Water conservancy project model dynamic construction and simulation method based on digital twinning

Abstract

The present application relates to the technical field of digital twinning of water conservancy projects, in particular to a method for dynamic construction and simulation of a water conservancy project model based on digital twinning, comprising: acquiring multi-source perception data such as time sequence of physical field, geometric structure image and environmental image in a monitoring period of the water conservancy project, performing synchronous space-time registration and multi-modal fusion, constructing a dynamic perception field with unified space-time reference in a virtual space, and integrating physical field quantity, geometric topology and environmental characteristics. A growable digital twin is built with the engineering design blueprint as the initial structure, and the internal state, external topography and operation boundary of the digital twin are synchronously and dynamically evolved by the dynamic perception field. A simulation deduction engine takes the real-time state of the digital twin as the initial condition, loads hydrological, structural and hydraulic rules, carries out multi-time sequence step deduction, and generates a trajectory. The method can eliminate space-time deviation of multi-source data, realize dynamic coupling of virtual models and physical entities, and ensure that the simulation deduction conforms to the actual change state of the project.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology for water conservancy projects, and in particular to a method for dynamic construction and simulation of water conservancy project models based on digital twins. Background Technology

[0002] Current simulation modeling work for water conservancy projects typically employs a discrete data acquisition model, using sensors to acquire physical field data, inspection equipment to collect structural images, and satellite remote sensing to acquire environmental images. Each type of data is stored and analyzed independently after acquisition, and the virtual scene is built using only a single type of data as the basis for modeling, without collaborative processing of multiple types of sensory data. The operating parameters, structural morphology, and changes in the external environment of the physical entity of the water conservancy project are isolated, making it impossible to form a complete sensory scene corresponding to the physical entity within the virtual space. The model construction relies on engineering design drawings to form a fixed structure and lacks the ability to adjust to changes in the physical entity.

[0003] Multi-source sensing data suffers from misalignment and bias in the spatiotemporal dimensions, making it impossible to fully integrate and present physical field quantities, geometric topology, and environmental features under the same spatiotemporal reference. Digital twin models can only maintain their initial construction state and cannot synchronously adapt and adjust their internal state, external morphology, and operational boundaries based on real-time sensing data. Simulation and extrapolation stages use fixed model parameters as the starting point for computation, allowing only limited-scale simulation calculations. They cannot rely on real-time updated twin states to extrapolate future states over continuous time, resulting in a persistent gap between dynamic adaptation and forward-looking extrapolation between the virtual model and the physical entity. Summary of the Invention

[0004] The purpose of this invention is to address the shortcomings of existing technologies by proposing a dynamic construction and simulation method for water conservancy engineering models based on digital twins.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a method for dynamic construction and simulation of water conservancy engineering models based on digital twins, comprising:

[0006] The system acquires multi-source sensing data streams generated at the target water conservancy project site within a preset monitoring period. The multi-source sensing data streams include physical field time-series sequences captured by sensors, geometric structure images collected by inspection equipment, and environmental images acquired by satellite remote sensing.

[0007] The multi-source sensing data stream is synchronously spatiotemporally registered and multimodal fused to generate a dynamic sensing field in virtual space that has a unified spatiotemporal reference with the physical entity. The dynamic sensing field includes physical field quantities that reflect the physical state, geometric topology that characterizes the geometric shape, and environmental features that describe the environment.

[0008] The dynamic sensing field is used to drive a growable digital twin in real time. The initial structure of the digital twin is defined by the design blueprint of the hydraulic engineering project. Under the drive of the dynamic sensing field, the internal state of the digital twin is updated in real time according to the physical field quantity, the external morphology is dynamically adjusted according to the geometric topology, and the operating boundary evolves synchronously according to the environmental characteristics.

[0009] While the digital twin is evolving, a simulation engine is activated. The simulation engine uses the current state of the digital twin as the initial condition, loads preset hydrological, structural and hydraulic simulation rules, performs state extrapolation calculations for the digital twin at multiple future time steps, and generates a state extrapolation trajectory.

[0010] As a further aspect of the present invention, synchronous spatiotemporal registration and multimodal fusion are performed on the multi-source sensing data stream, including:

[0011] Establish timestamps and spatial coordinate systems for physical field time series, geometric structure images, and environmental images, respectively;

[0012] Align all data source timestamps to a unified time base and unify the spatial coordinate system with the preset global coordinate system of the virtual space;

[0013] Under a unified spatiotemporal reference, physical field quantities are mapped to spatial locations corresponding to geometric topology, and environmental images are used as background constraints to fuse and generate a dynamic perception field containing numerical, geometric, and semantic information.

[0014] As a further aspect of the present invention, a growable digital twin is driven in real time using the dynamic sensing field, including:

[0015] Read the design blueprint files of water conservancy projects, and initialize and construct a digital twin in virtual space with design parameters as the skeleton, including component attributes, connection relationships and material properties;

[0016] Continuously read updated physical field quantities from the dynamic sensing field and assign the physical field quantities to the physical state attributes of the corresponding positions in the digital twin;

[0017] By continuously reading the changing geometric topology from the dynamic sensing field, the external contour and internal structure of the digital twin undergo corresponding deformation and growth.

[0018] Continuously read environmental features from the dynamic sensing field and dynamically adjust the boundaries of the virtual environment in which the digital twin is located.

[0019] As a further aspect of the present invention, the simulation engine is activated, wherein the simulation engine uses the current state of the digital twin as the initial condition, loads preset hydrological, structural, and hydraulic simulation rules, and performs state extrapolation calculations for the digital twin at multiple future time steps, including:

[0020] From the real-time status of the digital twin, the stress, displacement, water level, flow rate and pressure of each component are extracted as the initial values ​​for the simulation.

[0021] By invoking hydrological simulation rules, the basin runoff and channel evolution are calculated based on the current flow and water level.

[0022] The structural simulation rules are invoked to calculate the redistribution of internal forces and deformation of the components based on the current stress and displacement.

[0023] By invoking hydraulic simulation rules, the head loss and velocity field in the pipe network or open channel are calculated based on the current pressure and flow rate.

[0024] The calculation results of the three rules are coupled and iterated within the same time step to update the predicted state of the digital twin at the same time step, and used as the input for the next time step. This process is repeated for multiple steps to form a state inference trajectory.

[0025] The construction steps of the simulation engine include:

[0026] Define the basic computing framework of the simulation engine. The basic computing framework includes a state variable register for storing the current simulation moment, a knowledge base module for storing hydrological simulation rules, a knowledge base module for storing structural simulation rules, a knowledge base module for storing hydraulic simulation rules, and a core computing unit for performing multi-rule coupled iterative calculations.

[0027] The geometric topology, initial distribution of physical field quantities, and environmental feature boundaries of the digital twin are used as a spatial grid. The spatial grid is discretized to generate a computational grid model that includes node coordinates and cell connection relationships.

[0028] The computational grid model, state variable register, and various knowledge base modules are associated and bound with the core computing unit, enabling the core computing unit to call corresponding rules from the knowledge base modules based on the topological relationship of the computational grid model, and to iteratively update the variables in the state variable register, thereby completing the initial construction of the simulation inference engine.

[0029] As a further aspect of the present invention, it also includes:

[0030] Extract key evolution nodes from the state deduction trajectory. When the deduction reaches a key evolution node, determine whether the state of the digital twin has reached a preset warning threshold.

[0031] At critical evolutionary nodes where the state simulation trajectory reaches the warning threshold, a preset combination of perturbation factors is applied to the structure, load, and boundary conditions of the digital twin to generate a perturbation scenario.

[0032] The disturbance scenario, which includes a combination of disturbance factors, is re-inputted into the simulation engine to start a new round of state simulation and calculate the response trajectory of the digital twin under the disturbance scenario.

[0033] By comparing the state projection trajectory in the undisturbed scenario with the response trajectory in the disturbed scenario, the difference sequence of state parameters between the state projection trajectory in the undisturbed scenario and the response trajectory in the disturbed scenario is analyzed to form a resilience assessment map.

[0034] Based on the aforementioned resilience assessment map, the propagation path and amplification nodes of the perturbation factor combination in the digital twin are analyzed in reverse to locate the key weak link that has the greatest impact on the overall stability.

[0035] Based on the information of the key weak links, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space. These optimization instructions are used to guide the maintenance or control decisions of the physical hydraulic engineering project.

[0036] As a further aspect of the present invention, the extraction of key evolution nodes from the state deduction trajectory includes:

[0037] In the state simulation trajectory, monitoring thresholds for multiple state parameters related to engineering safety are set;

[0038] When the rate of change of any state parameter in the state deduction trajectory exceeds its corresponding rate of change threshold, this moment will be recorded as a potential critical evolution node.

[0039] Near potential critical evolutionary nodes, check whether the changing trends of multiple related state parameters are consistent. If the trends are consistent, then the critical evolutionary node is confirmed as a critical evolutionary node.

[0040] At critical evolutionary nodes where the state deduction trajectory reaches the warning threshold, a preset combination of perturbation factors is applied to the structure, load, and boundary conditions of the digital twin to generate a perturbation scenario, specifically including:

[0041] At the virtual moment corresponding to the critical evolution node that triggers the warning threshold, the trajectory of the paused state is simulated;

[0042] Select a set of disturbance factors from the preset disturbance factor library, which include seismic wave loads, extreme rainfall processes, or structural material degradation.

[0043] The selected disturbance factors are superimposed onto the structural properties, external loads, or environmental boundary conditions of the digital twin in the form of time functions or spatial distribution functions, respectively, to form a data snapshot of the disturbance scenario.

[0044] As a further aspect of the present invention, the step of re-inputting the disturbance scenario containing the combination of disturbance factors into the simulation engine to start a new round of state simulation includes:

[0045] The data snapshot of the disturbance scenario is used as the new initial conditions;

[0046] Restart the simulation engine and re-execute the coupled iterative calculation from the virtual time corresponding to the key evolution node;

[0047] During the calculation process, the influence of the disturbance factor will continue to affect the simulation update process of the digital twin until it is extrapolated to the preset future time point, generating a new response trajectory.

[0048] As a further aspect of the present invention, the step of comparing the state projection trajectory in the undisturbed scenario with the response trajectory in the disturbed scenario, analyzing the difference sequence of state parameters between the state projection trajectory in the undisturbed scenario and the response trajectory in the disturbed scenario, and forming a resilience assessment map includes:

[0049] At the same point in time, the values ​​of the same state parameter in the state deduction trajectory and the response trajectory are compared point by point;

[0050] Calculate the absolute difference of the state parameters at each time point to form a difference sequence of state parameters changing over time;

[0051] All the monitored state parameters and their corresponding difference sequences are plotted into a two-dimensional map, namely the toughness assessment map, with time as the horizontal axis, parameter type as the vertical axis, and the degree of difference as the color depth.

[0052] As a further aspect of the present invention, based on the resilience assessment map, the propagation path and amplification nodes of the perturbation factor combination in the digital twin are analyzed in reverse, including:

[0053] In the resilience assessment map, the region with the deepest color depth, i.e. the greatest difference, is identified and marked as the initial influence node;

[0054] Starting from the initial influencing node, trace the adjacent nodes that are physically or hydraulically directly connected to the initial influencing node in the topology of the digital twin in the reverse time direction;

[0055] By examining the changing trends of the degree of difference between adjacent nodes at corresponding time points in the resilience assessment map, the propagation path of the disturbance from the source to the final point of impact can be constructed.

[0056] Along the propagation path, nodes where the degree of difference suddenly increases are identified and marked as perturbation amplification nodes.

[0057] As a further aspect of the present invention, based on the information of the key weak links, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space, including:

[0058] The location information of key weak links, the type of components they belong to, and the disturbance amplification factor are encapsulated;

[0059] Based on the operation and maintenance specifications and design standards of water conservancy projects, generate specific adjustment schemes for reinforcement structure parameters or modification schemes for scheduling rules;

[0060] The parameter adjustment scheme or scheduling rule modification scheme is converted into optimization instructions that can be recognized and executed by the simulation engine.

[0061] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0062] By synchronously registering and multimodal fusion the time-series physical field data captured by sensors, the geometric structure images collected by inspection equipment, and the environmental images acquired by satellite remote sensing, a dynamic sensing field with a unified spatiotemporal reference can be constructed in virtual space. Physical field quantities, geometric topology, and environmental features are integrated and presented within the sensing field, eliminating the spatiotemporal biases of multi-source data. The three types of sensing information form a mutually matching overall scene in virtual space, and the virtual sensing content maintains a dimensionally consistent correspondence with the actual state of physical entities. The dynamic sensing field continuously receives real-time incoming sensing data streams, and various information within the sensing field is updated synchronously with the operating state of physical entities. The sensing scene in virtual space always closely matches the real-time changing form of physical entities.

[0063] A scalable digital twin is built using the design blueprint of a water conservancy project as the initial structure. Dynamic sensing fields drive the twin's internal state to update based on physical field quantities, its external morphology to adjust based on geometric topology, and its operational boundaries to evolve based on environmental characteristics. The structure and state of the digital twin follow the physical entity, achieving multi-dimensional synchronous changes. The simulation engine uses the real-time evolution of the digital twin as the initial condition, loads preset hydrological, structural, and hydraulic simulation rules, and performs state extrapolation calculations for the water conservancy project at multiple future time steps. The extrapolation process continuously advances based on real-time updated model parameters, ultimately forming a coherent state extrapolation trajectory. The starting point of the simulation calculation remains consistent with the real-time state of the physical entity, and the extrapolation content closely matches the dynamic operational patterns of the water conservancy project. Attached Figure Description

[0064] Figure 1 This is a flowchart of the dynamic construction and simulation method for water conservancy engineering models based on digital twins as described in this invention;

[0065] Figure 2 A flowchart illustrating how to drive a growable digital twin using a dynamically sensed field;

[0066] Figure 3 A flowchart illustrating the methodology for resilience assessment and identification of critical weaknesses;

[0067] Figure 4 Line graph comparing digital twin disturbance simulations of water conservancy projects;

[0068] Figure 5 Comparison of principal stresses before and after digital twin optimization of a water conservancy project. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0070] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0071] See Figure 1 The system acquires multi-source sensing data streams generated at the target water conservancy project site within a preset monitoring period. These data streams include time-series sequences of physical fields captured by sensors, geometric structure images collected by inspection equipment, and environmental images acquired by satellite remote sensing. These multi-source sensing data streams are synchronously spatiotemporally registered and fused using multimodal methods to generate a dynamic sensing field in virtual space with a unified spatiotemporal reference to the physical entity. This field includes physical field quantities reflecting the physical state, geometric topology characterizing the geometric shape, and environmental features describing the environment. Subsequently, the generated dynamic sensing field is used to drive a growable digital twin in real time. The initial structure of the digital twin is defined by the design blueprint of the water conservancy project. Driven by the dynamic sensing field, the internal state of the digital twin is updated in real time according to the physical field quantities, the external morphology is dynamically adjusted according to the geometric topology, and the operating boundary evolves synchronously according to environmental features. While the digital twin is dynamically evolving, a simulation engine is launched. This simulation engine uses the real-time current state of the digital twin as the initial condition, loads preset hydrological, structural and hydraulic simulation rules, and performs state extrapolation calculations for the digital twin at multiple future time steps, thereby generating a predictive state extrapolation trajectory.

[0072] In one embodiment of the present invention, independent timestamps and spatial coordinate systems are established for the physical field time series, geometric structure images, and environmental images, respectively. All timestamps from different data sources are aligned to a unified high-precision time reference, and all spatial coordinate systems are uniformly transformed to a preset global coordinate system within the virtual space. After completing the unified spatiotemporal reference alignment, physical field quantities are mapped to specific spatial locations corresponding to the geometric topology in the virtual space. Simultaneously, environmental images are used as background constraints. These numerical, geometric, and semantic information are fused to generate a dynamic perception field containing multi-dimensional information. See also... Figure 2 The system reads the original design blueprints of the water conservancy project and initializes a digital twin in virtual space using the design parameters as the basic framework. This digital twin includes component attributes, connection relationships, and material properties. It continuously reads updated physical field quantities from a dynamic sensing field and assigns these quantities to the physical state attributes of the corresponding locations on the digital twin. It also continuously reads changing geometric topologies from the dynamic sensing field, driving the external contours and internal structures of the digital twin to deform and grow accordingly. Finally, it continuously reads environmental features from the dynamic sensing field and dynamically adjusts the boundaries of the virtual environment in which the digital twin exists.

[0073] In practical implementation, a concrete gravity dam hydraulic engineering project is used as an example scenario. A dynamic construction and simulation method for hydraulic engineering models based on digital twins is implemented. The multi-source sensing data stream includes physical field time series captured by piezometers and strain gauges deployed on the dam body, geometric structure images of the dam surface and corridor collected by UAV inspection equipment, and environmental images of the reservoir area obtained by meteorological satellite remote sensing. The physical field time series includes pore water pressure time series and structural strain time series. The geometric structure images include point cloud data and texture photos of the dam surface. The environmental images include surface temperature and vegetation index images. Independent time stamps and spatial coordinate systems are established for the physical field time series, geometric structure images, and environmental images, respectively. The time stamp of the physical field time series is derived from the built-in clock of the sensors, and the spatial coordinate system is the engineering local coordinate system with the dam site center as the origin. The time stamp of the geometric structure images is derived from the UAV flight log, and the spatial coordinate system is a vision-based model coordinate system. The time stamp of the environmental images is derived from the satellite overpass time, and the spatial coordinate system is a geographic latitude and longitude coordinate system. Align all data source timestamps to a unified network time protocol time base, and transform the spatial coordinate system to a preset global Cartesian coordinate system within the virtual space. The alignment process involves coordinate transformation calculations, where the transformation formula from the geographic latitude and longitude coordinate system to the global Cartesian coordinate system is as follows:

[0074]

[0075] in: Indicates latitude, Indicates longitude. Indicates altitude, The function represents three-dimensional coordinates in the global Cartesian coordinate system. This represents the standard transformation relationship from geodetic coordinates to rectangular coordinates. Under a unified spatiotemporal reference, physical field quantities such as pore water pressure and structural strain are mapped to the spatial grid node positions corresponding to the geometric topology of the dam body. Environmental images such as surface temperature and vegetation index are used as background constraints to fuse and generate a dynamic sensing field containing numerical, geometric, and semantic information. Each spatial location in the dynamic sensing field is associated with physical state quantities, geometric coordinates, and environmental feature values.

[0076] In some embodiments, the design blueprint file of a concrete gravity dam is read. This blueprint file includes dam section dimensions, concrete zonal elastic modulus, and gallery layout information. In virtual space, using the design parameters as a framework, a digital twin is initialized, containing component attributes, connection relationships, and material properties. Component attributes include dam segment numbers, connection relationships include construction joints between dam segments, and material properties include concrete unit weight. Updated pore water pressure physical field quantities are continuously read from the dynamic sensing field and assigned to the uplift pressure physical state attribute at the corresponding location of the dam foundation in the digital twin. Updated structural strain physical field quantities are also continuously read from the dynamic sensing field and assigned to the stress physical state attribute at the corresponding location of the dam body in the digital twin. The changing dam surface geometry and topology are continuously read from the dynamic sensing field. These changes are manifested as surface erosion or scouring represented by point cloud data, driving corresponding deformation and growth of the dam's external contour and internal gallery structure within the digital twin. The reservoir's water level environmental characteristics are continuously read from the dynamic sensing field. These characteristics are obtained through inversion from remote sensing images. The virtual environmental boundary of the digital twin is dynamically adjusted; this boundary represents the reservoir's water storage area. In the spatiotemporal registration stage, the unaligned original physical field time series is compared with the aligned series. The original series has sensor time stamps that drift by a few seconds, causing a mismatch with the image acquisition time. After alignment, all data reference a unified time base, ensuring a strict correspondence between the physical field data and the geometric image at the same moment in virtual space. In the coordinate system transformation stage, the latitude and longitude coordinates before and after the transformation are compared with the global Cartesian coordinates after the transformation. The coordinate values ​​of the same feature point differ before and after the transformation, but their spatial locations are uniquely corresponding, ensuring the consistency of spatial locations across multiple data sources. Optionally, during the fusion and generation of the dynamic sensing field, if the spatial grid nodes of the geometric topology do not completely coincide with the physical field sampling points when mapping physical field quantities to the geometric topology, a spatial interpolation algorithm is used to interpolate the physical field quantities onto the grid nodes. The spatial interpolation algorithm can be either the inverse distance weighted method or the Kriging interpolation method. In some embodiments, when driving the deformation of the digital twin, the continuously read geometric topology change data is compared with the historical topology data of the previous period. When the offset of the dam surface coordinates represented by the point cloud data exceeds a set tolerance, the position update of the outer contour mesh nodes of the digital twin is triggered, thereby simulating the scour deformation of the dam body. It can be understood that the initial construction of the digital twin is entirely based on the design blueprint, while the subsequent physical state update, geometric shape adjustment, and environmental boundary evolution are entirely driven by the multi-source fusion data continuously input from the dynamic sensing field, realizing the synchronous growth and evolution of the digital model on the physical entity.

[0077] In one embodiment of the present invention, parameters such as stress, displacement, water level, flow rate, and pressure of each component are extracted from the real-time state of the digital twin as initial values ​​for the simulation calculation. Preset hydrological simulation rules are invoked to calculate the watershed confluence and river evolution process based on the current flow rate and water level data. Preset structural simulation rules are invoked to calculate the redistribution of internal forces and structural deformation of components based on the current stress and displacement data. Preset hydraulic simulation rules are invoked to calculate the head loss and velocity field distribution in the pipe network or open channel based on the current pressure and flow rate data. The calculation results of the hydrological, structural, and hydraulic simulation rules are coupled and iterated within the same time step to update the predicted state of the digital twin at that time step. This updated state is then used as the input condition for the next time step, and this process is repeated for multiple steps to form a complete state extrapolation trajectory. The construction of the simulation extrapolation engine requires first defining its basic computational framework. This framework includes a state variable register for storing the current simulation moment, a knowledge base module for storing the hydrological, structural, and hydraulic simulation rules respectively, and a core computational unit for performing multi-rule coupled iterative calculations. The geometric topology, initial distribution of physical field quantities, and environmental feature boundaries of the digital twin are discretized as a spatial grid to generate a computational grid model containing node coordinates and unit connection relationships. This computational grid model, state variable register, and various knowledge base modules are then associated and bound to the core computing unit. This allows the core computing unit to call simulation rules from the corresponding knowledge base modules based on the topological relationships of the computational grid model, iteratively update the state variables in the state variable register, and complete the initial construction of the simulation engine.

[0078] In the specific implementation, a digital twin of a concrete gravity dam is used as the example scenario to construct and run the simulation engine. From the real-time state of the digital twin, the stress, displacement, water level, flow rate, and pressure of each component are extracted as initial values ​​for the simulation. The extracted stress is the principal stress value at the dam heel and toe; the extracted displacement is the horizontal displacement value at the dam crest; the extracted water level is the upstream reservoir water level; the extracted flow rate is the inlet flow rate of the spillway; and the extracted pressure is the water pressure value of the inner wall of the channel. The Saint-Venant equations solver, pre-integrated in the hydrological simulation rules knowledge base module, is invoked to calculate the watershed confluence and river evolution process based on the current spillway inlet flow rate and upstream reservoir water level, thus simulating the change of downstream water level over time. The finite element analysis program, pre-integrated in the structural simulation rules knowledge base module, is invoked to calculate the redistribution of internal forces and deformation of the components based on the current principal stress values ​​and horizontal displacement values ​​at the dam crest, thus simulating the development of the dam's stress and displacement fields. The computational fluid dynamics program, pre-integrated in the hydraulic simulation rule knowledge base module, calculates the head loss and velocity field within the pipe network or open channel based on the current water pressure value on the inner wall of the corridor and the flow rate of the spillway, thus deduce the flow regime within the corridor. The calculation results from the hydrological simulation rules, structural simulation rules, and hydraulic simulation rules are coupled and iterated within the same time step. The coupling process satisfies the data exchange and equilibrium conditions between physical fields. The convergence criterion for the coupling residual within one time step can be expressed as:

[0079]

[0080] in: This represents the water level and flow residual vector obtained from the hydrological simulation rules. This represents the stress and displacement residual vectors calculated using structural simulation rules. This represents the pressure and velocity residual vectors obtained from hydraulic simulation calculations. This represents a coupled iterative operator used to determine whether the multiphysics residuals satisfy the convergence condition. After satisfying the convergence condition, the predicted state of the digital twin at that time step is updated, and this updated state is used as the input for the next time step. This process is repeated for multiple time steps, forming a state projection trajectory spanning several hours or days into the future. The construction steps of the simulation engine include defining the basic computational framework of the simulation engine. This framework includes state variable registers for storing the water level, stress, displacement, pressure, and flow rate at the current simulation moment; a hydrological simulation rule knowledge base module for storing the Saint-Venant equations solution program; a structural simulation rule knowledge base module for storing the finite element analysis program; a hydraulic simulation rule knowledge base module for storing the computational fluid dynamics program; and core computational units for performing multi-rule coupled iterative calculations. The dam's geometric topology, initial stress distribution, and reservoir water environment boundary of the digital twin are used as a spatial mesh. The spatial mesh is discretized using tetrahedral elements to generate an integrated dam-watershed computational mesh model containing node coordinates and element connection relationships. The computational grid model, state variable register, and various knowledge base modules are associated and bound with the core computing unit. This enables the core computing unit to call the Saint-Venant equation solver from the hydrological simulation rule knowledge base module, the finite element analysis program from the structural simulation rule knowledge base module, and the computational fluid dynamics program from the hydraulic simulation rule knowledge base module, based on the topological relationship of the computational grid model. This allows iterative updates of the state variables in the state variable register, completing the initialization and construction of the simulation engine.

[0081] In some embodiments, the initial values ​​stored in the state variable register are derived from the instantaneous state dynamically updated by the digital twin. Compared to traditional simulations that only use static initial values ​​from the design blueprint, the derivation using dynamic initial values ​​more closely approximates the actual starting conditions of the physical entity. When extracting initial values ​​for stress derivation from the digital twin, comparing the extracted instantaneous principal stress values ​​with the allowable stress values ​​in the design specifications reveals that the instantaneous principal stress values ​​may be higher or lower than the allowable values ​​due to actual load effects, resulting in different starting points for subsequent structural simulation derivations. When calling hydrological simulation rules to calculate watershed confluence, the input is the current measured flow rate value provided by the digital twin. Compared to calculations using historical average flow rates, calculations using the current measured flow rate value more accurately reflect the impact of instantaneous inflow conditions on the evolution process. In coupled iterative calculations, each iteration requires exchanging boundary data between the calculation programs of hydrological, structural, and hydraulic simulation rules. For example, the downstream water level calculated by the hydrological simulation rules is passed to the structural simulation rules as the uplift pressure boundary of the dam foundation, while the dam deformation calculated by the structural simulation rules is passed to the hydraulic simulation rules to update the flow boundary geometry. Optionally, when generating the computational grid model, the grid is locally refined in areas of stress concentration and drastic flow field changes in the dam body. Compared with a uniform grid, a locally refined grid can more accurately capture gradients and deformations under the same computational resources. In some embodiments, when the core computing unit schedules multi-rule coupled iterative calculations, a sequential coupling and fixed-point iteration method is adopted. Compared with a fully coupled method, the sequential coupling method reduces the complexity of the solution, but the iteration step size needs to be controlled to ensure stability. It can be understood that the simulation calculation of the simulation engine starts entirely from the real-time dynamic state of the digital twin. The loaded simulation rules are preset, encapsulated physical law calculation programs. Through the coupled iteration of multiple rules within a unified time step, continuous and coupled simulation of future multi-time step states is achieved, generating a state simulation trajectory. Optionally, the state simulation trajectory not only includes the curves of each state parameter changing with time, but also includes the stress cloud map, displacement cloud map, and velocity vector field distribution of the digital twin as a whole at the end of the simulation.

[0082] In one embodiment of the present invention, see [reference] Figure 3The process involves extracting key evolution nodes from the generated state deduction trajectory. When these key evolution nodes are reached, it is determined whether the state of the digital twin has reached a preset warning threshold. At the key evolution nodes where the state deduction trajectory reaches the warning threshold, preset perturbation factor combinations are applied to the structure, load, and boundary conditions of the digital twin to form a new perturbation scenario. The perturbation scenario containing the perturbation factor combination is then re-inputted into the simulation engine to start a new round of state deduction, calculating the response trajectory of the digital twin under this perturbation scenario. By comparing the original state deduction trajectory in the undisturbed scenario with the response trajectory in the perturbation scenario, the differences in state parameters between the two trajectories are analyzed to form a resilience assessment map. Based on the resilience assessment map, the propagation path and amplification nodes of the perturbation factor combination in the digital twin are analyzed in reverse to locate the key weak links that have the greatest impact on the overall state stability. Based on the information of the key weak links, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space. These instructions can be used to guide the maintenance or control decisions of physical hydraulic engineering projects. When extracting critical evolution nodes, monitoring thresholds for multiple state parameters related to engineering safety are set in the state simulation trajectory. When the rate of change of any state parameter in the trajectory exceeds its corresponding rate of change threshold, this moment is recorded as a potential critical evolution node. Near the potential critical evolution node, the consistency of the change trends of multiple related state parameters is checked; if the trends are consistent, the node is confirmed as a critical evolution node. The specific operation for generating the disturbance scenario is to pause the state simulation trajectory at the virtual moment corresponding to the critical evolution node that triggers the warning threshold. A set of disturbance factors, including seismic wave loads, extreme rainfall processes, or structural material degradation, is selected from a pre-set disturbance factor library. The selected disturbance factors are then superimposed onto the structural properties, external loads, or environmental boundary conditions of the digital twin, respectively, in the form of a time function or a spatial distribution function, forming a data snapshot of the disturbance scenario.

[0083] In practical implementation, the state simulation trajectory generated based on the digital twin of the concrete gravity dam is further analyzed and subjected to disturbance tests. Several monitoring thresholds related to engineering safety are set in the state simulation trajectory, including the principal stress threshold at the dam heel, the horizontal displacement threshold at the dam crest, and the downstream river level threshold. When the rate of change of any state parameter in the state simulation trajectory exceeds its corresponding rate of change threshold, this moment is recorded as a potential critical evolution node. For example, when the time-varying rate of change of the principal stress at the dam heel exceeds 0.1 MPa per second, this moment is recorded as a potential critical evolution node. Near the potential critical evolution node, the consistency of the changing trends of multiple related state parameters is checked. If the principal stress at the dam heel, the principal stress at the dam crest, and the vertical displacement of the dam all show an accelerating growth trend near the potential critical evolution node, then this potential critical evolution node is confirmed as a critical evolution node. These critical evolution nodes that meet the consistency conditions are extracted from the state simulation trajectory. When the simulation reaches these critical evolution nodes, it is determined whether the state of the digital twin has reached a preset warning threshold, which is a warning value lower than the monitoring threshold. At critical evolutionary nodes where the state-deduction trajectory reaches the warning threshold, a preset combination of perturbation factors is applied to the structure, loads, and boundary conditions of the digital twin to generate a perturbation scenario. Specifically, at the virtual moment corresponding to the critical evolutionary node that reaches the warning threshold, the calculation process of the state-deduction trajectory is paused. From a preset perturbation factor library, a set of perturbation factors containing seismic wave loads, extreme rainfall processes, or structural material degradation is selected. For example, a seismic wave load with a peak ground acceleration of 0.2g and a material degradation factor with a 20% decrease in the elastic modulus of concrete are selected as a combination. The selected perturbation factors are superimposed onto the structural properties, external loads, or environmental boundary conditions of the digital twin in the form of time functions or spatial distribution functions, respectively. For seismic wave loads, they are treated as the basic excitation acceleration in the form of a time function. Superimposed on the dam foundation boundary, for material degradation, the spatial distribution function of the concrete elastic modulus E is modified as follows: A data snapshot of the disturbance scenario is generated. The disturbance scenario, including the combination of seismic wave load and material degradation factors, is re-input into the simulation engine to initiate a new round of state simulation, calculating the response trajectory of the digital twin under the disturbance scenario. By comparing the original state simulation trajectory under the undisturbed scenario with the response trajectory under the disturbed scenario, the differences in the dam heel stress, dam crest displacement, and downstream water level parameters between the two trajectories are analyzed to form a toughness assessment map. Based on the generated toughness assessment map, the propagation path and amplification nodes of the combination of seismic wave load and material degradation factors in the digital twin dam structure are analyzed in reverse, locating the key weak links that have the greatest impact on the overall state stability. Based on the location information and attributes of the key weak links, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space. The structural optimization instructions may be to increase the reinforcement ratio, and the control strategy optimization instructions may be to lower the flood limit water level. The optimization instructions are used to guide the maintenance or control decisions of the physical hydraulic engineering.

[0084] In some embodiments, when checking the consistency of the trend of state parameter changes, the correlation of the numerical sequences of multiple related state parameters within a time window before and after key evolution nodes is calculated, and the correlation coefficients of each parameter sequence with a preset baseline change pattern are compared. If the correlation coefficients all exceed 0.7, the trend is determined to be consistent. When selecting a combination of perturbation factors, the historical impact intensity records of different factor combinations are compared from the perturbation factor library, and a factor combination that has historically caused a significant response in similar structures is selected, such as the "seismic load + material freeze-thaw degradation" combination that has historically caused dam cracking. When applying perturbation factors, the perturbation superposition process can be described by a general function:

[0085]

[0086] in: Indicates the total load or property after the disturbance is applied. Indicates the original load or property. Let represent the influence function of the i-th perturbation factor, which can be a function of time t and spatial location S, and n represent the number of perturbation factors applied. It can be understood that a snapshot of the perturbation scenario fully preserves the state of the digital twin at a specific virtual moment and all superimposed perturbation conditions. Compared to a snapshot of the original projected trajectory at the same moment, the snapshot of the perturbation scenario includes additional load functions and modified material parameters. Optionally, the seismic wave loads in the perturbation factor library contain multiple artificial waves with different spectral characteristics. Comparing the selection of seismic wave loads with different spectral characteristics will generate different load-time functions. This influences the shape of the subsequent response trajectory. In some embodiments, when locating critical weak points, the structural component numbers corresponding to the areas showing the greatest differences in the toughness assessment map are compared with the design drawings of the digital twin. The safety reserve coefficient of the component in the design is compared, and the component with the lowest safety reserve coefficient is located as the critical weak point. It can be understood that extracting key evolution nodes from the state deduction trajectory is based on the rate of change of state parameters and the consistency of multi-parameter trends. Compared with relying solely on a single threshold exceedance criterion, combining the rate of change and consistency criteria can more sensitively and reliably identify the moment when the state undergoes an important transition. Optionally, the process of generating tuning instructions requires access to an external engineering knowledge base, which contains reinforcement design specifications for various component failure modes. Compared with suggestions based solely on simulation results, tuning instructions generated in conjunction with specifications have stronger engineering executability.

[0087] In one embodiment of the present invention, a data snapshot of the perturbation scenario is used as the new initial condition. The simulation engine is restarted, and the coupled iterative calculation is re-executed starting from the virtual time corresponding to the key evolution node. During the calculation, the influence of the applied perturbation factor will continue to affect the simulation update process of the digital twin until it is extrapolated to a preset future time point, generating a new response trajectory. At the same time point, the values ​​of the same state parameter in the original state extrapolation trajectory and the response trajectory are compared point by point. The absolute difference of the state parameter at each time point is calculated to form a difference sequence of the state parameter changing over time. The difference sequences corresponding to all monitored state parameters are plotted into a two-dimensional map with time as the horizontal axis, parameter type as the vertical axis, and the degree of difference as the color depth, thus obtaining the resilience assessment map.

[0088] In practical implementation, taking the disturbance test triggered by a key evolution node of a concrete gravity dam digital twin as an example, at the virtual moment corresponding to the key evolution node that reaches the warning threshold, the original state simulation trajectory under the undisturbed scenario is paused. A snapshot of the disturbance scenario data, including a combination of seismic wave load and material degradation factor, is used as the new initial condition. The snapshot records the dam foundation boundary conditions superimposed with a peak ground acceleration of 0.2g seismic wave function, as well as the dam concrete material parameters with a 20% decrease in elastic modulus. The simulation engine is restarted, starting from the virtual moment T_k corresponding to the key evolution node, and the coupled iterative calculation of hydrological, structural, and hydraulic simulation rules is re-executed. During the calculation, the seismic wave load, as a time function of the base acceleration input, continuously affects the structural dynamic equations, and the material degradation factor, as a spatially distributed function modifying material property parameters, continuously affects the constitutive relation. The influence of these disturbance factors will continue to affect the simulation update process of the digital twin until the simulation reaches the preset future time point. This generates a new time interval. , The response trajectory is compared with the state derivation trajectory in the undisturbed scenario and the response trajectory in the perturbed scenario. At the same time point, the values ​​of the same state parameter in the two trajectories are compared point by point, for example, comparing... , , The principal stress at the dam heel, the horizontal displacement at the dam crest, and the downstream river level are measured at each time point. The absolute differences of the state parameters at each time point are calculated, forming a difference sequence of state parameters over time. The difference sequence of the principal stress at the dam heel is as follows: The difference sequence of horizontal displacement on the dam crest is as follows The sequence of differences in downstream river water levels is as follows The formula for calculating the absolute difference is:

[0089]

[0090] in: Indicates at time The absolute difference of the state parameters, This represents the value of the state parameter in the response trajectory under the perturbation scenario at time t_i. Indicates at time The values ​​of the same state parameter in the state projection trajectory under a undisturbed scenario. The difference sequences corresponding to all monitored state parameters are plotted with time on the x-axis and parameter type on the y-axis, representing the degree of difference. To determine the color depth, a two-dimensional graph is plotted, namely the toughness assessment graph. The darker the color of the color block in the graph, the greater the difference in the parameter caused by the disturbance at that moment.

[0091] In some embodiments, when the simulation engine is restarted, the core computing unit no longer reads the original state derivation trajectory from the state variable register. Instead of using the variable values ​​at any given time step, the initial values ​​for the next round of calculations are read from a snapshot of the disturbed scenario data, after adding the disturbance factor. Comparing these two initial values, the initial stress field and boundary conditions under the disturbed scenario have changed. In the coupled iterative calculation for generating the response trajectory, the seismic excitation input received by the structural simulation rules within each time step is a function of time. In contrast to the zero seismic excitation input in the undisturbed calculation, this results in different terms on the right-hand side of the structural dynamic response equation at each time step. Optionally, when calculating the absolute difference of state parameters, in addition to calculating the absolute difference ΔV, the relative rate of change can also be calculated, but the plotting of the toughness assessment map specifies the use of the absolute difference as the mapping basis for color depth. It can be understood that the response trajectory is extrapolated from time T_k. Compared with the full-cycle extrapolation starting from the initial time, the extrapolation starting from key evolution nodes focuses on specific stages after the disturbance occurs, resulting in higher computational efficiency. In some embodiments, to clearly demonstrate the differences, multi-parameter data at a representative time can be tabulated for comparison, see Table 1.

[0092] Table 1: Comparison of State Parameter Differences under Disturbance Scenarios

[0093] State parameter type Unperturbed trajectory values Disturbance trajectory values absolute difference Principal stress at the dam heel (MPa) 2.5 3.1 0.6 Horizontal displacement of the dam crest (mm) 15.2 18.7 3.5 Downstream river level (m) 102.3 102.8 0.5

[0094] As can be understood, a resilience assessment map is a tool that visually integrates multiple differential sequences across time and parameter dimensions. By comparing the curves of each differential sequence analyzed individually, the map can more intuitively reveal the spatiotemporal distribution patterns of different parameters affected by disturbances at different times. Optionally, the color depth mapping can employ linear or logarithmic mapping. Compared to linear mapping, logarithmic mapping can more clearly display the hierarchy between smaller differences.

[0095] See Figure 4 This is a line graph comparing digital twin disturbance simulations of a hydraulic engineering project, focusing on the temporal changes of core parameters under undisturbed and disturbed scenarios after the critical evolution node Tk. Under undisturbed conditions, the principal stress at the dam heel increases slowly and linearly with minimal fluctuations; under disturbed conditions, the principal stress at the dam heel increases rapidly and linearly with a significant increase. Under undisturbed conditions, the horizontal displacement at the dam crest increases slowly and linearly with minimal fluctuations; under disturbed conditions, the horizontal displacement at the dam crest increases rapidly and linearly with a significant increase. Earthquake combined with material degradation significantly accelerates the growth rate of the principal stress at the dam heel and the horizontal displacement at the dam crest, and the difference continues to widen over time. This visually verifies the significant impact of disturbance factors on the structural safety of gravity dams, providing direct temporal comparison evidence for toughness assessment. The trend of widening differences over time suggests the need to take reinforcement or control measures as early as possible after a disturbance to avoid further accumulation of structural safety risks.

[0096] In one embodiment of the present invention, the region with the deepest color depth, i.e., the greatest difference, is identified in the resilience assessment map and marked as the initial impact node. Starting from the initial impact node, the adjacent nodes that are physically or hydraulically directly connected to the initial impact node are traced in the topology of the digital twin along the reverse time direction. The trend of the degree of difference of these adjacent nodes at the corresponding time points in the resilience assessment map is examined to construct the propagation path of the disturbance from the source to the final impact point. On the identified propagation path, nodes where the degree of difference suddenly increases are further identified and marked as disturbance amplification nodes. The location information of the key weak link, the type of the component to which it belongs, and the disturbance amplification factor are encapsulated. Combined with the operation and maintenance specifications and design standards of hydraulic engineering, specific reinforcement structure parameter adjustment schemes or scheduling rule modification schemes are generated. The parameter adjustment schemes or scheduling rule modification schemes are converted into optimization instructions that can be recognized and executed by the simulation engine.

[0097] In practical implementation, the region with the deepest color depth, indicating the greatest difference, is identified in the toughness assessment map. The map shows that the color block corresponding to the principal stress parameter at the dam heel is the darkest in the later stages of the simulation. This region is marked as the initial influence node, corresponding to a specific unit near the foundation at the bottom of the dam body. Starting from the dam heel unit corresponding to the initial influence node, tracing back in time, adjacent nodes that are physically or hydraulically directly connected to the initial influence node are tracked in the topology of the digital twin. Physically directly connected nodes include the adjacent dam body unit above and the adjacent bedrock contact surface unit below. The changing trends of the degree of difference between these adjacent nodes at corresponding time points in the toughness assessment map are examined. The color block of the adjacent dam body unit above is already darker at earlier time points, while the color block of the bedrock contact surface unit below is lighter at the same time. This constructs a propagation path of disturbance from the foundation contact surface to the dam body and then amplified and concentrated at the dam heel. On the identified propagation path, nodes where the degree of difference suddenly increases were further identified. On the path from the bedrock contact surface node to the initial influence node of the dam heel, the difference value of the dam heel node shows a significant jump relative to the difference value of its directly upstream node. The dam heel node is marked as a disturbance amplification node. The identification condition of the disturbance amplification node can be expressed as the calculation of the amplification factor:

[0098]

[0099] in: This represents the amplification factor of the difference from node i-1 to node i. This represents the absolute difference in the state parameters of node i at the corresponding time point in the resilience assessment map. This represents the absolute difference in state parameters at node i-1 at the corresponding time point in the resilience assessment map. Greater than the preset threshold At that time, node i is marked as the disturbance amplification node. Based on the key weak link information located by the above analysis, the location information of the key weak link, the type of its component, and the disturbance amplification factor are encapsulated. The location information is the dam heel unit number, the component type is concrete dam body, and the disturbance amplification factor is the calculated value. Value. Combining the operation and maintenance specifications for concrete dams and the design standards for gravity dams in water conservancy projects, specific reinforcement structure parameter adjustment schemes or scheduling rule modification schemes are generated. For the dam heel unit, the reinforcement scheme is to increase the grade of concrete in that area or add prestressed anchors, and the scheduling scheme is to further reduce the reservoir operating water level during extreme load warnings. The parameter adjustment schemes or scheduling rule modification schemes are converted into optimization instructions that can be recognized and executed by the simulation engine. The optimization instructions can be to modify the material elastic modulus parameters of specified units in the digital twin, or to modify the upstream water level time series in the simulation boundary conditions.

[0100] In some embodiments, when identifying initial impact nodes in a resilience assessment map, the difference values ​​of all color patch units in the map are compared, and the units with the highest difference values ​​(top 5%) are defined as the areas with the deepest color depth. Compared to relying solely on visual judgment, quantitative ranking can more accurately locate initial impact nodes. When tracing the propagation path, the comparison is made between tracing only along spatial topological connections and considering hydraulic connections. For example, in spillway structures, disturbances may be transmitted through water flow, so adjacent nodes of pipe connections need to be examined simultaneously. It can be understood that disturbance amplification nodes are locations where the degree of difference increases non-linearly along the propagation path. Compared to the gradual changes in difference values ​​at other nodes along the propagation path, the difference value curve at amplification nodes shows a clear inflection point. Optionally, an amplification factor threshold... The settings can be based on historical data or engineering experience. Compared with using a fixed threshold, using a dynamic threshold related to the average response level of the structure may be more reasonable. In some embodiments, when generating a strengthening structural parameter adjustment scheme, it is necessary to access the design code provisions in an external knowledge base. Compared with simply relying on the magnification factor, the scheme generated by combining the code's library of strengthening measures for stress exceeding limits is more engineering feasible. It can be understood that the generation of tuning instructions is a key step in transforming engineering analysis conclusions into machine-executable operations. Compared with only providing an analysis report, generating tuning instructions that can directly drive simulation forms a closed loop from diagnosis to intervention. Optionally, after inputting the tuning instructions into the simulation engine, a round of simulation can be restarted to verify the tuning effect. By comparing the response trajectory before tuning, the new trajectory after tuning is used to evaluate the effectiveness of the intervention measures.

[0101] See Figure 5This is a comparison chart of principal stresses before and after digital twin optimization of a hydraulic engineering project, used to verify the control effect of structural optimization measures on the evolution of principal stresses. From 0 to 60 hours before optimization, the principal stresses fluctuated slowly, remaining generally stable within 10 MPa. After optimization, the principal stresses were basically the same as before, remaining stable within 10 MPa. From 60 to 100 hours, the principal stresses continued to increase slowly, eventually reaching approximately 10 MPa. After optimization, the principal stresses increased exponentially, approaching 780 MPa after 100 hours. After 100 hours of optimization, the principal stresses were close to 800 MPa, far exceeding the safety threshold of conventional concrete dams (usually <20 MPa), requiring immediate suspension of the optimization scheme and recalculation. This visually demonstrates the nonlinear coupling relationship between disturbance, optimization, and stress response. After optimization, the principal stresses abruptly changed from slow linear growth to exponential growth, revealing the mechanical instability mechanism of the structure under specific boundary conditions.

[0102] The above are merely preferred embodiments of the present invention and are 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 that can be applied to 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 method for dynamic construction and simulation of hydraulic engineering models based on digital twins, characterized in that, include: The system acquires multi-source sensing data streams generated at the target water conservancy project site within a preset monitoring period. The multi-source sensing data streams include physical field time-series sequences captured by sensors, geometric structure images collected by inspection equipment, and environmental images acquired by satellite remote sensing. The multi-source sensing data stream is synchronously spatiotemporally registered and multimodal fused to generate a dynamic sensing field in virtual space that has a unified spatiotemporal reference with the physical entity. The dynamic sensing field includes physical field quantities that reflect the physical state, geometric topology that characterizes the geometric shape, and environmental features that describe the environment. The dynamic sensing field is used to drive a growable digital twin in real time. The initial structure of the digital twin is defined by the design blueprint of the hydraulic engineering project. Under the drive of the dynamic sensing field, the internal state of the digital twin is updated in real time according to the physical field quantity, the external morphology is dynamically adjusted according to the geometric topology, and the operating boundary evolves synchronously according to the environmental characteristics. While the digital twin is evolving, a simulation engine is activated. The simulation engine uses the current state of the digital twin as the initial condition, loads preset hydrological, structural and hydraulic simulation rules, performs state extrapolation calculations for the digital twin at multiple future time steps, and generates a state extrapolation trajectory.

2. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 1, characterized in that, Synchronous spatiotemporal registration and multimodal fusion of the multi-source sensing data stream include: Establish timestamps and spatial coordinate systems for physical field time series, geometric structure images, and environmental images, respectively; Align all data source timestamps to a unified time base and unify the spatial coordinate system with the preset global coordinate system of the virtual space; Under a unified spatiotemporal reference, physical field quantities are mapped to spatial locations corresponding to geometric topology, and environmental images are used as background constraints to fuse and generate a dynamic perception field containing numerical, geometric, and semantic information.

3. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 2, characterized in that, The method of using the dynamic sensing field to drive a growable digital twin in real time includes: Read the design blueprint files of water conservancy projects, and initialize and construct a digital twin in virtual space with design parameters as the skeleton, including component attributes, connection relationships and material properties; Continuously read updated physical field quantities from the dynamic sensing field and assign the physical field quantities to the physical state attributes of the corresponding positions in the digital twin; By continuously reading the changing geometric topology from the dynamic sensing field, the external contour and internal structure of the digital twin undergo corresponding deformation and growth. Continuously read environmental features from the dynamic sensing field and dynamically adjust the boundaries of the virtual environment in which the digital twin is located.

4. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 3, characterized in that, The simulation engine is activated, using the current state of the digital twin as initial conditions, loading preset hydrological, structural, and hydraulic simulation rules, and performing state extrapolation calculations for the digital twin at multiple future time steps, including: From the real-time status of the digital twin, the stress, displacement, water level, flow rate and pressure of each component are extracted as the initial values ​​for the simulation. By invoking hydrological simulation rules, the basin runoff and channel evolution are calculated based on the current flow and water level. The structural simulation rules are invoked to calculate the redistribution of internal forces and deformation of the components based on the current stress and displacement. By invoking hydraulic simulation rules, the head loss and velocity field in the pipe network or open channel are calculated based on the current pressure and flow rate. The calculation results of the three rules are coupled and iterated within the same time step to update the predicted state of the digital twin at the same time step, and used as the input for the next time step. This process is repeated for multiple steps to form a state inference trajectory. The construction steps of the simulation engine include: Define the basic computing framework of the simulation engine. The basic computing framework includes a state variable register for storing the current simulation moment, a knowledge base module for storing hydrological simulation rules, a knowledge base module for storing structural simulation rules, a knowledge base module for storing hydraulic simulation rules, and a core computing unit for performing multi-rule coupled iterative calculations. The geometric topology, initial distribution of physical field quantities, and environmental feature boundaries of the digital twin are used as a spatial grid. The spatial grid is discretized to generate a computational grid model that includes node coordinates and cell connection relationships. The computational grid model, state variable register, and various knowledge base modules are associated and bound with the core computing unit, enabling the core computing unit to call corresponding rules from the knowledge base modules based on the topological relationship of the computational grid model, and to iteratively update the variables in the state variable register, thereby completing the initial construction of the simulation inference engine.

5. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 4, characterized in that, Also includes: Extract key evolution nodes from the state deduction trajectory. When the deduction reaches a key evolution node, determine whether the state of the digital twin has reached a preset warning threshold. At critical evolutionary nodes where the state simulation trajectory reaches the warning threshold, a preset combination of perturbation factors is applied to the structure, load, and boundary conditions of the digital twin to generate a perturbation scenario. The disturbance scenario, which includes a combination of disturbance factors, is re-inputted into the simulation engine to start a new round of state simulation and calculate the response trajectory of the digital twin under the disturbance scenario. By comparing the state projection trajectory in the undisturbed scenario with the response trajectory in the disturbed scenario, the difference sequence of state parameters between the state projection trajectory in the undisturbed scenario and the response trajectory in the disturbed scenario is analyzed to form a resilience assessment map. Based on the aforementioned resilience assessment map, the propagation path and amplification nodes of the perturbation factor combination in the digital twin are analyzed in reverse to locate the key weak link that has the greatest impact on the overall stability. Based on the information of the key weak links, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space. These optimization instructions are used to guide the maintenance or control decisions of the physical hydraulic engineering project.

6. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 5, characterized in that, The extraction of key evolution nodes from the state deduction trajectory includes: In the state simulation trajectory, monitoring thresholds for multiple state parameters related to engineering safety are set; When the rate of change of any state parameter in the state deduction trajectory exceeds its corresponding rate of change threshold, this moment will be recorded as a potential critical evolution node. Near potential critical evolutionary nodes, check whether the changing trends of multiple related state parameters are consistent. If the trends are consistent, then the critical evolutionary node is confirmed as a critical evolutionary node. At critical evolutionary nodes where the state deduction trajectory reaches the warning threshold, a preset combination of perturbation factors is applied to the structure, load, and boundary conditions of the digital twin to generate a perturbation scenario, specifically including: At the virtual moment corresponding to the critical evolution node that triggers the warning threshold, the trajectory of the paused state is simulated; Select a set of disturbance factors from the preset disturbance factor library, which include seismic wave loads, extreme rainfall processes, or structural material degradation. The selected disturbance factors are superimposed onto the structural properties, external loads, or environmental boundary conditions of the digital twin in the form of time functions or spatial distribution functions, respectively, to form a data snapshot of the disturbance scenario.

7. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 6, characterized in that, The step of re-inputting the disturbance scenario, which includes combinations of disturbance factors, into the simulation engine to initiate a new round of state simulation includes: The data snapshot of the disturbance scenario is used as the new initial conditions; Restart the simulation engine and re-execute the coupled iterative calculation from the virtual time corresponding to the key evolution node; During the calculation process, the influence of the disturbance factor will continue to affect the simulation update process of the digital twin until it is extrapolated to the preset future time point, generating a new response trajectory.

8. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 7, characterized in that, The process involves comparing the state projection trajectory in an undisturbed scenario with the response trajectory in a disturbed scenario, analyzing the differences in state parameters between the two scenarios, and forming a resilience assessment map, including: At the same point in time, the values ​​of the same state parameter in the state deduction trajectory and the response trajectory are compared point by point; Calculate the absolute difference of the state parameters at each time point to form a difference sequence of state parameters changing over time; All the monitored state parameters and their corresponding difference sequences are plotted into a two-dimensional map, namely the toughness assessment map, with time as the horizontal axis, parameter type as the vertical axis, and the degree of difference as the color depth.

9. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 8, characterized in that, Based on the aforementioned resilience assessment map, the propagation path and amplification nodes of the perturbation factor combination in the digital twin are analyzed in reverse, including: In the resilience assessment map, the region with the deepest color depth, i.e. the greatest difference, is identified and marked as the initial influence node; Starting from the initial influencing node, trace the adjacent nodes that are physically or hydraulically directly connected to the initial influencing node in the topology of the digital twin in the reverse time direction; By examining the changing trends of the degree of difference between adjacent nodes at corresponding time points in the resilience assessment map, the propagation path of the disturbance from the source to the final point of impact can be constructed. Along the propagation path, nodes where the degree of difference suddenly increases are identified and marked as perturbation amplification nodes.

10. The method for dynamic construction and simulation of water conservancy engineering models based on digital twins according to claim 9, characterized in that, Based on the information regarding the key weaknesses, optimization instructions for the structure or control strategy of the digital twin are generated in virtual space, including: The location information of key weak links, the type of components they belong to, and the disturbance amplification factor are encapsulated; Based on the operation and maintenance specifications and design standards of water conservancy projects, generate specific adjustment schemes for reinforcement structure parameters or modification schemes for scheduling rules; The parameter adjustment scheme or scheduling rule modification scheme is converted into optimization instructions that can be recognized and executed by the simulation engine.

Images