A dam safety monitoring method based on digital twinning

Abstract

The application belongs to the technical field of dam safety early warning, and provides a dam safety monitoring method based on digital twinning to solve the problems of weak prediction ability and early warning lag of existing monitoring methods. The method comprises the following steps: deploying multiple types of sensors to collect multi-dimensional physical state data; based on BIM, Internet of Things and finite elements, a digital twinning model containing a multi-physical field coupling simulation sub-model and a data-driven correction model is established; inputting historical data to drive the model to predict the output state trend sequence and the state data value at a certain time; performing difference analysis to construct an error vector, calculating and correcting the deviation index; calculating the structure health score, and if the score exceeds the threshold, an early warning is given. The method has the advantages of improving prediction ability, enhancing adaptability, realizing comprehensive evaluation, and providing a system-ahead decision basis for safety management.

Description

Technical Field

[0001] This invention belongs to the field of dam safety early warning technology, and in particular relates to a dam safety monitoring method based on digital twins. Background Technology

[0002] Dams are critical water infrastructure, and their long-term operational status directly affects the safety of life and property in downstream areas. Existing dam safety monitoring methods primarily rely on real-time acquisition of physical parameters such as displacement, stress, and water level, and anomaly detection based on static thresholds. These methods suffer from weak predictive capabilities, delayed early warnings, and poor coupling with the physical structure, making them ill-suited for monitoring the evolution of dam structural states in complex environments. In recent years, digital twin technology has been gradually applied in industrial and engineering fields. By constructing virtual models that map to the actual structure, it enables visualization, predictability, and feedback control of the system's operational status. Introducing digital twins into dam safety monitoring can integrate finite element simulation with real-time data-driven approaches, establishing predictive models oriented towards structural evolution. The difference between the model output and actual monitoring data reflects the structural health level. However, existing methods still lack systematicity in multiphysics coupled simulation, adaptive parameter correction, and quantitative analysis of deviation indicators, making it difficult to establish unified health assessment standards. Summary of the Invention

[0003] In view of the technical problems existing in the background art, the present invention proposes a dam safety monitoring method based on digital twin.

[0004] To achieve the above objectives, the technical solution adopted by the present invention includes the following steps:

[0005] S1. Deploy multiple types of sensors in the dam structure and its surrounding environment. The sensors include displacement sensors, stress sensors, and water level monitors, which are used to collect multi-dimensional physical state data of the dam during operation.

[0006] S2. Establish a digital twin model based on BIM, IoT and finite element method. The digital twin model includes a multi-physics coupled simulation sub-model of the dam and data-driven correction.

[0007] Among them, the multi-physics field coupled simulation sub-model of the dam is used to reflect the dynamic evolution of each monitoring index over time under normal conditions;

[0008] Data-driven correction is used to adaptively fit and dynamically correct the parameters of the simulation sub-model by combining historical data;

[0009] S3. Acquire historical monitoring data and input it into the digital twin model to drive the digital twin model to make prediction outputs, the outputs including a first output and a second output;

[0010] The first output is to predict and generate a future preset cycle state trend sequence, and extract the volatility intensity factor and trend derivative factor from the trend sequence;

[0011] The second output is to predict and generate the state data value at a certain moment within a preset future period;

[0012] S4. Perform a difference analysis between the predicted state data value of the second output and the historical stable state data, and construct a multi-dimensional error mapping vector;

[0013] S5. Based on the error mapping vector, calculate the displacement deviation index, stress deviation index and water level deviation index respectively. Use the first output fluctuation intensity factor and trend derivative factor to correct the displacement deviation index, stress deviation index and water level deviation index to obtain the corrected displacement deviation correction index, stress deviation correction index and water level deviation correction index.

[0014] S6. Calculate the structural health score based on the displacement deviation correction index, stress deviation correction index and water level deviation correction index, and compare the structural health score with the preset safety threshold. If the score is higher than the threshold, it is determined that there is a structural anomaly risk in the dam and the corresponding early warning information is output.

[0015] Preferably, the construction process of the multiphysics coupled simulation sub-model of the dam in step S2 includes the following steps:

[0016] A finite element discrete model reflecting the dam structure was established, the dam structure was spatially meshed, and material parameters and boundary conditions were assigned to each element.

[0017] A coupled multiphysics solution module is constructed, and the response coupling between multiple fields is achieved by setting the boundary conditions and loading paths for multi-field interactions;

[0018] Set historical typical working conditions as simulation input conditions, execute the initial simulation, output the theoretical response values ​​of each monitoring point, and compare them with the measured historical data to extract residual features;

[0019] The extracted residual features are used to fit and adjust the material constitutive parameters and boundary coupling coefficients in the model, thus completing the initialization and adjustment of the simulation sub-model.

[0020] Preferably, the data-driven correction step of adaptive fitting and dynamic correction of the parameters of the simulation sub-model based on historical data includes:

[0021] The measured status data of the dam during multiple historical operating cycles were obtained, including stress, displacement, and seepage pressure time series data, and the data were normalized.

[0022] The residual sequences between the historical measured data and the corresponding predicted outputs of the simulation sub-models are calculated respectively, and the residual fluctuation curves are constructed based on the sliding time window;

[0023] For the monitoring area in the residual fluctuation curve where the offset trend is greater than the set threshold, a residual feedback function is constructed to map the residual mean, rate of change and local volatility into parameter correction factors;

[0024] The parameter correction factor is applied to the corresponding physical unit parameters in the simulation sub-model to complete the parameter fine-tuning.

[0025] Preferably, the residual feedback function is implemented as follows: ,in As a weighting factor, The baseline reference value is obtained by fitting historical data. These represent the residual mean, the residual rate of change, and the residual standard deviation, respectively.

[0026] Preferably, in step S3, driving the digital twin model to make predictions, the output including a first output and a second output, includes the following steps:

[0027] Historical monitoring data is input into the digital twin model to drive multiphysics response calculations in the simulation solution, thereby obtaining the current structural state;

[0028] Based on the current structural state, a multi-time predicted state sequence within a preset future period is generated using a sliding prediction window. The predicted state sequence is represented as follows: ,in This represents the predicted physical index state value at time t+i, where T is the prediction period length.

[0029] The predicted state sequence is used as the first output, from which the volatility intensity factor and trend derivative factor are extracted.

[0030] The trend derivative factor is calculated as follows: ,in represents the trend derivative factor, and j represents the j-th physical index;

[0031] The fluctuation intensity factor is calculated as follows: ,in, Indicates volatility intensity factor. This represents the predicted value of the j-th physical index at time t+i.

[0032] Finally, a preset target time point t+k is selected from the predicted state sequence, and the corresponding predicted vector x(t+k) is extracted as the second output, which is used to perform multi-dimensional difference comparison with the historical stable state vector.

[0033] Preferably, the construction of the multidimensional error mapping vector in step S4 is implemented by using the multidimensional state data vector of the future time predicted by the second output. Reference value vector under historical stable operating conditions By subtracting each dimension, the error vector is obtained. The error vector includes displacement deviation, stress deviation, and water level deviation.

[0034] Preferably, step S5, which involves obtaining the displacement deviation index, stress deviation index, and water level deviation index based on the error mapping vector, includes:

[0035] The exponent of the difference between the current predicted displacement and the historical stable displacement reference value in the error mapping vector is defined as the displacement deviation index. ;in, The displacement deviation index. ;

[0036] Set stress safety threshold Calculate the stress deviation index ,in, The stress deviation index, ;

[0037] Calculate the water level deviation index ,in, This refers to the water level deviation index. ,in, This represents the standard reference water level.

[0038] Preferably, the calculation method for correcting the displacement deviation index, stress deviation index, and water level deviation index using the first output fluctuation intensity factor and trend derivative factor to obtain the corrected displacement deviation index, stress deviation index, and water level deviation index is as follows: ,in, ,in This is to avoid division by zero. Represented as the j-th type deviation correction index, including , This represents the intermediate term used for correction.

[0039] Preferably, the structural health score in step S6 is calculated as the normalized value of the product of the displacement deviation correction index, the stress deviation correction index, and the water level deviation correction index.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] 1. Based on BIM and finite element method, a multi-physics coupled simulation sub-model is constructed, which integrates stress and seepage dynamic interaction. Through historical working condition simulation and residual correction parameters, the dynamic evolution law of monitoring indicators can be accurately simulated, breaking through the limitations of traditional single threshold judgment and improving the ability to predict the trend of structural state.

[0042] 2. By constructing a residual feedback function using multi-cycle measured data, the residual characteristics are mapped to parameter correction factors, dynamically fine-tuning the material constitutive and boundary coefficients, automatically compensating for parameter drift caused by environmental disturbances, solving the problem of long-term prediction error accumulation in static models, and enhancing adaptability under complex working conditions.

[0043] 3. Construct a multi-dimensional error vector, combine fluctuation intensity and trend derivative factor to correct the deviation index, and use a normalized score to comprehensively assess the health status. When the score exceeds the threshold, output early warning information containing abnormal types in real time, breaking through the limitations of independent parameter analysis and realizing the transformation from lagging response to trend prediction, providing a systematic and advanced decision-making basis for safety management. Attached Figure Description

[0044] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0045] Figure 1 This is a flowchart illustrating the overall structure of a dam safety monitoring method based on digital twins. Detailed Implementation

[0046] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described below in conjunction with the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0047] Numerous specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways than those described herein, and therefore the invention is not limited to the specific embodiments disclosed in the following specification.

[0048] In practice, traditional dam safety monitoring methods primarily rely on real-time acquisition of physical parameters such as displacement, stress, and water level, and anomaly detection based on static thresholds. However, this method has significant drawbacks: firstly, its predictive ability is weak, only allowing for simple comparison of current data and failing to effectively predict the future evolution of the dam's structural state; secondly, its early warning is delayed, often only issuing an alarm after the abnormal state has reached a certain scale, missing the optimal time for intervention. Therefore, this invention proposes a dam safety monitoring method based on digital twins, the specific implementation process of which is as follows: Figure 1 As shown.

[0049] Firstly, to comprehensively and in real-time acquire physical state information during dam operation and provide a solid data foundation for subsequent construction and analysis of the digital twin model, a scheme of deploying multiple types of sensors on the dam structure and its surrounding environment was adopted. Specifically, multiple types of sensors, including displacement sensors, stress sensors, and water level monitors, were deployed on the dam structure and its surrounding environment to collect multi-dimensional physical state data of the dam during operation. Through the coordinated deployment of multiple types of sensors, multi-dimensional data acquisition of the dam's operating status was achieved, enabling real-time and comprehensive reflection of the dam's physical state under different operating conditions.

[0050] Next, in order to accurately simulate the dynamic evolution of various monitoring indicators of the dam under normal conditions and to truly reflect the response characteristics of the dam structure under the coupling effect of multiple physics fields, a digital twin model was constructed using a technical solution based on BIM, IoT and finite element method. The digital twin model includes a multi-physics field coupling simulation sub-model of the dam and a data-driven correction. The multi-physics field coupling simulation sub-model of the dam is used to reflect the dynamic evolution of various monitoring indicators under normal conditions, while the data-driven correction is used to adaptively fit and dynamically correct the parameters of the simulation sub-model by combining historical data.

[0051] The construction process of the multiphysics coupled simulation sub-model of the dam includes the following steps: establishing a finite element discrete model reflecting the dam structure entity; dividing the dam structure into spatial meshes and assigning material parameters and boundary conditions to each element; constructing a coupled multiphysics solution module, achieving response coupling between multiple fields by setting multi-field interaction boundary conditions and loading paths; setting historical typical working conditions as simulation input conditions, executing initial simulation, outputting the theoretical response values ​​of each monitoring point, and comparing them with measured historical data to extract residual features; and using the extracted residual features to fit and adjust the material constitutive parameters and boundary coupling coefficients in the model to complete the initial adjustment of the simulation sub-model. Specifically, firstly, the solid geometric structure is established based on the dam design drawings, construction records, and BIM model, and the dam structure is spatially discretized using the finite element modeling method. Specifically, the overall structure is divided into multiple calculation units according to the stress characteristics of the dam structure, and adaptive element types are selected for mesh generation. Subsequently, material constitutive property parameters are assigned to each finite element element, and boundary conditions reflecting the boundary stress state are set, including foundation constraint boundaries, water pressure and seepage pressure boundaries, etc. Based on the established finite element mesh of the structure, a multiphysics solution including stress field and seepage field is constructed. By defining the interaction relationship between the boundaries of the fields, the loading path and coupling formula between different physical fields are established. At the same time, an explicit joint solution strategy is adopted to iterate the influence of each field variable synchronously during the solution process, and a time stepping mechanism is introduced to simulate the dynamic response of the structure during long-term operation. Then, representative historical typical working conditions are selected as simulation input conditions. These working conditions may include changes in operating water level and dam load at a certain period. These time series data are used as boundary loading inputs to drive the constructed multiphysics simulation model for initial simulation. During the simulation, the system will output the theoretical response values ​​of all monitoring points at each time node, including indicators such as displacement, stress, and pore water pressure. To facilitate subsequent deviation analysis, these theoretical prediction values ​​will be aligned with the historical measured data of the same period point by point, and the difference will be compared and the residual sequence will be recorded. Finally, the residual sequence between the aforementioned simulation output and measured values ​​is used as input to adaptively adjust the model parameters. First, the mean, rate of change, and local fluctuation rate of the residuals at each monitoring point are extracted to construct the residual feature matrix. Then, the constitutive parameters of the material and the boundary coupling coefficients are fitted and optimized using this characteristic matrix. The residual feedback function is then used to map these parameters to corresponding correction factors. Correction methods can include exponential decay, normal modulation, or weighted averaging. After adjustment, the simulation is run again. If the residuals are within the set convergence threshold range, the model initialization and adjustment are considered complete.

[0052] After the sub-model is constructed, data-driven corrections are performed on the parameters of the simulation sub-model using adaptive fitting and dynamic correction based on historical data. Specifically, data sequences covering typical hydrological, meteorological, and operational conditions from multiple historical operating cycles of the dam are first selected. This includes extracting continuous time series data of stress, displacement, and seepage pressure according to sensor type, forming three types of one-dimensional multi-period measured datasets. To reduce the interference of dimensional differences between different physical quantities on subsequent residual calculations, each type of measured data is normalized using Min-Max based on its historical maximum and minimum values, so that its value range is uniformly mapped to the [0,1] interval. After data normalization, each type of measured data is differentially analyzed point-by-point with the theoretical output value predicted by the digital twin simulation model at the corresponding time point to obtain a residual sequence of the same dimension. To more comprehensively reflect the dynamic fluctuation characteristics of the residuals over time, a sliding window mechanism is used to divide the residual sequence into multiple equal-length sub-sequences, and the residual mean, standard deviation, and derivative change trends within each window are calculated, thereby constructing a residual fluctuation curve reflecting the offset trend over time. For monitoring areas where the residual fluctuation curve shows a significant offset trend, i.e., the mean of consecutive multi-window intervals deviates from a set threshold, a residual feedback function is established to jointly map the residual mean, rate of change, and local volatility into a model correction factor. The specific implementation method is as follows: ,in For model correction factors, These represent the residual mean, rate of change, and local volatility, respectively. The weighting coefficients are used. Finally, the correction factors output by the residual feedback function are applied to the constitutive parameters and boundary conditions of the corresponding physical units in the simulation sub-model.

[0053] Next, to obtain predictive information on the future operating status of the dam and provide forward-looking data support for safety assessment and early warning, a scheme is adopted that inputs historical monitoring data into a digital twin model to drive the model to make predictive outputs. Historical monitoring data is acquired and input into the digital twin model, driving the model to make predictive outputs. The outputs include a first output and a second output. The first output is to predict and generate a future preset period state trend sequence, extracting the fluctuation intensity factor and trend derivative factor from the trend sequence. The second output is to predict and generate the state data value at a certain moment within the future preset period. Specifically, historical monitoring data is input into the digital twin model, driving multiphysics response calculations in the simulation solution to obtain the current structural state.

[0054] Based on the current structural state, a multi-time predicted state sequence within a preset future period is generated using a sliding prediction window. The predicted state sequence is represented as follows: ,in The predicted physical index state value at time t+i represents the predicted state value, where T is the prediction period length. The predicted state sequence is used as the first output, from which the fluctuation intensity factor and trend derivative factor are extracted. The trend derivative factor is calculated as follows: ,in The trend derivative factor is represented by j, which represents the j-th physical index; the fluctuation intensity factor is calculated as follows: ,in, Indicates volatility intensity factor. The first output represents the predicted value of the j-th physical index at time t+i. Finally, a preset target time point t+k is selected from the predicted state sequence, and the corresponding predicted vector x(t+k) is extracted as the second output, used for multi-dimensional difference comparison with the historical stable state vector. Through the above prediction output scheme, dynamic prediction and multi-dimensional feature extraction of the dam's future state are achieved. The first output provides the state trend sequence and feature factors within a preset future period, comprehensively reflecting the evolution trend and stability of the dam's state. The second output provides accurate predictions for specific time points, facilitating comparative analysis with historical stable states, providing rich information for subsequent difference analysis and early warning, and significantly improving the predictive capability and foresight of the monitoring system.

[0055] To quantitatively analyze the difference between the predicted future state of the dam and its historical stable state, and to reveal the degree of anomaly in the dam's structural state, a difference analysis was performed between the predicted state data values ​​from the second output and the historical stable state data, constructing a multidimensional error mapping vector. Based on this error mapping vector, the displacement deviation index, stress deviation index, and water level deviation index were calculated respectively. Using the fluctuation intensity factor and trend derivative factor from the first output, the displacement deviation index, stress deviation index, and water level deviation index were corrected to obtain the corrected displacement deviation index, stress deviation index, and water level deviation index. Specifically, the multidimensional state data vector of the predicted future time from the second output was first... Reference value vector under historical stable operating conditions By subtracting each dimension, the error vector is obtained. The error vector includes displacement deviation, stress deviation, and water level deviation. The exponent of the difference between the current predicted displacement and the historical stable displacement reference value in the error mapping vector is defined as the displacement deviation index. ;in, The displacement deviation index. Set stress safety threshold Calculate the stress deviation index ,in, The stress deviation index, ; Calculate the water level deviation index ,in, This refers to the water level deviation index. ,in, The standard reference water level is used as the reference. Finally, the fluctuation intensity factor and trend derivative factor from the first output are used to correct the displacement deviation index, stress deviation index, and water level deviation index, resulting in the corrected displacement deviation correction index, stress deviation correction index, and water level deviation correction index. The calculation method for these indices is as follows: ,in, ,in This is to avoid division by zero. Represented as the j-th type deviation correction index, including , The term represents the intermediate term used for correction. By correcting the deviation index, the dynamic characteristics of the dam's state trend are fully considered. The fluctuation intensity factor reflects the stability of the state, while the trend derivative factor reflects the rate of state change. The introduction of both allows the deviation index to not only reflect the current degree of deviation but also to be dynamically adjusted based on future state trend changes.

[0056] Finally, to comprehensively assess the health status of the dam structure and provide timely early warning of structural anomaly risks, a structural health score is calculated based on the displacement deviation correction index, stress deviation correction index, and water level deviation correction index. This score is then compared to a preset safety threshold. If the score exceeds the threshold, a structural anomaly risk is identified, and a corresponding early warning is issued. Specifically, the structural health score is calculated as the normalized value of the product of the displacement deviation correction index, stress deviation correction index, and water level deviation correction index. Normalization unifies the score range to the [0,1] interval, facilitating the setting of safety thresholds and comparisons of health status at different times. When the calculated structural health score exceeds the preset safety threshold, a structural anomaly risk is identified, and a corresponding early warning is issued. The early warning information may include detailed information such as the anomaly type, severity, time, and location, providing decision-making support for dam management departments to take timely maintenance measures and eliminate safety hazards. Through the calculation of the structural health score and the establishment of the early warning mechanism, a quantitative and comprehensive assessment of the dam's structural health status is achieved. This method overcomes the limitations of existing methods that rely on a single static threshold for judgment. It can comprehensively consider the deviations and dynamic trends of multiple physical indicators to form a unified health assessment standard, thereby improving the scientificity and accuracy of anomaly judgment.

[0057] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A dam safety monitoring method based on digital twins, characterized in that, Includes the following steps: S1. Deploy multiple types of sensors in the dam structure and its surrounding environment. The sensors include displacement sensors, stress sensors, and water level monitors, which are used to collect multi-dimensional physical state data of the dam during operation. S2. Establish a digital twin model based on BIM, IoT and finite element method. The digital twin model includes a multi-physics coupled simulation sub-model of the dam and data-driven correction. Among them, the multi-physics field coupled simulation sub-model of the dam is used to reflect the dynamic evolution of each monitoring index over time under normal conditions; Data-driven correction is used to adaptively fit and dynamically correct the parameters of the simulation sub-model by combining historical data; S3. Acquire historical monitoring data and input it into the digital twin model to drive the digital twin model to make prediction outputs, the outputs including a first output and a second output; The first output is to predict and generate a future preset cycle state trend sequence, and extract the volatility intensity factor and trend derivative factor from the trend sequence; The second output is to predict and generate the state data value at a certain moment within a preset future period; S4. Perform a difference analysis between the predicted state data value of the second output and the historical stable state data, and construct a multi-dimensional error mapping vector; S5. Based on the error mapping vector, calculate the displacement deviation index, stress deviation index and water level deviation index respectively. Use the first output fluctuation intensity factor and trend derivative factor to correct the displacement deviation index, stress deviation index and water level deviation index to obtain the corrected displacement deviation correction index, stress deviation correction index and water level deviation correction index. S6. Calculate the structural health score based on the displacement deviation correction index, stress deviation correction index and water level deviation correction index, and compare the structural health score with the preset safety threshold. If the score is higher than the threshold, it is determined that there is a structural anomaly risk in the dam and the corresponding early warning information is output. The implementation of driving the digital twin model to make predictions in step S3, including the first output and the second output, includes the following steps: Historical monitoring data is input into the digital twin model to drive multiphysics response calculations in the simulation solution, thereby obtaining the current structural state; Based on the current structural state, a multi-time predicted state sequence within a preset future period is generated using a sliding prediction window. The predicted state sequence is represented as follows: ,in This represents the predicted physical index state value at time t+i, where T is the prediction period length. The predicted state sequence is used as the first output, from which the volatility intensity factor and trend derivative factor are extracted. The trend derivative factor is calculated as follows: ,in represents the trend derivative factor, and j represents the j-th physical index; The fluctuation intensity factor is calculated as follows: ,in, Indicates volatility intensity factor. This represents the predicted value of the j-th physical index at time t+i. Finally, a preset target time point t+k is selected from the predicted state sequence, and the corresponding predicted vector x(t+k) is extracted as the second output, which is used to perform multi-dimensional difference comparison with the historical stable state vector. Step S5, which involves obtaining the displacement deviation index, stress deviation index, and water level deviation index based on the error mapping vector, includes the following: The exponent of the difference between the current predicted displacement and the historical stable displacement reference value in the error mapping vector is defined as the displacement deviation index. ;in, The displacement deviation index. ; Set stress safety threshold Calculate the stress deviation index ,in, The stress deviation index, ; Calculate the water level deviation index ,in, This refers to the water level deviation index. ,in, Represents the standard reference water level; The calculation method for correcting the displacement deviation index, stress deviation index, and water level deviation index using the first output fluctuation intensity factor and trend derivative factor is as follows: ,in, ,in This is to avoid division by zero. Represented as the j-th type deviation correction index, including , This represents the intermediate term used for correction.

2. The dam safety monitoring method based on digital twins according to claim 1, characterized in that, The construction process of the multiphysics coupled simulation sub-model of the dam in step S2 includes the following steps: A finite element discrete model reflecting the dam structure was established, the dam structure was spatially meshed, and material parameters and boundary conditions were assigned to each element. A coupled multiphysics solution module is constructed, and the response coupling between multiple fields is achieved by setting the boundary conditions and loading paths for multi-field interactions; Set historical typical working conditions as simulation input conditions, execute the initial simulation, output the theoretical response values ​​of each monitoring point, and compare them with the measured historical data to extract residual features; The extracted residual features are used to fit and adjust the material constitutive parameters and boundary coupling coefficients in the model, thus completing the initialization and adjustment of the simulation sub-model.

3. The dam safety monitoring method based on digital twins according to claim 2, characterized in that, The data-driven correction step, which combines historical data to adaptively fit and dynamically correct the parameters of the simulation sub-model, includes: The measured status data of the dam during multiple historical operating cycles were obtained, including stress, displacement, and seepage pressure time series data, and the data were normalized. The residual sequences between the historical measured data and the corresponding predicted outputs of the simulation sub-models are calculated respectively, and the residual fluctuation curves are constructed based on the sliding time window; For the monitoring area in the residual fluctuation curve where the offset trend is greater than the set threshold, a residual feedback function is constructed to map the residual mean, rate of change and local volatility into parameter correction factors; The parameter correction factor is applied to the corresponding physical unit parameters in the simulation sub-model to complete the parameter fine-tuning.

4. The dam safety monitoring method based on digital twins according to claim 3, characterized in that, The residual feedback function is implemented as follows: ,in As a weighting factor, The baseline reference value is obtained by fitting historical data. These represent the residual mean, the residual rate of change, and the residual standard deviation, respectively.

5. A dam safety monitoring method based on digital twins according to claim 1, characterized in that, The construction of the multidimensional error mapping vector in step S4 is achieved by using the multidimensional state data vector of the future time predicted by the second output. Reference value vector under historical stable operating conditions By subtracting each dimension, the error vector is obtained. The error vector includes displacement deviation, stress deviation, and water level deviation.

6. The dam safety monitoring method based on digital twin according to claim 1, characterized in that, The structural health score in step S6 is calculated as the normalized value of the product of the displacement deviation correction index, the stress deviation correction index, and the water level deviation correction index.

Images