A deep foundation anti-floating risk assessment method based on monitoring data

Abstract

The application provides a deep foundation pit anti-floating risk assessment method based on monitoring data, and relates to the technical field of deep foundation pit anti-floating. The method comprises the following steps: carrying out layered multi-source heterogeneous monitoring on the deep foundation pit, obtaining a water pressure deformation data set and extracting spatial node and time sequence parameters; calculating the spatial distribution gradient of buoyancy and the dynamic resistance threshold; then establishing a local mechanical coordinate system in the effective stress coupling area, extracting a structure deformation mutation sequence from the data set; subsequently constructing a space-time coupling graph, generating a nonlinear attenuation path covering a weak area through a feature fusion algorithm; finally extracting a local anti-floating risk set along the path, generating a global anti-floating evolution graph through time domain smoothing and three-dimensional space mapping, and outputting an evaluation result containing three-dimensional positioning and automatic hardware control instructions. The application realizes space-time linkage analysis of foundation water pressure and structure deformation, and significantly improves the fine early warning and entity control ability of anti-floating risk under complex working conditions of deep foundation pit.

Description

Technical Field

[0001] This invention relates to the field of deep foundation pit anti-buoyancy technology, and in particular to a method for assessing the anti-buoyancy risk of deep foundation pits based on monitoring data. Background Technology

[0002] As underground space development moves towards deeper and more complex hydrogeological environments, the anti-buoyancy stability of deep foundation pits is directly related to the overall structural safety of underground engineering projects. Traditional anti-buoyancy design and assessment of deep foundation pits are mainly based on Archimedes' principle and Terzaghi's effective stress principle. This involves calculating the anti-buoyancy dead loads such as the self-weight of components, the weight of the overburden, and the frictional resistance of the side walls, based on the static groundwater level or anti-buoyancy design water level obtained from preliminary surveys, and then deriving the static anti-buoyancy safety factor. In actual construction and operation and maintenance phases, the industry generally adopts the method of deploying water level observation wells, vibrating wire pore water pressure gauges, and structural settlement observation markers on-site to obtain discrete groundwater hydrological parameters and foundation pit deformation characteristics.

[0003] In recent years, monitoring of deep foundation pit engineering has been gradually shifting from traditional manual fixed-point surveys to automated network monitoring. Existing automated monitoring systems, relying on sensor networks, have initially achieved continuous acquisition of time-series data on single-hole water levels, single-point earth pressure, and single-point structural displacement. At the data processing level, some cutting-edge research has begun to introduce time-series prediction algorithms to perform trend fitting and extrapolation on monitoring data of single types of pore water pressure or groundwater level elevations, aiming to achieve a certain degree of early warning of potential conditions through the historical evolution patterns of single variables.

[0004] However, the failure of buoyancy resistance in deep foundation pits is essentially a nonlinear evolution process involving the coupling of multiple fields: seepage, stress, and displacement. Existing assessment methods based on monitoring data mainly focus on the direct comparison of extreme values ​​of single-dimensional indicators with absolute warning thresholds, failing to consider the spatial distribution gradient characteristics of buoyancy at the foundation caused by the dynamic changes in complex confined water heads. Furthermore, throughout the entire construction cycle from earthwork excavation to the capping of the main structure, the buoyancy resistance dead load of the foundation pit itself exhibits dynamic and non-stationary changes as the construction progresses. Existing methods often use empirically fixed static buoyancy resistance allowable values ​​as the sole evaluation benchmark, failing to couple the real-time deformation abrupt changes of the main structure with the non-uniform water and soil pressure at the foundation in a spatiotemporal dimension, making it difficult to truly characterize the overall and local buoyancy resistance mechanical response mechanism of the foundation pit. Summary of the Invention

[0005] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a method for assessing the buoyancy risk of deep foundation pits based on monitoring data. This invention solves the problems of existing technologies, such as the lack of spatiotemporal coupling analysis of multi-source heterogeneous monitoring data, static lag in resistance assessment thresholds, and difficulty in accurately characterizing local non-uniform buoyancy risk under dynamic working conditions.

[0006] To achieve the above objectives, the present invention provides the following solution: A method for assessing the buoyancy risk of deep foundation pits based on monitoring data includes: Based on the preset monitoring frequency, the target deep foundation pit is monitored in a layered, multi-source heterogeneous manner to obtain a water pressure deformation dataset containing multiple monitoring sections, and to extract the spatial node coordinates and corresponding time series parameters of all the monitoring sections. The buoyancy spatial distribution gradient of each monitoring section is determined based on the spatial node coordinates and the corresponding time series parameters, and the current construction progress is obtained. Calculate the dynamic resistance threshold based on the current construction progress and the buoyancy spatial distribution gradient. Within the effective stress coupling region of each monitoring section, a spatiotemporal feature evolution node set is generated based on the dynamic resistance threshold, and a local mechanical coordinate system is established for each effective stress coupling region; Under the local mechanical coordinate system, structural deformation mutation sequences are extracted from the hydraulic deformation dataset according to the spatiotemporal feature evolution node set; Spatiotemporal coupling maps are constructed on all the structural deformation mutation sequences, and a feature fusion algorithm is used to generate nonlinear decay paths in the spatiotemporal coupling maps to cover all the effective stress coupling regions; The local anti-buoyancy risk set on the spatiotemporal coupling map is extracted sequentially according to the nonlinear attenuation path, and the local anti-buoyancy risk set is smoothed in the temporal domain and mapped in space to obtain the global anti-buoyancy evolution map. Based on the global anti-buoyancy evolution map and the dynamic resistance threshold, risk interval matching is performed to obtain the anti-buoyancy risk assessment result of the target deep foundation pit.

[0007] The present invention discloses the following technical effects: This invention provides a method for assessing the uplift risk of deep foundation pits based on monitoring data. By constructing a spatiotemporal coupled map and deeply integrating multi-source heterogeneous monitoring data, this invention breaks through the physical separation of traditional single-index assessments, realizing spatiotemporal linkage analysis of foundation water pressure and structural deformation, effectively solving the problem of inaccurate characterization of local non-uniform uplift risk. Secondly, by introducing a dynamic resistance threshold calculation mechanism that is synchronously mapped with real-time construction conditions, this invention completely overcomes the assessment lag and error caused by static fixed thresholds, enabling the resistance benchmark to adapt to the dynamic evolution of the project progress. Finally, based on a global uplift evolution map generated by a nonlinear attenuation path, this invention accurately achieves dynamic matching of risk intervals and three-dimensional positioning of uplift weak points, significantly improving the refined management and early warning capabilities of uplift risk in deep foundation pits under complex dynamic conditions. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 A flowchart of a deep foundation pit anti-buoyancy risk assessment method based on monitoring data is provided for an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0012] like Figure 1 As shown, this invention provides a method for assessing the buoyancy risk of deep foundation pits based on monitoring data, including: Step 100: Based on the preset monitoring frequency, perform layered multi-source heterogeneous monitoring of the target deep foundation pit, obtain a water pressure deformation dataset containing multiple monitoring sections, and extract the spatial node coordinates and corresponding time series parameters of all the monitoring sections. Step 200: Determine the buoyancy spatial distribution gradient of each monitoring section based on the spatial node coordinates and the corresponding time series parameters, and obtain the current construction progress. Step 300: Calculate the dynamic resistance threshold based on the current construction progress and the buoyancy spatial distribution gradient; Step 400: Within the effective stress coupling region of each monitoring section, generate a set of spatiotemporal feature evolution nodes based on the dynamic resistance threshold, and establish a local mechanical coordinate system for each effective stress coupling region; Step 500: Extract structural deformation mutation sequences from the hydraulic deformation dataset according to the spatiotemporal feature evolution node set in the local mechanical coordinate system; Step 600: Construct a spatiotemporal coupling map on all the structural deformation mutation sequences, and use a feature fusion algorithm to generate a nonlinear decay path in the spatiotemporal coupling map to cover all the effective stress coupling regions; Step 700: Extract the local anti-buoyancy risk set on the spatiotemporal coupling map sequentially according to the nonlinear attenuation path, and perform temporal smoothing and spatial mapping on the local anti-buoyancy risk set to obtain the global anti-buoyancy evolution map; Step 800: Based on the global anti-buoyancy evolution map and the dynamic resistance threshold, risk interval matching is performed to obtain the anti-buoyancy risk assessment result of the target deep foundation pit.

[0013] Furthermore, the specific implementation process of step 100 is as follows: This embodiment first uses the vertical geological profile characteristics of the target deep foundation pit as a physical anchoring benchmark, dividing the pit into multiple groundwater level elevations along its depth. Within each groundwater level elevation, a heterogeneous sensor array is deployed. Specifically, vibrating wire pore water pressure gauges are pre-embedded in the bedrock or soil at the bottom of the pit and outside the surrounding cutoff wall. Simultaneously, large-range flexible inclinometers and surface strain gauges are closely attached to the main structure's base plate and load-bearing support piles at the corresponding elevation. Through this spatially coupled and interwoven arrangement of physical entities, this embodiment can accurately capture the transient mechanical response of the microscopic seepage field fluctuations in the deep soil and the macroscopic displacement of the main structure, thus providing high-fidelity underlying hardware support for the subsequent construction of a multi-dimensional feature data base.

[0014] Subsequently, this embodiment triggers the aforementioned heterogeneous sensor group to perform high-frequency synchronous acquisition according to a preset monitoring frequency, obtaining the original water pressure and deformation sequences at various groundwater level elevations. To completely eliminate the mechanical phase lag caused by the physical differences in sampling periods of heterogeneous sensing devices, this embodiment utilizes an underlying communication gateway to perform nanosecond-level timestamp alignment of the original water pressure and deformation sequences based on a unified absolute clock source, and performs spatial topology interpolation and noise reduction filtering on the data streams within the same elevation along the horizontal extension direction. The data after deep registration and fusion jointly generate a water pressure and deformation dataset containing multiple monitoring sections, ensuring that each independent monitoring section can accurately map the physical state of groundwater-soil coupling stress at its corresponding elevation, achieving strict isomorphism of heterogeneous monitoring data in both physical space and time dimensions.

[0015] Finally, this embodiment deeply analyzes the underlying data link topology of each independent data stream in the hydraulic deformation dataset, extracting the three-dimensional physical deployment positions of the heterogeneous sensor group in the absolute coordinate system of the foundation pit construction, which serve as spatial node coordinates representing all monitoring sections. Based on this physical spatial reference, this embodiment dynamically slides time windows into the hydraulic deformation dataset according to the extracted spatial node coordinates, accurately identifying the key sampling timestamps most severely affected by earthwork excavation unloading or forced drainage disturbances, as well as the time-series evolution step size directly reflecting the rate of structural deformation evolution, and then combining them to generate corresponding time-series parameters. This process effectively filters out steady-state redundant background data in the conventional engineering environment, condensing the complex underlying physical monitoring signals into core spatiotemporal parameters that directly characterize the dynamic degradation of the foundation's anti-buoyancy performance, greatly reducing redundant computing power consumption and effectively avoiding the divergence risk of subsequent nonlinear coupling calculations.

[0016] Furthermore, in this embodiment, when dividing the groundwater level elevation, the static groundwater level elevation extracted from the geological survey in the early stage of the project is used as the initial zero surface. This is combined downwards with the abrupt changes in water head of each confined aquifer and the highest water level fluctuation envelope during seasonal wet and dry seasons to perform physical depth slicing. When parsing the topology network, this embodiment does not extract conventional wireless communication routing links, but rather constructs a spatial adjacency physical topology map based on the physical support lattice of the deep foundation pit's main retaining structure. This map uses the physical three-dimensional deployment coordinates of each heterogeneous sensor as vertices and the physical paths of structural force transmission as solid edges. For the spatiotemporal alignment and data cleaning stages, this embodiment sets clear bottom-level operational boundaries. First, the Laida criterion is used to dynamically remove hardware spikes and abrupt noise points exceeding three standard deviations from the original water pressure and deformation sequences. Then, strictly using this spatial adjacency physical topology map as a constraint boundary, the Kriging spatial interpolation algorithm is used to map discrete sensor single-point data to continuous corresponding elevation level physical grid nodes, thereby fusing them to generate a water pressure deformation dataset. This process transforms abstract data processing into feature fusion with corresponding physical spaces, completely eliminating blind spots in data distribution and ensuring the continuous operability of the dataset in the spatial dimension.

[0017] Regarding the specific deployment and synchronization mechanism of the heterogeneous sensor group, this embodiment deploys matrix nodes along the outer water-stop curtain and the inner core load-bearing lattice columns of the target deep foundation pit according to the preset engineering load span. A mechanism combining high-frequency polling and hardware interrupt signals is used at the bottom layer of the data acquisition terminal to force all sensing nodes to complete the data latching of the current state quantity within a microsecond-level physical time difference, achieving synchronization of the real physical cross-section under harsh engineering conditions. When dividing the time window and extracting the temporal evolution step length, this embodiment binds the length range of the time window to the actual physical earthwork excavation step depth of the foundation pit, explicitly limiting the operation cycle of a single earthwork excavation process or a single layer of main slab pouring to a complete dynamic time window. Within this dynamic time window, this embodiment employs an adaptive resampling strategy based on the current geological disturbance intensity. During quiet periods without large machinery operations, it uses a long-term evolution step size on the order of hours for mean resampling. However, during periods of abrupt changes in conditions, such as severe unloading of the foundation pit or exposure to extreme rainstorms, it automatically switches to a short-term evolution step size on the order of minutes and uses cubic spline interpolation for local feature resampling. This dynamic optimization strategy not only provides clear engineering operation boundaries for time parameters but also enables the subsequent risk assessment link to capture transient anti-buoyancy distortions at the moment of change in operating conditions with zero delay.

[0018] Furthermore, the specific implementation process of step 200 is as follows: This embodiment first extracts the nodal water pressure characteristic values ​​of each monitoring section at a specific time section from the water pressure deformation dataset based on the spatial node coordinates and corresponding time series parameters. In the real geological environment of the confined aquifer at the bottom of a deep foundation pit, the physical readings of pore water pressure alone cannot be directly equivalent to the buoyancy force acting on the foundation structure. Therefore, this embodiment introduces preset foundation bearing parameters, specifically including the effective porosity of the foundation soil, the saturated unit weight of the soil, and the confined head reduction coefficient. The discrete nodal water pressure characteristic values ​​are converted into equivalent values ​​based on the effective stress principle, thereby accurately separating the coupled components of static water pressure and excess pore water pressure, and obtaining the discrete set of foundation buoyancy for each monitoring section. This physical-mechanical conversion process effectively eliminates the false increment of local effective stress in the foundation soil caused by the forced drainage of the foundation pit, and transforms the original electrical signals of the bottom sensors into engineering boundary parameters that truly reflect the actual buoyancy thrust of the foundation pit bottom slab.

[0019] Subsequently, this embodiment uses the coordinates of all extracted spatial nodes as hard geometric constraint vertices and employs the Delaunay triangulation algorithm to construct a triangular mesh model of the monitoring section on a two-dimensional continuous plane containing the outer contour of the foundation pit slab. This model serves as a physical mesh base that rigorously characterizes the topological transmission relationship of the underlying water pressure. On this physical spatial mesh architecture, this embodiment performs spatial differentiation operations in the form of finite differences along each topological connection edge of the triangular mesh model for the extracted discrete set of buoyancy at the base. By calculating the three-dimensional spatial rate of change of the buoyancy scalar values ​​between adjacent physical nodes, this embodiment derives a continuous vector field containing the direction of buoyancy extrema and the rate of water pressure mutation, i.e., the spatial distribution gradient of buoyancy at each monitoring section. This meshed spatial differentiation mechanism completely breaks through the isolated physical blind zone of traditional single-hole water level over-limit alarms, reconstructing the discrete water pressure characteristics of the base points into a continuous non-uniform buoyancy field. This enables precise location of weak zones in the engineering field where groundwater runoff is obstructed or local water sac ruptures in confined strata cause a surge in base water pressure.

[0020] While acquiring the underlying buoyancy driving field, this embodiment directly connects to a pre-set foundation pit construction management terminal via an engineering digital interface to extract the real-time excavation depth elevation and structural pouring load parameters of the target deep foundation pit at the current physical time scale. To bring the macroscopic construction data down to the mechanical calculation surface, this embodiment divides the entire target deep foundation pit into multiple three-dimensional physical construction sections, converting the real-time excavation depth elevation into the physical volume and gravity loss equivalent of the unloaded soil, and simultaneously visualizing the structural pouring load parameters as the incremental counterweight of the bottom slab and sidewall concrete entities that have reached their design strength. Based on the aforementioned unloading volume of the earthwork entities and the incremental physical counterweight of the concrete, this embodiment performs dynamic force mapping for each working stage, fusing and generating a current construction progress that accurately matches the current net weight state of the foundation pit structure. This materialization mapping method transforms the originally coarse engineering work schedule into a time-varying benchmark for anti-buoyancy constant load that can be directly used for mechanical equilibrium deduction, ensuring that the resistance calculation for risk assessment is strictly anchored to the actual physical construction evolution stage of the foundation pit.

[0021] Furthermore, in this embodiment, when extracting the node water pressure characteristic value, it does not use the instantaneous physical glitch reading, which is easily affected by the high-frequency vibration of the rotary drilling rig or crawler crane on site. Instead, it extracts the steady-state mean water pressure within the current time window after Kalman filtering and smoothing, as well as the transient extreme value used to capture the sudden breach of the local water sac in the confined aquifer. The two are combined according to a preset weight to form a physical characteristic benchmark characterizing the real hydrological environment of the node. When performing equivalent conversion to obtain the discrete set of buoyancy of the base, this embodiment uses the physical space of the foundation pit floor as the carrier and generates the corresponding Voronoi diagram based on the coordinates of each spatial node to accurately define the physical effective pressure-bearing area of ​​each heterogeneous sensor on the foundation pit floor. Subsequently, this embodiment directly multiplies the node water pressure characteristic value, the confined head reduction coefficient, and the dedicated physical effective pressure-bearing area. This physical conversion process directly transforms the water pressure scalar per unit area into the concentrated load component of the buoyancy of the base acting on the Thiessen polygon grid block, thus solidifying the calculation basis for the uneven anti-buoyancy load of deep foundation pits from the underlying logic of engineering statics.

[0022] Regarding the generation mechanism of the triangular mesh model and its two-dimensional continuous plane, this embodiment independently constructs a corresponding two-dimensional continuous plane for each independently divided monitoring section (i.e., each specific groundwater level elevation), ensuring that the water pressure topology networks of deep confined water and shallow unconfined water do not interfere with each other in physical space. When defining the physical boundary of the triangular mesh model, this embodiment directly imports the measured three-dimensional inner contour line of the continuous underground wall or interlocking pile water-stop curtain surrounding the target deep foundation pit, using this contour line as the outer rigid closed boundary to forcibly constrain the topological extension range of the internal spatial nodes. This layer-by-layer independent mesh generation strategy based on the physical boundary of the real engineering support structure not only completely eliminates the invalid numerical divergence generated by the finite difference algorithm in the blind zone of the soil outside the foundation pit, but also enables the subsequently calculated buoyancy spatial distribution gradient to accurately fit the physical weak zone at the junction of the foundation pit bottom slab and the surrounding water-stop curtain, providing a precise spatial entity anchor point for finely capturing the uneven heave of the bottom slab caused by local minor leakage of the seepage prevention curtain.

[0023] Furthermore, the specific implementation process of step 300 is as follows: Based on the current construction progress, this embodiment first extracts the self-weight parameters and sidewall soil friction parameters of the foundation pit structure that perfectly match the current physical cross-section. Regarding the extraction of the structural self-weight parameters, this embodiment abandons the traditional method of relying solely on the original design blueprint. Figure 1Instead of relying on static empirical methods, this approach uses real-time inversion to convert the concrete pouring volume and the actual hoisting measurement of large steel supports at the target deep foundation pit into the equivalent vertical gravity load per unit area on each monitoring section. This load equivalent is directly controlled by the actual physical operation nodes of on-site material allocation. For obtaining the sidewall overburden skin friction parameters, this embodiment deploys a strain sensor array in the physical sliding zone of the actual contact interface between the outer support piles and the soil to capture the active earth pressure degradation phenomenon caused by soil unloading, thereby obtaining the real-time evolving average effective sidewall pull-out skin friction. This mechanism of dynamically tracking the actual assembly of internal components and the release of physical stress in the external soil provides dynamic dead load foundation data for subsequent resistance calculations that are not divorced from the engineering physical site.

[0024] Subsequently, in this embodiment, the self-weight parameters of the foundation pit structure and the frictional resistance parameters of the side wall soil are substituted into the three-dimensional force physical space of the target deep foundation pit, and static balance superposition including the anti-buoyancy safety reserve coefficient is performed to calculate the initial anti-buoyancy dead load of the target deep foundation pit in the current construction stage.

[0025] In its specific implementation, this embodiment goes beyond simple vertical mechanical algebraic summation. Instead, it divides the entire foundation pit into multiple interconnected physical stress sub-regions and performs a coordinated static integration of the real-time physical structural gravity above the bottom slab in each region, the effective pull-out lateral friction of the external support piles, and the ultimate pull-out bearing capacity of the pre-set regional physical pull-out piles or anti-buoyancy anchors. This process of regionally superimposing the actual working efficiency of discrete physical components eliminates the illusion of local soil buoyancy reaction caused by uneven settlement of the foundation pit bottom slab. This ensures that the calculated initial anti-buoyancy dead load can truly represent the total physical ultimate bearing capacity of the entire deep foundation pit against upward deformation under absolutely ideal uniform water pressure.

[0026] After obtaining the ideal uniform load baseline, this embodiment further extracts the extreme water pressure distribution difference of each monitoring section within the current physical time window based on the buoyancy spatial distribution gradient. This distribution difference is not a simple scalar range, but rather a physical vector magnitude characterizing the spatial distance gradient between the highest pressure head node and the adjacent weakest permeable node of the foundation pit bottom slab. Simultaneously, this embodiment utilizes pre-set engineering structural modeling software to construct a three-dimensional solid finite element mesh containing the actual reinforcement ratio and concrete physical strength grade for the actually poured bottom slab and supporting beams of the target deep foundation pit. By applying a unit nodal moment to this mesh and performing static push-over inversion, a structural stiffness matrix that strictly corresponds to the geometric dimensions of the currently stressed physical components and characterizes the bending and shear deformation resistance of the bottom slab is extracted. Combined with this stiffness matrix, this embodiment further quantifies the nonlinear bending moment surge and local shear stress concentration effect caused by the extreme water pressure distribution difference at the weak sections of the foundation pit bottom slab, thereby constructing a non-uniform resistance reduction coefficient to directly characterize the degree of attenuation of the overall effective anti-buoyancy capacity of the foundation pit main structure caused by spatially non-uniform water pressure.

[0027] This embodiment directly multiplies the initial anti-buoyancy load with the non-uniform resistance reduction coefficient, which characterizes the overall structural physical deformation degradation, to obtain the dynamic resistance threshold of the target deep foundation pit. This multiplication and reduction process is physically equivalent to the premature disintegration and failure of the overall rigid anti-buoyancy system of the foundation pit due to excessively large local non-uniform pressure head causing micro-cracks or plastic hinges to form first in the weakest area of ​​the foundation slab. This dynamic resistance threshold completely breaks away from the ideal state of traditional anti-buoyancy calculations that assumes a perpetually uniform distribution of base water pressure, deviating from actual engineering conditions. It directly converts the structural eccentric stress distortion caused by real physical engineering emergencies such as localized ground seepage and uneven pumping from dewatering wells into a dynamic physical weakening of the total foundation pit resistance. This greatly enhances the risk assessment model's early warning sensitivity for chain collapse accidents caused by localized ultimate failure of deep foundation pits.

[0028] The calculation expression for the initial anti-buoyancy dead load is as follows: ; in, The initial anti-buoyancy dead load for the target deep foundation pit during the current construction phase; The resultant force of the self-weight of the structure participating in the anti-buoyancy during the current construction phase (including the self-weight of the already poured structure and the construction period dead load that can be included). The resultant force of total uplift frictional resistance provided to the sidewall-soil interface during the current construction phase; The formula for calculating the non-uniform resistance reduction factor is as follows: ; This is the reduction factor for uneven resistance; The equivalent stiffness matrix that matches the structural system at the current construction stage; For matrix The smallest eigenvalue; For matrix The largest eigenvalue; It is the difference between the maximum and minimum values ​​of pore water pressure at nodes within the same monitoring section (or the same effective stress coupling region); It is the arithmetic mean of the pore water pressure at nodes within the same monitoring section (or the same effective stress coupling region).

[0029] Furthermore, the specific implementation process of step 400 is as follows: This embodiment first defines the physically effective stress coupling region within each monitoring section. In terms of specific division rules, this embodiment abandons the conventional method of roughly grouping measuring points according to a fixed grid. Instead, it uses the post-cast strip or solid settlement joint reserved in the actual pouring of the main structure's bottom slab of the target deep foundation pit as the geometrically rigid boundary, and performs intersection calculations based on the physical seepage influence radius of the confined aquifer. On this basis, this embodiment further isolates the edge ineffective zone that has lost its bearing capacity due to lateral plastic extrusion of the soil caused by unloading disturbance of surrounding strata. The remaining solid block of the bottom slab, with continuous seepage pressure and a complete structural force transmission path, is the effective stress coupling region. Under this physical spatial constraint, this embodiment decomposes the previously calculated dynamic resistance threshold according to the actual amortized physical counterweight ratio of each block, and equivalently maps it to the interior of each effective stress coupling region, thereby establishing a regional-level anti-buoyancy mechanical baseline.

[0030] After completing the regional mechanical mapping, this embodiment further calculates the local resistance redundancy of the coordinates of each spatial node within the effective stress coupling region. Specifically, this embodiment extracts the effective value of the actual local dynamic resistance allocated to each spatial node as the numerator, and uses the transient buoyancy load currently borne by the node as the denominator for division, thus directly quantifying the node's anti-buoyancy safety margin using this physical ratio. Simultaneously, based on the structural ultimate limit state bearing capacity standard in current engineering specifications, this embodiment strictly sets the physical strain ratio that leads to the critical state of plastic micro-crack initiation in the foundation slab concrete as a preset safety threshold. Through comprehensive comparison, this embodiment accurately selects the coordinates of spatial nodes with local resistance redundancy lower than this preset safety threshold, extracting these nodes, which are already close to the edge of ultimate cracking in their physical stress state, as locally weak spatial points, thereby directly locking down the true source of stress threat in the complex blind area of ​​the foundation pit foundation slab.

[0031] For the extracted local weak points, this embodiment performs multidimensional feature clustering and temporal dimensional expansion based on the corresponding time series parameters. In terms of the physical composition of the feature vectors, this embodiment not only selects the three-dimensional geometric absolute coordinates of each weak point, but also deeply integrates the node physical deformation evolution rate and transient water pressure mutation increment derived from the time series parameters, jointly constructing a multidimensional feature vector characterizing the dynamic degradation of the physical state. In terms of clustering algorithm application, this embodiment uses a density-based noisy spatial clustering algorithm to perform unsupervised aggregation of the multidimensional feature vectors. This entity algorithm mechanism does not require manual preset or subjective intervention in the number of clusters, and can adaptively divide the entity region clusters with coordinated stress deterioration based entirely on the uneven physical aggregation density of the actual water pressure distribution of the base. Subsequently, this embodiment traces back multiple physical evolution observation cycles from the current time window to perform temporal state dimensional expansion of each cluster, transforming the originally static isolated weak points into a spatiotemporal feature evolution node set containing a complete stress deterioration history trajectory, providing a continuous data foundation for accurately capturing the dynamic expansion trend of hidden cracks in the base.

[0032] While generating the spatiotemporal feature evolution node set, this embodiment extracts the geometric centroid position of the outer contour of each effective stress coupling region entity, and derives the maximum principal stress direction vector based on the buoyancy spatial distribution gradient obtained in the previous steps. In terms of the specific physical deduction mechanism, this embodiment uses the buoyancy spatial distribution gradient as an external driving force boundary condition, directly acting on the pre-constructed physical finite element model of the foundation pit slab entity. Through the analysis of the elastic mechanical spatial stress tensor, the internal force response distribution matrix of each entity micro-element under the non-uniform water buoyancy push of the slab is calculated. This embodiment further extracts the micro-element cross-section with the most severe bending and torsional responses, and determines the direction perpendicular to the physical section normal where the shear stress is zero on this cross-section as the maximum principal stress direction vector. This deduction calculation process directly transforms the macroscopic buoyancy water pressure gradient characteristics into the microscopic mechanical dominant elongation direction that causes physical tearing of the concrete slab, thoroughly ensuring the rigorous availability of the mechanical vector source and the consistency with the on-site physical stress state.

[0033] After obtaining the aforementioned core physical parameters, this embodiment uses the extracted geometric centroid location as the origin of the absolute physical coordinate system and the derived maximum principal stress direction vector as the principal axis of the local coordinate system. A dedicated local mechanical coordinate system is established for each independent effective stress coupling region subject to force transmission. This local mechanical coordinate system completely overcomes the engineering defects of traditional global orthogonal geodetic coordinate systems in foundation pit construction, which cannot accurately reflect local non-uniform torsional deformation. Its principal axis is completely parallel to the physical propagation path of the foundation structure most prone to tensile failure and through-cracks. Subsequent matrix operations by uniformly replacing the spatiotemporal feature evolution node set with this local mechanical coordinate system effectively filter out background displacement noise unrelated to the principal force direction, directionally amplify and capture minute deformation abrupt changes along the principal stress direction, greatly improving the evaluation model's accuracy in predicting sudden water inrush or brittle fracture failure of the foundation pit floor slab and enhancing its physical perception sensitivity.

[0034] Specifically, the formula for calculating local resistance redundancy is: ; in, For the effective stress coupling region, the first Local resistance redundancy of each spatial node; For the first The equivalent upward pressure at each node (the upward effect per unit area obtained by converting the monitored water pressure into pressure). For the first The representative area corresponding to each node; This represents the dynamic resistance threshold for the current construction phase. Equivalent stiffness matrix In and nodes The diagonal terms corresponding to the degrees of freedom; For a set of nodes The sum of the corresponding diagonal terms; This refers to the set of nodes participating in the calculation within the current effective stress coupling region; Number the nodes; For set Internal summation index.

[0035] Furthermore, the specific implementation process of step 500 is as follows: This embodiment first precisely maps the previously generated spatiotemporal feature evolution node set from the global geophysical three-dimensional coordinate system to the dedicated local mechanical coordinate system to obtain the relative spatial position of each evolution node in this local mechanical coordinate system. This coordinate translation and rotation matrix operation mechanism effectively avoids the interference of rigid body mean heave caused by the large-area unloading of soil in the foundation pit. Subsequently, based on the relative spatial position and the corresponding time series parameters, this embodiment directly extracts the three-dimensional deformation displacement of the spatiotemporal feature evolution node set from the hydraulic deformation dataset synchronously. In terms of data physical source, this three-dimensional deformation displacement is directly fed back by a large-range flexible inclinometer and a three-dimensional settlement node array embedded in the concrete base slab and support structure, and has already been preprocessed in the underlying data cleaning stage. In this extraction process, this embodiment strictly maintains that the direction of the three-dimensional deformation displacement is completely coincident with the principal and secondary axes of the local mechanical coordinate system. The macroscopic absolute displacement is deconstructed into the tensile elongation component along the direction of the local maximum principal stress and the shear torsion component perpendicular to the principal axis, thereby establishing a precise spatial vector reference for subsequently capturing the abrupt changes on a specific mechanical failure surface.

[0036] After obtaining the three-dimensional deformation displacement within the local space, this embodiment performs a first-order difference operation along the corresponding time series parameters to calculate the deformation rate of each evolution node and the deformation acceleration of adjacent time steps. Given that heterogeneous sensors in automated monitoring of deep foundation pits often employ adaptive resampling strategies under different operating conditions, resulting in timestamps in the time series parameters exhibiting non-uniform intervals, direct algebraic difference can easily lead to distortion of mechanical characteristics. Therefore, this embodiment introduces a non-uniform time series difference algorithm with a Gaussian smoothing kernel before performing the first-order difference, strictly dividing the three-dimensional deformation displacement increment between two adjacent physical sampling times by the corresponding actual physical time interval. This differential operation based on the actual irregular time step effectively eliminates pseudo-acceleration fluctuations caused by variations in the sensor's sleep-wake cycle interval, accurately restoring the true transient kinematic physical response characteristics of the base structure under complex non-uniformly distributed water buoyancy.

[0037] After extracting the aforementioned kinematic features, this embodiment further filters evolution nodes whose deformation acceleration exceeds a preset deformation abrupt change threshold. This preset deformation abrupt change threshold is not a fixed empirical constant, but rather an adaptive statistical confidence upper limit threshold calculated using a Bayesian update method based on the regional geological historical settlement database of the target deep foundation pit site and the current reinforcement stiffness of the foundation slab. Specifically, this embodiment extracts the acceleration background noise variance of each node during the quiet period of the working condition, and dynamically accumulates the structural elastoplastic strain safety margin matching the working condition during the active construction period of severe unloading of the foundation pit. The product of these two values ​​is then used as the preset deformation abrupt change threshold under the current working condition window. This adaptive threshold setting mechanism, driven by both physical working conditions and historical statistics, can ensure high sensitivity to real water pressure breakthrough events while significantly filtering out false deformation acceleration alarm signals induced by occasional construction live loads such as the passage of large crawler cranes.

[0038] For the selected evolution nodes that exceed the preset deformation mutation threshold, this embodiment concatenates their corresponding three-dimensional deformation displacement and deformation rate in chronological order to obtain the structural deformation mutation sequence. During the concatenation process, this embodiment sets strict physical boundary constraints on the sequence length and minimum duration. Specifically, sampling points continuously exceeding the threshold must have a duration greater than a complete structural physical natural vibration period, and the total sequence length must cover the complete mechanical process from the initial sharp increase in deformation rate to the local yield extreme. For discrete, isolated spike data that do not meet the duration requirement, this embodiment uses a wavelet soft thresholding denoising algorithm to remove them as high-frequency noise entities. This strict concatenation and denoising mechanism filters out transient signal spikes, ensuring that the generated structural deformation mutation sequence can realistically and completely characterize the timeline of irreversible physical failure in locally weak regions under the continuous action of non-uniform buoyancy, from elastic deformation to a surge in plastic nonlinear displacement.

[0039] Furthermore, the specific implementation process of step 600 is as follows: This embodiment first maps all previously extracted structural deformation mutation sequences into graph nodes of a spatiotemporally coupled graph. Regarding the specific mapping rules and node attribute definitions, this embodiment does not simply abstract a single numerical value into a node. Instead, it uses the absolute three-dimensional coordinates of the local weak points in the spatial space that generated the structural deformation mutation sequence as the physical geometric origin of the graph node. Simultaneously, the entire structural deformation mutation sequence, containing the nonlinear increase history of deformation rate and deformation acceleration over time, is encapsulated as the dynamic temporal feature vector attribute of the graph node. This physically materialized mapping mechanism ensures that each discrete node in the graph structure precisely corresponds to a real, micro-crack-propagating solid stress unit on the foundation pit floor, thus laying a data carrier with real engineering physical significance for subsequent global graph theory calculations.

[0040] After establishing the graph nodes, this embodiment extracts the spatial topological connectivity of each graph node across all effective stress coupling regions and connects adjacent graph nodes based on this spatial topological connectivity to generate initial graph edges. Regarding the connection criteria for spatial topological connectivity, this embodiment strictly adheres to the physical force transmission path of the actual steel mesh tied to the foundation pit floor slab and the shear force transmission influence radius of the concrete entity, setting a physical distance threshold that characterizes the effective force transmission limit of the structure. Only when two graph nodes are within the same effective stress coupling region and their spatial Euclidean distance is less than this physical distance threshold is this embodiment considered to indicate a real possibility of mechanical tear transmission between them, thus generating an initial graph edge connecting them. This graph edge generation strategy based on the stiffness transmission boundary of the solid structure effectively eliminates physically isolated deformation islands, ensuring that the constructed topological network strictly conforms to the actual physical load-bearing skeleton of the foundation pit.

[0041] After obtaining the initial graph edges, this embodiment calculates and assigns range weights to each initial graph edge based on the previously calculated buoyancy spatial distribution gradient. In this process, the range weight is not simply the difference between absolute water pressure scalars, but rather the absolute value of the actual confined head range calculated by integrating the buoyancy spatial distribution gradient along the physical connection direction of the generated initial graph edges. This range weight precisely quantifies the difference in the actual non-uniform water thrust gradient between two adjacent physical nodes experiencing abrupt deformation. Using this range weight as an attribute of the initial graph edges, this embodiment associates and combines graph nodes as vertex sets, initial graph edges as edge sets, and range weights as graph edge attributes to construct a complete spatiotemporal coupled graph. At this point, the spatiotemporal coupled graph is no longer an abstract data structure, but a holographic mapping of the physical and mechanical network interwoven with the non-uniform water pressure driving force and the local structural damage state of the foundation pit floor.

[0042] Based on the constructed spatiotemporal coupled graph, this embodiment employs a feature fusion algorithm to perform a heuristic search on graph nodes within the spatiotemporal coupled graph, fusing the structural deformation mutation sequences of each graph node with the adjacent range weights to calculate the local attenuation cost. In the specific implementation of the feature fusion algorithm and heuristic search, this embodiment uses the A* heuristic search algorithm with a physical penalty factor, performing a joint matrix operation on the peak deformation acceleration contained in the feature vector of the graph node and the range weights on the adjacent graph edges. This embodiment introduces an exponential stiffness attenuation function characterizing the tensile softening effect of concrete in this operation. When the structural deformation mutation sequence of a node surges and the adjacent range weights are too large, this attenuation function outputs an exponentially amplified physical and mechanical penalty cost, thus reflecting the nonlinear characteristics. The calculated local attenuation cost directly reflects the lower limit of physical energy consumption for the local slab to evolve from microcracks to through-type failure under the drive of water pressure gradient.

[0043] Based on the calculated local attenuation cost, this embodiment employs an improved minimum spanning tree algorithm to solve for the global minimum cost spanning tree connecting all graph nodes. A nonlinear attenuation path is then generated according to the branch structure of the global minimum cost spanning tree to cover all effective stress coupling regions. This nonlinear attenuation path is not an absolute straight line in the spatial geometric coordinate system, but rather a predicted crack propagation trajectory that meanders along the physically weakest surface where the tensile stiffness of the foundation slab degrades fastest and the nonlinear exponential attenuation work is greatest. By connecting all evolution nodes through this nonlinear attenuation path, this embodiment can accurately depict the optimal physical collapse channel that spreads outwards after a local surge occurs at the bottom of the foundation pit, using the global minimum anti-buoyancy failure energy work as a guide. This engineering effect completely overcomes the shortcomings of traditional single-point early warning systems, which cannot predict the direction of continuous global collapse evolution of the foundation pit, providing extremely accurate spatial entity targeting guidance for deploying pressure relief wells or increasing counterweights in advance for weak force transmission paths.

[0044] Specifically, the expression for calculating the local attenuation cost is as follows: ; in, Nodes in the spatiotemporal coupling graph With nodes The local attenuation cost between them; The weights are the edge ranges of the graph. For equivalent upward pressure field In node coordinates The spatial gradient vector at that location; For nodes 3D spatial coordinate vector; For nodes 3D spatial coordinate vector; For nodes With nodes The Euclidean distance; All edges in the initial graph edge set The maximum value; For nodes With nodes The time difference corresponding to the mutation event; All edges in the initial graph edge set The maximum value; For nodes The feature vector of the mutation sequence; For nodes The feature vector of the mutation sequence; All edges in the initial graph edge set The maximum value; For nodes, determine the timestamps of mutation events; For nodes The timestamp corresponding to the mutation event; Number the nodes in the graph; For nodes There are adjacent graph node numbers that are initially connected by graph edges.

[0045] Furthermore, the specific implementation process of step 700 is as follows: This embodiment first traverses the constructed spatiotemporal coupling graph along the generated nonlinear decay path to obtain the structural deformation mutation sequence and adjacent spatial topological relationships corresponding to each physical node on the nonlinear decay path. After obtaining the above features, this embodiment does not use a simple empirical threshold to subjectively discretize and classify the risk, but calculates the dynamic anti-buoyancy failure probability of each node on the nonlinear decay path based on the adjacent spatial topological relationships and structural deformation mutation sequences. In specific implementation, this embodiment introduces a Bayesian reliability update model based on physical mechanics priors, using the peak deformation acceleration of the node at the current time step as the real-time observed physical input of the likelihood function, and using the design ultimate flexural bearing capacity of the node under the constraints of the corresponding adjacent spatial topological relationships as the physical prior distribution benchmark. Through Markov chain Monte Carlo sampling for posterior probability inference, this embodiment directly outputs a dynamic anti-buoyancy failure probability between zero and one, which strictly characterizes the physical possibility of tensile yielding or microcrack penetration of the bottom concrete, completely solving the engineering pain point of traditional index evaluation lacking a clear physical boundary of mechanical failure.

[0046] After calculating the physical failure probability of each entity, this embodiment filters out nodes with a dynamic anti-buoyancy failure probability higher than a preset benchmark value, performs topological clustering based on entity connectivity, and extracts a local anti-buoyancy risk set. In this step, this embodiment employs a density-based spatial clustering algorithm with physical stiffness grid constraints, strictly anchoring the preset benchmark value to the reliability index conversion probability limit corresponding to the normal serviceability limit state in current engineering specifications. During clustering, the algorithm not only requires nodes to numerically exceed the benchmark but also mandates a continuous spatial topological connectivity relationship along the force transmission path of the base entity's reinforcing steel between high-probability nodes. This physical entity-based clustering criterion accurately eliminates isolated high-probability illusions caused by individual sensor hardware drift, merging and extracting the truly dangerous points that spread in a clustered state along the weakest physical section of the base slab structure into a local anti-buoyancy risk set, realizing the transformation of risk from single-point alarm to regional collaborative deterioration of the physical entity surface.

[0047] For the extracted local anti-buoyancy risk set, this embodiment extracts the dynamic risk fluctuation curve of the local anti-buoyancy risk set in the time dimension according to the corresponding time series parameters. Given that the passage of heavy-duty dump trucks or frequent start-stop of deep well dewatering pumps around the foundation pit site will induce a large amount of high-frequency disturbance stress, this embodiment uses a sliding time window algorithm to perform low-pass filtering on the dynamic risk fluctuation curve, obtaining a time-domain smoothed risk feature set. In terms of specific filtering parameter settings, this embodiment strictly aligns the physical window length of the sliding time window with the standard engineering workflow time of single-layer earthwork excavation of the target deep foundation pit basement, and sets the cutoff frequency of the low-pass filter to directly shield high-frequency components higher than the natural frequency of the foundation slab and the micro-oscillation frequency of the site environment. This customized filtering mechanism effectively filters out transient mechanical vibration noise unrelated to the long-term physical evolution of the actual soil and water pressure, enabling the finally extracted time-domain smoothed risk feature set to reflect the mechanical variation characteristics of the basement pushover caused by the trend of confined water level rise with extreme purity.

[0048] Subsequently, this embodiment reverse-maps the time-domain smoothed risk feature set to the global three-dimensional geodetic coordinate system of the target deep foundation pit, accurately obtaining discrete anti-buoyancy risk field nodes containing spatially distributed physical absolute coordinates and continuous time steps. Based on the above discrete anti-buoyancy risk field nodes, this embodiment uses a three-dimensional spatial interpolation algorithm to reconstruct risk values ​​and fit surfaces in the physical blind zone where no sensors are deployed, obtaining a global anti-buoyancy evolution map. In terms of the core reconstruction algorithm, this embodiment adopts a generalized kriging spatial interpolation algorithm that takes into account the anisotropy of the structure's physics, introducing the solid bending stiffness gradient caused by the difference in reinforcement ratio within the foundation pit slab as a covariate into the semivariogram function calculation. At the same time, this embodiment calculates residuals through cross-validation and sets mandatory physical error boundary constraints, requiring that the residuals of the reconstructed surface at each monitoring entity node must not exceed the nominal measurement accuracy of the sensors under the current working conditions. This entity-based fitting method strictly constrains the mathematical numerical divergence, ensuring that the interpolated surface fully conforms to the true physical warping envelope shape of the slab under non-uniform water buoyancy.

[0049] The final generated global anti-buoyancy evolution map is a three-dimensional continuous field model representing the spatial distribution trend and transmission gradient of the overall base buoyancy and dynamic resistance of the main structure in a non-equilibrium state in the deep foundation pit. Addressing the question of how a finite number of sparse monitoring sensors can support the engineering rigor verification of this three-dimensional continuous field model reconstruction, this embodiment establishes a mechanical assumption based on the continuous coordination of the physical deformation of the large-volume reinforced concrete base slab of the foundation pit. In terms of practical verification, this embodiment randomly selects redundant settlement physical verification points that are not involved in the modeling on-site, and periodically compares the predicted evolution value of these points output by the model with the actual high-precision physical leveling settlement results, continuously and dynamically calibrating the anisotropic variability function parameters of the aforementioned Kriging interpolation. This three-dimensional continuous field model intuitively reproduces the entire physical evolution of the base confined water breaking through the weak permeable layer and causing nonlinear local overburden damage to the main structure in a holographic three-dimensional digital space. This provides engineering commanders with a solid decision-making foundation with rigorous digital and mechanical support for targeted deployment of decompression and anti-buoyancy wells in the downstream physically weak areas of the evolution transmission gradient.

[0050] Specifically, the calculation expression for the dynamic anti-buoyancy failure probability is as follows: ; in, For nodes The probability of dynamic anti-buoyancy failure at the location; It is a natural exponential function; For nodes The equivalent upward pressure at the location; The representative area corresponding to each node; To map to nodes Local dynamic resistance capacity at the location; This represents the dynamic resistance threshold for the current construction phase. Equivalent stiffness matrix In and nodes The diagonal terms corresponding to the degrees of freedom; For a set of nodes The sum of the corresponding diagonal terms; This refers to the set of nodes participating in the calculation within the current effective stress coupling region; Number the nodes; For set Internal summation index.

[0051] Furthermore, the specific implementation process of step 800 is as follows: This embodiment first analyzes the previously constructed global anti-buoyancy evolution map in depth, accurately extracting the real-time anti-buoyancy load equivalent and local load abrupt change characteristics of the target deep foundation pit in the global three-dimensional spatial system. In terms of specific physical definitions and extraction paths, the real-time anti-buoyancy load equivalent does not come from the raw readings of a single sensor, but rather from the absolute mechanical value of the overall structural buoyancy lifting force obtained by surface integral calculation of the upward thrust water pressure vector field distributed in space along the three-dimensional reconstructed surface representing the physical bottom surface of the foundation plate in this embodiment. Simultaneously, this embodiment calculates the curvature and spatial derivative along different directions for this three-dimensional continuous field model, extracting the solid coordinate regions where the curvature deflects drastically and the spatial derivative extrema cluster as local load abrupt change characteristics. This analytical process rigorously transforms the abstract spatial surface trend in the map into the macroscopic total thrust and microscopic non-uniform shear failure sources directly acting on the foundation pit's solid structure, ensuring that the subsequent matching data base has a reliable mechanical and geometric entity mapping relationship.

[0052] After obtaining the aforementioned load characteristics, this embodiment performs gradient discretization on the previously determined dynamic resistance threshold according to a preset safety reduction ratio, thereby constructing a dynamic risk threshold interval containing multiple warning levels. In terms of discretization rules and level settings, this embodiment strictly adheres to the physical failure evolution stages of solid reinforced concrete flexural members, setting three rigid physical safety reduction ratio boundaries of 85%, 70%, and 50% sequentially downwards for the dynamic resistance threshold. Based on these physical strength reduction boundaries, this embodiment constructs a four-level progressive dynamic risk threshold interval, corresponding to the green interval where the structure is in a normal elastic working state, the yellow interval where microcracks in the base slab initiate but close, the orange interval where the main reinforcing steel reaches its yield strength, and the red interval where the base slab experiences through-fracture failure. This threshold discretization strategy, anchoring the true elastic-plastic development process of the material, completely abandons subjective experience limits set arbitrarily, giving each warning level a clear semantic meaning of physical entity failure.

[0053] Subsequently, this embodiment inputs the extracted real-time anti-buoyancy load equivalent into the aforementioned divided dynamic risk threshold interval for rigorous boundary matching and difference calculation, thereby obtaining the local anti-buoyancy failure probability under each global three-dimensional spatial system. In terms of the specific computational physical mechanism, this embodiment does not simply compare numerical values, but subtracts the lower physical threshold of the interval into the real-time anti-buoyancy load equivalent, and divides this mechanical difference by the total physical difference between the upper and lower limits of the current interval, performing nonlinear normalization calculation. This boundary matching and difference calculation mechanism essentially solves for the energy consumption evolution ratio of the current physical foundation's pushing force in a specific physical failure stage. Through this calculation, this embodiment smoothly transforms the complex absolute tonnage difference of the physical structure into a local anti-buoyancy failure probability between 0 and 1, representing the sliding rate of the local foundation structure towards the next severe physical failure state, providing a high-precision probabilistic scale for quantifying the local anti-buoyancy mechanical redundancy of deep foundation pits.

[0054] Based on the calculated local anti-buoyancy failure probabilities, this embodiment further utilizes extracted local load abrupt change features for spatial aggregation and weighted evaluation. In terms of physical aggregation logic, this embodiment uses the peak curvature gradient, which characterizes the degree of shear stress concentration in the local load abrupt change features, as an adaptive physical weighting factor, assigning higher weight amplification multipliers to failure probabilities located at the corners of the foundation pit floor slab, around load-bearing lattice columns, and other locations highly susceptible to punching shear failure. This embodiment performs weighted integration and global summation of the weighted amplified local anti-buoyancy failure probabilities within the entire three-dimensional foundation pit floor slab entity space, generating the final overall anti-buoyancy risk assessment result for the target deep foundation pit. This spatial weighting mechanism accurately reproduces the chain-like failure physical property in engineering practice where local water bladder breaches often trigger the instantaneous paralysis of the global anti-buoyancy system, enabling the final risk assessment to keenly capture and amplify the fatal threat posed by minor local hazards to the overall stability of the foundation pit.

[0055] The generated anti-buoyancy risk assessment results are specifically manifested as a comprehensive control instruction set including the overall anti-buoyancy safety level, the three-dimensional location of the most unfavorable local anti-buoyancy weak point, and the corresponding intervention strategy. To avoid the intervention strategy falling into the abstract engineering management category and being identified as a non-technical feature, the intervention strategy output in this embodiment is the underlying electrical signal and equipment start-stop logic sequence that directly triggers the automated mechanical hardware actions on the foundation pit. For the three-dimensional absolute coordinates of the most unfavorable local anti-buoyancy weak point, the comprehensive control instruction set in this embodiment directly generates execution code that can be recognized by the field programmable logic controller, automatically scheduling and controlling the automatically numbered automated depressurization and dewatering wells deployed around the coordinate area. By controlling the physical pumping rate of the variable frequency water pump in a closed loop, or automatically opening the preset physical depressurization and dewatering valves on the bottom plate, this embodiment transforms the risk assessment probability in the virtual space into a hardcore intervention action in the physical space in real time, realizing a millisecond-level automatic closed-loop machine response from deep foundation pit anti-buoyancy risk perception to physical entity depressurization and risk mitigation.

[0056] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0057] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for assessing the buoyancy risk of deep foundation pits based on monitoring data, characterized in that, include: Based on the preset monitoring frequency, the target deep foundation pit is monitored in a layered, multi-source heterogeneous manner to obtain a water pressure deformation dataset containing multiple monitoring sections, and to extract the spatial node coordinates and corresponding time series parameters of all the monitoring sections. The buoyancy spatial distribution gradient of each monitoring section is determined based on the spatial node coordinates and the corresponding time series parameters, and the current construction progress is obtained. Calculate the dynamic resistance threshold based on the current construction progress and the buoyancy spatial distribution gradient. Within the effective stress coupling region of each monitoring section, a spatiotemporal feature evolution node set is generated based on the dynamic resistance threshold, and a local mechanical coordinate system is established for each effective stress coupling region; Under the local mechanical coordinate system, structural deformation mutation sequences are extracted from the hydraulic deformation dataset according to the spatiotemporal feature evolution node set; Spatiotemporal coupling maps are constructed on all the structural deformation mutation sequences, and a feature fusion algorithm is used to generate nonlinear decay paths in the spatiotemporal coupling maps to cover all the effective stress coupling regions; The local anti-buoyancy risk set on the spatiotemporal coupling map is extracted sequentially according to the nonlinear attenuation path, and the local anti-buoyancy risk set is smoothed in the temporal domain and mapped in space to obtain the global anti-buoyancy evolution map. Based on the global anti-buoyancy evolution map and the dynamic resistance threshold, risk interval matching is performed to obtain the anti-buoyancy risk assessment result of the target deep foundation pit.

2. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, The method involves layered, multi-source heterogeneous monitoring of the target deep foundation pit based on a preset monitoring frequency, acquiring a water pressure deformation dataset containing multiple monitoring sections, and extracting the spatial node coordinates and corresponding time series parameters of all the monitoring sections, including: The target deep foundation pit is divided into multiple groundwater level elevations based on the vertical geological profile, and a heterogeneous sensor group is deployed in each of the groundwater level elevations. The heterogeneous sensor group is triggered to synchronously collect data according to the preset monitoring frequency, thereby obtaining the original water pressure sequence and deformation sequence of each groundwater level elevation. The original water pressure sequence and the original deformation sequence are spatiotemporally aligned and cleaned, and then fused to generate the water pressure deformation dataset containing multiple monitoring sections, wherein the monitoring section is a physical monitoring plane corresponding to the groundwater level elevation. The topology network of the heterogeneous sensor group corresponding to each independent data stream in the water pressure deformation dataset is analyzed, and the three-dimensional deployment position of the heterogeneous sensor group is extracted as the spatial node coordinates of all the monitoring sections. The hydraulic deformation dataset is divided into time windows based on the spatial node coordinates, and the corresponding sampling timestamps and time series evolution steps are extracted to obtain the corresponding time series parameters.

3. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, The step of determining the buoyancy spatial distribution gradient of each monitoring section based on the spatial node coordinates and the corresponding time series parameters, and obtaining the current construction progress, includes: Based on the spatial node coordinates and the corresponding time series parameters, extract the nodal water pressure characteristic values ​​of each monitoring section from the water pressure deformation dataset; Based on the node water pressure characteristic value and the preset base bearing pressure parameter, an equivalent conversion is performed to obtain the discrete set of base buoyancy of each monitoring section; Based on the spatial node coordinates, a triangular mesh model of the monitoring section is constructed, and the spatial derivative of the discrete set of the buoyancy of the base water is obtained on the triangular mesh model to obtain the spatial distribution gradient of buoyancy of each monitoring section. Connect to the preset foundation pit construction management terminal to extract the real-time excavation depth elevation and structural pouring load parameters of the target deep foundation pit; The current construction progress is obtained by mapping the working condition stages based on the real-time excavation depth and elevation and the structural pouring load parameters.

4. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 3, characterized in that, The triangular mesh model is a two-dimensional continuous plane with the coordinates of the spatial nodes as vertices, representing the topological relationship of the base water pressure.

5. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, The calculation of the dynamic resistance threshold based on the current construction progress and the buoyancy spatial distribution gradient includes: Based on the current construction progress, extract the self-weight parameters of the foundation pit structure and the sidewall soil friction parameters that match the current construction progress. The self-weight parameters of the foundation pit structure and the frictional resistance parameters of the sidewall soil are statically balanced and superimposed to calculate the initial anti-buoyancy dead load of the target deep foundation pit in the current construction stage. The extreme water pressure distribution difference of each monitoring section is extracted based on the buoyancy spatial distribution gradient, and the non-uniform resistance reduction coefficient is calculated based on the extreme water pressure distribution difference and the preset structural stiffness matrix. The dynamic resistance threshold of the target deep foundation pit is obtained by multiplying the initial anti-buoyancy load with the non-uniform resistance reduction coefficient.

6. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, Within the effective stress coupling region of each monitoring section, a spatiotemporal feature evolution node set is generated based on the dynamic resistance threshold, and a local mechanical coordinate system is established for each effective stress coupling region, including: The dynamic resistance threshold is mapped to the effective stress coupling region of each monitoring section, and the local resistance redundancy of each spatial node coordinate within the effective stress coupling region is calculated. The coordinates of spatial nodes whose local resistance redundancy is lower than a preset safety threshold are selected to extract spatial points with weak local resistance. Based on the corresponding time series parameters, multidimensional feature clustering and temporal dimension expansion are performed on the local weak resistance spatial points to obtain the spatiotemporal feature evolution node set. Extract the geometric centroid position of each effective stress coupling region and the maximum principal stress direction vector derived from the buoyancy spatial distribution gradient; A local mechanical coordinate system is established for each effective stress coupling region, with the geometric centroid position as the origin and the maximum principal stress direction vector as the principal axis.

7. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, The step of extracting structural deformation mutation sequences from the hydraulic deformation dataset according to the spatiotemporal feature evolution node set in the local mechanical coordinate system includes: The spatiotemporal feature evolution node set is mapped to the local mechanical coordinate system to obtain the relative spatial position of each evolution node in the local mechanical coordinate system; Based on the relative spatial position and the corresponding time series parameters, the three-dimensional deformation displacement of the spatiotemporal feature evolution node set is synchronously extracted from the hydraulic deformation dataset; The first-order difference operation is performed on the three-dimensional deformation displacement along the time series parameters to calculate the deformation rate of each evolution node and the deformation acceleration of adjacent time steps; The evolution nodes whose deformation acceleration exceeds a preset deformation mutation threshold are selected, and the three-dimensional deformation displacement and deformation rate corresponding to the current evolution node are concatenated and spliced ​​in time order to obtain the structural deformation mutation sequence.

8. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 7, characterized in that, A spatiotemporal coupling map is constructed on all the aforementioned structural deformation abrupt sequence, and a feature fusion algorithm is used to generate a nonlinear decay path in the spatiotemporal coupling map to cover all the aforementioned effective stress coupling regions, including: All the structural deformation mutation sequences are mapped to graph nodes, and the spatial topological connectivity of each graph node in all the effective stress coupling regions is extracted; The initial graph edges are generated by connecting adjacent graph nodes according to the spatial topological connectivity relationship, and the range weight of each initial graph edge is calculated based on the buoyancy spatial distribution gradient. The spatiotemporal coupled graph is constructed by associating and combining the graph nodes as the vertex set, the initial graph edges as the edge set, and the range weight as the graph edge attribute. A feature fusion algorithm is used to perform a heuristic search on the graph nodes in the spatiotemporal coupled graph, and the local attenuation cost is calculated by fusing the structural deformation mutation sequence of each graph node and the connected range weights. The global minimum cost spanning tree connecting all the graph nodes is solved based on the local attenuation cost, and a nonlinear attenuation path is generated according to the branch structure of the global minimum cost spanning tree to cover all the effective stress coupling regions.

9. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, Local anti-buoyancy risk sets are extracted sequentially from the spatiotemporal coupling map according to the nonlinear attenuation path, and the local anti-buoyancy risk sets are smoothed in the temporal domain and mapped spatially to obtain a global anti-buoyancy evolution map, including: Traverse the spatiotemporal coupling graph along the nonlinear decay path to obtain the structural deformation mutation sequence and adjacent spatial topological relationships corresponding to each node on the nonlinear decay path; Based on the adjacent spatial topological relationship and the structural deformation mutation sequence, the dynamic anti-buoyancy failure probability of each node on the nonlinear attenuation path is calculated, and the nodes with dynamic anti-buoyancy failure probability higher than the preset benchmark value are topologically clustered to extract the local anti-buoyancy risk set. Based on the corresponding time series parameters, the dynamic risk fluctuation curve of the local anti-buoyancy risk set in the time dimension is extracted, and the dynamic risk fluctuation curve is low-pass filtered by the sliding time window algorithm to obtain the time-domain smooth risk feature set. The time-domain smoothed risk feature set is inversely mapped to the global three-dimensional coordinate system of the target deep foundation pit to obtain discrete anti-buoyancy risk field nodes containing spatial distribution coordinates and continuous time steps; Based on the discrete anti-buoyancy risk field nodes, a three-dimensional spatial interpolation algorithm is used to reconstruct risk values ​​and fit surfaces in the blind area where no sensors are deployed, thereby obtaining the global anti-buoyancy evolution map. The global anti-buoyancy evolution map is a three-dimensional continuous field model that characterizes the spatial distribution trend and transmission gradient of the overall base buoyancy and dynamic resistance of the main structure of the target deep foundation pit in a non-equilibrium state.

10. The method for assessing the buoyancy risk of deep foundation pits based on monitoring data according to claim 1, characterized in that, The step of matching risk intervals based on the global anti-buoyancy evolution map and the dynamic resistance threshold to obtain the anti-buoyancy risk assessment result of the target deep foundation pit includes: The global anti-buoyancy evolution map is analyzed to extract the real-time anti-buoyancy load equivalent and local load mutation characteristics of the target deep foundation pit in the global three-dimensional space system; The dynamic resistance threshold is gradient discretized according to a preset safety reduction ratio to construct a dynamic risk threshold range containing multiple warning levels; The real-time anti-buoyancy load equivalent is input into the dynamic risk threshold range for boundary matching and difference calculation to obtain the local anti-buoyancy failure probability under each global three-dimensional spatial system. Based on the local load mutation characteristics, spatial aggregation and weighted evaluation of all local anti-buoyancy failure probabilities are performed to generate the anti-buoyancy risk assessment result of the target deep foundation pit. The anti-buoyancy risk assessment result is a comprehensive control instruction set that includes the overall anti-buoyancy safety level, the three-dimensional location of the most unfavorable local anti-buoyancy weak point, and the corresponding intervention strategy.

Images