil temperature (unit: °C) is collected in real time via an FBG fiber optic network or a temperature sensor. The reference temperature for SMA (unit: °C) is the deformation reference temperature (such as the phase change initiation temperature or the initial design temperature).

[0013] Preferably, in the hybrid model prediction unit, the STGCN-MHSA hybrid model encodes the roadbed topology as graph data, extracts local features through spatiotemporal convolution, and uses multi-head attention to weight key nodes.

[0014] Preferably, in the hybrid model prediction unit, the multi-objective particle swarm optimization algorithm (MOPSO) is set as a dual fitness function to simultaneously optimize the prediction error (MAE) and the model inference speed (FPS) to generate the Pareto optimal solution set.

[0015] Preferably, in the dynamic control module, the temperature control module initiates geothermal circulation in the permafrost region to maintain the roadbed temperature ≥0℃.

[0016] This invention also provides an adaptive multi-source fusion roadbed intelligent monitoring and dynamic control method, used to implement the above-mentioned adaptive multi-source fusion roadbed intelligent monitoring and dynamic control system, the specific steps of which are as follows:

[0017] Step S1: Real-time acquisition of pressure P, displacement δ, strain ε, and temperature H is achieved through miniature pressure sensors on the conical tube helical blades, Hall sensors in the magnetic servo module, and a distributed FBG fiber optic network, thus constructing a spatiotemporal alignment matrix. X E R N×T×C ,in N This represents the number of sensor nodes. T The time window length, C For feature dimensions;

[0018] Step S2: Use the DBSCAN clustering algorithm to remove outliers and perform normalization processing;

[0019] Step S3: Input the preprocessed data into the STGCN-MHSA hybrid model and output the settlement prediction value Y and risk level R;

[0020] Step S4: Based on the prediction results of the STGCN-MHSA hybrid model, adjust the local load distribution of the roadbed through the hydraulic support system or activate the temperature control module to suppress freeze-thaw settlement.

[0021] Preferably, in step S2, the normalization formula is as follows:

[0022] ;

[0023] in, These are the element values ​​of the original feature tensor. These are the normalized feature tensor element values. , The first c The mean and standard deviation of the dimensional features.

[0024] Preferably, in step S3, the calculation process of the STGCN-MHSA hybrid model is as follows:

[0025] Spatial graph convolution:

[0026] Define sensor topology G = (V, E)The adjacency matrix A is constructed using the threshold truncation method, as shown in the following formula:

[0027] ;

[0028] in, V It is a set of vertices, that is, a set of sensor nodes, where each node represents a physical monitoring point; E The set of edges represents the connections between nodes, indicating the spatial relationships between sensors. Define the edges connecting nodes (distance only) The nodes are connected. (This is the distance threshold, in meters). For nodes Euclidean distance, Distance threshold;

[0029] The Chebyshev polynomial approximation is used, and the formula is as follows:

[0030] ;

[0031] in, For spatial characteristic matrix, for Chebyshev polynomials To scale the Laplacian matrix, For trainable parameters, The order of the Chebyshev polynomial (controls the receptive field range);

[0032] Temporally dilated convolution:

[0033] Using expansion rate The convolutional kernel is used to extract temporal features, as shown in the following formula:

[0034] ;

[0035] in, To extract time features, This is the time offset. The size of the temporal convolution kernel. For convolution kernel weights, This is the spatial association feature matrix of sensor nodes extracted by the spatial graph convolution module;

[0036] Bullish Self-Attention Weighted:

[0037] Will Split into The formula is as follows:

[0038] ;

[0039] ;

[0040] ;

[0041] ;

[0042] in, For the first Each attention head output, , , For querying key-value matrices, for The transpose of the key matrix is ​​used to calculate query-key similarity. Key vector dimension (scaling factor) (used for stabilizing gradients) , For the attention calculation formula, , , The projection matrix;

[0043] The final result is calculated using the following formula:

[0044] ;

[0045] in, As a feature of fusion, For the first Each attention head output, It is a linear projection matrix (output weights). To splice the outputs of multiple attention heads;

[0046] The formula for predicting settlement results is as follows:

[0047] ;

[0048] in, For the first Each attention head output, For tensor flattening operations (preprocessing of input fully connected layers). , These are the parameters for the fully connected layer. This is the Sigmoid activation function.

[0049] Preferably, in step S3, the STGCN-MHSA hybrid model uses the NSGA-II dual fitness function to optimize the model hyperparameters, as shown in the following formula:

[0050] ;

[0051] in, The mean absolute error, The inverse of the model's inference speed. The actual settlement value (mm, from historical monitoring or on-site measurement) is given. The model inference speed (frames / second). This represents the mean absolute error.

[0052] Preferably, in step S4, based on the settlement prediction results, the dynamic control of the hydraulic support system is triggered, specifically as follows:

[0053] when At that time, the hydraulic support system is triggered:

[0054] ;

[0055] in For PID coefficients, The differential gain coefficient (kPa·s / mm, suppressing the rate of sedimentation change) is given. The settlement risk threshold (mm) is used. Adjust the load (kPa) of the hydraulic system to compensate for uneven settlement.

[0056] Therefore, this invention proposes an adaptive multi-source fusion roadbed intelligent monitoring and dynamic control system and method, the beneficial effects of which are as follows:

[0057] (1) This invention achieves comprehensive and high-precision monitoring of the roadbed through the collaborative operation of a composite anchoring structure, a magnetic servo module, and a fiber optic monitoring network. Actual testing shows that the monitoring accuracy of this system can reach ±0.3mm, which is 40% higher than that of traditional monitoring equipment. It also reduces the multidimensional settlement capture error by 50% compared to traditional equipment, enabling earlier and more accurate detection of subtle changes in the roadbed and providing sufficient time for subsequent adjustments.

[0058] (2) The hybrid prediction model in this invention is based on an improved spatiotemporal graph convolutional network (STGCN) and a multi-head attention mechanism (MHSA), which fully exploits the topological relationships and spatiotemporal features of the roadbed sensor network, greatly improving prediction accuracy. At the same time, the multi-objective particle swarm optimization (MOPSO) algorithm optimizes the model hyperparameters, improving inference speed while ensuring prediction accuracy. Experimental results show that the mean absolute error (MAE) of the prediction in this invention is reduced to 0.8 mm, and the inference speed is improved by 20%, enabling reliable prediction of future roadbed settlement trends and providing strong support for formulating response strategies in advance.

[0059] (3) This invention fully considers complex geological and extreme climatic conditions and can still operate normally in extreme environments of -40℃. In permafrost areas, the temperature control module initiates geothermal circulation to effectively maintain the roadbed temperature ≥0℃ and suppress freeze-thaw settlement; facing complex geological conditions such as high-fill roadbeds, the differentiated sensing units flexibly adapt to soil deformation to ensure stable data acquisition. Compared with traditional systems that frequently experience failures and data distortion in extreme environments, the adaptive multi-source fusion system of this invention demonstrates excellent stability and reliability.

[0060] (4) This invention not only possesses precise monitoring and prediction capabilities, but also innovatively introduces a dynamic control unit. When the predicted settlement exceeds the threshold, the hydraulic support system rapidly adjusts the local load distribution of the roadbed based on a precise algorithm; when freeze-thaw risks occur, the temperature control module responds promptly. In practical applications in permafrost regions, through active control, this system reduces maintenance costs by 30%, and the system's design life can reach more than 10 years, significantly reducing long-term operation and maintenance costs and significantly improving the economic and social benefits of road engineering.

[0061] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0062] Figure 1 This is a flowchart of an adaptive multi-source fusion roadbed intelligent monitoring and dynamic control method according to the present invention. Detailed Implementation

[0063] To make the technical solutions, advantages, and objectives of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below. The described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the described embodiments of the present invention without creative effort are within the protection scope of the present invention.

[0064] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning as understood by one of ordinary skill in the art to which this invention pertains.

[0065] Example 1

[0066] This invention proposes an adaptive multi-source fusion intelligent monitoring and dynamic control system for roadbed, which improves upon the shortcomings of previous roadbed settlement monitoring and control methods, including:

[0067] The differentiated sensing unit includes a composite anchoring structure, a magnetic servo module, and a fiber optic monitoring network. The composite anchoring structure comprises a tapered tube with staggered helical blades on its outer wall, and miniature pressure sensors integrated at the blade edges. The magnetic servo module consists of a magnetic sleeve and an elastic plate. The elastic plate embeds a shape memory alloy (SMA), which deforms when settlement exceeds a threshold, increasing the contact area with the roadbed. The sleeve contains a Hall effect sensor that dynamically captures displacement changes of the signal receiver at the end of the flexible pipe through magnetic attraction. The fiber optic monitoring network consists of distributed fiber optic grating sensors (FBGs) layered along the roadbed, used to collect strain, temperature, and humidity data in real time, facilitating precise monitoring and control of the roadbed's real-time condition.

[0068] Deformation of shape memory alloy SMA The calculation formula is as follows:

[0069] ;

[0070] in, α The coefficient of thermal expansion is β The stress response coefficient is... σ For soil stress; The soil temperature (unit: °C) is collected in real time via FBG fiber optic network or temperature sensor. The reference temperature for SMA (unit: °C) is the deformation reference temperature (such as the phase change initiation temperature or the initial design temperature).

[0071] The data transmission and processing unit uses the LoRaWAN protocol to upload multi-source data (pressure, displacement, strain) to the edge computing node, with a theoretical communication distance of greater than or equal to 10km. The data fusion layer in the edge computing node eliminates the time delay of sensor data through a spatiotemporal alignment algorithm and filters outliers based on DBSCAN clustering to generate a standardized feature matrix.

[0072] The hybrid model prediction unit constructs an STGCN-MHSA hybrid model based on an improved spatiotemporal graph convolutional network (STGCN) and a multi-head attention mechanism (MHSA). The hyperparameters of the STGCN-MHSA hybrid model are optimized using a multi-objective particle swarm optimization (MOPSO) algorithm. A standardized feature matrix is ​​input to predict settlement trends and risk levels. Specifically, the STGCN topology encodes the roadbed sensor network into a graph structure, extracting spatial correlation features through graph convolution. The MHSA mechanism dynamically weights key nodes in the time series. MOPSO uses prediction error (MAE) and inference speed (FPS) as dual fitness functions to generate a Pareto optimal solution set.

[0073] The dynamic control unit includes a hydraulic support system, a temperature control module, and a blockchain storage module. The hydraulic support system is used to adjust the local load distribution of the roadbed based on the prediction results to suppress uneven settlement. The temperature control module is used to initiate geothermal circulation in permafrost areas to maintain the roadbed temperature above or equal to 0°C and suppress freeze-thaw settlement. The blockchain storage module is used to upload all operation records to the blockchain for storage to ensure that the data is tamper-proof.

[0074] Example 2

[0075] like Figure 1 As shown, the present invention also provides an adaptive multi-source fusion roadbed intelligent monitoring and dynamic control method, used to implement the adaptive multi-source fusion roadbed intelligent monitoring and dynamic control system proposed in Embodiment 1, the steps of which are as follows:

[0076] S1. Using miniature pressure sensors on the conical spiral blades, Hall sensors in the magnetically assisted servo module, and a distributed FBG fiber optic network, pressure P, displacement δ, strain ε, and temperature H are collected in real time to construct a spatiotemporal alignment matrix. X E R N×T×C ,in N This represents the number of sensor nodes. T The time window length, C For feature dimensions;

[0077] S2. Use the DBSCAN clustering algorithm to remove outliers and perform normalization, as shown in the following formula:

[0078] ;

[0079] in, These are the element values ​​of the original feature tensor. These are the normalized feature tensor element values. , The first c The mean and standard deviation of the dimensional features;

[0080] S3. Input the preprocessed data into the STGCN-MHSA hybrid model, and output the predicted settlement value Y and risk level R; the calculation process of the STGCN-MHSA hybrid model is as follows:

[0081] Spatial graph convolution:

[0082] Define sensor topology G = (V, E) The adjacency matrix A is constructed using the threshold truncation method, as shown in the following formula:

[0083] ;

[0084] in, VIt is a set of vertices, that is, a set of sensor nodes, where each node represents a physical monitoring point; E The set of edges represents the connections between nodes, indicating the spatial relationships between sensors. Define the edges connecting nodes (distance only) The nodes are connected. (This is the distance threshold, in meters). For nodes Euclidean distance, Distance threshold;

[0085] The Chebyshev polynomial approximation is used, and the formula is as follows:

[0086] ;

[0087] in, For spatial characteristic matrix, for Chebyshev polynomials To scale the Laplacian matrix, For trainable parameters, The order of the Chebyshev polynomial (controls the receptive field range);

[0088] Temporally dilated convolution:

[0089] Using expansion rate The convolutional kernel is used to extract temporal features, as shown in the following formula:

[0090] ;

[0091] in, This is the time series feature matrix obtained after processing by a time feature extraction module (such as temporal dilated convolution, LSTM, etc.). The time offset represents the sliding position of the convolution kernel on the time series; The size of the temporal convolution kernel controls the window length for extracting temporal features. For convolution kernel weights, This is the spatial association feature matrix of sensor nodes extracted by the spatial graph convolution module;

[0092] Bullish Self-Attention Weighted:

[0093] Will Split into The formula is as follows:

[0094] ;

[0095] ;

[0096] ;

[0097] ;

[0098] in, For the output of the i-th attention head, , , For querying key-value matrices, for The transpose of the key matrix is ​​used to calculate query-key similarity. Key vector dimension (scaling factor) (used for stabilizing gradients) , For the attention calculation formula, , , The projection matrix;

[0099] The final result is output using the following formula:

[0100] ;

[0101] in, As a feature of fusion, For the h-th attention head output, It is a linear projection matrix (output weights). To splice the outputs of multiple attention heads;

[0102] The formula for predicting settlement results is as follows:

[0103] ;

[0104] in, For the output of the i-th attention head, For tensor flattening operations (preprocessing of input fully connected layers). , These are the parameters for the fully connected layer. This is the Sigmoid activation function.

[0105] The STGCN-MHSA hybrid model uses the NSGA-II dual fitness function to optimize the model's hyperparameters, as shown in the following formula:

[0106] ;

[0107] in, The mean absolute error, The inverse of the model's inference speed. The actual settlement value (mm, from historical monitoring or on-site measurement) is given. The model inference speed (frames / second). This represents the mean absolute error.

[0108] S4. Based on the prediction results of the STGCN-MHSA hybrid model, adjust the local load distribution of the roadbed through the hydraulic support system or activate the temperature control module to suppress freeze-thaw settlement. The specific operation is as follows:

[0109] when At that time, the hydraulic support system is triggered:

[0110] ;

[0111] in For PID coefficients, The differential gain coefficient (kPa·s / mm, suppressing the rate of sedimentation change) is given. The settlement risk threshold (mm) is used. Adjust the load (kPa) of the hydraulic system to compensate for uneven settlement.

[0112] The following specific implementation case will provide further explanation.

[0113] After deployment on highways in high-altitude permafrost regions, the STGCN-MHSA model predicted local settlement trends 72 hours in advance (with an error of 0.7 mm), triggering geothermal circulation to maintain the roadbed temperature at 1.2℃; the hydraulic system adjusted the load distribution, reducing uneven settlement by 62%.

[0114] It is worth noting that all contents not described in detail in this invention are existing technologies and are well known to those skilled in the art.

[0115] Therefore, this invention provides an adaptive multi-source fusion roadbed intelligent monitoring and dynamic control system and method. Through the synergy of multi-source heterogeneous sensor networks and advanced algorithms, it achieves high-precision monitoring of ±0.3mm (40% improvement over traditional methods), reduces the predicted MAE to 0.8mm and increases the inference speed by 20%, and can operate stably in extreme environments of -40℃. Its active control mechanism reduces maintenance costs by 30% in permafrost areas, and the system has a lifespan of over 10 years. It comprehensively solves the problems of low monitoring accuracy, poor prediction timeliness, and passive control in traditional methods, providing an efficient solution for roadbed health management in complex scenarios.

[0116] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. An adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation method, characterized in that, The specific steps are as follows: Step S1, through the micro pressure sensor on the conical pipe spiral blade, the hall sensor of the magnetic attraction servo module, the distributed FBG optical fiber network, real-time collection of pressure P, displacement δ, strain ε, temperature H, construction of space-time alignment matrix wherein N is the number of sensor nodes, T is the length of the time window, C is the feature dimension; Step S2, using DBSCAN clustering algorithm to eliminate outliers, and normalizing; Step S3, inputting the preprocessed data into the STGCN-MHSA hybrid model, outputting the settlement prediction value Y and the risk level r, wherein the STGCN-MHSA hybrid model calculation process is as follows: Spatial graph convolution: Defining a sensor topology The adjacency matrix A is constructed by thresholding, as follows: ； wherein, V is a set of vertices, E is a set of edges, is a defined node edge, is a node Euclidean distance of, is a distance threshold; Using Chebyshev polynomial approximation, the formula is as follows: ； wherein, is a spatial feature matrix, is is a Chebyshev polynomial of order is a scaled Laplacian matrix, is a trainable parameter, is a Chebyshev polynomial order, is a spatio-temporal alignment matrix; Time expansion convolution: An expansion rate is adopted The time sequence feature is extracted by using the convolution kernel with the expansion rate, and the formula is as follows: ； wherein, is a time feature extraction module, is a time offset, is a size of a time convolution kernel, is a convolution kernel weight, is a sensor node spatial correlation feature matrix extracted by a spatial graph convolution module; Multi-head self-attention weighting: will be described below. split into The size, formula as follows: ； ； ； ； wherein, is the th attention head output, , , is the query, key, value matrix, is the transpose of the key matrix used to compute query-key similarity, is the key vector dimension, , is the attention computation function, , , is the projection matrix; Calculate the final result, the formula is as follows: ； wherein, is a fused feature, is a linear projection matrix, is an output of the i-th attention head, is a linear projection matrix, is a concatenation of outputs of the plurality of attention heads; Settlement result prediction, the formula is as follows: ； wherein, is the output of the th attention head, is a tensor flattening operation, , are fully connected layer parameters, is a sigmoid activation function. Step S4, according to the STGCN-MHSA hybrid model prediction result, adjusting the local load distribution of the hydraulic support system or starting the temperature control module to suppress freeze-thaw settlement. 2.The self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation method according to claim 1, characterized in that, In step S2, the normalization formula is as follows: ； wherein, is the element value of the original feature tensor, is the normalized feature tensor element value, , are the mean and standard deviation of the first c dimensional feature, respectively. 3.The self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation method according to claim 1, characterized in that, In step S3, the STGCN-MHSA hybrid model uses the NSGA-II double fitness function to optimize the model hyperparameters, and the formula is as follows: ； wherein, is the mean absolute error, is the inverse of the model inference speed, is the true settling value, is the model inference speed, is the mean absolute error. 4.The self-adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation method according to claim 1, characterized in that, In step S4, according to the settlement prediction result, trigger the hydraulic support system dynamic regulation and control execution, specifically: When the hydraulic support system is triggered: ； wherein is a PID coefficient, is a derivative gain coefficient, is a settlement risk threshold, is a hydraulic system adjustment load amount.

5. An adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system, characterized in that, The system is used to realize the method of any one of claims 1-4, including a differentiated sensing unit, a data transmission and processing unit, a hybrid model prediction unit and a dynamic regulation and control unit; The differentiated sensing unit includes a composite anchoring structure, a magnetic attraction following module and an optical fiber monitoring network; the composite anchoring structure includes a tapered pipe, the outer wall of the tapered pipe is provided with staggered spiral blades, and the edge of the blade is integrated with a micro pressure sensor; the magnetic attraction following module is composed of a magnetic sleeve and an elastic plate, the sleeve is provided with a Hall sensor, and the elastic plate is provided with a shape memory alloy SMA; The optical fiber monitoring network is arranged along the roadbed layering distributed fiber Bragg grating sensor FBG; The data transmission and processing unit uses LoRaWAN wireless transmission protocol to upload multi-source data to edge computing node, and the data fusion layer in the edge computing node performs time and space alignment and outlier filtering on multi-source data; The hybrid model prediction unit constructs an STGCN-MHSA hybrid model based on an improved spatiotemporal graph convolution network STGCN and a multi-head attention mechanism MHSA, and uses a multi-objective particle swarm optimization algorithm MOPSO to optimize the hyperparameters of the STGCN-MHSA hybrid model to predict the settlement trend and risk level; The dynamic regulation and control unit includes a hydraulic support system, a temperature control module and a blockchain storage module. 6.The adaptive multi-source fusion subgrade intelligent monitoring and dynamic regulation system according to claim 5, characterized in that, In the differential sensing unit, the deformation amount of the shape memory alloy (SMA) The calculation formula is as follows: ； in, The coefficient of thermal expansion is The stress response coefficient is... For soil stress, For real-time monitoring of soil temperature, for SMA The reference temperature. 7.The adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system according to claim 5, characterized in that, In the hybrid model prediction unit, the STGCN-MHSA hybrid model encodes the roadbed topology structure into graph data, extracts local features through spatiotemporal convolution, and uses multi-head attention to weight key nodes. 8.The adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system according to claim 5, characterized in that, In the hybrid model prediction unit, the multi-objective particle swarm optimization algorithm MOPSO is set to a double fitness function, which synchronously optimizes the prediction error MAE and the model inference speed FPS to generate a Pareto optimal solution set. 9.The adaptive multi-source fusion roadbed intelligent monitoring and dynamic regulation system according to claim 5, characterized in that, In the dynamic regulation module, the temperature control module starts the geothermal cycle in the frozen soil area to maintain the roadbed temperature not less than 0℃.

Images