GNSS (Global Navigation Satellite System) real-time dynamic resolving system and method for automatically monitoring deformation of dam

Abstract

The invention relates to the technical field of sensor networks and state sensing, in particular to a GNSS real-time dynamic resolving system and method for dam deformation automatic monitoring. The method comprises the following steps: automatically forming and updating a resolving reference role by dynamically sensing the operation state of a GNSS monitoring node; constructing a state vector based on the multi-dimensional state information, performing normalization and credibility evaluation, and adaptively generating a reference node candidate set; residual quantization, environmental disturbance correction and self-correction are carried out through a reference node time sequence evolution model; the self-correction reference field and observation data are integrated through cross-epoch coupling calculation, real-time continuous and steady displacement calculation is achieved, and node credibility is generated; the resolving result is coupled with the dam structure mechanical constraint, and closed-loop verification and abnormal node dynamic correction are carried out; and a multi-model fusion prediction system is constructed to realize future displacement prediction, risk assessment and graded early warning. According to the invention, automation, intelligence and stability of dam deformation monitoring are improved.

Description

Technical Field

[0001] This invention relates to the field of sensor networks and state perception technology, specifically to a GNSS real-time dynamic calculation system and method for automated monitoring of dam deformation. Background Technology

[0002] With the continuous improvement of the positioning accuracy and reliability of the Global Navigation Satellite System, its application in the automated monitoring of structural deformation in major engineering projects such as dams is becoming increasingly widespread. Current monitoring technology can achieve continuous observation of displacement at the millimeter level, while engineering practice has placed higher demands on the long-term stability, environmental adaptability and reliability of the monitoring system.

[0003] Chinese invention patent CN116123982B discloses an automated observation method for a GNSS-based benchmark network for dam vertical displacement monitoring. It includes the following steps: Step 1: Site selection, instrument and observation pier selection; Step 2: GNSS data preprocessing and scheme optimization; in automated GNSS vertical displacement monitoring, determining the optimal observation period and correcting frequent small cycle slips; Step 3: Baseline calculation and network adjustment methods; in automated GNSS vertical displacement monitoring, models such as tropospheric delay correction and thermal expansion effect correction are applied to baseline calculation, and prior elevation difference information is introduced in network adjustment to improve the accuracy and reliability of the baseline solution.

[0004] In complex operating environments, maintaining the long-term stability of the solution benchmark, effectively integrating structural mechanical behavior with real-time observation data, and establishing a closed-loop monitoring system from real-time perception to trend prediction have become important directions for further improving the intelligence level of dam safety monitoring. Therefore, developing a real-time dynamic solution system that can adaptively maintain the reference frame, integrate multi-source information, and has intelligent feedback adjustment capabilities is of great significance for achieving more reliable, accurate, and forward-looking dam safety monitoring. Summary of the Invention

[0005] The purpose of this invention is to address the problems existing in the background technology by proposing a GNSS real-time dynamic calculation system and method for automated monitoring of dam deformation.

[0006] The technical solution of this invention: a GNSS real-time dynamic calculation method for automated monitoring of dam deformation, comprising the following specific implementation steps: S1. Dynamically perceive the operating status and calculation behavior of each monitoring node in the GNSS monitoring network, and automatically form and dynamically update the calculation reference role based on the multi-dimensional status information of the node's observation continuity, calculation stability and environmental response. S2. Construct a time-series evolution model of reference nodes, perform statistical analysis and quantification of the relative coordinate residuals between selected reference node sets, introduce environmental disturbance factors for correction, generate self-correcting coordinate increments to update reference node coordinates, and conduct an overall credibility assessment of the updated reference field. S3. Perform cross-epoch coupled solution, construct multi-epoch state vectors and use the evolution trend of the reference field to predict the state, couple the real-time observation node data with the self-calibrating reference node coordinates to construct a dynamic gain matrix, update the node state vectors and calculate the solution residuals, and generate node solution credibility. S4. Couple the multi-epoch GNSS solution with the dam's structural mechanical constraints, construct a structural mechanical constraint model and calculate the displacement residuals. Combine the node credibility to identify abnormal nodes and perform weighted least squares closed-loop correction. At the same time, dynamically update the node weights and evaluate the consistency of the corrected displacement across the entire network. S5. Based on the corrected node displacement sequence, historical trend data and environmental disturbance information, a multi-model fusion prediction system is constructed to predict future displacements. The risk index of future node displacement exceeding the threshold is calculated and the network-wide risk index is generated by weighting according to the node credibility. The warning level is divided according to the index, and the prediction results are fed back to the node management and solution process in a closed loop.

[0007] Preferably, in step S1, the operational status and calculation behavior of each monitoring node in the dynamic sensing GNSS monitoring network specifically include: The observation integrity, signal quality, changes in solution results, and response to environmental disturbances of each GNSS monitoring node are continuously collected and quantified to construct a unified node operation status vector. Standardize and smooth the state components in the original state vector of the node to obtain state feature values ​​under a unified scale, and introduce the solution continuity consistency feature to construct the core index of solution stability. A node comprehensive credibility function is constructed based on multi-dimensional state features, which integrates multi-dimensional state features into a unified credibility evaluation, and adaptively generates a candidate set of reference nodes based on the overall credibility distribution of the network. Continuously monitor changes in node credibility. When the credibility of a reference node falls below the minimum threshold or the credibility of a non-reference node exceeds the average credibility of the reference node, trigger automatic adjustment and update of the solution role.

[0008] Preferably, the node running state vector is as follows: The node operating state vector includes observation continuity index, observation quality stability index, short-term consistency index of solution results, environmental disturbance response index, and historical role reliability index. Among them, the observation continuity index reflects the node's ability to stably provide effective raw GNSS observation data; the observation quality stability index is derived from a comprehensive assessment of carrier phase noise, multipath effects, and signal fluctuations; the short-term consistency index of the solution results is used to measure the stability of the node's results during continuous epoch solution processes; the environmental disturbance response index is used to describe the node's sensitivity to external environmental factors; and the historical role reliability index is used to describe the degree of contribution of the node to the overall solution stability when it served as a reference node during historical operation.

[0009] Preferably, in step S2, constructing the reference node temporal evolution model specifically includes: The relative coordinate residuals between nodes in the reference node set are calculated within a sliding time window, and a time-weighted moving average is applied to the residual sequence to quantify the small evolution trend of the reference field. An environmental disturbance factor matrix containing environmental disturbance variables such as temperature, wind speed, humidity, and shading type is introduced, and the environmental disturbance correction vector is coupled with the reference node residual sequence to eliminate unstructured offsets. Based on the corrected residual sequence, a self-correcting coordinate increment is generated for each reference node, and the reference node coordinates are updated accordingly. An overall credibility assessment is performed on the updated reference field. When the overall evolution credibility of the reference field obtained from the assessment is lower than a set threshold, a reference node reselection or weight reallocation is triggered.

[0010] Preferably, in step S3, the cross-epoch coupled solution specifically includes: Based on the self-calibrating reference node set and its temporal evolution parameters, the coordinates of all monitoring nodes are organized into a multi-epoch state vector according to the time series. Construct a cross-epoch prediction model and use the state transition matrix to predict the current epoch state by applying the historical displacement and the evolution trend of the reference field. An observation equation is constructed to couple real-time observation node data with self-calibrating reference node coordinates, and a dynamic coupling gain matrix is ​​constructed. Among them, the observation noise covariance matrix is ​​dynamically adjusted according to the reliability of the node's historical solution; The node state vector is updated using the gain matrix, and the solution residual sequence is calculated. Time-weighted analysis is performed on the residual sequence to evaluate the node solution stability. The solution reliability of each node is generated based on the residual evolution, and this solution reliability is fed back to dynamically adjust the allocation of reference nodes and observation nodes.

[0011] Preferably, in step S4, coupling the multi-epoch GNSS solution results with the dam structural mechanical constraints specifically includes: The dam is decomposed into several structural units, and elastic mechanical constraint equations are defined to associate nodal displacement vectors with structural constraint conditions. Different weighting coefficients are introduced for different dam sections and key monitoring points to form weighted structural constraints. Calculate the residuals between the solved displacements of each node and the weighted structural constraints, and perform weighted analysis based on the node solution reliability to identify abnormal nodes; The identified abnormal nodes are corrected using the weighted least squares method, where the correction weight of a node is calculated by combining its node confidence and residual size. Calculate the network-wide corrected displacement consistency index. If the index is lower than the set threshold, trigger the reference node update and node weight redistribution, and use the corrected displacement as the input to the reference field dynamic evolution model.

[0012] Preferably, in step S5, constructing the multi-model fusion prediction system specifically includes: Collect the corrected displacement sequence and node credibility of the entire network to form a sliding time window historical dataset, and extract key feature indicators such as short-term displacement velocity, displacement acceleration and residual variance from it, and weight the features in combination with node credibility; For each node, a multi-model prediction system is constructed, which includes a weighted ARMA model, a structural mechanics coupling model, and a machine learning nonlinear model. The weighted ARMA model uses historical node displacements and residuals within a sliding time window to predict short-term cyclical fluctuations. The structural mechanics coupling model predicts long-term trends based on structural constraints and environmental loads; Machine learning nonlinear models use LSTM to capture nonlinear historical patterns; The prediction results of the three types of models are dynamically weighted according to their historical prediction errors and then fused to obtain the node fusion prediction displacement result.

[0013] Preferably, the machine learning nonlinear model is a long short-term memory network model, which is used to capture the nonlinear and complex time-dependent patterns in the historical data of node displacement.

[0014] Preferably, the indicators used to classify early warning levels include: Based on the fusion prediction displacement results of the nodes and their corresponding safe displacement thresholds determined by dam design, historical deformation statistics and material properties, the risk index of future displacement exceeding the threshold of the nodes is calculated. The risk indicators of all nodes in the network are weighted and averaged according to node credibility to generate the network-wide risk indicators. Based on the numerical range of risk indicators across the entire network, warning levels are divided into green, yellow, orange, and red.

[0015] The technical solution of this invention: A GNSS real-time dynamic calculation system for automated monitoring of dam deformation, which is used to execute the above-mentioned GNSS real-time dynamic calculation method for automated monitoring of dam deformation, comprising: The node monitoring and status awareness module is used to perceive the operating status and data quality of each GNSS monitoring node in real time, and realize the node credibility calculation and dynamic allocation and adjustment of reference roles based on multi-dimensional status assessment. The multi-epoch real-time dynamic solution module is used to perform cross-epoch coupled solution based on the self-calibrated reference field and observation data, update the nodal displacement state and generate solution reliability. The structural consistency verification and constraint fusion module is used to integrate the dam structural mechanical constraint model, perform consistency verification on the calculated displacements, and identify and correct abnormal node displacements. The Future Displacement Prediction and Risk Assessment module is used to predict node displacement based on historical data and multi-model fusion strategies, calculate risk indicators, and implement graded early warning. The closed-loop feedback and intelligent control module is used to feed back the prediction and risk assessment results to the aforementioned modules, dynamically adjust the monitoring strategy, and generate monitoring reports and control instructions.

[0016] Compared with the prior art, the above-mentioned technical solution of the present invention has the following beneficial technical effects: This invention designs a GNSS real-time dynamic solution system and method for automated monitoring of dam deformation. By constructing an adaptive, closed-loop GNSS deformation monitoring system for dams, it achieves intelligent operation of the entire process from data acquisition and real-time solution to prediction and early warning. Firstly, by dynamically sensing node status and autonomously assigning reference roles, the dependence on fixed reference stations is eliminated, significantly improving the reliability and long-term stability of the solution benchmark. Secondly, by introducing a multi-dimensional state assessment and credibility mechanism, the solution process possesses self-identification and self-adjustment capabilities, effectively resisting environmental disturbances and single-point anomalies, ensuring the continuity and stability of the solution results. Robustness; by integrating structural mechanics constraints and multi-epoch solution data, the monitoring results not only have mathematical accuracy but also conform to the actual structural behavior of the dam, enhancing the engineering interpretability and credibility of deformation data; based on the displacement prediction and risk assessment mechanism of multi-model fusion, it can realize early warning and graded response of abnormal trends, providing forward-looking decision support for dam safety management; overall, the system forms a complete closed loop of perception, calculation, verification, prediction, and feedback, promoting the evolution of dam monitoring from passive recording to proactive early warning and adaptive control, significantly improving the level of automation, system robustness, and engineering practical value. Attached Figure Description

[0017] Figure 1 This is a flowchart of a GNSS real-time dynamic solution method for automated monitoring of dam deformation proposed in this invention; Figure 2 This is a system architecture diagram of a GNSS real-time dynamic calculation system for automated monitoring of dam deformation proposed in this invention. Detailed Implementation

[0018] Example 1, as Figure 1 As shown, the present invention proposes a GNSS real-time dynamic solution method for automated monitoring of dam deformation, the specific implementation steps of which are as follows: S1. By continuously sensing the operational status and calculation behavior of each monitoring node in the dam's GNSS (Global Navigation Satellite System) monitoring network, the original calculation reference construction process, which relied on manual experience and fixed reference stations, is transformed into a computable and evolvable data processing flow. Without pre-setting reference stations, based on multi-dimensional status information such as observation continuity, calculation stability, and environmental response, the calculation reference role is automatically formed and dynamically updated. This provides a stable, reliable, and long-term operational spatial reference system for subsequent real-time dynamic calculations. The specific implementation process is as follows: S11. By continuously collecting and quantifying the observation integrity, signal quality, changes in solution results, and environmental disturbance response of each GNSS monitoring node, a unified node operating state vector is constructed, transforming the node operating state from discrete empirical judgment to structured data expression. Specifically: During the operation of the dam deformation monitoring system, the raw observation behavior and solution feedback information of each GNSS monitoring node are continuously collected, and the node's operating status is comprehensively characterized within a sliding time window. To avoid distortion from a single indicator, the node status is described as a multi-source fused state vector. ; ; in, This represents the comprehensive operational status vector of the i-th GNSS monitoring node within the time window t, used to uniformly describe the status characteristics of the node in terms of observation, calculation, and environmental response. This represents the observation continuity index of node i within the time window t, which reflects the node's ability to stably provide effective raw GNSS observation data within the statistical period. The observation quality stability index of node i is derived from a comprehensive assessment of carrier phase noise, multipath effects, and signal fluctuations, and is used to reflect the reliability of the node's observation data. The short-term consistency index of the solution results of node i is used to measure the stability of the result changes of the node in the continuous epoch solution process. It is a key parameter that directly reflects the reliability of the solution. This represents the environmental disturbance response index of node i, used to describe the node's sensitivity to external environmental factors such as temperature changes, wind loads, and changes in shading. The historical role reliability index of node i is used to describe the contribution of this node to the overall solution stability when it serves as a reference node in the historical operation process; N represents the total number of GNSS epochs that should theoretically be collected within the statistical time window, which is determined by the sampling frequency and the time window length. This represents the observation validity discrimination value of node i in the kth epoch, where the value is 1 when the observations in this epoch are complete and the solution conditions are met, and 0 otherwise. S12. Based on the node state vector, normalization and trend smoothing are performed on different state characteristics, and stability indicators reflecting the continuity of the solution are extracted as a key focus. This extends node evaluation from the signal level to the solution effect level, directly serving the deformation monitoring target. Specifically: After obtaining the original state vectors of the nodes, in order to eliminate the influence of different dimensions and fluctuation amplitudes on the comprehensive judgment, the state components are standardized and trend smoothed to obtain state feature values ​​under a unified scale: ; Based on this, to accurately characterize the impact of node participation on the continuity of results, a solution continuity consistency feature is introduced to construct a core index for solution stability: ; in, This represents the original state value of node i in the j-th state dimension within the time window t, where j corresponds to different state indices such as observation continuity and solution stability. Represents the original state value The standardized state characteristic values ​​are used to eliminate the influence of different dimensions and numerical scales between different indicators; This represents the mean of the j-th state dimension across all nodes in the current monitoring network, used to describe the overall level of this state indicator; This represents the standard deviation of the j-th state dimension across all nodes in the current monitoring network, used to describe the dispersion of this state index among nodes; This represents the spatial coordinates of node i obtained by GNSS calculation in the kth epoch. These coordinates are used to reflect the calculation status of the node in that epoch. The vector norm is used to calculate the spatial variation between the coordinate solutions of adjacent epochs, thereby quantifying the stationarity of the solution results. Indicates core indicators of stability; S13. By constructing a comprehensive credibility function, multi-dimensional state features are integrated into a unified credibility evaluation, and a candidate set of reference nodes is adaptively generated based on the overall credibility distribution of the network. This ensures that the solution reference originates from the objective results of the network's operating state rather than a fixed configuration. Specifically: Based on multidimensional state features, a node comprehensive credibility function is constructed: ; Based on the credibility distribution, nodes are divided into multiple credibility level intervals, and nodes with high credibility levels are selected to form a candidate set of reference roles: ; in, represents the overall credibility score of node i within the time window t, used to quantify whether the node is suitable to assume the role of solution reference; m represents the number of state feature dimensions participating in the overall credibility calculation, and its value is equal to the number of state indicators contained in the state vector. This represents the weight coefficient of the j-th state feature in the comprehensive credibility calculation. This weight comes from the statistical analysis results of the degree of influence of the state feature on the solution error in the historical running data. This represents the set of candidate reference nodes automatically generated within time t, and the nodes in this set are preferentially assigned to the GNSS solution reference role. This represents the average comprehensive credibility score of all monitored nodes within the current time window, used to reflect the overall credibility level of the network. The standard deviation of the overall credibility score of nodes within the current time window is used to describe the dispersion of credibility among nodes; This represents an adaptive adjustment factor used to control the size of the candidate set of reference nodes. Its value is dynamically adjusted based on the number of nodes in the monitored network and the stability distribution. S14. By continuously monitoring changes in node credibility, an automatic adjustment and update mechanism for the solution role is established. The reference system is reconstructed promptly when node states deteriorate or the environment changes, thereby ensuring the stability, self-recovery capability, and solution reliability of the monitoring network during long-term operation. Specifically: After the candidate set of reference roles is formed, it is automatically assigned as the solution reference role for the current time period, while the remaining nodes participate in the relative solution as observation roles. At the same time, the confidence evolution of each node is continuously monitored during the real-time solution process, and a role update is triggered when the following conditions are detected: ; Then perform role restructuring: ; in, This represents the minimum credibility threshold for a node to participate in the reference role, and is set based on the credibility level of the reference node before it became unstable in historical operation. This represents the set of non-reference nodes, i.e., the set of observation nodes.

[0019] S2. By constructing a time-series evolution model of the reference node, dynamic self-calibration and long-term stable maintenance of the reference system are achieved, providing a reliable foundation for subsequent multi-epoch dynamic solutions. Through residual quantization, environmental disturbance correction, self-calibration increment generation, and closed-loop reliability assessment, a closed-loop continuous evolution process is formed to ensure that the reference field adapts to time and environmental conditions. The specific implementation process is as follows: S21. Perform statistical analysis of the relative coordinate residuals between the reference nodes, and quantify the minute evolution trend of the reference field using a weighted average with a sliding time window. This provides a quantitative basis for subsequent environmental correction and self-calibration, ensuring that the reference field dynamically reflects the stability and long-term structural changes between nodes. Specifically: For the set of reference nodes selected in step S1 Calculate the relative coordinate residuals between nodes within the sliding time window: ; Subsequently, a time-weighted moving average was applied to the residual sequence to quantify the minute evolutionary trends of the reference field: ; ; in, This represents the relative position residual between node i and node j at the kth epoch, reflecting the small offset between reference nodes. It is calculated from the reference node solution coordinates output in step S1 and the initial reference coordinates. The initial reference coordinates of node i are used as a static benchmark for comparison. The weight represents the weighted moving average weight, indicating that the importance of an epoch decays over time. This represents the weighted average residual between node i and node j within the time window; This represents the smoothing adjustment coefficient, which controls the rate of weight decay and is set through parameter tuning based on historical data or experience. S22. Introduce an environmental disturbance factor matrix to quantify the impact of temperature, wind speed, and obstruction on GNSS observations. Introduce disturbance corrections into the reference field residual sequence to eliminate non-structural offsets, ensuring that the reference field evolution truly reflects long-term structural change trends. Specifically: Considering the long-term, minor offset effects of the surrounding environment (such as changes in temperature, wind, humidity, and obstruction) on GNSS signals, an environmental disturbance factor matrix is ​​introduced.

[0020] The perturbation factor is then coupled with the reference node residual sequence: ; in, represents the environmental disturbance correction vector of node i at time t; L represents the number of environmental disturbance types, selected according to the monitoring environment, such as temperature, wind speed, humidity, occlusion, etc. This represents the value of the l-th type of environmental disturbance variable at time t; The sensitivity coefficient for type l environmental disturbances is obtained through historical data regression or statistical analysis. This represents the corrected inter-node residuals, after removing the effects of environmental disturbances; S23. Based on the corrected residual sequence, generate the self-calibration coordinate increment for each reference node and update the reference node coordinates to achieve continuous evolution of the reference system over time, maintain long-term structural stability, and prevent the accumulation of local anomalies from affecting the overall reference field. Specifically: Based on the calculated residuals and corrections, a self-correcting coordinate increment is generated for each reference node: ; Update reference node coordinates: ; in, This represents the self-correcting coordinate increment of node i at time t; This represents the number of reference nodes in the set, i.e., the set of highly reliable reference nodes output in step S1. This represents the updated reference coordinates of node i, derived from the old reference coordinates. It is obtained by adding the self-correcting increment; S24. Conduct an overall reliability assessment of the updated reference field, establish a closed-loop feedback by quantifying the consistency between the residuals and the self-calibration results, and automatically trigger node weight adjustment or reference node replacement when the reliability falls below a threshold to ensure the long-term stability and reliability of the reference system. Specifically: After generating the self-calibrating reference node coordinates, a reliability assessment is further performed on the entire reference field evolution process: ; like This triggers the reselection of reference nodes or the redistribution of weights, ensuring that the overall reliability of the reference field's evolution remains at a high level. in, It represents the overall reliability of the reference field evolution and measures the stability of the reference system in the current time window. It is calculated by correcting the residuals between all reference nodes. This indicates the minimum threshold for the set reference field confidence level.

[0021] S3. Through cross-epoch coupling calculation, the self-calibrated reference field generated in step S2 is integrated with the observation node data to achieve real-time, continuous, and robust calculation of dam node displacements. At the same time, node calculation reliability is generated and fed back to node role management to achieve long-term adaptive closed-loop optimization. The specific implementation process is as follows: S31. Construct a multi-epoch state vector, organize the three-dimensional coordinates of all monitoring nodes according to the time series, predict the current epoch state using historical displacement and reference field evolution trends, provide prior constraints for subsequent coupled solutions, and achieve continuity and long-term stability of node solutions. Specifically: Based on the self-calibrating reference node set output in step S2 And its temporal evolution parameters, representing the coordinates of all nodes in the dam monitoring network as a multi-epoch state vector: ; Construct a cross-epoch prediction model that uses historical displacement and reference field evolution trends to predict the current epoch state: ; in, The node state vector at the k-th epoch contains the coordinates of all monitored nodes; M represents the number of epochs within the sliding time window, which determines the cross-epoch coupling length. This represents the predicted state vector for the k-th epoch, which is predicted based on the state of the previous epoch and the evolution of the reference field. It represents the state transition matrix, that is, the evolution relationship of the node state between epochs, which can be dynamically adjusted to reflect the behavior trend of the dam structure. This represents an external input vector used to introduce changes in environmental or structural loads (such as the effects of temperature gradients and wind loads on displacement). Represents the process noise vector, with a zero-mean Gaussian distribution, quantifying unobservable minute perturbations or modeling errors; This represents the state vector of the (k-1)th epoch after observation and update; S32. Construct observation equations, couple real-time observation node data with self-calibrating reference node coordinates to form a dynamic gain matrix. By adjusting the observation noise covariance, adaptive node weights are achieved, ensuring that the solution simultaneously considers both short-term accuracy and long-term stability of the reference field constraints. Specifically: For each epoch k, the observation node data and the reference node self-calibration coordinates are incorporated into the observation equation: ; Then, the dynamic coupling gain matrix is ​​constructed. : ; in, This represents the observation vector at the k-th epoch, which includes the GNSS measurement coordinates of the observation node and the self-calibrated coordinates of the reference node; The observation matrix represents the mapping of node state vectors to the observation space. This represents the observation noise vector, reflecting GNSS signal errors, multipath effects, and local interference. This represents the Kalman gain matrix, which determines the contribution of the observations to the state update; Let represent the prior prediction error covariance matrix of the k-th epoch, which characterizes the uncertainty of the predicted state and is generated by the state transition process; The observation noise covariance matrix is ​​dynamically adjusted based on the historical solution confidence of the nodes, with the weights increasing when the node confidence is high. S33. Update the node state vector using the gain matrix and calculate the solution residual sequence; evaluate the node solution stability through time-weighted residual analysis, identify long-term offsets or local anomalies, and provide a basis for node weight adjustment and solution reliability generation, achieving continuous and robust multi-epoch dynamic solution, specifically: Under the influence of the gain matrix, the state vector is updated in real time: ; Simultaneously calculate the solution residuals: ; Time-weighted analysis of the residual sequence: ; ; in, This represents the state vector updated in the k-th epoch. This represents the residual of the i-th node at the k-th epoch; This represents the weighted average of the node residuals over time, used to analyze the trend of node status. This represents a weighting factor, making recent epochs have a greater impact; S34. Generate the solution reliability of each node based on the residual evolution and provide feedback. This allows for dynamic adjustment of the allocation of reference and observation nodes, achieving long-term closed-loop optimization and ensuring stable and highly reliable solution results. It also supports adaptive handling of abnormal nodes, improving overall reliability. Specifically: After completing the state update, the system generates the solution confidence level for each node: ; The solution confidence level is fed back to step S1 to dynamically adjust the allocation of reference nodes and observation nodes. For nodes with low confidence level, measures can be taken such as reducing their weight, delaying their participation in the solution, or replacing them with nodes with high confidence level to ensure the overall stability of the solution results. in, This represents the reliability of the solution for the i-th node. The closer it is to 1, the more reliable the solution is. Its range is [0,1]. This represents the residual difference between nodes, used to measure the degree to which a node deviates from the overall trend.

[0022] S4. Couple the multi-epoch GNSS solution results from step S3 with the dam's structural mechanical constraints. Through closed-loop verification, abnormal node identification, and dynamic correction, achieve highly reliable and stable displacement monitoring, and generate a network-wide consistency assessment and node weight adjustment to realize a long-term adaptive monitoring closed loop. The specific implementation process is as follows: S41. Construct a structural mechanics constraint model, forming a constraint matrix by relating the displacements of each node of the dam to structural stiffness, material elasticity, and environmental load. Use weighted coefficients to reflect the sensitivity of different dam sections, providing a physical benchmark for subsequent residual calculations and displacement corrections. Specifically: The dam is decomposed into several structural units (the entire dam body, dam sections, dam crest, dam base, etc.), with monitoring nodes within each unit. Associated with elasticity models; The constraint equations for elasticity are defined as follows: ; Different weighting coefficients are introduced for different dam sections. To reflect the differences in local structural sensitivity, a weighted structural constraint is formed, with high-sensitivity areas or key monitoring points assigned high weights to ensure that the correction process focuses on key structural locations. ; Where S represents the dam structural mechanical constraint matrix, which is used to associate the nodal displacement vector with the structural constraint conditions; This represents the target vector of the force / displacement constraint at the node, which characterizes the theoretical displacement that the dam structure is allowed to undergo in the current epoch. The node weighting coefficient represents the sensitivity of different nodes to structural constraints and is set according to the material properties, local structural sensitivity, and importance of the dam section. Represents the weighted structure constraint matrix; Represents a diagonal matrix; S42. Calculate the residuals between the displacements of each node and the structural constraints, and perform a weighted analysis based on the node reliability from step 3 to identify anomalous nodes, distinguish between noise disturbances and actual structural deviations, and provide a basis for closed-loop correction. Specifically: Calculate the residuals between the solution displacement and the structural constraints: ; Perform statistical analysis on the residuals of each node, and combine it with the node confidence level output in step S3. : ; When the node confidence level is low, the residual is amplified to identify potential abnormal nodes; Nodal residuals When the value exceeds the set threshold, it is marked as an abnormal node, triggering a local correction. in, This represents the nodal displacement residual vector, characterizing the deviation between the solved displacement and the structural constraints; This represents the weighted residual, which is used to adjust the residual size in conjunction with the confidence level of node S3 in step S3. S43. Perform weighted least squares closed-loop correction on the identified abnormal nodes, and dynamically update the node weights. Use high-confidence nodes for local reference and reduce the weights of low-confidence nodes to achieve adaptive optimization of node roles. Specifically: The identified anomalous nodes are corrected by loop closure using weighted least squares method: ; ; ; Nodes with low weight have little impact on the correction results, while high-confidence nodes dominate the correction. Update node role management: reduce the weight of abnormal nodes to decrease their contribution to subsequent calculations; high-confidence nodes can temporarily serve as local references to enhance the stability of corrections; in, represents the node displacement state vector after closed-loop correction, that is, the node displacement state vector of the i-th node after closed-loop correction in the k-th epoch; W represents the node weight matrix. The corrected weight for node i is calculated by combining the node's credibility and the residual size. This represents the magnitude of the residual at node i in the current epoch, i.e., the deviation between the solved displacement and the structural constraints. This represents the weighting coefficient of historical credibility in the weighting calculation; This represents the adjustment coefficient of the residuals to the weight calculation; S44. Evaluate the consistency of the corrected displacement across the entire network, determine the overall network status through consistency indices, and feed the correction results back to the reference field dynamic evolution model to achieve long-term stability maintenance and trend anomaly labeling, providing a basis for subsequent security assessments. Specifically: Calculate the overall network displacement consistency index: ; ; like If the value is less than the set threshold, a reference node update or node weight redistribution will be triggered. Long-term maintenance strategy: Use the corrected displacement as input for the reference field update in step S2 to further improve the dynamic evolution model; mark the abnormal nodes in the long-term trend for subsequent safety assessment and maintenance decision-making. in, This represents the overall network displacement consistency index, with a value range of [0,1]. This represents the self-calibrated coordinates of reference field node i after correction; This represents the self-calibrated coordinates of the reference field node j after correction; This represents the average distance between reference field nodes, used for normalization to ensure... .

[0023] S5. By integrating the node displacements corrected in step S4, multi-epoch historical trend data, and environmental disturbance information, a multi-model fusion prediction system is constructed to achieve future displacement prediction, anomaly risk assessment, and graded early warning for dam nodes. Simultaneously, the prediction results are fed back in a closed loop to the node management and solution system, enabling real-time dynamic early warning and intelligent control, forming a complete, highly reliable monitoring closed loop. The specific implementation process is as follows: S51. Collect the corrected displacement sequences and node confidence scores from the entire network to form a sliding time window dataset. Extract short-term trends, long-term accelerations, residual variances, and confidence-weighted indices to provide quantitative inputs and feature foundations for subsequent multi-model predictions. Specifically: Collect the full network corrected displacement sequence output in step S4 With node credibility This forms a historical dataset within a sliding time window: ; Extract key feature indicators: Short-term displacement velocity: ; Displacement acceleration: ; Residual variance: ; Measure the fluctuation range of nodes, combined with a credibility weighting: ; in, This represents the historical displacement sequence and confidence set within the sliding time window of node i, with a time window length of M. This represents the short-term displacement velocity of node i in epoch k, i.e., the current deformation rate. Indicates the time interval between epochs; This represents the acceleration of node i in epoch k, reflecting the second-order trend of displacement change; This represents the residual variance of node i within the time window, measuring the magnitude of node fluctuation. The weighted index representing the speed characteristic of node i is adjusted by combining the node's credibility. S52. Construct a multi-model fusion system comprising a weighted ARMA model, a structural mechanics coupling model, and a machine learning prediction model. By adaptively adjusting weights based on historical errors, this system achieves a fusion of short-term accuracy prediction and long-term trend prediction for nodes, balancing nonlinear pattern recognition and engineering feasibility. Specifically: For each node, a multi-model prediction system is constructed, including: ARMA model: Predicts short-term periodic fluctuations by using historical node displacements and residuals within a sliding time window. ; Structural mechanics coupled model: Predicting long-term trends based on structural constraints and environmental loads (such as water level, temperature, and wind load) in step S4: ; Machine learning nonlinear models: using LSTM to capture nonlinear historical patterns: ; The three types of prediction results are then merged according to dynamic weights: ; ; in, This represents the short-term predicted displacement of node i in the next epoch, generated by a weighted autoregressive moving average model. This represents the autoregressive coefficient, obtained by fitting historical data. This represents the moving average coefficient, i.e., the weight of the historical impact of the residuals; This represents the prediction residual of node i at epoch k-q+1; This indicates that the structural mechanics coupled model predicts displacement. This indicates external environmental disturbances, such as changes in water level, temperature gradients, and wind loads. This represents a mapping function that combines the structural stiffness matrix and nodal constraints. It represents the prediction results of machine learning models and captures nonlinear and complex patterns; , and Indicates the weights for multi-model prediction; This indicates the historical prediction error of each model; This represents the fused predicted displacement result of node i in the next epoch; S53. Combine predicted displacement with safe displacement thresholds to calculate node risk indicators, and generate network-wide risk by weighting nodes according to their reliability. Classify warning levels into green, yellow, orange, and red based on these indicators to achieve dynamic quantitative warnings and potential anomaly identification. Specifically: Calculate the risk indicator of future node displacement exceeding the threshold: ; Generate network-wide risk indicators by weighting nodes based on their trustworthiness: ; Early warning levels are determined based on the indicators: ; in, This indicates the risk index of node i's future displacement exceeding the threshold. The safe displacement threshold of node i is determined by dam design, historical deformation statistics, and material properties. This represents a credibility-weighted average of all network risk indicators. S54. The predicted displacement and risk indicators are fed back in a closed loop to node management, multi-epoch calculation, and structural consistency verification. Node roles and weights are dynamically adjusted to generate intelligent early warning reports and engineering handling suggestions, achieving proactive control and long-term monitoring closed-loop maintenance. Specifically: The predicted displacement and risk indicators are fed back to steps S1-S4 to achieve closed-loop optimization. Step S1 (Node Management): Nodes with high risk can have their roles or weights temporarily adjusted; Step S3 (Multi-epochal solution): Future trend information can be used for predictive correction; Step S4 (Structural Consistency Verification): Risk nodes are given priority in mechanical constraint correction; Automatically generate early warning reports, including: the predicted future displacement curve of the node; the location of the abnormal node and the potential risk level; and engineering treatment suggestions (such as increasing the monitoring frequency or local reinforcement).

[0024] Example 2, as Figure 2 As shown, the present invention proposes a GNSS real-time dynamic calculation system for automated monitoring of dam deformation, which is used to execute the GNSS real-time dynamic calculation method for automated monitoring of dam deformation proposed in Embodiment 1. It includes: a node monitoring and state perception module, a multi-epoch real-time dynamic calculation module, a structural consistency verification and constraint fusion module, a future displacement prediction and risk assessment module, and a closed-loop feedback and intelligent control module.

[0025] The node monitoring and status awareness module is responsible for real-time status awareness and management of each GNSS monitoring node of the dam, including node operation status, data quality and reliability assessment; the module dynamically adjusts the solution priority and participation weight of different nodes through an adaptive role allocation mechanism to realize the intelligent construction of the monitoring network; the module outputs historical data records, real-time displacement data and node credibility of each node, providing basic data and weight information for subsequent solution and prediction. The multi-epoch real-time dynamic solution module performs dynamic solution with a sliding time window based on node status and historical data. By performing multi-epoch fusion and weighted processing on node observations, it achieves high-precision correction of the entire network displacement. The module can automatically identify abnormal observations and adjust node weights while maintaining the real-time performance and continuity of the solution. The output is the current corrected displacement result for each node and the updated node confidence level, providing input for structural consistency verification. The structural consistency verification and constraint fusion module combines the geometric constraints of the dam with the corrected displacement data to verify the consistency of the multi-epoch solution results. By analyzing the mutual constraints and physical constraints between nodes, the module automatically corrects abnormal deviations and local drifts, ensuring that the solution results of the entire network conform to the structural characteristics of the dam. The output is the corrected displacement data and the final credibility of each node, providing a reliable basis for prediction and risk analysis. The future displacement prediction and risk assessment module utilizes historical correction data and node reliability, combined with a multi-model fusion strategy (statistical model, structural mechanics model, and machine learning model), to make short-term and long-term predictions of the future displacement of each node. Simultaneously, the module calculates dynamic risk indicators based on the predicted node displacement and safety thresholds, and generates network-wide risk assessment results and early warning levels, providing a basis for closed-loop control decisions. This module enables the quantification of dam deformation trends and early warning of abnormal behavior. The closed-loop feedback and intelligent control module feeds back the prediction results and risk levels to the node management, multi-epoch calculation and structural verification modules to achieve closed-loop control of the system. The module can automatically generate monitoring reports, early warning notifications and node optimization scheduling schemes, and also supports adjusting the data acquisition frequency, node roles and key monitoring areas.

[0026] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A GNSS real-time dynamic solution method for automated monitoring of dam deformation, characterized in that, The specific implementation steps include the following: S1. Dynamically perceive the operating status and calculation behavior of each monitoring node in the GNSS monitoring network, and automatically form and dynamically update the calculation reference role based on the multi-dimensional status information of the node's observation continuity, calculation stability and environmental response. S2. Construct a time-series evolution model of reference nodes, perform statistical analysis and quantification of the relative coordinate residuals between selected reference node sets, introduce environmental disturbance factors for correction, generate self-correcting coordinate increments to update reference node coordinates, and conduct an overall credibility assessment of the updated reference field. S3. Perform cross-epoch coupled solution, construct multi-epoch state vectors and use the evolution trend of the reference field to predict the state, couple the real-time observation node data with the self-calibrating reference node coordinates to construct a dynamic gain matrix, update the node state vectors and calculate the solution residuals, and generate node solution credibility. S4. Couple the multi-epoch GNSS solution with the dam's structural mechanical constraints, construct a structural mechanical constraint model and calculate the displacement residuals. Combine the node credibility to identify abnormal nodes and perform weighted least squares closed-loop correction. At the same time, dynamically update the node weights and evaluate the consistency of the corrected displacement across the entire network. S5. Based on the corrected node displacement sequence, historical trend data and environmental disturbance information, a multi-model fusion prediction system is constructed to predict future displacements. The risk index of future node displacement exceeding the threshold is calculated and the network-wide risk index is generated by weighting according to the node credibility. The warning level is divided according to the index, and the prediction results are fed back to the node management and solution process in a closed loop.

2. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 1, characterized in that, In step S1, the operational status and calculation behavior of each monitoring node in the dynamic sensing GNSS monitoring network specifically include: The observation integrity, signal quality, changes in solution results, and response to environmental disturbances of each GNSS monitoring node are continuously collected and quantified to construct a unified node operation status vector. Standardize and smooth the state components in the original state vector of the node to obtain state feature values ​​under a unified scale, and introduce the solution continuity consistency feature to construct the core index of solution stability. A node comprehensive credibility function is constructed based on multi-dimensional state features, which integrates multi-dimensional state features into a unified credibility evaluation, and adaptively generates a candidate set of reference nodes based on the overall credibility distribution of the network. Continuously monitor changes in node credibility. When the credibility of a reference node falls below the minimum threshold or the credibility of a non-reference node exceeds the average credibility of the reference node, trigger automatic adjustment and update of the solution role.

3. The GNSS real-time dynamic solution method for automated monitoring of dam deformation according to claim 2, characterized in that, The node's running state vector is specifically as follows: The node operating state vector includes observation continuity index, observation quality stability index, short-term consistency index of solution results, environmental disturbance response index, and historical role reliability index. Among them, the observation continuity index reflects the node's ability to stably provide effective raw GNSS observation data; the observation quality stability index is derived from a comprehensive assessment of carrier phase noise, multipath effects, and signal fluctuations; the short-term consistency index of the solution results is used to measure the stability of the node's results during continuous epoch solution processes; the environmental disturbance response index is used to describe the node's sensitivity to external environmental factors; and the historical role reliability index is used to describe the degree of contribution of the node to the overall solution stability when it served as a reference node during historical operation.

4. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 3, characterized in that, Step S2, constructing the reference node temporal evolution model specifically includes: The relative coordinate residuals between nodes in the reference node set are calculated within a sliding time window, and a time-weighted moving average is applied to the residual sequence to quantify the small evolution trend of the reference field. An environmental disturbance factor matrix containing environmental disturbance variables such as temperature, wind speed, humidity, and shading type is introduced, and the environmental disturbance correction vector is coupled with the reference node residual sequence to eliminate unstructured offsets. Based on the corrected residual sequence, a self-correcting coordinate increment is generated for each reference node, and the reference node coordinates are updated accordingly. An overall credibility assessment is performed on the updated reference field. When the overall evolution credibility of the reference field obtained from the assessment is lower than a set threshold, a reference node reselection or weight reallocation is triggered.

5. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 4, characterized in that, In step S3, the cross-epoch coupled solution specifically includes: Based on the self-calibrating reference node set and its temporal evolution parameters, the coordinates of all monitoring nodes are organized into a multi-epoch state vector according to the time series. Construct a cross-epoch prediction model and use the state transition matrix to predict the current epoch state by applying the historical displacement and the evolution trend of the reference field. An observation equation is constructed to couple real-time observation node data with self-calibrating reference node coordinates, and a dynamic coupling gain matrix is ​​constructed. Among them, the observation noise covariance matrix is ​​dynamically adjusted according to the reliability of the node's historical solution; The node state vector is updated using the gain matrix, and the solution residual sequence is calculated. Time-weighted analysis is performed on the residual sequence to evaluate the node solution stability. The solution reliability of each node is generated based on the residual evolution, and this solution reliability is fed back to dynamically adjust the allocation of reference nodes and observation nodes.

6. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 5, characterized in that, Step S4, coupling the multi-epoch GNSS solution results with the dam structural mechanical constraints, specifically includes: The dam is decomposed into several structural units, and elastic mechanical constraint equations are defined to associate nodal displacement vectors with structural constraint conditions. Different weighting coefficients are introduced for different dam sections and key monitoring points to form weighted structural constraints. Calculate the residuals between the solved displacements of each node and the weighted structural constraints, and perform weighted analysis based on the node solution reliability to identify abnormal nodes; The identified abnormal nodes are corrected using the weighted least squares method, where the correction weight of a node is calculated by combining its node confidence and residual size. Calculate the network-wide corrected displacement consistency index. If the index is lower than the set threshold, trigger the reference node update and node weight redistribution, and use the corrected displacement as the input to the reference field dynamic evolution model.

7. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 6, characterized in that, Step S5, constructing a multi-model fusion prediction system specifically includes: Collect the corrected displacement sequence and node credibility of the entire network to form a sliding time window historical dataset, and extract key feature indicators such as short-term displacement velocity, displacement acceleration and residual variance from it, and weight the features in combination with node credibility; For each node, a multi-model prediction system is constructed, which includes a weighted ARMA model, a structural mechanics coupling model, and a machine learning nonlinear model. The weighted ARMA model uses the historical displacements and residuals of nodes within a sliding time window to predict short-term cyclical fluctuations. The structural mechanics coupling model predicts long-term trends based on structural constraints and environmental loads; Machine learning nonlinear models use LSTM to capture nonlinear historical patterns; The prediction results of the three types of models are dynamically weighted according to their historical prediction errors and then fused to obtain the node fusion prediction displacement result.

8. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 7, characterized in that, The machine learning nonlinear model is a long short-term memory network model, used to capture the nonlinear and complex time-dependent patterns in the historical data of node displacements.

9. The GNSS real-time dynamic calculation method for automated monitoring of dam deformation according to claim 8, characterized in that, The specific indicators used to classify early warning levels include: Based on the fusion prediction displacement results of the nodes and their corresponding safe displacement thresholds determined by dam design, historical deformation statistics and material properties, the risk index of future displacement exceeding the threshold of the nodes is calculated. The risk indicators of all nodes in the network are weighted and averaged according to node credibility to generate the network-wide risk indicators. Based on the numerical range of risk indicators across the entire network, warning levels are divided into green, yellow, orange, and red.

10. A GNSS real-time dynamic calculation system for automated monitoring of dam deformation, used to execute the GNSS real-time dynamic calculation method for automated monitoring of dam deformation as described in any one of claims 1 to 9, characterized in that, include: The node monitoring and status awareness module is used to perceive the operating status and data quality of each GNSS monitoring node in real time, and realize the node credibility calculation and dynamic allocation and adjustment of reference roles based on multi-dimensional status assessment. The multi-epoch real-time dynamic solution module is used to perform cross-epoch coupled solution based on the self-calibrated reference field and observation data, update the nodal displacement state and generate solution reliability. The structural consistency verification and constraint fusion module is used to integrate the dam structural mechanical constraint model, perform consistency verification on the calculated displacements, and identify and correct abnormal node displacements. The Future Displacement Prediction and Risk Assessment Module is used to predict node displacement based on historical data and multi-model fusion strategies, calculate risk indicators, and implement graded early warning. The closed-loop feedback and intelligent control module is used to feed back the prediction and risk assessment results to the aforementioned modules, dynamically adjust the monitoring strategy, and generate monitoring reports and control instructions.

Images