Integrated foundation pit support monitoring and early warning system and method

Abstract

The invention discloses an integrated foundation pit support monitoring and early warning system and method. A multi-source sensor array, a distributed data processing unit and an intelligent early warning module form a cooperative work framework; the multi-source sensor array is deployed in a stress concentration node area of a foundation pit enclosure wall body and an internal steel support system by adopting a multi-level nested structure, and the data processing unit carries out space-time registration and feature fusion processing on a multi-source heterogeneous monitoring data flow, and constructs a dynamic safety margin evaluation model containing a steel support axial force-bending moment coupling effect; and the intelligent early warning module generates a multi-level risk early warning signal based on the output result of the dynamic safety margin evaluation model. According to the method, the axial force / bending moment / temperature parameters are synchronously captured through asymmetrically arranged multi-source sensor arrays, cooperative warning of threshold triggering and deep learning trend prediction is achieved in combination with a two-stage early warning mechanism, continuous monitoring during communication interruption is guaranteed by deploying edge computing nodes, and the risk sensing precision and the advanced early warning capacity of a deep foundation pit supporting system are improved.

Description

Technical Field

[0001] This invention relates to the field of safety monitoring technology for deep foundation pit engineering, specifically to an integrated foundation pit support monitoring and early warning system and method. Background Technology

[0002] With the acceleration of urbanization, deep foundation pit projects are developing towards ultra-deep and ultra-large scales, and the stability of their support structures directly affects the safety of surrounding buildings. As the core load-bearing component of deep foundation pits, the internal steel support system is prone to risks such as sudden changes in axial force and eccentric instability under dynamic earth pressure, construction loads, and environmental factors. Traditional monitoring methods are insufficient to achieve real-time and accurate perception of the stress state of the steel support and to provide early warning of risks, necessitating the development of intelligent monitoring and early warning technologies. Current technologies have shortcomings:

[0003] 1. Blind Spots in Multi-Dimensional Stress Monitoring of Steel Supports: Conventional sensor deployment methods are limited to single-point axial force measurement and cannot simultaneously capture the coupling effects of bending moment, temperature, and vibration. Key parameters such as stress gradient distribution in the anchorage zone at the end of the steel support and the mid-span region, as well as bending moment changes under eccentric loads, have monitoring blind spots, resulting in the inability to effectively identify hidden damage to the support system.

[0004] 2. Data distortion defects due to environmental interference: The existing system lacks a collaborative mechanism to suppress environmental interference. The non-uniform temperature field of the steel support caused by solar radiation leads to strain measurement drift. The soil pressure changes caused by precipitation infiltration and the noise from construction vibrations are superimposed, causing the monitoring data to deviate from the actual mechanical state and reducing the reliability of early warning.

[0005] 3. Failure of the Delayed Single-Level Early Warning Mechanism: The single-level alarm mode, which relies on fixed threshold triggers, cannot adapt to the dynamic risk evolution in deep foundation pit construction. Steel support instability often involves a gradual process such as axial force redistribution and bending moment exceeding limits. Traditional systems lack the ability to predict trends before the critical state, resulting in early warnings lagging behind the actual development of the danger.

[0006] 4. Emergency Response Deficiencies of Centralized Architecture: Centralized architectures based on cloud processing heavily rely on continuous communication. When a network outage occurs at the deep foundation pit site, the channels for uploading monitoring data and issuing commands become blocked, the system loses its real-time response capability, and cannot guarantee emergency handling of sudden risks to steel supports during communication failures. Summary of the Invention

[0007] The purpose of this invention is to provide an integrated foundation pit support monitoring and early warning system and method to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] The integrated foundation pit support monitoring and early warning system is composed of a multi-source sensor array, a distributed data processing unit and an intelligent early warning module; the multi-source sensor array is arranged in a multi-level nested structure at the stress concentration node area of the foundation pit retaining wall and the internal steel support system, the multi-source sensor includes a high-precision displacement sensor, a fiber strain sensor and a wideband vibration sensor, the multi-dimensional deformation of the retaining structure is continuously collected through the high-precision displacement sensor network, the pre-embedded fiber strain sensor group is used to synchronously obtain the axial pressure distribution gradient, the bending moment action range and the temperature drift parameter of the steel support, and the vibration frequency spectrum characteristics of the surrounding ground of the foundation pit are captured in real time based on the wideband vibration sensor array; the data processing unit implements space-time registration and feature fusion processing on the multi-source heterogeneous monitoring data stream, and constructs a dynamic safety margin evaluation model containing the axial force-bending moment coupling effect of the steel support; the intelligent early warning module generates multi-level risk early warning signals based on the output results of the dynamic safety margin evaluation model, and internally embeds a steel support instability risk special prediction mechanism.

[0010] Further, the sensor cluster in the multi-source sensor array covers multiple key stress nodes from the end anchorage area to the mid-span area of the steel support structure, the non-linear distribution characteristics of the axial pressure, the dynamic change amount of the eccentricity and the vibration response spectrum of the connection node are synchronously obtained through the staggered layout of the strain sensing units, wherein a temperature compensation module is embedded in each sensing node to eliminate the measurement deviation caused by environmental temperature difference, and the vibration monitoring frequency band covers a wide frequency domain range from the natural frequency of the structure to the construction disturbance frequency.

[0011] Further, the data processing unit executes a multi-physical field coupling analysis algorithm, eliminates the monitoring data drift effect caused by solar radiation through a steel support temperature-stress decoupling module, separates the environmental background vibration and the abnormal structure vibration components by using an adaptive frequency domain filtering technology, dynamically corrects the critical safety threshold interval under the combined action of the axial force and the bending moment, and establishes an overall stability evolution map of the support system.

[0012] Further, the intelligent early warning module includes a two-stage early warning architecture operating in parallel, a first-stage threshold early warning mechanism triggers a local sound-light alarm when a single-point monitoring parameter exceeds a preset safety tolerance interval, a second-stage trend early warning mechanism analyzes the multi-dimensional coupling relationship among the time sequence characteristics of the steel support axial force change rate, the load transmission difference rate of adjacent support points and the displacement acceleration value of the retaining structure through a deep learning model, predicts the instability evolution path and generates an early warning instruction in a time window.

[0013] Further, the input data of the deep learning model fuses a space-time correlation feature matrix of multiple sensors, including a steel support stress redistribution dynamic mode, a soil pressure gradient change trend caused by foundation pit dewatering and a construction machinery vibration energy transmission path, and outputs continuous prediction values of the instability probability distribution and the critical failure time interval of the steel support system through a convolutional recurrent neural network.

[0014] Furthermore, it also includes a 3D visualization platform, which maps real-time monitoring data to the foundation pit BIM model and dynamically renders the thermal map distribution of the steel support stress field. When an early warning is triggered, it automatically generates a topological diffusion vector animation of high-risk areas and overlays the matching degree analysis results of historical similar working conditions to form a visualized emergency decision map.

[0015] Furthermore, it also includes the construction of an edge computing node cluster architecture, which enables raw data preprocessing and real-time identification of abnormal patterns at the sensor network layer, autonomously executes early warning decision-making processes when communication is interrupted through a locally stored dynamic security margin assessment model, and initiates a multi-level disaster recovery mechanism to ensure monitoring continuity.

[0016] This invention provides another technical solution: an integrated foundation pit support monitoring and early warning method, based on an integrated foundation pit support monitoring and early warning system, comprising the following steps:

[0017] S1: Sensor network deployment and initialization: In the monitoring of deep foundation pit projects, tilt displacement sensor arrays are deployed every 10-15 meters on the inner side of the retaining wall. Simultaneously, pre-embedded fiber optic strain sensor groups are installed at the end anchor points, key cross sections and connection nodes of the steel support system inside the foundation pit in an asymmetric topology. At the same time, broadband vibration sensors are deployed on the ground surface around the foundation pit.

[0018] S2: Real-time acquisition and fusion of multi-source data: The data processing unit executes the following steps every 5 seconds:

[0019] S201: Spatiotemporal registration algorithm: Time synchronization: The asynchronous data streams of tilt displacement sensor, fiber optic strain sensor and broadband vibration sensor are uniformly aligned to the same timestamp reference through interpolation algorithm;

[0020] Spatial mapping: Based on the precise coordinates in the BIM model, the data of each measurement point is bound to the three-dimensional spatial position to realize the linkage and visualization of data and physical structure;

[0021] S202: Solving the mechanical state of steel supports, including:

[0022] Axial force calculation: For n compensated strain measurement points ε1 to ε2 of a single support. n The axial pressure is calculated based on the material mechanics formula: F = E × A × (∑ε n ) / n, where E is the elastic modulus of steel and A is the cross-sectional area of ​​the support;

[0023] Bending moment calculation: The bending stress is derived using the extreme difference of strain gauges arranged symmetrically: M = E × I × |ε max -ε minI, I is the cross-sectional moment of inertia, d is the distance between symmetric measuring points; the bending degree of the support is directly reflected by the strain gradient;

[0024] S203: Data fusion and early warning triggering: correlate the calculated axial force, bending moment, displacement, and vibration spectrum: if the axial force mutation exceeds the threshold or the bending moment distribution is abnormal, trigger the support instability early warning; when the vibration energy suddenly increases in the 80Hz frequency band, correlate the construction machinery action, distinguish abnormal vibration from normal operation, and superimpose all results in real time on the BIM model to generate a structure health degree thermal map and a historical trend curve;

[0025] S3: Dynamic safety margin assessment: build a steel support instability risk model:

[0026] Safety margin S = [F_critical - F_actual] / F_critical + a·(1 - |M / M_limit|), where F_critical is the critical axial force, dynamically calculated according to the support specification and installation angle; M_limit is the allowable bending moment, calculated from the material yield strength x sectional coefficient; a is the axial force-bending moment coupling weight coefficient, taking 0.2-0.5; when S<0.3, activate the early warning analysis;

[0027] S4: Deep learning trend early warning: input historical time series data into the convolutional recurrent network CRNN to build a feature matrix:

[0028] Time window Axial force change gradient ΔF / Δt Adjacent support load difference rate δ = (F1 - F2) / F1 Displacement acceleration a t-60 min 0.12 kN / s 0.08 0.05 mm / s² ... ... ... ...

[0029] Instability probability prediction: P_failure = Sigmoid(W1·Conv(ΔF) + W2·LSTM(δ) + W3·a), where W1-W3 are network weights, Sigmoid outputs a probability value of 0-1, and P_failure>0.7 triggers a secondary early warning;

[0030] S5: Visual emergency response: convert the calculation results of the steel support instability risk model into an operational engineering language through a three-dimensional visualization platform, real-time locate risks through a thermal map, predict chain reactions through soil mechanics simulation, output emergency construction methods through case matching, and form a closed-loop management and control system of "perception-prediction-decision".

[0031] Further, the temperature compensation algorithm of the optical fiber strain sensor in S1 is as follows: collect the initial ambient temperature value T0, and real-time correct the measurement value through the compensation formula ε_corrected = ε_raw - K·(T_current - T0), K takes the material adaptation coefficient of 0.8-1.2με / ℃;

[0032] The baseline calibration method of deploying broadband vibration sensors in S1 is as follows: record the background vibration spectrum V_base in the non-construction state, establish a background noise benchmark template, and provide a basis for comparison for subsequent vibration event identification.

[0033] Further, the processing steps of the three-dimensional visualization platform in S5 are as follows:

[0034] S501: BIM visualization and risk grading rendering, divided into:

[0035] Heat map dynamic mapping: real-time update of steel support color state according to safety margin S value;

[0036] Red early warning S<0.3: highlight the instability risk support, and superimpose a flashing warning frame;

[0037] Yellow alert 0.3≤S≤0.6: label stress concentration area and historical change curve;

[0038] Green safety S>0.6: semi-transparent display to highlight key attention objects;

[0039] Data linkage: when clicking on a high-risk support, the associated sensor real-time data panel automatically pops up;

[0040] S502: risk diffusion dynamic simulation, including:

[0041] Soil pressure conduction simulation: based on the improved Coulomb-Rankine theory to build a deformation transmission model:

[0042]

[0043] Where, k is the compression coefficient of passive zone soil, which is valued according to geological exploration data, β is the stiffness attenuation factor of the enclosure wall, P soil is the dynamic soil pressure;

[0044] Animation generation: simulate the deformation trend of the enclosure wall after the failure of the steel support in BIM, and predict the influence radius and the risk of adjacent building settlement;

[0045] S503: intelligent disposal scheme pushing, including:

[0046] Case library matching engine: compare the key parameters of the historical database, as shown in the following table:

[0047] Matching dimension Weight Example threshold Safety margin S change slope 35% S drop > 0.2 in 72 hours Vibration energy sudden change 25% Energy surge 3 times in 20-50 Hz band Soil type 20% Soft clay / sand alternating strata Excavation depth-support spacing 20% 20 m deep @ 4 m support spacing

[0048] Treatment scheme output: scheme A matching degree > 90%: immediately add φ609x16mm temporary steel support within 3m on both sides of risk support; scheme B matching degree 85%-90%: start sleeve valve pipe grouting reinforcement, and reduce the excavation rate by 50%; additional measures: generate a two-dimensional code associated with the construction briefing video and push it to the mobile terminal device on site.

[0049] Compared with the prior art, the beneficial effects of the present application are:

[0050] 1、The integrated foundation pit support monitoring and early warning system and method, through the asymmetric topology arrangement strategy of the multi-source sensor array, synchronously captures the axial stress distribution gradient, bending moment action range and temperature drift parameter of the deep foundation pit steel support system, combines the vibration sensor network perception construction disturbance frequency spectrum characteristics, realizes the multidimensional collaborative perception of the mechanical state of the supporting structure; secondly, based on the multi-physical field coupling analysis algorithm of the data processing unit, the temperature-stress decoupling mechanism is used to eliminate the measurement distortion caused by environmental temperature difference, and the adaptive frequency domain filtering is used to separate the abnormal vibration component of the structure, and the monitoring data reliability under complex working conditions is improved.

[0051] 2、The integrated foundation pit support monitoring and early warning system and method, relying on the double-stage triggering architecture of the intelligent early warning module, on the basis of real-time alarm of fixed threshold, introduces a deep learning driven instability trend prediction model, analyzes the space-time correlation characteristics of the steel support axial force change rate, load transfer difference rate and enclosure structure displacement acceleration, and identifies the progressive instability risk in advance; at the same time, the edge computing node cluster is constructed, and when the communication is interrupted, the abnormal diagnosis and early warning decision ability is maintained through the localization lightweight model, forming a whole cycle safety protection closed loop from accurate perception, anti-interference analysis, early warning to disaster recovery emergency. BRIEF DESCRIPTION OF DRAWINGS

[0052] Figure 1 It is the system framework diagram of the present application.

[0053] Figure 2 It is the steel support sensor arrangement detail drawing of the present application.

[0054] Figure 3 It is the double-stage early warning mechanism flow chart of the present application.

[0055] Figure 4 It is the edge computing disaster recovery flow chart of the present application. DETAILED DESCRIPTION

[0056] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by a person of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0057] Please refer to Figures 1-4 In the embodiments of the present application, an integrated foundation pit support monitoring and early warning system is provided, which is composed of a multi-source sensor array, a distributed data processing unit and an intelligent early warning module to form a cooperative work architecture. The multi-source sensor array is deployed in the stress concentration node area of the foundation pit retaining wall and the internal steel support system in a multi-level nested structure. The multi-source sensor includes a high-precision displacement sensor, a fiber strain sensor and a wideband vibration sensor. The multi-dimensional deformation of the retaining structure is continuously collected by the high-precision displacement sensor network. The pre-embedded fiber strain sensor group is used to synchronously obtain the axial pressure distribution gradient, the bending moment action range and the temperature drift parameters of the steel support. The wideband vibration sensor array is used to capture the vibration frequency spectrum characteristics of the surrounding ground of the foundation pit in real time. The data processing unit performs space-time registration and feature fusion processing on the multi-source heterogeneous monitoring data stream, and constructs a dynamic safety margin evaluation model containing the axial force-bending moment coupling effect of the steel support. The intelligent early warning module generates multi-level risk early warning signals based on the output results of the dynamic safety margin evaluation model, and internally embeds a steel support instability risk special prediction mechanism.

[0058] In the above embodiment, the sensor cluster in the multi-source sensor array covers multiple key stress nodes from the end anchorage area of the steel support structure to the midspan area. The strain sensing unit arranged in a staggered manner is used to synchronously obtain the nonlinear distribution characteristics of the axial pressure, the dynamic change amount of the eccentricity and the vibration response spectrum of the connecting node. A temperature compensation module is embedded in each sensing node to eliminate the measurement deviation caused by the environmental temperature difference. The vibration monitoring frequency band covers a wide frequency domain range from the natural frequency of the structure to the construction disturbance frequency.

[0059] In the above embodiment, the data processing unit performs a multi-physical field coupling analysis algorithm, eliminates the monitoring data drift effect caused by the solar radiation through a steel support temperature-stress decoupling module, separates the environmental background vibration and the abnormal structure vibration components by using an adaptive frequency domain filtering technology, dynamically corrects the critical safety threshold interval under the combined action of the axial force and the bending moment, and establishes an overall stability evolution map of the support system.

[0060] In the above embodiment, the intelligent early warning module comprises a two-stage early warning architecture operating in parallel, a first-stage threshold early warning mechanism triggers a local audible and visual alarm when a single-point monitored parameter exceeds a preset safety tolerance interval, and a second-stage trend early warning mechanism analyzes the multi-dimensional coupling relationship of the time series characteristics of the steel support axial force change rate, the adjacent support point load transmission difference rate, and the enclosure displacement acceleration value through a deep learning model, predicts the instability evolution path, and generates an advanced time window warning instruction. The input data of the deep learning model fuses the spatiotemporal correlation feature matrix, including the steel support stress redistribution dynamic mode, the soil pressure gradient change trend caused by foundation pit dewatering, and the construction machinery vibration energy transmission path, and outputs the continuous prediction values of the steel support system instability probability distribution and the critical failure time interval through a convolutional recurrent neural network.

[0061] The integrated foundation pit support monitoring and early warning system in the embodiment of the application further comprises a three-dimensional visualization platform for mapping real-time monitoring data to a foundation pit BIM model and dynamically rendering a steel support stress field thermodynamic diagram distribution, and automatically generating a high-risk area topology diffusion vector animation when an early warning is triggered, and superimposing a historical similar working condition matching degree analysis result to form a visual emergency decision graph.

[0062] In addition, the integrated foundation pit support monitoring and early warning system in the embodiment of the application further comprises a constructed edge computing node cluster architecture, which realizes raw data preprocessing and abnormal pattern real-time identification at the sensor network layer, autonomously executes an early warning decision process through a locally stored dynamic safety margin evaluation model in the event of communication interruption, and starts a multi-level disaster recovery mechanism to ensure monitoring continuity.

[0063] Embodiment one:

[0064] In order to further better explain the above system, the application further provides an integrated foundation pit support monitoring and early warning method, which adopts a conventional construction monitoring mode, is suitable for the deep foundation pit excavation stage, the support system bears dynamic soil pressure, and the steel support axial force mutation risk needs to be monitored in real time, and the specific steps are as follows:

[0065] S1: In the deep foundation pit engineering monitoring, an array of inclined displacement sensors is arranged every 10-15 meters on the inner side of the enclosure wall, and pre-buried optical fiber strain sensors are installed at the end anchorage points, key cross sections and connection nodes of the steel support system inside the foundation pit in an asymmetric topology structure (optimized weak point capturing capability) to ensure that each steel support covers at least 8 measuring points; at the same time, a wideband vibration sensor is deployed on the ground surface around the foundation pit, and the frequency response range covers 0.5-80Hz to capture the construction disturbance spectrum. After starting the self-calibration program:

[0066] 1. The temperature compensation algorithm of the fiber optic strain sensor is as follows: The initial ambient temperature value T0 is acquired, and the measured value is corrected in real time using the compensation formula ε_corrected = ε_raw - K·(T_current - T0) (K is a material adaptation coefficient of 0.8-1.2με / ℃) to eliminate temperature change interference;

[0067] 2. The baseline calibration method for deploying broadband vibration sensors is as follows: Record the background vibration spectrum V_base under no-construction conditions to establish a background noise benchmark template, providing a comparison basis for subsequent vibration event identification. This scheme achieves synchronous and accurate monitoring of structural response and ground vibration through multi-source sensor collaborative deployment and intelligent calibration mechanism, providing high-confidence data support for deep foundation pit safety control.

[0068] S2: Real-time Acquisition and Fusion of Multi-Source Data: To address the real-time processing requirements of deep foundation pit monitoring data streams, the data processing unit executes the following steps every 5 seconds:

[0069] S201: Spatiotemporal registration algorithm: Time synchronization: The asynchronous data streams of the tilt displacement sensor (sampling rate 1Hz), fiber optic strain sensor (10Hz), and broadband vibration sensor (100Hz) are uniformly aligned to the same timestamp reference (such as UTC millisecond level) through an interpolation algorithm to eliminate timing misalignment;

[0070] Spatial mapping: Based on the precise coordinates in the BIM model, the data of each measuring point is bound to the three-dimensional spatial location (such as the positioning of measuring points of the retaining wall and the coordinates of support nodes), so as to realize the linkage and visualization of data and physical structure;

[0071] S202: Solving the mechanical state of steel supports, including:

[0072] Axial force calculation: For n compensated strain measurement points ε1 to ε2 of a single support. n The axial pressure is calculated based on the material mechanics formula: F = E × A × (∑ε n ) / n, where E is the elastic modulus of steel and A is the cross-sectional area of ​​the support;

[0073] Bending moment calculation: The bending stress is derived using the extreme difference of strain gauges arranged symmetrically: M = E × I × |ε max -ε min | / d, where I is the moment of inertia of the cross section and d is the spacing between symmetrical measuring points; the strain gradient directly reflects the bending degree of the support.

[0074] S203: Data fusion and early warning trigger: Correlation analysis of calculated axial force, bending moment, displacement and vibration spectrum: If the axial force mutation exceeds the threshold value, or the bending moment distribution is abnormal (such as asymmetric load), the support instability early warning is triggered; when the vibration energy suddenly increases in the 80Hz frequency band, the construction machinery action (such as piling) is correlated to distinguish abnormal vibration from normal operation. All results are superimposed in real time on the BIM model to generate a structure health degree heat map and a historical trend curve.

[0075] The above S2 process completes multi-source data fusion and mechanical interpretation within 5 seconds, provides quantitative basis for dynamic adjustment of foundation pit support parameters (such as supplemental support, pressure reduction grouting), and synchronously pushes to the engineering decision end through the visual interface to realize "monitoring-analysis-intervention" closed-loop management.

[0076] S3: Dynamic safety margin evaluation: Construct a steel support instability risk model:

[0077] Safety margin S = [F_critical - F_actual] / F_critical + a · (1 - |M / M_limit|), where F_critical is the critical axial force, which is dynamically calculated according to the support specification and installation angle; M_limit is the allowable bending moment, which is calculated by the material yield strength x sectional coefficient; a is the axial force-bending moment coupling weight coefficient, taking 0.2-0.5; when S < 0.3, the early warning analysis is activated;

[0078] S4: Deep learning trend early warning: By inputting historical time series data into the convolution recurrent network CRNN, a feature matrix is constructed:

[0079] Time window Axial force change gradient ΔF / Δt Adjacent support load difference rate δ = (F1 - F2) / F1 Displacement acceleration a t-60 min 0.12 kN / s 0.08 0.05 mm / s² ... ... ... ...

[0080] Instability probability prediction: P_failure = Sigmoid(W1·Conv(ΔF) + W2·LSTM(δ) + W3·a), where W1-W3 are network weights, Sigmoid outputs probability value 0-1, P_failure > 0.7 triggers secondary early warning;

[0081] S5: Visual emergency response: In the three-dimensional visualization platform, the system based on the calculation results of the steel support instability risk model, real-time driving the following automatic response process:

[0082] S501: BIM visualization and risk grading rendering, divided into:

[0083] Heat map dynamic mapping: Real-time update of steel support color state according to safety margin S value;

[0084] Red early warning S < 0.3: Highlight the instability risk support and superimpose the flashing warning frame;

[0085] Yellow alert 0.3≤S≤0.6: Mark stress concentration area and historical change curve;

[0086] Green safety S>0.6: Semi-transparent display to highlight key objects of attention;

[0087] Data linkage: When clicking on high-risk support, automatically pop up the real-time data panel of associated sensors;

[0088] S502: Risk diffusion dynamic simulation, including:

[0089] Soil pressure conduction simulation: Based on the improved Coulomb-Rankine theory, a deformation transmission model is constructed:

[0090]

[0091] Where k is the compression coefficient of the passive zone soil, which is assigned according to geological exploration data, β is the stiffness attenuation factor of the enclosure wall, P soil is the dynamic soil pressure;

[0092] Animation generation: Simulate the deformation trend of the enclosure wall after the failure of the steel support in BIM (such as "kicking" deformation or waist beam fracture process), predict the influence radius and the risk of adjacent building settlement;

[0093] S503: Intelligent disposal scheme pushing, including:

[0094] Case library matching engine: Compare the key parameters of the historical database (≥1000 foundation pit cases), as shown in the following table:

[0095] Matching dimension Weight Example threshold Safety margin S change slope 35% S drop > 0.2 in 72 hours Vibration energy sudden change 25% Energy surge 3 times in 20-50 Hz band Soil type 20% Soft clay / sand alternating strata Excavation depth-support spacing 20% 20 m deep @ 4 m support spacing

[0096] Disposal scheme output: Scheme A matching degree >90%: Immediately add φ609x16mm temporary steel support within 3m on both sides of the risk support; Scheme B matching degree 85%-90%: Start sleeve valve pipe grouting reinforcement, and reduce the excavation rate by 50%; Additional measures: Generate a two-dimensional code associated with the construction briefing video and push it to the mobile terminal device on site.

[0097] The above three-dimensional visualization platform converts abstract data into operational engineering language, locates risks in real time through heat maps, predicts chain reactions through soil mechanics simulation, and outputs emergency construction methods through case matching, forming a "perception-prediction-decision" closed-loop management and control system. When yellow and above warning is triggered, the system records the disposal response time (from alarm to scheme signing ≤8 minutes) simultaneously, providing a full-cycle digital archive for subsequent accident liability tracing and process optimization.

[0098] Example two:

[0099] The application also provides an integrated foundation pit support monitoring and early warning method, which adopts a rainstorm working condition monitoring mode, is suitable for a scenario that deep heavy rain causes groundwater level to rise and steel supports bear asymmetric loads, and specific steps are as follows.

[0100] Step 1: Active suppression of environmental interference

[0101] 1. Water level-soil pressure coupling monitoring: The groundwater level sensor transmits the buried depth H_w in real time.

[0102] Calculate the equivalent additional soil pressure ΔP: ΔP = γ_w·(H_w-H_base) + K_a·γ_sat·ΔH (γ_w: water specific gravity, γ_sat: saturated soil specific gravity, K_a: active soil pressure coefficient;

[0103] 2. Rain noise filtering processing: Wavelet packet decomposition is performed on the vibration signal: V_clean = V_raw - Σ(DetailCoeff_i) | i∈[4,6] high-frequency components.

[0104] Step 2: Temperature-stress full decoupling analysis: In the sunshine temperature difference strain compensation of the deep foundation pit steel support, the system realizes accurate decoupling of the mechanical response through the following steps:

[0105] Each optical fiber strain sensor node synchronously collects real-time temperature data Tx, combines the daily average temperature mean as a reference datum Tref, and calculates the temperature change ΔT. Then, double decoupling calculation is performed: first, the thermal output interference generated by the temperature drift of the sensor itself is deducted (modified by the material thermal output coefficient β calibrated in the laboratory), and then the physical deformation effect of the steel due to thermal expansion and cold contraction is eliminated (based on the inherent linear expansion characteristics of the steel for reverse calibration). This algorithm separates the thermal interference component in the original strain measurement value from the real deformation of the structure, ensures that the axial force and bending moment calculation results are not affected by the day and night temperature difference or one-sided sunshine shadow, and especially solves the false strain deviation of the east-west support caused by the temperature difference of more than 20℃ between the sunny surface and the shady surface in the morning in summer (such error has caused an axial force error of 12%). The corrected real strain data are fed back to the safety margin model in real time to prevent temperature interference from triggering the early warning system, and a daily temperature-strain decoupling report is generated, which marks the maximum correction value point and the correction amplitude (the correction amount can reach 150-300με in a typical scenario), and provides a high-confidence data basis for long-term service performance evaluation of the support system.

[0106] Step 3: Asymmetric load early warning: In response to the potential risk of eccentric loading of the steel support system (such as unilateral stacking or loose connection nodes), the system performs the following diagnostic and response mechanism: First, calculate the imbalance degree U of the bending moment of the nodes on both sides of the support in real time, defined as the absolute deviation proportion of the difference between the left and right bending moments and the average value of the design allowable bending moment. When the U value is continuously out of limit (> 0.4) for 10 minutes, automatically trace back the installation parameter library of the support: obtain the measured value of the installation inclination θ (accuracy ± 0.5°) and the thickness of the connecting plate (allowable tolerance check), and check the actual bending resistance based on the yield strength and cross-sectional characteristics of the steel. If the analysis finds that the current bending moment has exceeded 80% of the bending resistance of the support, a first-level structure alarm is triggered - at this time, the system simultaneously performs three actions:

[0107] 1. Accident simulation: dynamically display the eccentric fracture path of the support in the BIM model (the typical failure mode is the shear deformation of the connecting plate bolt → flange buckling);

[0108] 2. Emergency intervention: automatically lock the 50-ton hydraulic jacks on both sides of the support, and push the real-time compensation force value (initial compensation force = F_actual×sinθ);

[0109] 3. Causal tracing: correlate the construction log on the same day, and screen for inducing factors (such as eccentric overloading of the earthwork, uneven settlement leading to differences in soil hardness).

[0110] This diagnostic process deeply couples the theoretical mechanics model (reduced inclination effective bending resistance) with construction big data (node plate thickness deviation > 2mm case library), successfully warning a support instability accident in a subway deep foundation pit in Shanghai caused by the eccentric load of the earthwork ramp (connection plate tearing occurred 24 minutes after the monitoring showed that the eccentricity was out of limit).

[0111] Example Three:

[0112] In order to further better explain the above system, the application also provides an integrated foundation pit support monitoring and early warning method, which adopts a communication interruption edge computing mode, which is suitable for the scene of self-maintaining monitoring of edge nodes in the case of network failure on site, and the specific steps are as follows:

[0113] Step 1: Localized data preprocessing: in the edge computing node of the deep foundation pit, the system realizes real-time diagnosis and risk pre-identification of high-frequency data through the following process:

[0114] 1. Data intelligence dimension reduction and feature extraction: For the original strain data stream of each steel support (10Hz sampling rate), a sliding time window (window length 30 seconds, step length 5 seconds) is used to perform fast Fourier transform (FFT) to extract the amplitude features (A1-A5) of the top 5 dominant harmonic components and the fundamental frequency value f_base, generating a 6-dimensional feature vector [A1, A2, A3, A4, A5, f_base]. Dimension reduction value: only 200 bytes need to be uploaded per 10 seconds (originally 5KB), meeting the narrowband Internet of Things transmission requirements; while retaining key dynamic characteristics (such as the downward shift of the fundamental frequency indicating stiffness attenuation);

[0115] 2. Dynamic outlier detection and pattern recognition: Mahalanobis distance model: based on historical normal operating data (≥500 hours) in the training phase, the feature vector mean μ and covariance matrix Σ are established; the normalized Mahalanobis distance value D_M of the current feature vector X is calculated in real time (reflecting its deviation in the statistical space).

[0116] Abnormal criterion: when D_M>3.0 (equivalent to 99.7% confidence interval), the following actions are triggered:

[0117] Hierarchical diagnosis: if A1 increases by 200% and f_base decreases by >15%, it is determined that the connection node is loose (bolt pre-tightening force is lost);

[0118] Temperature difference compensation check: automatically associate sensor temperature to exclude false strain spectrum distortion caused by solar gradient temperature difference.

[0119] 3. Edge real-time response mechanism: when marked as abnormal, the node performs within 500ms:

[0120] Pre-disposal instruction: automatically increase the sampling rate of the associated measuring point to 100Hz (for 1 minute), capture transient impact signals;

[0121] Risk map update: push the abnormal support position and frequency spectrum features to the three-dimensional platform, triggering BIM model red flash alarm;

[0122] Cross-domain verification: compare the displacement increment of the surrounding retaining wall at the same horizontal position (confirm risk transmission if >3mm / 10min);

[0123] Model self-evolution: update the covariance matrix Σ based on new data every 24 hours, adapting to the changes in mechanical properties at different stages of foundation pit excavation (such as the drift of the fundamental frequency caused by the increase of support axial force).

[0124] The edge computing architecture achieved a key breakthrough in a 28m deep foundation pit project in Tianjin: when the west side support developed micro-cracks due to construction vehicle rolling (initial crack <0.1mm), the system issued an early warning through a 120% increase in A5 amplitude and a Mahalanobis distance D_M=5.2 (47 hours earlier than traditional manual inspection to identify the problem), avoiding support fracture accidents after jack axial force compensation, with potential losses recovered by a single early warning exceeding 8 million yuan. The edge node uses a feature space model with continuous learning to make implicit risks explicit, providing a millisecond-level safety barrier for super-deep foundation pit construction.

[0125] Step 2: Lightweight model prediction: a simplified LSTM network deployed on the edge node:

[0126] Input sequence: feature vector group [V1, V2,..., V 12 ] (1 hour long) every 5 minutes.

[0127] Prediction logic:

[0128] 1. Memory cell update: C_t = f_t ⊙ C_{t-1} + i_t ⊙ g_t (f_t: forget gate output, i_t: input gate, g_t: candidate state).

[0129] 2. Output prediction value: S_next = σ(W_o · [h_{t-1}, V_t] + b_o) (σ: Sigmoid activation function, W_o / b_o: output layer parameters), if S_next <0.25, activate local audible and visual alarms.

[0130] Step 3: Disaster recovery data synchronization: after network recovery, the edge computing node performs the following key operations to ensure monitoring continuity and data integrity:

[0131] 1. Intelligent data compression transmission: selective backtracking mechanism: only upload complete time period raw data (including strain raw waveform, temperature time series, and full frequency spectrum) marked as abnormal (e.g. Mahalanobis distance D_M>3.0) during network interruption. For periods without triggering alarms, retain dimension-reduced feature vectors [A1-A5, f_base] and timestamp summaries, reducing bandwidth occupancy by 98%.

[0132] Resume transmission optimization: check cloud reception status in time windows to avoid repeated transmission (e.g. automatically skip data from 2023-08-05T14:00:00 to 14:03:00 that has been confirmed received), and ensure transmission data integrity through HMAC-SHA256 signature.

[0133] 2. Cloud model dynamic correction:

[0134] Edge weight delta update: Download the latest version of the covariance matrix Σ (integrate the recent 100 similar foundation pit data) and the feature space mean μ trained by the global big data in the cloud, replace the old local parameters, and dynamically adapt the out-of-control detection threshold to the seasonal softening effect of the soil layer (such as the decrease of the clay internal friction angle in the rainy season, resulting in the shift of the fundamental frequency).

[0135] Emergency patch pushing: If the cloud analysis shows that a certain type of abnormal pattern (such as A5 sudden increase accompanied by f_base < 5Hz) is strongly related to support fracture, remotely issue diagnostic rules to edge nodes to enhance local prediction sensitivity (D_M threshold from 3.0 to 2.5).

[0136] 3. Autonomous operation period traceability report: Data inversion verification: Compare local alarms during network interruption and cloud diagnosis conclusions after recovery (such as edge warning of support node loosening, verified by cloud finite element simulation with a consistency of >92%), generate false / missed alarm analysis report, and optimize edge model confidence.

[0137] This network interruption processing mechanism plays a key role in the deep foundation pit project in the core area of Beijing CBD: In July 2023, heavy rainfall caused communication interruption for 17 hours, and the edge node successfully captured the abnormal shift of the fundamental frequency of the east support system (f_base from 12.3Hz to 9.1Hz), and after the network was restored, the data uploaded by the cloud was verified and confirmed as waist beam embedded corrosion (field excavation verified corrosion depth >2mm), and grouting reinforcement was started 28 hours in advance. The system through the "local decision-making-cloud verification-model evolution" closed loop, still guarantees monitoring zero interruption, early warning zero false alarm, provides a resilient perception paradigm for smart construction site in harsh working conditions.

[0138] The above describes only the preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, any skilled person in the art can make equivalent replacement or change according to the technical solution and the inventive concept of the present application within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. An integrated foundation pit support monitoring and early warning system, characterized in that, A collaborative working architecture is formed by a multi-source sensor array, a distributed data processing unit and an intelligent early warning module; the multi-source sensor array is arranged in a stress concentration node area of a foundation pit retaining wall and an internal steel support system in a multi-level nested structure, and includes high-precision displacement sensors, fiber strain sensors and wide-band vibration sensors; the multi-dimensional deformation of the retaining structure is continuously collected through a high-precision displacement sensor network, the gradient of the axial pressure distribution, the bending moment action range and the temperature drift parameters of the steel support are synchronously obtained by using the embedded fiber strain sensor group, and the vibration frequency spectrum characteristics of the surrounding ground of the foundation pit are captured in real time based on the wide-band vibration sensor array; the data processing unit implements time and space registration and feature fusion processing on the multi-source heterogeneous monitoring data stream, and constructs a dynamic safety margin evaluation model containing the coupling effect of the axial force and bending moment of the steel support; the intelligent early warning module generates multi-level risk warning signals based on the output results of the dynamic safety margin evaluation model, and internally embeds a steel support instability risk special prediction mechanism.

2. The integrated foundation pit support monitoring and early warning system of claim 1, wherein: The sensor cluster in the multi-source sensor array covers multiple key stress nodes from the end anchorage area to the mid-span area of the steel support structure, and synchronously obtains the nonlinear distribution characteristics of the axial pressure, the dynamic change amount of the eccentricity and the vibration response spectrum of the connecting node through the staggered layout of the strain sensing units, wherein a temperature compensation module is embedded in each sensing node to eliminate the measurement deviation caused by environmental temperature difference, and the vibration monitoring frequency band covers a wide frequency range from the natural frequency of the structure to the construction disturbance frequency.

3. The integrated foundation pit support monitoring and early warning system of claim 1, wherein: The data processing unit executes a multi-physical field coupling analysis algorithm, eliminates the monitoring data drift effect caused by solar radiation through a steel support temperature-stress decoupling module, separates the environmental background vibration and the abnormal structure vibration components by using an adaptive frequency domain filtering technology, dynamically corrects the critical safety threshold interval under the combined action of the axial force and the bending moment, and establishes an overall stability evolution map of the support system.

4. The integrated excavation bracing monitoring and early warning system and method of claim 1, wherein: The intelligent early warning module includes a two-stage early warning architecture operating in parallel, a first-stage threshold early warning mechanism triggers a local sound-light alarm when a single-point monitoring parameter exceeds a preset safety tolerance interval, and a second-stage trend early warning mechanism analyzes the multi-dimensional coupling relationship among the time sequence characteristics of the steel support axial force change rate, the load transmission difference rate of adjacent support points and the displacement acceleration value of the retaining structure by using a deep learning model, predicts the instability evolution path and generates an advanced time window warning instruction.

5. The integrated foundation pit support monitoring and early warning system of claim 4, wherein: The input data of the deep learning model fuse the time and space correlation feature matrix of multiple sensors, including the dynamic mode of steel support stress redistribution, the change trend of soil pressure gradient caused by foundation pit dewatering and the vibration energy transmission path of construction machinery, and the continuous prediction values of the instability probability distribution and the critical failure time interval of the steel support system are output by a convolutional recurrent neural network.

6. The integrated excavation bracing monitoring and warning system of claim 1, wherein: A three-dimensional visualization platform is further included for mapping real-time monitoring data to a foundation pit BIM model and dynamically rendering a steel support stress field heat map distribution, generating a high-risk area topology diffusion vector animation automatically when an early warning is triggered, and forming a visual emergency decision map by superimposing the analysis results of historical similar working conditions.

7. The integrated excavation bracing monitoring and warning system of claim 1, wherein: Also included is the constructed edge computing node cluster architecture, which realizes raw data preprocessing and abnormal pattern real-time identification at the sensor network layer, autonomously executes early warning decision-making process and starts multi-level disaster recovery mechanism to ensure monitoring continuity when communication is interrupted through the locally stored dynamic safety margin evaluation model.

8. An integrated foundation pit support monitoring and early warning method, based on the integrated foundation pit support monitoring and early warning system of claim 6, characterized in that: The method comprises the following steps: S1: sensor network deployment and initialization: in deep foundation pit engineering monitoring, an array of inclination displacement sensors is arranged every 10-15 meters on the inner side of the retaining wall, and pre-buried optical fiber strain sensor groups are installed in an asymmetric topological structure at the end anchor points, key cross sections and connecting nodes of the steel support system inside the foundation pit, and wideband vibration sensors are deployed on the ground around the foundation pit; S2: real-time acquisition and fusion of multi-source data: the following steps are executed by the data processing unit every 5 seconds: S201: time and space registration algorithm: time synchronization: synchronize the asynchronous data streams of the inclination displacement sensor, the optical fiber strain sensor and the wideband vibration sensor to the same timestamp reference through the interpolation algorithm; Space mapping: based on the accurate coordinates in the BIM model, the data of each measuring point is bound to the three-dimensional space position to realize the linkage visualization of data and physical structure; S202: steel support mechanical state solution, including: Axial force calculation: for single support n compensation after strain measuring point ε1 to ε n , based on the formula of material mechanics to calculate the axial pressure: F = E × A × (∑ε n ) / n, E is the elastic modulus of steel, A is the support cross-sectional area; Bending moment calculation: Bending stress is derived by using the extreme difference of symmetrically arranged strain measuring points: M = E x I x |ε max -ε min | / d, I is the section moment of inertia, and d is the distance between symmetric measuring points; the bending degree of the support is directly reflected by the strain gradient; S203: data fusion and early warning triggering: correlation analysis is performed on the solved axial force, bending moment, displacement and vibration spectrum: if the axial force mutation exceeds the threshold value or the bending moment distribution is abnormal, the support instability early warning is triggered; when the vibration energy suddenly increases in the 80Hz frequency band, the construction machinery action is associated to distinguish abnormal vibration from normal operation, and all results are superimposed on the BIM model in real time to generate a structure health degree thermal map and a historical trend curve; S3: dynamic safety margin evaluation: construct a steel support instability risk model: Safety margin S = [F_critical-F_actual] / F_critical+α·(1- |M / M_limit|), wherein F_critical is the critical axial force, which is dynamically calculated according to the support specification and installation angle; M_limit is the allowable bending moment, which is calculated by the material yield strength x sectional coefficient; α is the axial force-bending moment coupling weight coefficient, which is taken as 0.2-0.5; when S<0.3, the early warning analysis is activated; S4: deep learning trend early warning: by inputting historical time series data into the convolutional recurrent network CRNN, a feature matrix is constructed: Instability probability prediction: P_failure = Sigmoid( W1·Conv(ΔF) + W2·LSTM(δ) + W3·a ), wherein W1-W3 are network weights, Sigmoid outputs probability value 0-1, and P_failure>0.7 triggers the second level early warning; S5: visual emergency response: the calculation results of the steel support instability risk model are converted into an operable engineering language through a three-dimensional visualization platform, the risk is located in real time through a thermal map, the chain reaction is predicted through soil mechanics simulation, and the emergency construction method is output through case matching, forming a closed-loop management and control system of "perception-prediction-decision".

9. The integrated foundation pit support monitoring and early warning method of claim 8, wherein: The temperature compensation algorithm of the fiber optic strain sensor in S1 is as follows: collect the initial ambient temperature value T0, and real-time correct the measurement value through the compensation formula ε_corrected =ε_raw - K·(T_current - T0), K takes the material adaptation coefficient of 0.8-1.2 με / ℃; The baseline calibration method of the broadband vibration sensor deployed in S1 is as follows: record the background vibration spectrum V_base in the non-construction state, establish a background noise benchmark template, and provide a basis for comparison for subsequent vibration event identification.

10. The integrated foundation pit support monitoring and early warning method of claim 8, wherein: The processing steps of the three-dimensional visualization platform in S5 are as follows: S501: BIM visualization and risk grading rendering, divided into: Thermal map dynamic mapping: real-time update steel support color state according to safety margin S value; Red early warning S<0.3: highlight unstable risk support and superimpose flashing warning frame; Yellow warning 0.3≤S≤0.6: mark stress concentration area and historical change curve; Green safety S>0.6: semi-transparent display to highlight key attention objects; Data linkage: when clicking on high-risk support, automatically pop up the associated sensor real-time data panel; S502: risk diffusion dynamic simulation, including: Soil pressure conduction simulation: based on the improved Coulomb-Rankine theory to build a deformation transmission model:

11. wherein, k is the passive zone soil compression coefficient, which is valued according to geological exploration data, β is the containment wall stiffness decay factor, P soil is the dynamic soil pressure; Animation generation: simulate the deformation trend of the enclosing wall after the failure of the steel support in BIM, and predict the influence radius and the risk of adjacent building settlement; S503: intelligent disposal scheme pushing, including: Case library matching engine: compare the key parameters of the historical database, as shown in the following table: Disposal scheme output: scheme A matching degree>90%: immediately add φ609×16mm temporary steel support within 3m on both sides of the risk support; scheme B matching degree 85%-90%: start sleeve valve pipe grouting reinforcement, and reduce the excavation rate by 50%; additional measures: generate a two-dimensional code associated with the construction briefing video and push it to the mobile terminal device on site.

Images