Perception monitoring system and method for construction of underground diaphragm wall of adjacent sensitive building

Abstract

The invention discloses a sensing and monitoring system for construction of an underground diaphragm wall of an adjacent sensitive building. The sensing and monitoring system comprises a data acquisition unit, an edge processing unit, a prediction and evaluation unit and a linkage control unit, wherein the data acquisition unit is composed of distributed optical fiber strain monitoring, three-dimensional laser radar monitoring and acoustic emission monitoring and is used for respectively acquiring strain data, three-dimensional point cloud data of a groove wall and an acoustic emission signal; the edge processing unit performs time synchronization, space registration, filtering noise reduction and feature extraction on the multi-source data to form fused feature data and output the position and range of an abnormal region; the prediction evaluation unit performs short-term prediction of sensitive building deformation based on the fused feature data, and outputs a risk evaluation result in combination with a finite element mechanical analysis model; and the linkage control unit outputs a deviation correction control instruction, a grouting control instruction and a support triggering instruction according to a risk assessment result, so that sensing-prediction-regulation closed-loop treatment is formed.

Description

Technical Field

[0001] This invention relates to the field of underground engineering monitoring technology, and in particular to a sensing and monitoring system and method for the construction of underground continuous walls near sensitive buildings. Background Technology

[0002] Diaphragm walls are commonly used in projects such as subway stations, deep foundation pit support, and underground structure waterproofing. In construction scenarios near existing factories, historical buildings, important pipelines, and other sensitive structures, the trenching and slurry wall construction processes can cause stress release and redistribution in the soil, easily inducing settlement, tilting, and differential deformation of the surface and surrounding buildings. At the same time, risks such as local instability of the trench wall, hole collapse, and over-excavation can develop rapidly in a short period of time, posing high requirements for project safety and the protection of surrounding buildings.

[0003] Current on-site risk control methods typically rely on mud wall protection, segmented trenching, process control, and manual inspections combined with point-based monitoring. Monitoring primarily employs discrete point monitoring methods such as leveling points, inclinometers, and vibrating wire sensors, coupled with fixed threshold alarms. However, these methods have limitations in spatial coverage and response speed, making it difficult to continuously capture early signals such as continuous spatial deformation, trench wall geometric anomalies, and micro-fractures during construction, easily leading to monitoring blind spots and response delays. Furthermore, static thresholds and manual interpretation carry a high risk of false alarms and missed alarms in dynamic construction environments, and post-warning response measures often depend on human experience, hindering the formation of rapid, standardized, and coordinated responses.

[0004] Furthermore, in the construction of underground continuous walls near sensitive buildings, risk evolution often manifests as the coupling of multiple types of information: for example, signals such as strain changes, trench wall morphology changes, and acoustic emission activities are heterogeneous in time and space (different sampling frequencies, coordinate references, and data formats). Without alignment and correlation analysis under a unified time reference and a unified construction coordinate system, even if strain monitoring, three-dimensional morphology monitoring, and acoustic emission monitoring are deployed separately, it is difficult to locate anomalies in specific areas, let alone support the subsequent corrective, grouting, and support actions to form a closed-loop linkage. Summary of the Invention

[0005] This invention aims to at least partially solve one of the technical problems in related technologies. To this end, one objective of this invention is to propose a sensing and monitoring system and method for the construction of underground continuous walls near sensitive buildings. This system integrates distributed fiber optic sensing, lidar scanning, and acoustic emission monitoring technologies to construct a multi-dimensional sensing network, enabling precise location of abnormal areas and forming a proactive protection mechanism encompassing sensing, prediction, and control throughout the entire process.

[0006] In a first aspect, the present invention proposes a construction sensing and monitoring system for underground continuous walls near sensitive buildings, comprising:

[0007] Trenching equipment is used to excavate trenches to form underground continuous walls;

[0008] Grouting equipment is used to inject grout into the grouting reinforcement area to reinforce the soil or fill voids;

[0009] Support actuators are used to carry out support operations or adjust support parameters.

[0010] It also includes: a data acquisition unit, an edge processing unit, a prediction and evaluation unit, and a linkage control unit;

[0011] The data acquisition unit includes a distributed fiber optic strain monitoring subunit, a three-dimensional lidar monitoring subunit, and an acoustic emission monitoring subunit, which respectively output strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals;

[0012] The edge processing unit is connected to the data acquisition unit and is used to perform time synchronization, spatial registration, filtering and noise reduction and feature extraction on strain data, three-dimensional point cloud data of the tank wall and acoustic emission signals to obtain fused feature data, and output the abnormal area results based on the fused feature data.

[0013] The prediction and assessment unit is connected to the edge processing unit to output the deformation prediction results of sensitive buildings based on the fused feature data, and to output the risk assessment results based on the finite element mechanical analysis model.

[0014] The linkage control unit is connected to the prediction and evaluation unit, and is also connected to the trenching equipment, grouting equipment, and support actuator. It is used to output correction control commands, grouting control commands, and support trigger commands based on the risk assessment results.

[0015] Preferably:

[0016] The distributed optical fiber strain monitoring subunit includes multiple sets of BOTDR distributed optical fibers. These multiple sets of BOTDR distributed optical fibers are laid along the top of the guide wall and cover the side closest to the sensitive building, and are used to output continuous strain data along the laying direction.

[0017] The 3D lidar monitoring subunit is mounted on the mounting bracket of the trenching equipment and faces the trench wall, and is used to output the 3D point cloud data of the trench wall;

[0018] The acoustic emission monitoring subunit includes multiple acoustic emission sensors, which are deployed around the foundation of sensitive buildings to output acoustic emission signals.

[0019] Preferably, the edge processing unit includes a time synchronization module, a spatial registration module, a filtering and noise reduction module, a feature extraction module, and an anomaly region determination module;

[0020] The time synchronization module generates a unified time reference and adds a unified timestamp to the multi-source data formed by strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals to complete time synchronization;

[0021] The spatial registration module maps multi-source data to a unified construction coordinate system to complete spatial registration;

[0022] The filtering and noise reduction module performs filtering and noise reduction on the multi-source data after spatial registration;

[0023] The feature extraction module extracts strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the filtered and denoised multi-source data and forms fused feature data.

[0024] The anomaly region determination module outputs the spatial location and spatial range of the anomaly region based on the fused feature data;

[0025] The linkage control unit determines the early warning level based on the risk assessment results, and outputs grouting control commands to the grouting equipment when the early warning level meets the grouting triggering conditions, outputs correction control commands to the trenching equipment when the early warning level meets the correction triggering conditions, and outputs support triggering commands to the support actuator when the early warning level meets the support triggering conditions.

[0026] Secondly, the present invention proposes a method for sensing and monitoring the construction of underground diaphragm walls near sensitive buildings, which includes any of the above-mentioned schemes of the sensing and monitoring system for underground diaphragm walls near sensitive buildings. The method steps are as follows:

[0027] S1. Data Acquisition: During the construction of the diaphragm wall, distributed optical fiber strain data, trench wall three-dimensional point cloud data, and acoustic emission signals are collected.

[0028] S2. Time Synchronization: Generate a unified time reference and add a unified timestamp to strain data, 3D point cloud data of the tank wall, and acoustic emission signals to complete time synchronization;

[0029] S3. Spatial Registration: Mapping strain data, 3D point cloud data of the trench wall, and acoustic emission signals to a unified construction coordinate system to complete spatial registration;

[0030] S4. Filtering and noise reduction: Filter and reduce noise on the registered data to obtain denoised data;

[0031] S5. Feature Extraction and Fusion: Extract strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the noise reduction data and form fused feature data;

[0032] S6. Abnormal Region Determination: Determine abnormal regions based on fused feature data, and output the spatial location and spatial range of the abnormal regions;

[0033] S7. Prediction, assessment and linkage control: Based on the fusion of feature data, output the deformation prediction results of sensitive buildings; based on the finite element mechanical analysis model, output the risk assessment results; and based on the risk assessment results, output the correction control command, grouting control command and support trigger command.

[0034] Preferably, in step S2, the unified time reference output by the time synchronization module is used as the reference to resample the multi-source data and complete the alignment of the same sampling period to form time synchronization data.

[0035] Preferably, in step S3, a unified construction coordinate system is established based on the construction measurement benchmark. The positioning results of the acoustic emission signal are mapped to the coordinate system where the three-dimensional point cloud data of the trench wall is located using calibration parameters. The distributed optical fiber strain data is also mapped to the same coordinate system to form spatial registration data.

[0036] Preferably, in step S6:

[0037] Calculate the verticality deviation of the tank wall based on the 3D point cloud data of the tank wall and determine the area exceeding the limit;

[0038] Calculate the strain gradient and determine the gradient over-limit region based on distributed optical fiber strain data;

[0039] Calculate the acoustic emission energy based on the acoustic emission signal and determine the energy excess region;

[0040] By spatially superimposing the oversized areas under a unified construction coordinate system, the spatial location and spatial range of the abnormal areas can be obtained.

[0041] Preferably, in step S7:

[0042] A time-series prediction model was used to make short-term predictions of the deformation of sensitive buildings, and the deformation prediction results were obtained.

[0043] The deformation prediction results are used as the boundary condition input for the finite element mechanical analysis model to obtain the stress response results of the structure and soil.

[0044] The risk assessment results are determined based on the stress response results and the abnormal area results, and the risk assessment results are output to the linkage control unit.

[0045] Preferably, the risk assessment results are expressed through risk indicators. Quantification is performed, and the risk indicator R satisfies:

[0046]

[0047] in, These are the strain gradient weighting coefficients. This is the verticality deviation weighting coefficient. For acoustic emission energy weighting coefficients, These are the characteristic values ​​of the strain gradient within the anomalous region. This represents the characteristic value of the verticality deviation of the tank wall within the abnormal area. The characteristic value of acoustic emission energy in the anomalous region, based on The warning level is determined by comparing the result with the preset threshold.

[0048] Preferably, the correction control command in step S7 is:

[0049]

[0050] in, For a moment Corrective control quantity; For a moment Verticality deviation; This is the proportionality coefficient; The integral time constant; The differential time constant; For integration variables; For the current moment; and based on the control quantity Output correction control commands to the trenching equipment.

[0051] The beneficial effects of this invention are:

[0052] (1) Multidimensional perception fusion significantly reduces monitoring blind spots and lag risks: Simultaneously acquire strain data, three-dimensional morphological data of the tank wall and acoustic emission signals, and perform fusion analysis to achieve complementary coverage of three types of risk signs: deformation, geometric anomaly and micro-fracture activity;

[0053] (2) Spatiotemporal alignment and edge preprocessing to ensure that multi-source data can be correlated, located, and judged more stably: multi-source data are annotated with a unified timestamp and time synchronization is completed by using a unified time reference. At the same time, multi-source data are mapped to a unified construction coordinate system to complete spatial registration. Filtering and noise reduction and feature extraction are performed to form fused feature data, so that data from different sources can be correlated and analyzed under the same coordinate and time axis, reducing the interference of construction noise on the judgment and improving the stability and reproducibility of the early warning results;

[0054] (3) Output location and range of abnormal area to enable precise implementation of treatment measures and avoid generalized work stoppage: Based on the verticality deviation of the trench wall, strain gradient and acoustic emission energy, the over-limit area is determined respectively, and spatial superposition is performed under a unified construction coordinate system to obtain the spatial location and spatial range of the abnormal area. This result can be directly used to determine the object of correction, the grouting reinforcement area and the support layout location, thereby improving the pertinence of intervention and construction continuity.

[0055] (4) Prediction, assessment and linkage control closed loop, reduce reliance on human experience and improve the foresight of risk control: The system outputs the deformation prediction results of sensitive buildings and outputs the risk assessment results based on the finite element mechanical analysis model. In the method, the deformation prediction results are used as the boundary condition input of the finite element analysis to obtain the stress response and form the risk assessment. The linkage control unit outputs control commands to the trenching equipment, grouting equipment and support execution mechanism accordingly, reducing reliance on human experience and improving the foresight of risk control. Attached Figure Description

[0056] In the attached diagram:

[0057] Figure 1 This is a block diagram of the monitoring and sensing system for the construction of underground continuous walls near sensitive buildings proposed in this invention;

[0058] Figure 2 This is a flowchart of the construction sensing and monitoring system for underground continuous walls near sensitive buildings proposed in this invention. Detailed Implementation

[0059] Reference Figure 1 and Figure 2 A monitoring system for the construction of underground continuous walls near sensitive buildings, comprising:

[0060] Trenching equipment is used to excavate trenches to form underground continuous walls;

[0061] Grouting equipment is used to inject grout into the grouting reinforcement area to reinforce the soil or fill voids;

[0062] Support actuators are used to carry out support operations or adjust support parameters.

[0063] It also includes: a data acquisition unit, an edge processing unit, a prediction and evaluation unit, and a linkage control unit;

[0064] The data acquisition unit includes a distributed fiber optic strain monitoring subunit, a three-dimensional lidar monitoring subunit, and an acoustic emission monitoring subunit, which respectively output strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals;

[0065] Specifically, the distributed optical fiber strain monitoring subunit includes multiple sets of BOTDR distributed optical fibers. These multiple sets of BOTDR distributed optical fibers are laid along the top of the guide wall and cover the side closest to the sensitive building, and are used to output continuous strain data along the laying direction.

[0066] The 3D lidar monitoring subunit is mounted on the mounting bracket of the trenching equipment and faces the trench wall, and is used to output the 3D point cloud data of the trench wall;

[0067] The acoustic emission monitoring subunit includes multiple acoustic emission sensors, which are deployed around the foundation of sensitive buildings to output acoustic emission signals.

[0068] The edge processing unit is connected to the data acquisition unit and is used to perform time synchronization, spatial registration, filtering and noise reduction and feature extraction on strain data, three-dimensional point cloud data of the tank wall and acoustic emission signals to obtain fused feature data, and output the abnormal area results based on the fused feature data.

[0069] Specifically, the edge processing unit includes a time synchronization module, a spatial registration module, a filtering and noise reduction module, a feature extraction module, and an anomaly region determination module;

[0070] The time synchronization module generates a unified time reference and adds a unified timestamp to the multi-source data formed by strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals to complete time synchronization;

[0071] The spatial registration module maps multi-source data to a unified construction coordinate system to complete spatial registration;

[0072] The filtering and noise reduction module performs filtering and noise reduction on the multi-source data after spatial registration;

[0073] The feature extraction module extracts strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the filtered and denoised multi-source data and forms fused feature data.

[0074] The anomaly region determination module outputs the spatial location and spatial range of the anomaly region based on the fused feature data;

[0075] The linkage control unit is connected to the prediction and evaluation unit, and is also connected to the trenching equipment, grouting equipment, and support actuator. It is used to output correction control commands, grouting control commands, and support trigger commands based on the risk assessment results.

[0076] Specifically, the linkage control unit determines the warning level based on the risk assessment results, and outputs grouting control commands to the grouting equipment when the warning level meets the grouting triggering conditions, outputs correction control commands to the trenching equipment when the warning level meets the correction triggering conditions, and outputs support triggering commands to the support actuator when the warning level meets the support triggering conditions.

[0077] In this embodiment, the first Each type of data is represented as a sequence on its respective timeline. ,in For the local clock of this type of sensor, the first Each sampling time. The time synchronization module will... Mapped to a unified time axis and at a unified discrete time upsampling This ensures that different data sources are on the same Top alignment. Resampling is expressed using linear interpolation as follows:

[0078]

[0079] in, For the first under a unified time base discrete time intervals and To and Two adjacent mapping sampling times. Based on linear interpolation, to further eliminate phase deviation caused by differences in sampling frequencies between different devices, this embodiment introduces an optimal time axis transformation. Let the characteristic sequences of fiber strain and radar point cloud on the same time axis be respectively... and Define the time scaling matrix With time translation vector An optimization model is established with the objective of minimizing the sum of squared errors:

[0080]

[0081] in, The number of sampling points used for alignment. and These are the optimal time scaling matrix and the optimal time translation vector, obtained by solving using least squares. and The optimal time axis transformation is constructed to transform the time axis of the radar point cloud sequence to be consistent with the time axis of the fiber strain sequence, thereby forming time-synchronized data and providing a consistent time reference for subsequent spatial registration and feature fusion.

[0082] After completing time alignment, to ensure that anomalies can be located to specific spatial areas, the spatial registration module establishes a unified construction coordinate system based on the construction survey benchmark. And to each sensor coordinate system The rigid body transformation is calibrated. For any rigid body transformation in the first... Sensor coordinate system Points represented in the middle , its in The coordinates in the equation satisfy:

[0083]

[0084] in, For rotation matrix, This is a translation vector. This transformation is used to map the entire 3D LiDAR point cloud to... For distributed fiber strain data, the fiber arc length coordinates are... Mapped to Space curves in Thus strain Assign the value to the corresponding spatial location; for the acoustic emission signal, first perform acoustic emission event localization to obtain the event location. The coordinates in the data are then compared with the point cloud and fiber optic data. Subsequent fusion.

[0085] Acoustic emission event localization employs a time-difference positioning method. Let the first... Each acoustic emission sensor in The coordinates in are The location of the acoustic emission event is The speed of sound wave propagation is The time of the event was Then the first Arrival time of each sensor satisfy:

[0086]

[0087] For any chosen reference sensor 1, construct the time difference It can be eliminated get:

[0088]

[0089] The time difference equations from multiple sensors are combined into an overdetermined system of equations, and the event location is obtained by solving the least squares method. Then, within a unified time window, the event energy is calculated and mapped to... The spatial neighborhood is used to form acoustic emission localization results and energy distribution, which serve as inputs for subsequent feature fusion and anomaly region identification.

[0090] After obtaining the registered multi-source data, the filtering and denoising module uses an improved Kalman filter for online denoising. The true state to be estimated is defined as... Define the measured quantity as Establish a discrete state-space model:

[0091]

[0092] in, Here is the state transition matrix. For the observation matrix, For process noise, To observe noise. To adapt to noise variations caused by construction conditions, the process noise covariance is... Covariance of observation noise The prediction and update process of the improved Kalman filter is dynamically updated based on the sliding window statistics.

[0093]

[0094]

[0095]

[0096]

[0097]

[0098] in, For state estimation, The state estimation error covariance, Kalman gain, noise reduction output This serves as input to the feature extraction module, thereby ensuring the stability of subsequent feature calculations.

[0099] In the feature extraction stage, this embodiment first establishes a correlation description between the strain field gradient tensor and the point cloud displacement. Let the displacement vector of any point in three-dimensional space be... The displacement gradient is Strain tensor under linear small deformation conditions for:

[0100]

[0101] Within the elastic range, this embodiment uses a linear elastic constitutive relation to express the strain tensor. With stress tensor This establishes a computable bridging relationship between point cloud displacement and stress response.

[0102]

[0103] in, The elastic matrix is ​​represented by the colon ":", which indicates double contraction. This relationship allows the strain calculated from the point cloud displacement to be further transformed into a stress index, maintaining consistency with the finite element solution in terms of dimensions and physical meaning.

[0104] Furthermore, the strain field gradient tensor is defined as follows: This is used to characterize the rate of change of strain in space. It represents the strain output from the BOTDR along the fiber optic routing direction. Let the strain after noise reduction be... One-dimensional strain gradient is obtained by spatial difference. ,in A fixed spatial interval along the fiber optic cable routing direction. An increase in the absolute value of the strain gradient characterizes strain concentration and abrupt deformation, and the maximum absolute value within the candidate region of anomalies is taken as the strain gradient feature. For 3D LiDAR point clouds, the denoised point cloud is denoted as a set. With depth For layered variables, the centerline of the trench wall is fitted at each depth layer, and its tilt angle relative to the vertical direction is calculated to obtain the verticality deviation feature of the trench wall. The angle of inclination can be determined by the slope of the projection of the centerline onto the horizontal plane. get, For acoustic emission signals, calculate the energy characteristics within a unified time window:

[0105]

[0106] in, For the energy calculation window length, This is the denoised acoustic emission waveform. A fused feature vector is constructed based on the above features. It provides a unified numerical input for anomaly area identification, deformation prediction, and risk assessment.

[0107] In the fusion decision-making stage, this embodiment employs a two-level fusion approach to ensure consistency between positioning accuracy and risk assessment. First, in feature-level fusion, the point cloud geometric deviation and strain field gradient tensor are aligned in a unified construction coordinate system, and the acoustic emission event energy is mapped to the same spatial grid. Subsequently, in decision-level fusion, to reduce false alarms caused by mistriggered signals from a single sensor, the DS evidence theory is used to fuse the risk confidence levels of the three data sources. The information from fiber optics, radar, and acoustic emission is constructed as basic probability assignment functions. , and The basic probability assignments after fusion are obtained using Dempster's combination rule. :

[0108] On the edge computing gateway side, the fused feature vector under the unified construction coordinate system is discretized into a multi-channel feature tensor on the spatial grid. And input it into a convolutional neural network to output an anomaly probability map. The basic form of convolution is shown below:

[0109]

[0110] in, The convolution kernel coefficient matrix, For bias terms, For activation function, This is the output feature map of the convolution. (Anomaly probability map) The spatial location and extent of the abnormal region are obtained through thresholding and connected component analysis, and are used as one of the inputs for decision-level fusion, together with the DS evidence fusion output, to enter the prediction, evaluation and linkage control stage.

[0111]

[0112] in, The conflict coefficient, and Let be the basic probability assignment function for the two evidence sources to be fused; then, by fusing the three types of evidence sources sequentially according to the same rules, we obtain... The sum of probabilities corresponding to the set of risk propositions is defined as the comprehensive risk probability. The anomaly probability map of the fused output and Together, they serve as inputs to the prediction and evaluation unit and the linkage control unit, ensuring that the abnormal area identification results and risk level determination have consistent probabilistic semantics.

[0113] In the prediction and assessment phase, this embodiment uses a time-series prediction model to make short-term predictions of the deformation of sensitive buildings. The sensitive buildings are projected at a uniform time point. The deformation amount is defined as Deformation rate is defined as and using fused feature vectors The prediction map is constructed using exogenous input. To derive the neural network structure from the recursive relation, the general form of the loop recursion is first given. ,in For the input vector, This is the hidden state. To alleviate the gradient vanishing problem in long sequence training, this embodiment uses a Long Short-Term Memory (LSTM) network to store the hidden states. With hidden state The information flow is separated and controlled by gating, and its gating equation is as follows:

[0114] ,

[0115]

[0116]

[0117]

[0118] in, For the Sigmoid function, For Hadamard product, , and These are the parameters to be trained. The output layer provides the predicted deformation. ,in To predict the step size, this embodiment uses Huber loss and weight decay. The prediction model is trained and outputs the deformation prediction results and confidence intervals within a set future time period, thereby providing boundary inputs for subsequent finite element mechanical analysis and providing a forward-looking basis for risk classification.

[0119] The Huber loss is used to maintain a squared penalty in the small error region and a linear penalty in the large error region, and its definition is as follows:

[0120]

[0121] in, For prediction error, The threshold value is used for segmentation. Using Huber loss can reduce the impact of occasional outliers on the update of prediction model parameters, thus ensuring that the prediction output remains stable even during sudden changes in construction disturbances.

[0122] In the finite element method (FEM) analysis phase, the deformation prediction results are used as boundary condition inputs to obtain the stress response of the structure and soil. This is based on the equilibrium equations of continuum mechanics. By combining geometric equations and constitutive relations and performing finite element discretization, the static equations are obtained:

[0123]

[0124] in, Here is the stiffness matrix. Let be the nodal displacement vector. This represents the equivalent load vector. To reflect the time-varying characteristics of construction disturbance, the dynamic equations are established in this embodiment:

[0125]

[0126] in, For the quality matrix, The damping matrix is ​​given. The dynamic equations are integrated in the time domain using the implicit Newmark-β method, with β=0.25 chosen to ensure stability. Equivalent stress contour maps and stress indices for key components are output based on an adaptive hexahedral mesh. .Will The control points of the sensitive building foundation were applied as displacement boundary conditions, and the solution was obtained. Together with the results of abnormal areas, they are used to form the risk assessment results.

[0127] The Newmark-β method uses a recursive relationship between displacement and velocity to complete the time-domain integration. The recursive relationship is shown below:

[0128]

[0129]

[0130] in, The integration step size is... For velocity vectors, This is the acceleration vector.

[0131] In the risk quantification and early warning classification stage, the strain gradient characteristics within the abnormal region are... Verticality deviation characteristics Acoustic emission energy characteristics Risk indicators are obtained by weighted fusion. :

[0132]

[0133] in, These are the strain gradient weighting coefficients. This is the verticality deviation weighting coefficient. Let be the acoustic emission energy weighting coefficients, and let the sum of the weighting coefficients be 1, as shown in the following formula:

[0134] The warning level is determined by comparing the risk indicators with the preset thresholds, and the warning level is used as the input of the linkage control unit, thereby realizing a verifiable mapping relationship between the risk assessment output and the control command generation.

[0135] During the linkage control phase, the linkage control unit outputs grouting control commands to the grouting equipment, correction control commands to the trenching equipment, and a support trigger command to the support actuator when the support triggering condition is met. The correction control command is derived using a PID control law. Let the reference value for the trench wall verticality deviation be... The deviation was measured in real time as Then the control error is PID control quantity Defined as:

[0136]

[0137] in, This is the proportionality coefficient. The integral time constant is... The differential time constant is Let it be the integration variable. The mapping is used to adjust the attitude of the trenching equipment to achieve verticality correction. The grouting control commands are derived using a fuzzy control model. The input variables are defined as strain gradient characteristics. Acoustic emission energy characteristics Establish membership functions for each input and form rule triggering strength. The rule output is a representative value of the grouting volume. The grouting volume command is obtained by defuzzifying the fuzzy logic using the centroid method. :

[0138]

[0139] in, The number of rules is [number]. The support triggering condition uses deterministic logic; the linkage control unit responds when the early warning level reaches the support level and the stress index [is met / is determined]. When the preset threshold is exceeded, a support trigger command is output and the support execution mechanism is started, so that the support action is triggered when the conditions are met and the trigger criteria are reproducible.

[0140] As another embodiment of this application, this embodiment proposes a method for sensing and monitoring the construction of underground diaphragm walls near sensitive buildings, including any of the above-mentioned schemes of the sensing and monitoring system for underground diaphragm walls near sensitive buildings. The method steps are as follows:

[0141] S1. Data Acquisition: During the construction of the diaphragm wall, distributed optical fiber strain data, trench wall three-dimensional point cloud data, and acoustic emission signals are collected.

[0142] S2. Time Synchronization: Generate a unified time reference and add a unified timestamp to strain data, 3D point cloud data of the tank wall, and acoustic emission signals to complete time synchronization;

[0143] Specifically, using the unified time reference output by the time synchronization module as a benchmark, multi-source data is resampled and aligned to the same sampling period to form time-synchronized data.

[0144] S3. Spatial Registration: Mapping strain data, 3D point cloud data of the trench wall, and acoustic emission signals to a unified construction coordinate system to complete spatial registration;

[0145] Specifically, a unified construction coordinate system is established based on the construction measurement benchmark. The positioning results of the acoustic emission signal are mapped to the coordinate system of the three-dimensional point cloud data of the trench wall using calibration parameters. The distributed optical fiber strain data is also mapped to the same coordinate system to form spatial registration data.

[0146] S4. Filtering and noise reduction: Filter and reduce noise on the registered data to obtain denoised data;

[0147] S5. Feature Extraction and Fusion: Extract strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the noise reduction data and form fused feature data;

[0148] S6. Abnormal Region Determination: Determine abnormal regions based on fused feature data, and output the spatial location and spatial range of the abnormal regions;

[0149] Specifically, the verticality deviation of the tank wall is calculated based on the three-dimensional point cloud data of the tank wall, and the area exceeding the limit is determined;

[0150] Calculate the strain gradient and determine the gradient over-limit region based on distributed optical fiber strain data;

[0151] Calculate the acoustic emission energy based on the acoustic emission signal and determine the energy excess region;

[0152] By spatially superimposing the oversized areas under a unified construction coordinate system, the spatial location and spatial range of the abnormal areas can be obtained.

[0153] S7. Prediction, assessment and linkage control: Based on the fusion of feature data, output the deformation prediction results of sensitive buildings; based on the finite element mechanical analysis model, output the risk assessment results; and based on the risk assessment results, output the correction control command, grouting control command and support trigger command.

[0154] Specifically, a time-series prediction model is used to make short-term predictions of the deformation of sensitive buildings, and the deformation prediction results are obtained.

[0155] The deformation prediction results are used as the boundary condition input for the finite element mechanical analysis model to obtain the stress response results of the structure and soil.

[0156] The risk assessment results are determined based on the stress response results and the abnormal area results, and the risk assessment results are output to the linkage control unit.

[0157] Risk assessment results are obtained through risk indicators. Quantification is performed, and the risk indicator R satisfies:

[0158]

[0159] in, These are the strain gradient weighting coefficients. This is the verticality deviation weighting coefficient. For acoustic emission energy weighting coefficients, These are the characteristic values ​​of the strain gradient within the anomalous region. This represents the characteristic value of the verticality deviation of the tank wall within the abnormal area. The characteristic value of acoustic emission energy in the anomalous region, based on The warning level is determined by comparing the result with the preset threshold.

[0160] The corrective control command is as follows:

[0161]

[0162] in, For a moment Corrective control quantity; For a moment The perpendicularity deviation; is the proportionality coefficient; is the integral time constant; is the derivative time constant; is the integral variable; is the current time; and based on the control quantity outputs a deviation correction control instruction to the trench forming equipment.

[0163] In order to more clearly illustrate the solutions and effects of this implementation, a practical case is used as an example:

[0164] I. Multi-dimensional sensing network deployment

[0165] The BOTDR distributed optical fiber is arranged in a ring along the top of the guide wall at a spacing of 0.5 m. The monitoring range covers the building foundations on both sides of the NQ25 - NQ45 trench section, and the strain monitoring accuracy is ±10 με. The guide wall adopts a "丨丨" type reinforced concrete structure (height 1.5 m, wing width 1.0 m). The optical fiber is closely attached to the concrete surface of the guide wall and fixed through embedded buckles.

[0166] The 3D lidar is installed on the top of the Jintai SC50A trench forming machine, with a scanning frequency of 20 Hz and a perpendicularity detection accuracy of ±2 mm. For the 5 m wide trench section (NQ38 - NQ42), the double-grab trench forming process is adopted. After each grab advance, a 120° scan is triggered to generate a trench wall point cloud model (point spacing ≤2 mm), and the perpendicularity deviation is calculated in real time. When the deviation > 1 / 300, the PID deviation correction is triggered.

[0167] The acoustic emission probes are arranged in a densified grid of 5 m × 5 m around the building foundation, with a frequency response range of 1 kHz - 1 MHz and a sensitivity ≥70 dB. 12 groups of probes are buried in the foundation of XX Engineering Company, and an energy threshold of 800 mV·ms is set. When the cumulative value exceeds the limit, the grouting pump is started.

[0168] II. Edge computing and data fusion

[0169] The edge gateway is built-in with an improved Kalman filtering algorithm to dynamically denoise the optical fiber strain data (sampling rate 1 Hz), and the signal-to-noise ratio is increased to 40 dB. A lightweight CNN model (8-bit integer quantization) is deployed to process the 128×128×6-dimensional fusion data tensor and generate a heat map of the abnormal area within 5 seconds.

[0170] The spatio-temporal alignment module uses dual-mode timing of Beidou / GPS (accuracy ±1 μs) to correlate the lidar point cloud data with the BOTDR strain field through the elastic constitutive equation.

[0171] III. Dynamic prediction and closed-loop control

[0172] The LSTM model takes historical 30-minute settlement rate (resolution 0.01 mm / s), acoustic emission energy (integral value of 20-100 kHz), and strain gradient data as input, and outputs a settlement prediction for the next 30 minutes. The model is trained on a local soft soil dataset, and the prediction error is ≤0.2 mm.

[0173] The finite element simulation module integrates BIM geological parameters (elastic modulus 15-30MPa, Poisson's ratio 0.3-0.35) and generates stress cloud maps using adaptive hexahedral meshes (minimum size 50mm). When the proportion of the red area (stress ≥ 120% of the design value) is >15%, the steel support system (spacing ≤ 2m) is triggered.

[0174] The PID controller parameters for the trenching machine are set to Kp=0.8, Ti=8s, and the response time is ≤8s. For 6m wide trench sections, skip-trenching construction is implemented (interval ≥5 sections), and the hydraulic valve opening adjustment Δ=0.12×vertical deviation (mm).

[0175] IV. System Linkage Verification

[0176] The trial trenching stage (NQ43 images) shows that multi-source data fusion reduced the hole collapse positioning error from 35cm to 8cm using traditional methods, and the false alarm rate from 12% to 3.8%.

[0177] During construction, when the cumulative acoustic emission energy reaches 850 mV·ms, the fuzzy controller outputs a grouting pressure of 0.42 MPa (error ±2.7%), which is 60% more efficient than manual decision-making.

[0178] The digital twin platform overlays LSTM predictions and finite element stress cloud maps to generate risk heat maps to guide construction: a total of 23 Level I warnings, 7 Level II warnings, and 1 Level III warning were triggered, and the maximum settlement of the building was controlled within 3.2 mm.

Claims

1. A monitoring system for the construction of underground diaphragm walls near sensitive buildings, comprising: Trenching equipment is used to excavate trenches to form underground continuous walls; Grouting equipment is used to inject grout into the grouting reinforcement area to reinforce the soil or fill voids; Support actuators are used to carry out support operations or adjust support parameters. Its features include: a data acquisition unit, an edge processing unit, a prediction and evaluation unit, and a linkage control unit; The data acquisition unit includes a distributed fiber optic strain monitoring subunit, a three-dimensional lidar monitoring subunit, and an acoustic emission monitoring subunit, which respectively output strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals; The edge processing unit is connected to the data acquisition unit and is used to perform time synchronization, spatial registration, filtering and noise reduction and feature extraction on strain data, three-dimensional point cloud data of the tank wall and acoustic emission signals to obtain fused feature data, and output the abnormal area results based on the fused feature data. The prediction and assessment unit is connected to the edge processing unit to output the deformation prediction results of sensitive buildings based on the fused feature data, and to output the risk assessment results based on the finite element mechanical analysis model. The linkage control unit is connected to the prediction and evaluation unit, and is also connected to the trenching equipment, grouting equipment, and support actuator. It is used to output correction control commands, grouting control commands, and support trigger commands based on the risk assessment results.

2. The monitoring system for the construction of underground continuous walls near sensitive buildings according to claim 1, characterized in that: The distributed optical fiber strain monitoring subunit includes multiple sets of BOTDR distributed optical fibers. These multiple sets of BOTDR distributed optical fibers are laid along the top of the guide wall and cover the side closest to the sensitive building, and are used to output continuous strain data along the laying direction. The 3D lidar monitoring subunit is mounted on the mounting bracket of the trenching equipment and faces the trench wall, and is used to output the 3D point cloud data of the trench wall; The acoustic emission monitoring subunit includes multiple acoustic emission sensors, which are deployed around the foundation of sensitive buildings to output acoustic emission signals.

3. The monitoring system for the construction of underground continuous walls near sensitive buildings according to claim 1, characterized in that: The edge processing unit includes a time synchronization module, a spatial registration module, a filtering and noise reduction module, a feature extraction module, and an anomaly region determination module; The time synchronization module generates a unified time reference and adds a unified timestamp to the multi-source data formed by strain data, three-dimensional point cloud data of the tank wall, and acoustic emission signals to complete time synchronization; The spatial registration module maps multi-source data to a unified construction coordinate system to complete spatial registration; The filtering and noise reduction module performs filtering and noise reduction on the multi-source data after spatial registration; The feature extraction module extracts strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the filtered and denoised multi-source data and forms fused feature data. The anomaly region determination module outputs the spatial location and spatial range of the anomaly region based on the fused feature data; The linkage control unit determines the early warning level based on the risk assessment results, and outputs grouting control commands to the grouting equipment when the early warning level meets the grouting triggering conditions, outputs correction control commands to the trenching equipment when the early warning level meets the correction triggering conditions, and outputs support triggering commands to the support actuator when the early warning level meets the support triggering conditions.

4. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings, characterized in that: The method of the monitoring system for the construction of underground diaphragm walls near sensitive buildings, as described in any one of claims 1-3, comprises the following steps: S1. Data Acquisition: During the construction of the diaphragm wall, distributed optical fiber strain data, trench wall three-dimensional point cloud data, and acoustic emission signals are collected. S2. Time Synchronization: Generate a unified time reference and add a unified timestamp to strain data, 3D point cloud data of the tank wall, and acoustic emission signals to complete time synchronization; S3. Spatial Registration: Mapping strain data, 3D point cloud data of the trench wall, and acoustic emission signals to a unified construction coordinate system to complete spatial registration; S4. Filtering and noise reduction: Filter and reduce noise on the registered data to obtain denoised data; S5. Feature Extraction and Fusion: Extract strain gradient features, trench wall verticality deviation features, and acoustic emission energy features from the noise reduction data and form fused feature data; S6. Abnormal Region Determination: Determine abnormal regions based on fused feature data, and output the spatial location and spatial range of the abnormal regions; S7. Prediction, assessment and linkage control: Based on the fusion of feature data, output the deformation prediction results of sensitive buildings; based on the finite element mechanical analysis model, output the risk assessment results; and based on the risk assessment results, output the correction control command, grouting control command and support trigger command.

5. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, In step S2, the unified time reference output by the time synchronization module is used as the reference to resample the multi-source data and complete the alignment of the same sampling period to form time synchronization data.

6. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, In step S3, a unified construction coordinate system is established based on the construction measurement benchmark. The positioning results of the acoustic emission signal are mapped to the coordinate system where the three-dimensional point cloud data of the trench wall is located using calibration parameters. The distributed optical fiber strain data is also mapped to the same coordinate system to form spatial registration data.

7. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, In step S6: Calculate the verticality deviation of the tank wall based on the 3D point cloud data of the tank wall and determine the area exceeding the limit; Calculate the strain gradient and determine the gradient over-limit region based on distributed optical fiber strain data; Calculate the acoustic emission energy based on the acoustic emission signal and determine the energy excess region; By spatially superimposing the oversized areas under a unified construction coordinate system, the spatial location and spatial range of the abnormal areas can be obtained.

8. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, In step S7: A time-series prediction model was used to make short-term predictions of the deformation of sensitive buildings, and the deformation prediction results were obtained. The deformation prediction results are used as the boundary condition input for the finite element mechanical analysis model to obtain the stress response results of the structure and soil. The risk assessment results are determined based on the stress response results and the abnormal area results, and the risk assessment results are output to the linkage control unit.

9. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, Risk assessment results are obtained through risk indicators. Quantify and risk indicators satisfy: ； in, These are the strain gradient weighting coefficients. This is the verticality deviation weighting coefficient. For acoustic emission energy weighting coefficients, These are the characteristic values ​​of the strain gradient within the anomalous region. This represents the characteristic value of the verticality deviation of the tank wall within the abnormal area. The characteristic value of acoustic emission energy in the anomalous region, based on The warning level is determined by comparing the result with the preset threshold.

10. A method for sensing and monitoring the construction of underground continuous walls near sensitive buildings according to claim 4, characterized in that, The correction control command in step S7 is: ； in, For a moment Corrective control quantity; For a moment Verticality deviation; This is the proportionality coefficient; The integral time constant; The differential time constant; For integration variables; For the current moment; and based on the control quantity Output correction control commands to the trenching equipment.

Images