Slope excavation dynamic monitoring and early warning method based on digital twinning

Abstract

The invention discloses a side slope excavation dynamic monitoring and early warning method based on digital twinning. The method comprises the following steps that terrain point cloud, rock and soil layering and construction progress are obtained and preprocessed; three-dimensional geometric twinning and physical twinning are established, finite element simulation is adopted, boundary push-back and pore pressure rebalance are completed, and a twinning base line is formed; an observation residual error and a physical conservation residual error are calculated, and a high-consistency twin and assimilation sample set is obtained through joint updating; applying a topological constraint sliding surface evolver and implementing adaptive mesh refinement and interface reconstruction of a critical zone; and constructing a displacement and osmotic pressure uncertainty conical domain based on the assimilation sample set and the critical region strengthening model, and outputting a grading early warning result. According to the method, a double-consistency assimilation and topological constraint sliding surface evolution method is adopted, high-consistency simulation and quantitative early warning of the whole slope excavation process are achieved, and the method has the advantages of being high in accuracy, high in stability and good in traceability.

Description

Technical Field

[0001] This invention relates to the field of dynamic simulation of construction processes, and in particular to a method for dynamic monitoring and early warning of slope excavation based on digital twins. Background Technology

[0002] Current slope monitoring and early warning systems mostly employ multi-source sensing combined with 3D geometric modeling and finite element simulation. Based on material parameter assignment, mesh generation, and boundary condition settings, they obtain the stress-displacement-pore pressure field, often supplemented by strength reduction or limit equilibrium assessments for stability. Parameters are then updated and corrected through cyclic calibration using historical-real-time data or Kalman-type methods, forming a closed-loop process of "monitoring-simulation-early warning." However, this process relies heavily on observational fitting in the data assimilation stage, making it difficult to simultaneously ensure strict compliance with both mechanical and seepage mechanism constraints.

[0003] However, these methods generally focus on observation fitting and lack a dual-consistency assimilation layer that combines observation consistency with physical conservation consistency, making them prone to "pseudo-fitting". Slip zone identification often relies on thresholds or geometric search and does not introduce a topologically constrained slip surface evolver that coordinates connectivity preservation, curvature continuity and interface reconstruction. Early warning is often based on empirical thresholds and lacks quantitative support for uncertainty cone domains constructed by assimilation samples, slip surface feasible regions and critical region scale factors, as well as threshold crossing probabilities, making it difficult to achieve stable and traceable hierarchical early warning.

[0004] Therefore, how to provide a dynamic monitoring and early warning method for slope excavation based on digital twins is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a dynamic monitoring and early warning method for slope excavation based on digital twins. This invention adopts a dual consistency assimilation and topological constraint slip surface evolution method to achieve highly consistent simulation and quantitative early warning of the entire slope excavation process, which has the advantages of high accuracy, strong stability and good traceability.

[0006] A method for dynamic monitoring and early warning of slope excavation based on digital twin according to an embodiment of the present invention includes the following steps:

[0007] Acquire topographic point clouds, soil and rock stratification, and construction progress; complete time alignment and spatial registration; form a standardized multi-source input set and establish a phased work condition library.

[0008] Based on the standardized multi-source input set and stage working condition library, a three-dimensional geometric twin and a physical twin are established. Finite element simulation is used to complete boundary back-pushing and pore pressure rebalancing to form a twin baseline.

[0009] The double consistency-topology-constrained excavation twin simulation method is adopted. The observation residual and physical conservation residual are calculated in the double consistency assimilation layer and the material properties, boundary conditions and state variables are jointly updated to obtain the assimilated high consistency twin and the assimilated sample set.

[0010] Topological constraint sliding surface evolvers are applied to highly consistent twins in the topological constraint sliding surface evolution layer, and adaptive mesh refinement and interface reconstruction are implemented in the critical region to obtain a sliding surface feasible region and critical region enhancement model that conforms to topological constraints.

[0011] In the uncertainty early warning layer, a cone domain of displacement and seepage pressure uncertainty is constructed based on the assimilation sample set and the critical zone reinforcement model. The threshold crossing probability is calculated, and the graded early warning results are output.

[0012] Optionally, the generation of the stage operating condition library specifically includes:

[0013] The system acquires topographic point clouds, soil and rock layers, and construction progress and performs integrity checks. It generates a progress timeline with clear start and end times and corresponding work layers for the construction progress, forming an initial data set containing time markers, spatial coordinates, and layer attributes.

[0014] Time alignment is performed on the initial dataset, a unified time axis is selected as the global time reference, and time offset estimation and interpolation resampling are performed on the original time series of each monitoring channel to obtain a unified time axis sequence;

[0015] Spatial registration of terrain point cloud and soil layer is performed under a unified spatial coordinate system. Based on three-dimensional rigid body transformation, the original point cloud coordinates are rotated and translated to a unified coordinate set, so that the registered point cloud and the layer interface are consistent with the reference in position, orientation and scale, and the spatial registration result is output.

[0016] The unified time axis sequence and spatial registration results are normalized. For multi-source data such as displacement, dip angle, seepage pressure and stratification attributes, dimensionless removal, centering and variance normalization are performed respectively to generate a normalized multi-source input set.

[0017] Based on the construction progress timeline and standardized multi-source input set, each time point on the unified timeline is assigned to a unique stage index according to the start and end interval of the stage it falls into. The excavation location, working layer, support configuration and external hydrological conditions of each stage are recorded as working condition entries, forming a time series data cluster, spatial element set and working condition description set aggregated by stage. The output is a stage working condition library containing stage index, stage time range, stage spatial range and stage working condition entries.

[0018] Optionally, the generation of the twin baseline specifically includes:

[0019] Based on the stage spatial range of the standardized multi-source input set and stage working condition library, a topographic triangulation network and soil stratification voxel division are generated, and weak interlayers and potential structural surfaces are independently partitioned and identified to form a three-dimensional geometric twin containing grid topology, layer boundary and partition attributes.

[0020] Assign material properties, contact types of layered interfaces, and drainage boundary types to each partition of the three-dimensional geometric twin. Establish the initial value and constraint set of the physical twin. Map the unloading phase and replenishment phase in the stage condition library to mechanical load sequence and seepage condition sequence, respectively, to obtain physical constraints and condition input.

[0021] Based on the element contribution and boundary conditions of the discrete equations assembled from the mesh and partition list, the mechanical solvers for the stress field and displacement field, as well as the seepage solvers for the seepage field and pore pressure field, are established. The solution sequence, convergence criteria, and stage advancement strategy are determined to form a finite element simulation scheme.

[0022] The initial field results are obtained by executing the finite element simulation scheme. The mechanical equilibrium satisfies the equilibrium relationship that the overall stiffness formed by the mesh assembly is equal to the external load vector after being multiplied by the displacement field in each degree of freedom. The mechanical solver outputs the initial displacement field, initial stress field and initial reaction force set according to the equilibrium relationship.

[0023] Based on the correspondence between the initial displacement field and the observations in the stage condition library, boundary backtracking is performed. The displacement values ​​consistent with the monitoring points are extracted from the initial displacement field through the extraction mapping of the observation positions. The displacement residuals are compared point by point with the observed displacements at the same time to form displacement residuals. The residuals at each position are weighted using the observation weights. The weighted residuals are backtracked to the model boundary and load degrees of freedom through the transpose of the extraction mapping to obtain the boundary correction amount and update the boundary conditions and equivalent external loads.

[0024] The initial pore pressure is re-estimated based on the seepage conditions and updated boundary conditions, and then advanced to a stage steady state on a unified time axis. The seepage solver, based on the contribution of the water storage coefficient matrix to the change of pore pressure over time, the contribution of the water conductivity matrix to the seepage flux, and the influence of the consolidation coupling term on the interaction between displacement and pore pressure, superimposes the source and sink terms, and determines that the stage convergence condition is met to obtain the rebalanced pore pressure field and seepage field, which are consistent with the updated displacement field and stress field.

[0025] During the twin baseline generation stage, finite element simulations are re-executed based on updated boundary conditions and rebalanced pore pressure and seepage fields. The displacement field, stress field, pore pressure field, and reaction force set of each stage are summarized according to the time sequence and spatial range of the stage condition library to form the twin baseline.

[0026] Optionally, the generation of the highly consistent twin and its assimilated sample set specifically includes:

[0027] Based on the double consistency-topology constraint excavation twin simulation method, the displacement field, stress field, pore pressure field and reaction force set of the twin baseline are read in the double consistency assimilation layer. The observed displacement, observed pore pressure and their corresponding observation positions consistent with the unified time axis are extracted from the stage working condition library. The observation mapping relationship and observation weight are established to form an input set to be assimilated, which includes model prediction data and measured observation data.

[0028] The observation residual vector is calculated based on the input set to be assimilated, and an observation consistency term is constructed. The observation consistency term is a quadratic measure with the observation residual as the independent variable and the observation weight as the coefficient.

[0029] Based on the mechanical equilibrium condition and seepage continuity condition of the twin baseline, the physical conservation residual combination vector is calculated. It is composed of mechanical equilibrium residual and mass conservation residual in parallel components, and a physical consistency term is constructed. The physical consistency term is a quadratic measure with physical conservation residual as independent variable and physical constraint weight as coefficient.

[0030] The dual-consistency assimilation joint objective is formed by the parallel structure of observation consistency terms and physical consistency terms. The joint update is performed according to a fixed update order, including three steps: state variable update, boundary condition update, and material property update, and an intermediate twin is output.

[0031] For the intermediate twins at the current stage of the unified time axis, predictions are performed and the observation residuals and physical conservation residuals are recalculated. Judgment is made based on the normalized ratio of the observation residuals relative to the observation data and the normalized ratio of the physical conservation residuals relative to the physical reference scale. If either ratio is higher than the corresponding threshold, joint updates are performed according to a fixed update order, and the update results of the state variables, boundary conditions and material properties in this round are recorded. When both ratios are not higher than the corresponding thresholds, the iteration ends and multiple sets of intermediate twin samples are obtained.

[0032] The model states, boundary conditions, and material properties of multiple intermediate twin samples are collected and summarized to form an assimilation sample set;

[0033] The statistical mean values ​​of displacement, stress and pore pressure are calculated for each degree of freedom of the assimilated sample set to generate the mean state field. Three constraints are applied to the mean state field: mechanical conservation error, seepage continuity error and slip surface topological connectivity. Local results that do not meet the constraints are screened out. The material parameters and boundary conditions corresponding to the screened high consistency state field are backfilled into the twin master model to reconstruct a high consistency twin.

[0034] Optionally, the generation of the slip surface feasible region and critical region enhancement model specifically includes:

[0035] The displacement field, stress field and pore pressure field of the assimilated high-consistency twin are read in the topologically constrained slip surface evolution layer. A joint index field for slip identification is constructed. Candidate high gradient bands are formed with shear strain, plastic strain energy density and pore pressure gradient as components. The initial mask of the candidate slip band is output.

[0036] Connectivity labeling is performed on the initial mask of the candidate slip zone. Based on the connectivity preservation rules of the topology-constrained slip surface evolver, adjacent segments are merged according to voxel or cell adjacency, isolated noise patches are eliminated, and a set of connected components and their dominant paths are generated.

[0037] The curvature distribution and normal consistency distribution of the connected component set are calculated. Based on the curvature threshold and normal consistency threshold, the high-frequency fluctuation segment is smoothed and pruned, the dominant path that meets the threshold condition is retained, and a curvature-constrained slip path sketch is generated.

[0038] Based on the curvature-constrained slip path sketch, a slip band region is generated within its normal band. The topology correction at the bifurcation and merging points is performed in combination with the connectivity preservation rule, and the topology-constrained slip band template is output.

[0039] Perform adaptive mesh refinement of the critical region within the coverage area of ​​the topologically constrained slip zone template. Determine the refinement level according to the error estimation and gradient index. Implement local refinement within the zone and at the zone boundary while maintaining a consistent transition with the surrounding mesh. Output the refined mesh and mapping relationship of the critical region.

[0040] Based on the mesh refined in the critical zone, the interface is reconstructed at the slip zone boundary. The interface type, contact parameters and penetration reduction parameters are established according to the contact boundary processing strategy to form a set of interface elements.

[0041] On the set of mesh and interface elements after the critical region is refined, the displacement field, stress field and pore pressure field of the high consistency twin are used to perform consistency correction on the slip zone template, calculate the slip probability weight of each mesh element and perform thresholding accordingly, and output the slip surface feasible region that meets the topological constraints.

[0042] Taking the slip surface feasible region as the core area, the material properties, interface parameters and boundary conditions within it are locally reinforced and configured to establish a critical zone reinforcement model that includes a fine mesh, interface elements and local parameter sets. The reinforcement model is then structurally associated with a highly consistent twin.

[0043] Optionally, the generation of the tiered early warning results specifically includes:

[0044] The assimilation sample set, slip surface feasible region and critical zone reinforcement model are read in the uncertainty early warning layer, and the observed displacement and observed seepage pressure of the corresponding stage are obtained in the time sequence of the stage working condition library to form the stage input for uncertainty assessment.

[0045] Using the feasible region of the slip surface as the spatial weight basis, spatial weighted statistics are performed on the displacement and osmotic pressure samples of the assimilated sample set within its coverage area to obtain the displacement set statistics and osmotic pressure set statistics.

[0046] The mesh refinement level, interface element distribution and local parameter set are extracted from the critical zone reinforcement model. The critical zone scale factor is calculated, the spatial weight and threshold sensitivity in the sliding feasible region are adjusted, and the critical zone weight field containing spatial weight and scale adjustment is output.

[0047] Based on displacement set statistics, pressure set statistics and critical zone weight field, with the set statistical mean as the center and the set covariance and critical zone weight field as the metric boundary, a cone domain of displacement and pressure uncertainty is constructed to constrain the joint value range of displacement and pressure under a unified time axis.

[0048] Based on the uncertainty of displacement and seepage pressure cone domain, the Monte Carlo traversal of the slip surface feasible domain is used to calculate the threshold crossing probability of displacement or seepage pressure exceeding the preset safety threshold of the stage working condition library within the cone domain boundary.

[0049] The threshold crossing probability is compared with the graded threshold range of the stage working condition library to generate graded early warning results. The corresponding sliding surface feasible domain spatial range, trigger sub-region and time window are marked in the early warning results.

[0050] The beneficial effects of this invention are:

[0051] This invention achieves a unified closed loop of virtual-real fusion, topological continuity, and quantified early warning by introducing a dual-consistency assimilation layer, a topologically constrained slip surface evolver, and an uncertainty cone domain construction mechanism into a digital twin slope simulation system. The dual-consistency assimilation layer establishes a parallel structure between observational consistency and physical conservation consistency, jointly updating state variables, boundary conditions, and material properties in a fixed order, enabling the twin to converge synchronously in the observation space and mechanical space, forming a statistically quantifiable assimilation sample set. The topologically constrained slip surface evolver, based on connectivity preservation rules, curvature thresholds, and interface reconstruction triggering criteria, coordinates slip path evolution, critical zone adaptive mesh refinement, and interface element reconstruction to generate a physically continuous slip surface feasible domain and critical zone reinforcement model. The uncertainty early warning layer constructs a displacement and seepage pressure uncertainty cone domain based on the assimilation sample set, slip surface feasible domain, and critical zone scale factor, quantifies the threshold crossing probability, and outputs graded early warning results. This method effectively avoids the problems of observational fitting distortion, slip surface fracture, and ambiguous early warning in existing technologies, achieving highly consistent simulation and interpretable dynamic early warning of the slope excavation process. Attached Figure Description

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0053] Figure 1 This is a flowchart of a dynamic monitoring and early warning method for slope excavation based on digital twins proposed in this invention;

[0054] Figure 2 This is a schematic diagram illustrating the generation of a highly consistent twin for a dynamic monitoring and early warning method for slope excavation based on digital twins proposed in this invention. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0056] refer to Figures 1-2 A method for dynamic monitoring and early warning of slope excavation based on digital twins includes the following steps:

[0057] Acquire topographic point clouds, soil and rock stratification, and construction progress; complete time alignment and spatial registration; form a standardized multi-source input set and establish a phased work condition library.

[0058] Based on the standardized multi-source input set and stage working condition library, a three-dimensional geometric twin and a physical twin are established. Finite element simulation is used to complete boundary back-pushing and pore pressure rebalancing to form a twin baseline.

[0059] The double consistency-topology-constrained excavation twin simulation method is adopted. The observation residual and physical conservation residual are calculated in the double consistency assimilation layer and the material properties, boundary conditions and state variables are jointly updated to obtain the assimilated high consistency twin and the assimilated sample set.

[0060] Topological constraint sliding surface evolvers are applied to highly consistent twins in the topological constraint sliding surface evolution layer, and adaptive mesh refinement and interface reconstruction are implemented in the critical region to obtain a sliding surface feasible region and critical region enhancement model that conforms to topological constraints.

[0061] In the uncertainty early warning layer, a cone domain of displacement and seepage pressure uncertainty is constructed based on the assimilation sample set and the critical zone reinforcement model. The threshold crossing probability is calculated, and the graded early warning results are output.

[0062] In this embodiment, the generation of the stage working condition library specifically includes:

[0063] The system acquires topographic point clouds, soil and rock layers, and construction progress and performs integrity checks. It generates a progress timeline with clear start and end times and corresponding work layers for the construction progress, forming an initial data set containing time markers, spatial coordinates, and layer attributes.

[0064] Time alignment is performed on the initial dataset, a unified time axis is selected as the global time reference, and time offset estimation and interpolation resampling are performed on the original time series of each monitoring channel to obtain a unified time axis sequence;

[0065] Spatial registration of terrain point cloud and soil layer is performed under a unified spatial coordinate system. Based on three-dimensional rigid body transformation, the original point cloud coordinates are rotated and translated to a unified coordinate set, so that the registered point cloud and the layer interface are consistent with the reference in position, orientation and scale, and the spatial registration result is output.

[0066] The unified time axis sequence and spatial registration results are normalized. For multi-source data such as displacement, dip angle, seepage pressure and stratification attributes, dimensionless removal, centering and variance normalization are performed respectively to generate a normalized multi-source input set.

[0067] Based on the construction progress timeline and standardized multi-source input set, each time point on the unified timeline is assigned to a unique stage index according to the start and end interval of the stage it falls into. The excavation location, working layer, support configuration and external hydrological conditions of each stage are recorded as working condition entries, forming a time series data cluster, spatial element set and working condition description set aggregated by stage. The output is a stage working condition library containing stage index, stage time range, stage spatial range and stage working condition entries.

[0068] In this embodiment, the generation of the twin baseline specifically includes:

[0069] Based on the stage spatial range of the standardized multi-source input set and stage working condition library, a topographic triangulation network and soil stratification voxel division are generated, and weak interlayers and potential structural surfaces are independently partitioned and identified to form a three-dimensional geometric twin containing grid topology, layer boundary and partition attributes.

[0070] Assign material properties, contact types of layered interfaces, and drainage boundary types to each partition of the three-dimensional geometric twin. Establish the initial value and constraint set of the physical twin. Map the unloading phase and replenishment phase in the stage condition library to mechanical load sequence and seepage condition sequence, respectively, to obtain physical constraints and condition input.

[0071] Based on the element contribution and boundary conditions of the discrete equations assembled from the mesh and partition list, the mechanical solvers for the stress field and displacement field, as well as the seepage solvers for the seepage field and pore pressure field, are established. The solution sequence, convergence criteria, and stage advancement strategy are determined to form a finite element simulation scheme.

[0072] The initial field results are obtained by executing the finite element simulation scheme. The mechanical equilibrium satisfies the equilibrium relationship that the overall stiffness formed by the mesh assembly is equal to the external load vector after being multiplied by the displacement field in each degree of freedom. The mechanical solver outputs the initial displacement field, initial stress field and initial reaction force set according to the equilibrium relationship.

[0073] Based on the correspondence between the initial displacement field and the observations in the stage condition library, boundary backtracking is performed. The displacement values ​​consistent with the monitoring points are extracted from the initial displacement field through the extraction mapping of the observation positions. The displacement residuals are compared point by point with the observed displacements at the same time to form displacement residuals. The residuals at each position are weighted using the observation weights. The weighted residuals are backtracked to the model boundary and load degrees of freedom through the transpose of the extraction mapping to obtain the boundary correction amount and update the boundary conditions and equivalent external loads.

[0074] The initial pore pressure is re-estimated based on the seepage conditions and updated boundary conditions, and then advanced to a stage steady state on a unified time axis. The seepage solver, based on the contribution of the water storage coefficient matrix to the change of pore pressure over time, the contribution of the water conductivity matrix to the seepage flux, and the influence of the consolidation coupling term on the interaction between displacement and pore pressure, superimposes the source and sink terms, and determines that the stage convergence condition is met to obtain the rebalanced pore pressure field and seepage field, which are consistent with the updated displacement field and stress field.

[0075] During the twin baseline generation stage, finite element simulations are re-executed based on updated boundary conditions and rebalanced pore pressure and seepage fields. The displacement field, stress field, pore pressure field, and reaction force set of each stage are summarized according to the time sequence and spatial range of the stage condition library to form the twin baseline.

[0076] In this embodiment, the generation of the highly consistent twin and its assimilated sample set specifically includes:

[0077] Based on the double consistency-topology constraint excavation twin simulation method, the displacement field, stress field, pore pressure field and reaction force set of the twin baseline are read in the double consistency assimilation layer. The observed displacement, observed pore pressure and their corresponding observation positions consistent with the unified time axis are extracted from the stage working condition library. The observation mapping relationship and observation weight are established to form an input set to be assimilated, which includes model prediction data and measured observation data.

[0078] The observation residual vector is calculated based on the input set to be assimilated, and an observation consistency term is constructed. The observation consistency term is a quadratic measure with the observation residual as the independent variable and the observation weight as the coefficient.

[0079] Based on the mechanical equilibrium condition and seepage continuity condition of the twin baseline, the physical conservation residual combination vector is calculated. It is composed of mechanical equilibrium residual and mass conservation residual in parallel components, and a physical consistency term is constructed. The physical consistency term is a quadratic measure with physical conservation residual as independent variable and physical constraint weight as coefficient.

[0080] The dual-consistency assimilation joint objective is formed by the parallel structure of observation consistency terms and physical consistency terms. The joint update is performed according to a fixed update order, including three steps: state variable update, boundary condition update, and material property update, and an intermediate twin is output.

[0081] Specifically, the state variable update is obtained by incrementally solving the gradient direction of the joint objective with respect to the model state, and then superimposed on the current displacement field and pore pressure field. The boundary condition update is obtained by incrementally solving the gradient direction of the joint objective with respect to the boundary parameters, and then the boundary displacement, boundary load and drainage boundary are updated. The material property update is obtained by incrementally solving the gradient direction of the joint objective with respect to the material parameters, and then the elastic modulus, cohesion, internal friction angle and permeability coefficient are updated.

[0082] For the intermediate twins at the current stage of the unified time axis, predictions are performed and the observation residuals and physical conservation residuals are recalculated. Judgment is made based on the normalized ratio of the observation residuals relative to the observation data and the normalized ratio of the physical conservation residuals relative to the physical reference scale. If either ratio is higher than the corresponding threshold, joint updates are performed according to a fixed update order, and the update results of the state variables, boundary conditions and material properties in this round are recorded. When both ratios are not higher than the corresponding thresholds, the iteration ends and multiple sets of intermediate twin samples are obtained.

[0083] The model states, boundary conditions, and material properties of multiple intermediate twin samples are collected and summarized to form an assimilation sample set;

[0084] The statistical mean values ​​of displacement, stress and pore pressure are calculated for each degree of freedom of the assimilated sample set to generate the mean state field. Three constraints are applied to the mean state field: mechanical conservation error, seepage continuity error and slip surface topological connectivity. Local results that do not meet the constraints are screened out. The material parameters and boundary conditions corresponding to the screened high consistency state field are backfilled into the twin master model to reconstruct a high consistency twin.

[0085] In this embodiment, the generation of the slip surface feasible region and critical region enhancement model specifically includes:

[0086] The displacement field, stress field and pore pressure field of the assimilated high-consistency twin are read in the topologically constrained slip surface evolution layer. A joint index field for slip identification is constructed. Candidate high gradient bands are formed with shear strain, plastic strain energy density and pore pressure gradient as components. The initial mask of the candidate slip band is output.

[0087] Connectivity labeling is performed on the initial mask of the candidate slip zone. Based on the connectivity preservation rules of the topology-constrained slip surface evolver, adjacent segments are merged according to voxel or cell adjacency, isolated noise patches are eliminated, and a set of connected components and their dominant paths are generated.

[0088] The curvature distribution and normal consistency distribution of the connected component set are calculated. Based on the curvature threshold and normal consistency threshold, the high-frequency fluctuation segment is smoothed and pruned, the dominant path that meets the threshold condition is retained, and a curvature-constrained slip path sketch is generated.

[0089] Based on the curvature-constrained slip path sketch, a slip band region is generated within its normal band. The topology correction at the bifurcation and merging points is performed in combination with the connectivity preservation rule, and the topology-constrained slip band template is output.

[0090] Perform adaptive mesh refinement of the critical region within the coverage area of ​​the topologically constrained slip zone template. Determine the refinement level according to the error estimation and gradient index. Implement local refinement within the zone and at the zone boundary while maintaining a consistent transition with the surrounding mesh. Output the refined mesh and mapping relationship of the critical region.

[0091] Based on the mesh refined in the critical zone, the interface is reconstructed at the slip zone boundary. The interface type, contact parameters and penetration reduction parameters are established according to the contact boundary processing strategy to form a set of interface elements.

[0092] On the set of mesh and interface elements after the critical region is refined, the displacement field, stress field and pore pressure field of the high consistency twin are used to perform consistency correction on the slip zone template, calculate the slip probability weight of each mesh element and perform thresholding accordingly, and output the slip surface feasible region that meets the topological constraints.

[0093] Taking the slip surface feasible region as the core area, the material properties, interface parameters and boundary conditions within it are locally reinforced and configured to establish a critical zone reinforcement model that includes a fine mesh, interface elements and local parameter sets. The reinforcement model is then structurally associated with a highly consistent twin.

[0094] In this embodiment, the generation of the graded early warning result specifically includes:

[0095] The assimilation sample set, slip surface feasible region and critical zone reinforcement model are read in the uncertainty early warning layer, and the observed displacement and observed seepage pressure of the corresponding stage are obtained in the time sequence of the stage working condition library to form the stage input for uncertainty assessment.

[0096] Using the feasible region of the slip surface as the spatial weight basis, spatial weighted statistics are performed on the displacement and osmotic pressure samples of the assimilated sample set within its coverage area to obtain the displacement set statistics and osmotic pressure set statistics.

[0097] The mesh refinement level, interface element distribution and local parameter set are extracted from the critical zone reinforcement model. The critical zone scale factor is calculated, the spatial weight and threshold sensitivity in the sliding feasible region are adjusted, and the critical zone weight field containing spatial weight and scale adjustment is output.

[0098] Based on displacement set statistics, pressure set statistics and critical zone weight field, with the set statistical mean as the center and the set covariance and critical zone weight field as the metric boundary, a cone domain of displacement and pressure uncertainty is constructed to constrain the joint value range of displacement and pressure under a unified time axis.

[0099] Based on the uncertainty of displacement and seepage pressure cone domain, the Monte Carlo traversal of the slip surface feasible domain is used to calculate the threshold crossing probability of displacement or seepage pressure exceeding the preset safety threshold of the stage working condition library within the cone domain boundary.

[0100] The threshold crossing probability is compared with the graded threshold range of the stage working condition library to generate graded early warning results. The corresponding sliding surface feasible domain spatial range, trigger sub-region and time window are marked in the early warning results.

[0101] Example 1:

[0102] To verify the feasibility of this invention in practice, it was applied to the dynamic monitoring and early warning system for the entire process of slope excavation on a mountainous highway. This area has complex geological conditions, with significant soil and rock stratification, well-developed weak interlayers, and frequent groundwater activity. Conventional finite element method predictions deviate significantly from measured deformations, slip zone location identification is unstable, and early warning results are delayed with a high false alarm rate. Therefore, the digital twin-based dynamic monitoring and early warning method for slope excavation of this invention was introduced to uniformly model and dynamically update the data flow, model evolution, and risk assessment during the excavation process.

[0103] Multi-source sensing devices were deployed on-site to continuously collect topographic point cloud data, soil and rock stratification data, support parameters, and pore pressure monitoring data. All data, after time alignment, spatial registration, and normalization, were input into the stage-specific working condition library to drive the real-time evolution of the digital twin model. During the twin construction phase, through the collaborative establishment of geometric and physical twins, a precise reconstruction of rock mass stratification, interface contact, and pore pressure distribution was achieved. The initial displacement field and pore pressure field output from the finite element simulation were used to form the twin baseline after boundary backpropagation and rebalancing, providing accurate initial conditions for subsequent dual-consistency assimilation.

[0104] During excavation, the dual-consistency assimilation mechanism proposed in this invention is employed to calculate the observation residuals and physical conservation residuals separately, forming a joint structure of observation consistency terms and physical consistency terms. Through the synchronous updating of state variables, boundary conditions, and material properties, the virtual and real models achieve coordinated convergence in the observation and physical spaces. This process automatically generates an assimilation sample set and statistically forms a highly consistent twin, providing reliable input for the slip surface evolution.

[0105] Based on highly consistent twins, slip zone regions are identified in the topologically constrained slip surface evolution layer. By maintaining connectivity rules and curvature threshold constraints, the slip surface maintains continuous smoothness in its topological structure, eliminating spurious fractures and multi-branching problems common in traditional methods. After adaptive mesh refinement in the critical region, interface reconstruction generates a set of interface elements consistent with the actual sliding contact characteristics, forming a critical region reinforcement model, providing a locally high-resolution foundation for subsequent uncertainty analysis.

[0106] During the early warning phase, the system constructs a cone domain of displacement and pressure uncertainty based on the assimilated sample set and the feasible region of the slip surface. It adjusts the spatial weights using a critical zone scale factor to quantify the probability of potential instability areas. By calculating the threshold crossing probability, it automatically generates graded early warning results, displaying the possible evolution range of the slip surface, the risk level, and the time window. Practical application shows that this method can identify potential slip trends in advance, the predicted results match the measured deformation well, the early warning accuracy is significantly improved, and the virtual and real models maintain long-term stability and consistency.

[0107] In summary, this invention achieves deep integration of digital twins and on-site monitoring under complex geological conditions, breaking through the limitations of fitting distortion, slip surface topological fracture, and qualitative early warning in traditional simulation prediction. It achieves the unified goal of highly consistent simulation, topological constraint evolution, and quantitative early warning, providing a reliable and traceable intelligent technical path for slope safety monitoring.

[0108] Table 1. Performance comparison between digital twin-based dynamic monitoring and early warning methods for slope excavation and traditional methods.

[0109] As shown in Table 1, the dual consistency-topology-constrained excavation twin simulation method (DCA-TSE) of this invention significantly outperforms traditional methods in several key indicators. First, regarding displacement prediction and pore pressure prediction errors, the average errors of this invention are reduced to about one-third of those of the traditional finite element method. This indicates that the dual consistency assimilation mechanism effectively suppresses the error accumulation caused by simple observation fitting, enabling the virtual and real models to converge simultaneously in the observation space and physical space, resulting in prediction results that are closer to actual field measurements.

[0110] Secondly, regarding the accuracy of slip surface recognition and the integrity of topological connectivity, the DCA-TSE method, relying on a topologically constrained slip surface evolver, achieves geometric continuity and physical rationality of the slip surface through connectivity preservation and curvature threshold control. This reduces the slip surface spatial deviation from 1.35 meters to 0.32 meters and improves the connectivity integrity of the slip zone to over 97%. This improvement directly enhances the geometric representation quality of the slip region and avoids the false breakage problem commonly found in traditional threshold recognition.

[0111] Regarding the timeliness and accuracy of early warnings, the uncertainty cone domain constructed in this invention, based on an assimilated sample set and a critical zone scale factor, transforms risk assessment from a single indicator to a joint probability model. The quantitative calculation of the threshold crossing probability significantly improves the lead time and reliability of early warnings, providing stable early warning signals on average approximately 8 hours earlier than traditional methods, with an accuracy rate exceeding 96%, thus providing ample response time for construction scheduling.

[0112] Furthermore, this invention also demonstrates excellent performance in terms of virtual-real consistency index and computational convergence efficiency. Through hierarchical updates and constraint screening mechanisms, the model maintains a balance between physical conservation and observational consistency after multiple rounds of assimilation iterations, improving overall virtual-real consistency by approximately 13% and computational convergence efficiency by 2.2 times, indicating that it is not only more accurate but also more robust in its computational process.

[0113] In summary, the DCA-TSE method of this invention forms a fusion mode of data-driven and mechanism-driven approaches through three innovative links: dual consistency assimilation, topological constraint slip surface evolution, and uncertainty quantification. This significantly improves the accuracy, real-time performance, and interpretability of slope excavation monitoring and early warning, and provides highly reliable technical support for safety management in complex geological environments.

[0114] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for dynamic monitoring and early warning of slope excavation based on digital twins, characterized in that, Includes the following steps: Acquire topographic point clouds, soil and rock stratification, and construction progress; complete time alignment and spatial registration; form a standardized multi-source input set and establish a phased work condition library. Based on the standardized multi-source input set and stage working condition library, a three-dimensional geometric twin and a physical twin are established. Finite element simulation is used to complete boundary back-pushing and pore pressure rebalancing to form a twin baseline. The double consistency-topology-constrained excavation twin simulation method is adopted. The observation residual and physical conservation residual are calculated in the double consistency assimilation layer and the material properties, boundary conditions and state variables are jointly updated to obtain the assimilated high consistency twin and the assimilated sample set. Topological constraint sliding surface evolvers are applied to highly consistent twins in the topological constraint sliding surface evolution layer, and adaptive mesh refinement and interface reconstruction are implemented in the critical region to obtain a sliding surface feasible region and critical region enhancement model that conforms to topological constraints. In the uncertainty early warning layer, a cone domain of displacement and seepage pressure uncertainty is constructed based on the assimilation sample set and the critical zone reinforcement model. The threshold crossing probability is calculated, and the graded early warning results are output.

2. The method for dynamic monitoring and early warning of slope excavation based on digital twins according to claim 1, characterized in that, The generation of the stage working condition library specifically includes: The system acquires topographic point clouds, soil and rock layers, and construction progress and performs integrity checks. It generates a progress timeline with clear start and end times and corresponding work layers for the construction progress, forming an initial data set containing time markers, spatial coordinates, and layer attributes. Time alignment is performed on the initial dataset, a unified time axis is selected as the global time reference, and time offset estimation and interpolation resampling are performed on the original time series of each monitoring channel to obtain a unified time axis sequence; Spatial registration of terrain point cloud and soil layer is performed under a unified spatial coordinate system. Based on three-dimensional rigid body transformation, the original point cloud coordinates are rotated and translated to a unified coordinate set, so that the registered point cloud and the layer interface are consistent with the reference in position, orientation and scale, and the spatial registration result is output. The unified time axis sequence and spatial registration results are normalized. For multi-source data such as displacement, dip angle, seepage pressure and stratification attributes, dimensionless removal, centering and variance normalization are performed respectively to generate a normalized multi-source input set. Based on the construction progress timeline and standardized multi-source input set, each time point on the unified timeline is assigned to a unique stage index according to the start and end interval of the stage it falls into. The excavation location, working layer, support configuration and external hydrological conditions of each stage are recorded as working condition entries, forming a time series data cluster, spatial element set and working condition description set aggregated by stage. The output is a stage working condition library containing stage index, stage time range, stage spatial range and stage working condition entries.

3. The method for dynamic monitoring and early warning of slope excavation based on digital twins according to claim 1, characterized in that, The generation of the twin baseline specifically includes: Based on the stage spatial range of the standardized multi-source input set and stage working condition library, a topographic triangulation network and soil stratification voxel division are generated, and weak interlayers and potential structural surfaces are independently partitioned and identified to form a three-dimensional geometric twin containing grid topology, layer boundary and partition attributes. Assign material properties, contact types of layered interfaces, and drainage boundary types to each partition of the three-dimensional geometric twin. Establish the initial value and constraint set of the physical twin. Map the unloading phase and replenishment phase in the stage condition library to mechanical load sequence and seepage condition sequence, respectively, to obtain physical constraints and condition input. Based on the element contribution and boundary conditions of the discrete equations assembled from the mesh and partition list, the mechanical solvers for the stress field and displacement field, as well as the seepage solvers for the seepage field and pore pressure field, are established. The solution sequence, convergence criteria, and stage advancement strategy are determined to form a finite element simulation scheme. The initial field results are obtained by executing the finite element simulation scheme. The mechanical equilibrium satisfies the equilibrium relationship that the overall stiffness formed by the mesh assembly is equal to the external load vector after being multiplied by the displacement field in each degree of freedom. The mechanical solver outputs the initial displacement field, initial stress field and initial reaction force set according to the equilibrium relationship. Based on the correspondence between the initial displacement field and the observations in the stage condition library, boundary backtracking is performed. The displacement values ​​consistent with the monitoring points are extracted from the initial displacement field through the extraction mapping of the observation positions. The displacement residuals are compared point by point with the observed displacements at the same time to form displacement residuals. The residuals at each position are weighted using the observation weights. The weighted residuals are backtracked to the model boundary and load degrees of freedom through the transpose of the extraction mapping to obtain the boundary correction amount and update the boundary conditions and equivalent external loads. The initial pore pressure is re-estimated based on the seepage conditions and updated boundary conditions, and then advanced to a stage steady state on a unified time axis. The seepage solver, based on the contribution of the water storage coefficient matrix to the change of pore pressure over time, the contribution of the water conductivity matrix to the seepage flux, and the influence of the consolidation coupling term on the interaction between displacement and pore pressure, superimposes the source and sink terms, and determines that the stage convergence condition is met to obtain the rebalanced pore pressure field and seepage field, which are consistent with the updated displacement field and stress field. During the twin baseline generation stage, finite element simulations are re-executed based on updated boundary conditions and rebalanced pore pressure and seepage fields. The displacement field, stress field, pore pressure field, and reaction force set of each stage are summarized according to the time sequence and spatial range of the stage condition library to form the twin baseline.

4. The method for dynamic monitoring and early warning of slope excavation based on digital twins according to claim 1, characterized in that, The generation of the highly consistent twin and its assimilated sample set specifically includes: Based on the double consistency-topology constraint excavation twin simulation method, the displacement field, stress field, pore pressure field and reaction force set of the twin baseline are read in the double consistency assimilation layer. The observed displacement, observed pore pressure and their corresponding observation positions consistent with the unified time axis are extracted from the stage working condition library. The observation mapping relationship and observation weight are established to form an input set to be assimilated, which includes model prediction data and measured observation data. The observation residual vector is calculated based on the input set to be assimilated, and an observation consistency term is constructed. The observation consistency term is a quadratic measure with the observation residual as the independent variable and the observation weight as the coefficient. Based on the mechanical equilibrium condition and seepage continuity condition of the twin baseline, the physical conservation residual combination vector is calculated. It is composed of mechanical equilibrium residual and mass conservation residual in parallel components, and a physical consistency term is constructed. The physical consistency term is a quadratic measure with physical conservation residual as independent variable and physical constraint weight as coefficient. The dual-consistency assimilation joint objective is formed by the parallel structure of observation consistency terms and physical consistency terms. The joint update is performed according to a fixed update order, including three steps: state variable update, boundary condition update, and material property update, and an intermediate twin is output. For the intermediate twins at the current stage of the unified time axis, predictions are performed and the observation residuals and physical conservation residuals are recalculated. Judgment is made based on the normalized ratio of the observation residuals relative to the observation data and the normalized ratio of the physical conservation residuals relative to the physical reference scale. If either ratio is higher than the corresponding threshold, joint updates are performed according to a fixed update order, and the update results of the state variables, boundary conditions and material properties in this round are recorded. When both ratios are not higher than the corresponding thresholds, the iteration ends and multiple sets of intermediate twin samples are obtained. The model states, boundary conditions, and material properties of multiple intermediate twin samples are collected and summarized to form an assimilation sample set; The statistical mean values ​​of displacement, stress and pore pressure are calculated for each degree of freedom of the assimilated sample set to generate the mean state field. Three constraints are applied to the mean state field: mechanical conservation error, seepage continuity error and slip surface topological connectivity. Local results that do not meet the constraints are screened out. The material parameters and boundary conditions corresponding to the screened high consistency state field are backfilled into the twin master model to reconstruct a high consistency twin.

5. The method for dynamic monitoring and early warning of slope excavation based on digital twins according to claim 1, characterized in that, The generation of the smooth surface feasible region and critical region enhancement model specifically includes: The displacement field, stress field and pore pressure field of the assimilated high-consistency twin are read in the topologically constrained slip surface evolution layer. A joint index field for slip identification is constructed. Candidate high gradient bands are formed with shear strain, plastic strain energy density and pore pressure gradient as components. The initial mask of the candidate slip band is output. Connectivity labeling is performed on the initial mask of the candidate slip zone. Based on the connectivity preservation rules of the topology-constrained slip surface evolver, adjacent segments are merged according to voxel or cell adjacency, isolated noise patches are eliminated, and a set of connected components and their dominant paths are generated. The curvature distribution and normal consistency distribution of the connected component set are calculated. Based on the curvature threshold and normal consistency threshold, the high-frequency fluctuation segment is smoothed and pruned, the dominant path that meets the threshold condition is retained, and a curvature-constrained slip path sketch is generated. Based on the curvature-constrained slip path sketch, a slip band region is generated within its normal band. The topology correction at the bifurcation and merging points is performed in combination with the connectivity preservation rule, and the topology-constrained slip band template is output. Perform adaptive mesh refinement of the critical region within the coverage area of ​​the topologically constrained slip zone template. Determine the refinement level according to the error estimation and gradient index. Implement local refinement within the zone and at the zone boundary while maintaining a consistent transition with the surrounding mesh. Output the refined mesh and mapping relationship of the critical region. Based on the mesh refined in the critical zone, the interface is reconstructed at the slip zone boundary. The interface type, contact parameters and penetration reduction parameters are established according to the contact boundary processing strategy to form a set of interface elements. On the set of mesh and interface elements after the critical region is refined, the displacement field, stress field and pore pressure field of the high consistency twin are used to perform consistency correction on the slip zone template, calculate the slip probability weight of each mesh element and perform thresholding accordingly, and output the slip surface feasible region that meets the topological constraints. Taking the slip surface feasible region as the core area, the material properties, interface parameters and boundary conditions within it are locally reinforced and configured to establish a critical zone reinforcement model that includes a fine mesh, interface elements and local parameter sets. The reinforcement model is then structurally associated with a highly consistent twin.

6. The method for dynamic monitoring and early warning of slope excavation based on digital twins according to claim 1, characterized in that, The generation of the tiered early warning results specifically includes: The assimilation sample set, slip surface feasible region and critical zone reinforcement model are read in the uncertainty early warning layer, and the observed displacement and observed seepage pressure of the corresponding stage are obtained in the time sequence of the stage working condition library to form the stage input for uncertainty assessment. Using the feasible region of the slip surface as the spatial weight basis, spatial weighted statistics are performed on the displacement and osmotic pressure samples of the assimilated sample set within its coverage area to obtain the displacement set statistics and osmotic pressure set statistics. The mesh refinement level, interface element distribution and local parameter set are extracted from the critical zone reinforcement model. The critical zone scale factor is calculated, the spatial weight and threshold sensitivity in the sliding feasible region are adjusted, and the critical zone weight field containing spatial weight and scale adjustment is output. Based on displacement set statistics, pressure set statistics and critical zone weight field, with the set statistical mean as the center and the set covariance and critical zone weight field as the metric boundary, a cone domain of displacement and pressure uncertainty is constructed to constrain the joint value range of displacement and pressure under a unified time axis. Based on the uncertainty of displacement and seepage pressure cone domain, the Monte Carlo traversal of the slip surface feasible domain is used to calculate the threshold crossing probability of displacement or seepage pressure exceeding the preset safety threshold of the stage working condition library within the cone domain boundary. The threshold crossing probability is compared with the graded threshold range of the stage working condition library to generate graded early warning results. The corresponding sliding surface feasible domain spatial range, trigger sub-region and time window are marked in the early warning results.

Images