Devices and methods for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process.

Abstract

This invention discloses a device and method for monitoring and warning of pit bottom heave throughout the entire excavation process, belonging to the field of foundation pit engineering safety monitoring technology. The device's anchoring unit includes a cone head, a radially deployable umbrella-type support assembly, and a cylindrical guide rail. Reliable anchoring is achieved by inserting the cone head into a stable soil layer and deploying it in conjunction with the support plate. The measuring unit includes a protective tube that can slide axially along the cylindrical guide rail, a measuring rod fixed to the guide rail, a magnetic ring that moves synchronously with the protective tube, a magnetostrictive displacement sensor, a flexible cable, and an intelligent acquisition module. This invention enables high-precision, automated, and remote real-time monitoring and early warning of the entire foundation pit excavation process, facilitating timely warning of pit bottom heave risks.

Description

Technical Field

[0001] This invention relates to the field of foundation pit engineering safety monitoring technology, and more specifically, to a device and method for monitoring bottom heave and providing early warning of instability throughout the entire foundation pit excavation process. Background Technology

[0002] As a crucial component of underground space construction, foundation pit engineering must withstand the loads of surrounding strata and buildings to maintain the stability of the excavated area. With increasing excavation depth, the soil at the pit bottom undergoes elastic or plastic heave due to unloading and stress release, leading to intensified deformation of the support structure and, in severe cases, foundation pit instability. The heave displacement of the foundation pit directly reflects the stress release state of the soil, and its accurate monitoring is vital for construction safety, enabling timely detection of potential risks and guidance for support optimization. While current monitoring technologies encompass various methods such as ultrasonic sensing, fiber optic grating measurement, and laser positioning, significant limitations remain. First, most devices cannot achieve continuous automatic monitoring throughout the entire excavation process, making it difficult to capture the dynamic evolution of the heave. For example, while total stations paired with prism-based line-layout devices (CN216645366U) and devices based on laser transmitters and monitoring scales (CN116125037A) can observe changes in pit bottom height, the measurements rely on manual reading, making real-time continuous monitoring of the entire foundation pit excavation process impossible. Furthermore, the monitoring process requires manual operation or position adjustments, limiting the continuity and real-time nature of the monitoring. Secondly, existing solutions overemphasize the displacement of the support structure and the surrounding surface settlement, while neglecting the monitoring of soil heave at the pit bottom. Furthermore, early warning methods often rely on purely data-driven models, leading to low prediction accuracy in the early stages of construction due to missing samples. Alternatively, they rely solely on statistical methods to establish multi-indicator correlations, lacking constraints from geotechnical mechanisms, resulting in insufficient reliability of early warning results. Existing solutions primarily focus on monitoring and assessing the displacement and internal forces of the support structure (CN118862556A), surrounding surface settlement (CN119441741B), or surrounding sensitive structures (CN120450651A), neglecting the crucial role of soil heave at the pit bottom, which can also lead to pit instability. The attention given to this issue is relatively insufficient. In terms of early warning assessment efficiency, existing early warning methods, such as a dynamic early warning method for deep foundation pit construction risks, mostly adopt a purely data-driven model (CN119918940A). Such models show slow improvement in prediction accuracy when samples are missing in the early stages of construction. Alternatively, there is a deep foundation pit monitoring and early warning scheme based on multi-source feature fusion (CN119599217A). Although it strengthens the correlation between multiple indicators and establishes a risk assessment model through statistical means, its essence is still based on empirical correlation analysis of historical data. It lacks the constraints of underlying geotechnical mechanisms, and the correlation law may fail. The engineering reliability of the early warning results needs to be improved.More importantly, the measurement accuracy is constrained by three factors: First, the stability of the reference is a significant issue. For example, when an ultrasonic sensor is fixed to a crossbeam in the foundation pit, slight deflection or displacement of the crossbeam can cause the reference reference to drift. For instance, in the ultrasonic sensor-based device (CN206467699U), the ultrasonic sensor is fixed to the crossbeam in the foundation pit, but the crossbeam may experience slight deflection or displacement under construction or soil settlement conditions, leading to instability of the sensor reference reference and resulting in errors in the measured heave, thus affecting measurement accuracy. Second, the coupled motion between the monitoring device and the soil causes systematic errors. When the soil heaves, the device is moved as a whole, causing the measured value to reflect the absolute displacement of the device rather than the relative heave of the soil. Existing technologies have not yet solved the problem of the coupling motion between the device and the fixed reference. Independence issues exist. For example, when probe-type PCB electromagnetic induction devices (CN118563743A) and fiber optic displacement gauge devices (CN214657286U) are inserted into the soil, the monitoring devices are in close contact with the soil. When the soil heaves, the devices may be carried upwards as a whole, causing the measurement data to reflect the absolute displacement of the device rather than the true relative heave of the soil relative to the base. Thirdly, high-precision sensors such as magnetostrictive sensors are susceptible to temperature fluctuations and mechanical vibration interference in complex construction environments. Temperature changes cause circuit temperature drift, while vibration generates random noise. Existing solutions lack a real-time error calibration mechanism for the coupling effect of ambient temperature and construction vibration, resulting in a serious decrease in measurement reliability under dynamic construction conditions. Summary of the Invention

[0003] The purpose of this application is to provide a device and method for monitoring pit bottom heave and providing early warning of instability throughout the entire process of foundation pit excavation.

[0004] The present invention adopts the following solution:

[0005] A device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit includes: an anchoring unit and a measuring unit, wherein the anchoring unit and the measuring unit are connected axially front and rear; wherein... The anchoring unit includes a cone head, a radially expandable umbrella-type support assembly, and a cylindrical guide rail. The cone head is located at the front end of the umbrella-type support assembly and is used to insert into the stable soil layer below the bottom of the foundation pit for positioning. The umbrella-type support assembly is located at the front end of the cylindrical guide rail and is connected to the cone head. It can be radially expanded after being inserted into the stable soil layer to increase the contact area with the surrounding soil. The measuring unit includes a protective tube movably connected to the other end of the cylindrical guide rail, a measuring rod disposed inside the protective tube and fixedly connected to the cylindrical guide rail, a magnetic ring movably sleeved on the outer wall of the protective tube, a magnetostrictive displacement sensor located inside the protective tube for sensing changes in the position of the magnetic ring, a flexible cable electrically connected to the magnetostrictive displacement sensor, and an intelligent acquisition module; wherein, the protective tube is configured to have a torque-counteracting travel amount that slides along the axial direction of the cylindrical guide rail, so that the protective tube can slide relative to the cylindrical guide rail during soil heave, and eliminate the upward torque exerted on the measuring unit by the heave of the foundation pit, avoiding errors caused by the overall upward movement of the measuring unit; The magnetostrictive displacement sensor is adapted to detect and output the axial displacement of the magnetic ring relative to the measuring rod as the amount of pit bottom heave. The intelligent acquisition module is used to control the data acquisition of the magnetostrictive displacement sensor, process the soil displacement data, and realize data communication with the remote monitoring platform.

[0006] Furthermore, the umbrella-type support assembly is connected to the cylindrical guide rail via a spring, so that the umbrella-type support assembly is constrained and closed by external force during insertion, and automatically expands radially under the action of the spring after reaching a predetermined depth.

[0007] Furthermore, the umbrella-type support assembly includes an umbrella-type support plate, which adopts a sheet-like, arc-shaped, or hollowed-out reinforced structure, and has anti-slip ridges or rough surfaces on its outer surface to increase the frictional resistance with the soil and enhance the device's pull-out and shear stability.

[0008] Furthermore, the measuring unit includes a data acquisition section and at least one set of extension sections, the extension sections being adapted to connect to the upper and / or lower ends of the data acquisition section; both the data acquisition section and the extension sections are provided with protective pipes on their outer sides, and the end diameters of the data acquisition section and the extension sections are complementary and nested together, with locking nuts provided at the external connection points to ensure that the protective pipes on the data acquisition section and the extension sections undergo overall axial displacement when moved by the soil; wherein, the protective pipe of the data acquisition section is provided with a magnetic ring on its outer side, and the magnetostrictive displacement sensor is integrated inside, the magnetostrictive displacement sensor being connected to the intelligent acquisition module through the flexible cable, for sensing the displacement of the magnetic ring and processing the data.

[0009] Furthermore, the protective tube is sleeved on the outside of the cylindrical guide rail and maintains a radial gap of 0.5mm-20mm. The radial gap is filled with lubricating oil and covered with a water-resistant membrane material to reduce sliding resistance and achieve waterproof sealing.

[0010] Furthermore, the flexible cable is a spring conductor to accommodate the sliding displacement of the protective tube and the cylindrical guide rail.

[0011] This invention also provides a method for monitoring and predicting instability of the pit bottom during the entire excavation process, using the aforementioned device for monitoring and predicting instability of the pit bottom during the entire excavation process, and a remote monitoring platform communicatively connected to the intelligent acquisition module, comprising the following steps: S1. Drill holes before the foundation pit is excavated to ensure that the drilling depth reaches the deep stable soil layer that is not affected by the excavation of the foundation pit. S2. Inject lubricating oil at the connection between the cylindrical guide rail and the protective tube to fill the connection tightly with lubricating oil, and install a water-proof membrane material. S3. Wrap the bottom of the cone head of the anchoring unit with a bag of cement, and prepare to lower the anchoring unit and the bag of cement to the bottom of the burial hole simultaneously. S4. Connect the anchoring unit and the measuring unit in sequence and lower them synchronously. By controlling the lowering depth, make the magnetic ring accurately located at the preset depth of the foundation pit excavation surface, as the initial monitoring reference point. S5. Make the uppermost end of the measuring unit protrude more than 0.5m above the initial ground surface before the foundation pit is excavated; S6. After the cone is lowered into place, grout is injected into the bottom of the hole through the grouting pipe pre-installed in the hole, so that the cone end and the bagged cement are fixed with the stable soil layer, and the bottom anchor end is locked. S7. The intelligent acquisition module controls the measurement unit to read the initial value and perform zero calibration, and acquires the displacement output of the magnetostrictive displacement sensor during the excavation process, and calculates the axial displacement of the magnetic ring relative to the measuring rod to obtain the amount of pit bottom heave. S8. While acquiring displacement signals, simultaneously obtain ambient temperature and construction vibration parameters, and use the built-in error compensation model to correct the original uplift data in real time to eliminate measurement deviations caused by environmental interference. S9. The intelligent acquisition module uploads the compensated uplift monitoring data to the remote monitoring platform for trend prediction and early warning.

[0012] Furthermore, the intelligent acquisition module is used to receive sensor data, control waveguide wire pulse emission and response acquisition, process magnetic ring displacement data, and transmit it to a remote monitoring platform via a network. Simultaneously, the intelligent acquisition module incorporates a temperature sensing unit, a vibration sensing unit, and a processor unit, and embeds an error compensation model in the processor unit to correct measurement errors caused by ambient temperature and construction vibration in real time. The model establishment process includes: S81. Multi-source sample data acquisition: Before the device is buried in the foundation pit, an environmental interference sample library needs to be constructed by simulating the environment. The intelligent acquisition module performs time-frequency domain feature extraction on the acquired raw signals. First, the original acceleration sequence obtained by the vibration sensing unit is used. The process involves calculating the effective value of vibration that reflects the overall background energy of the environment. ,in And the peak factor reflecting the mechanical impact characteristics. ,in The maximum absolute peak value of the vibration signal within the sampling period; simultaneously, for the time-domain signal Perform a Fast Fourier Transform to obtain the spectrum function, as shown in the formula: ; in For time variables, To meet The imaginary unit, Angular frequency; Extract the angular frequency corresponding to the maximum amplitude of the spectral function. Calculate the main frequency To identify and locate frequency interference points that may cause resonance in magnetostrictive waveguide wires; Finally, the known calibration standard displacement in the laboratory will be determined. Sensor original displacement value Real-time ambient temperature The extracted vibration feature sets are then fused to construct the input feature vector. And based on the calculated measured deviation value An offline training database was established to serve as the prediction label. S82. Model Construction and Optimization Based on SVR: The SVR algorithm is used to model the nonlinear error law and solve the measurement deviation problem caused by the coupling of temperature drift and vibration noise. Using the radial basis function kernel to transform the low-dimensional input feature vector Mapping to a high-dimensional feature space to handle complex environmental interference, the formula is: ; in Parameters for controlling the mapping range of kernel functions, Support vector samples determined during the offline training phase of the model; Furthermore, by combining grid search Cross-validation, for the penalty factor Insensitive loss coefficient and kernel function parameters The formula for joint optimization is: ; in , For Lagrange multipliers, The total number of support vectors selected. This is the bias term; the model establishes the relationship between environmental characteristics and prediction bias. The precise mapping relationship between them; S83. Real-time compensation for uplift displacement: The trained and optimized model parameters are embedded in the embedded processor of the intelligent acquisition module; during the monitoring of the foundation pit excavation, the intelligent acquisition module synchronously reads the original displacement signals from the sensors. Real-time temperature and the vibration characteristic group after processing by the processor unit Input the real-time feature vector into the SVR decision function. The comprehensive prediction deviation under the current environmental conditions is calculated using a formula. The corrected output yields the final high-precision result of the pit bottom heave displacement. .

[0013] Furthermore, the remote monitoring platform is used to adjust and manage the sampling frequency, data content, and communication status of the measurement unit in real time. Simultaneously, the remote monitoring platform integrates a finite element model-driven foundation pit stability proxy model to predict and warn of instability risks during the foundation pit excavation process. The model establishment process includes: S91. Establishment of the Finite Element Model: Based on the actual geological report and support design, a three-dimensional finite element model reflecting the unloading characteristics of the foundation pit is established; considering the temporal nature of the foundation pit excavation, it is divided according to the construction conditions. The excavation sequence is as follows: Static soil parameters affecting the stability of the foundation pit are selected, including the soil elastic modulus. Cohesion internal friction angle The overall excavation sequence characterizing the project scale Current construction excavation sequence And the amount of bottom heave of the foundation pit corresponding to the current step sequence. Together as model input variables Assume random variables Following a specific probability distribution, it is generated using Latin hypercube sampling. A combination of static soil parameters was used to extract the state variables corresponding to each step of the process through finite element simulation. Together, we construct the input sample set ; each group of samples Input the finite element model to simulate the entire excavation process and extract the sequence of each excavation step. Deformation of key nodes that induce foundation pit instability This forms the corresponding output response sequence. This leads to the construction of a training sample database containing a "parameter-state-response" mapping relationship. ; S92. Establishing a spatiotemporal surrogate model for a CNN-BiLSTM hybrid neural network: Establishing the input variables The deformation time sequence of the entire foundation pit output The model uses a nonlinear mapping relationship between the input layer, a CNN feature extraction layer, a BiLSTM temporal prediction layer, and a fully connected output layer to replace the time-consuming finite element calculation. This represents the input variable under the t-th excavation step; Using size convolution kernel For the input vector Perform sliding convolution to extract feature maps ; As the convolution kernel traverses the entire The generated complete feature sequence ,in For input vectors The feature dimension is set to 6 here; convolutional feature maps Downsampling is performed: Max pooling is used as the specific implementation for feature dimensionality reduction to retain the most significant mechanical feature information and generate the output feature vector. and will As a BiLSTM unit in the Input at any moment; Forward propagation follows the excavation steps from arrive Calculate and capture the cumulative deformation effect of the foundation pit excavation, i.e. ; Backpropagation from excavation steps Calculation up to 1 captures the feedback effect of subsequent excavation and unloading on the current state, i.e. The formula expansion structure is the same as above, only the time direction is reversed; BiLSTM in step sequence The final output is from the forward hidden state. and backward hidden state It is pieced together, that is Finally, the predicted output is generated by mapping the deformation value to the fully connected layer. ; S93. Using mean squared error as a loss function to measure predicted deformation values. The actual value calculated by finite element method The difference between them is minimized by using the backpropagation algorithm to drive the update of network weights; S94. Reliability Risk Calculation: In the remote monitoring platform, a large-scale stochastic simulation is performed using a trained CNN-BiLSTM surrogate model to calculate the failure probability of the foundation pit under the current excavation conditions; assuming the foundation pit is in the [missing information]th ... The deformation safety warning threshold under the excavation sequence is Define function ,when This indicates that the foundation pit is in a reliable state, when This indicates that the foundation pit is in a failed state; generation A set of random parameter samples that conform to the statistical characteristics of the soil in the field. ,in Inputting the CNN-BiLSTM surrogate model allows for rapid prediction of the corresponding deformation response. The statistical prediction results satisfy Number of failure samples Calculate the failure probability ; S95. Current Risk Level Warning: Based on the calculated failure probability. Combined with reliability indicators Real-time early warning of the current excavation sequence of the foundation pit. It is the inverse function of the standard normal distribution function; set the target reliability threshold. ,like The remote monitoring platform automatically triggers an early warning signal, prompting the site to immediately stop excavation and take reinforcement measures; if If so, it indicates that the foundation pit is in a safe state; S96. Risk Classification and Early Warning for the Next Construction Step: After completing the risk classification and early warning for the current construction step, the remote monitoring platform will automatically predict the risk situation of the next construction step; known geological parameters include soil elastic modulus E, cohesion c, and internal friction angle. Total Excavation Sequence Time step sequence updated to , before Step bulge sequence [ , , ], obtained through polynomial fitting extrapolation Excavation sequence uplift volume ; Obtain the input variables under the (t+1)th excavation step sequence After reliability risk calculation, the following results were obtained. And in the risk classification and early warning of the next construction step, it is obtained and with reliability threshold Make judgments on early warning information.

[0014] Beneficial effects: This solution connects the conical head, radially expandable umbrella-type support assembly, and cylindrical guide rail. On the one hand, it allows the monitoring device to be stably inserted into the soil. On the other hand, by configuring the protective tube to have a torque that slides along the axial direction of the cylindrical guide rail to offset the travel amount, it eliminates the slight error in the overall rise of the monitoring device caused by the upward torque of the soil on the monitoring device, thereby making the monitoring more accurate. Attached Figure Description

[0015] Figure 1 This is an overall schematic diagram of a device for monitoring and instability early warning of pit bottom heave throughout the entire excavation process according to an embodiment of the present invention; Figure 2 This is a cross-sectional schematic diagram of a device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, according to an embodiment of the present invention. Figure 3 This is a partially enlarged schematic diagram of a device for monitoring and instability early warning of pit bottom heave throughout the entire excavation process according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the installation of a device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, according to an embodiment of the present invention. Figure 5 This is a schematic diagram of the soil after it has bulged, illustrating the device for monitoring and instability early warning of pit bottom heave during the entire excavation process of a foundation pit according to an embodiment of the present invention. Figure 6 This is a flowchart of a method for monitoring and instability warning of pit bottom heave during the entire excavation process of a foundation pit, according to an embodiment of the present invention.

[0016] Figure label: 1. Cone head; 2. Umbrella-type support assembly; 3. Cylindrical guide rail; 4. Protective tube; 5. Measuring rod; 6. Magnetic ring; 7. Magnetostrictive displacement sensor; 8. Flexible cable; 9. Intelligent acquisition module; 10. Radial clearance A; 11. Torque cancellation stroke S. Detailed Implementation

[0017] Combination Figures 1 to 5As shown, this embodiment provides a device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process. The device includes an anchoring unit and a measuring unit, which are connected axially front to back. The anchoring unit includes a cone head 1, a radially deployable umbrella-type support assembly 2, and a cylindrical guide rail 3. The cone head 1 is positioned at the front end of the umbrella-type support assembly 2 for insertion into a stable soil layer below the bottom of the pit for positioning. The umbrella-type support assembly 2 is detachably mounted on the front section of the cylindrical guide rail 3 and connected to the cone head 1, and can radially expand after insertion into the stable soil layer to increase the contact area with the surrounding soil. The measuring unit includes a protective tube 4 movably connected to the other end of the cylindrical guide rail 3, a measuring rod 5 disposed inside the protective tube 4 and fixedly connected to the cylindrical guide rail 3, a magnetic ring 6 movably sleeved on the outer wall of the protective tube 4, a magnetostrictive displacement sensor 7 located inside the protective tube 4 for sensing changes in the position of the magnetic ring 6, a flexible cable 8 electrically connected to the magnetostrictive displacement sensor 7, and an intelligent acquisition module 9; wherein, the protective tube 4 is configured with a torque-counteracting stroke S that slides along the axial direction of the cylindrical guide rail 3, so that the protective tube 4 can slide a certain displacement relative to the cylindrical guide rail 3 during soil heave, thereby eliminating the upward torque on the overall measuring unit. It also eliminates the upward torque exerted on the measuring unit by the pit heave, avoiding errors caused by the overall upward movement of the measuring unit; the magnetostrictive displacement sensor 7 is suitable for detecting and outputting the axial displacement of the magnetic ring 6 relative to the measuring rod 5 as the pit bottom heave amount; the intelligent acquisition module 9 is used to control the data acquisition of the magnetostrictive displacement sensor 7, process soil displacement data, and realize data communication with the remote monitoring platform.

[0018] The monitoring device in this embodiment aims to achieve accurate and continuous monitoring of the heave displacement of the soil at the bottom of the foundation pit. The overall structure of the device consists of an anchoring unit and a measuring unit, which are connected axially. The anchoring unit is primarily responsible for firmly fixing the device in the stable soil layer below the bottom of the foundation pit, providing a stable reference for the measuring unit. The measuring unit is responsible for sensing the relative displacement of the soil in real time and for data acquisition and transmission.

[0019] Specifically, the anchoring unit comprises a cone head 1, an umbrella-type support assembly 2, and a cylindrical guide rail 3. The cone head 1, located at the front end of the umbrella-type support assembly 2, is designed with a sharp shape to facilitate insertion into the stable soil layer below the bottom of the excavation pit for positioning during installation. For example, the cone head 1 can be a solid steel cone structure, driven into the soil layer by hammering or vibration. The umbrella-type support assembly 2 is detachably mounted at the front end of the cylindrical guide rail 3 and connected to the cone head 1. After the device is inserted into the stable soil layer, this assembly can expand radially outward to increase the contact area with the surrounding soil. For example, this assembly can consist of multiple hinged umbrella-type support metal plates, which remain closed during insertion into the soil and expand outward by external mechanical force or an internal mechanism after reaching a predetermined depth, thereby providing additional anchoring force and enhancing the pull-out stability of the device. The cylindrical guide rail 3 serves as the skeleton of the anchoring unit, with one end connected to the umbrella-type support assembly 2 and the other end movably connected to the measuring unit. The guide rail is typically made of high-strength metal materials, such as stainless steel or alloy steel pipes, to ensure its rigidity and stability in the soil. The cylindrical guide rail 3 not only supports the umbrella-type support assembly 2 but also provides a precise axial sliding path for the protective tube 4 in the measuring unit. In a preferred embodiment, the cone head 1 can be fixedly connected to the umbrella-type support assembly 2, which is connected to the cylindrical guide rail 3 via a detachable structure, such as a threaded connection. This allows for easy disassembly of the umbrella-type support assembly 2 and the cone head 1 after monitoring, enabling the cylindrical guide rail 3 and its upper portion to be recycled and reused. For subsequent work, only the lower-cost umbrella-type support assembly 2 and cone head 1 need to be replaced, improving the efficiency of the monitoring element and reducing costs.

[0020] The measuring unit comprises a protective tube 4, a measuring rod 5, a magnetic ring 6, a magnetostrictive displacement sensor 7, a flexible cable 8, and an intelligent acquisition module 9. The protective tube 4 is movably connected to the other end of the cylindrical guide rail 3, and houses the measuring rod 5 and the magnetostrictive displacement sensor 7. The protective tube 4 is configured with a torque-counteracting stroke S that slides along the axial direction of the cylindrical guide rail 3, allowing the protective tube 4 to slide relative to the cylindrical guide rail 3 during soil heave. For example, a certain radial gap A can be reserved between the protective tube 4 and the cylindrical guide rail 3, ensuring a low coefficient of friction between them. This allows the protective tube 4 to have upward sliding margin under the influence of the soil during heave, while the cylindrical guide rail 3 remains stationary in the stable soil layer. This eliminates the upward torque exerted on the measuring unit by the heave of the foundation pit, avoiding errors caused by the overall upward movement of the measuring unit.

[0021] The measuring rod 5 is installed inside the protective tube 4 and is fixedly connected to the cylindrical guide rail 3 of the anchoring unit. One end of the measuring rod 5 is firmly connected to the cylindrical guide rail 3, and the other end extends to the top of the protective tube 4. The measuring rod 5 serves as the waveguide wire of the magnetostrictive displacement sensor 7, providing a reference for the displacement measurement of the magnetic ring 6. The magnetic ring 6 is movably fitted onto the outer wall of the protective tube 4. The magnetic ring 6 is typically made of permanent magnet material, and its magnetic field can penetrate the wall thickness of the protective tube 4 and be sensed by the internal magnetostrictive displacement sensor 7. For example, the magnetic ring 6 can be designed as a ring structure, fitted onto the outside of the protective tube 4 by an elastic clamp or a low-friction material bushing, and move synchronously with the axial displacement of the protective tube 4. During measurement, its bottom surface contacts the soil plane, thus rising as the soil rises.

[0022] A magnetostrictive displacement sensor 7 is mounted on the pipe section, located inside the protective pipe 4, and is used to sense changes in the position of the magnetic ring 6. This sensor is typically arranged along the axial direction of the measuring rod 5. Its working principle is that when the magnetic field of the magnetic ring 6 interacts with the waveguide wire inside the sensor, a strain pulse is generated. By measuring the propagation time of this pulse, the axial displacement of the magnetic ring 6 relative to the measuring rod 5 is accurately calculated. This sensor is suitable for detecting and outputting the axial displacement of the magnetic ring 6 relative to the measuring rod 5, and using this displacement as the amount of pit bottom bulge. A flexible cable 8 is electrically connected to the magnetostrictive displacement sensor 7. This cable is used to transmit the displacement data acquired by the sensor and to power the sensor. To accommodate the relative sliding between the protective pipe 4 and the cylindrical guide rail 3, the cable needs to have sufficient flexibility or extensibility.

[0023] The intelligent acquisition module 9 is used to control the data acquisition and processing of soil displacement data from the magnetostrictive displacement sensor 7, and to achieve data communication with the remote monitoring platform. It can be located at the top of the acquisition tube section. This module integrates a microprocessor, memory, communication interface, and power management unit. For example, the intelligent acquisition module 9 can be configured as a low-power embedded system, periodically triggering the magnetostrictive displacement sensor 7 to perform measurements at a preset sampling frequency, and performing preliminary processing such as filtering and calibration on the acquired raw data. Then, the processed uplift data is uploaded to the remote monitoring platform via a wireless communication module (such as GPRS, LoRa, or NB-IoT). Preferably, the intelligent acquisition module 9 also includes a power supply component to provide power to the sensor and communication components.

[0024] The monitoring device in this embodiment effectively solves the measurement error problems caused by unstable reference and coupling movement between the device and the soil in traditional monitoring devices by firmly fixing the anchoring unit to a stable soil layer and allowing the protective tube 4 of the measuring unit to slide relative to the cylindrical guide rail 3. Therefore, this device can achieve continuous and high-precision monitoring of the bottom heave throughout the entire excavation process and provide reliable soil displacement data, providing data support for construction safety management and risk early warning in foundation pit engineering.

[0025] In a preferred embodiment, the umbrella-type support assembly 2, comprising an umbrella-type support plate and a spring connecting the umbrella-type support plate to the cylindrical guide rail 3, allows the umbrella-type support plate to be restrained and contracted by external force during insertion. After reaching a predetermined depth, it automatically expands radially under the action of the spring. The spring is a mechanical element that provides elastic restoring force, and its function is to store and release mechanical energy to achieve the automatic expansion of the umbrella-type support assembly 2. The spring can take various forms, such as compression springs, torsion springs, or leaf springs, and the specific selection can be optimized according to the structure of the umbrella-type support assembly 2, the required expansion force, and space constraints. For example, multiple compression springs can be evenly arranged between the umbrella-type support plate and the cylindrical guide rail 3. When the external constraint is released, the elastic potential energy of the spring is converted into kinetic energy, pushing the support plate to open outward. Alternatively, a torsion spring can be used to connect the pivot of the support plate, so that it tends to be in an expanded state when no external force is applied. During the process of inserting the device into the bottom of the foundation pit to stabilize the soil layer, the umbrella-type support assembly 2 needs to remain in a contracted state to reduce insertion resistance and smoothly pass through the narrow borehole. This "external constraint" utilizes the radial compression of the borehole wall to passively retract the support plate during insertion. Alternatively, a temporary mechanical locking mechanism, such as a shearable pin, a fusible wire, or a friction fit, can be designed to lock the support plate assembly in the retracted state before insertion and release it via a specific method after insertion. Once the device is precisely inserted to the predetermined depth of stable soil, the external constraint on the umbrella-shaped support plate is released. For example, when the support plate assembly disengages from the radial constraint of the borehole wall, or when the temporary locking mechanism is released, the elastic potential energy stored in the spring is released, driving the umbrella-shaped support plate to open radially outward. This automatic deployment mechanism ensures that the support plate can fully extend and form a stable contact with the surrounding soil, thereby achieving reliable anchoring.

[0026] Preferably, the umbrella-shaped support plate adopts a sheet-like, arc-shaped, or perforated reinforced structure, and its outer surface is provided with anti-slip ridges or a rough surface. The umbrella-shaped support plate can adopt a sheet-like structure, meaning its main body is a relatively flat plate or thin sheet, enhancing its resistance to bending and shearing through optimized material selection and thickness design. Alternatively, the support plate can adopt an arc-shaped structure, giving it a certain curvature, such as an arc or fan shape, to better conform to the shape of the soil at the bottom of the excavation pit, increasing the wrapping and contact tightness with the soil, while utilizing the bending stiffness of the arc structure itself to withstand greater radial support forces. Furthermore, the support plate can also adopt a perforated reinforced structure, by designing holes or a grid structure on the plate body, reducing weight while maintaining sufficient strength, and allowing some soil to be embedded within, forming a mechanical interlock, further enhancing the anchoring effect.

[0027] Meanwhile, to further enhance the anchoring effect, the outer surface of the umbrella-shaped support plate can be provided with anti-slip textures. These anti-slip textures can be a series of regular or irregular raised patterns formed by molding, casting, or machining, such as serrated, grid-like, wavy, or dotted protrusions. These textures can form a mechanical interlock with the soil when the soil applies upward or lateral shear force to the support plate, providing significant frictional resistance and effectively preventing the support plate from sliding relative to the soil. Alternatively, the outer surface of the support plate can be made into a rough surface, through sandblasting, etching, coating with high-friction materials, or using special surface treatment processes, giving it an uneven texture with multiple microscopic protrusions. This rough surface can increase the microscopic friction at the interface between the support plate and the soil, thereby improving the overall frictional resistance. Whether it is anti-slip textures or a rough surface, the purpose is to enhance the friction and interlocking between the support plate and the soil at both the microscopic and macroscopic levels by changing the physical properties of the interface between the support plate and the soil, thereby improving the pull-out and shear resistance of the device.

[0028] By employing the aforementioned technical solution, anti-slip textured or roughened surfaces are created on the outer surface of the umbrella-shaped support plate, and a sheet-like, arc-shaped, or hollowed-out reinforcing structure is used, significantly increasing the frictional resistance between the anchoring unit and the surrounding soil. When the soil at the bottom of the excavation pit heaves, these surface features can form a stronger mechanical interlock and frictional effect with the soil, effectively resisting the upward pull-out force and lateral shear force exerted by the soil on the support plate. This allows the anchoring unit to be more firmly fixed in the stable soil layer, effectively preventing the entire monitoring device from shifting upwards or tilting even under conditions of large soil heave force or complex soil conditions. This ensures that the displacement data collected by the magnetostrictive displacement sensor 7 accurately reflects the amount of heave at the bottom of the pit, avoiding measurement errors caused by anchoring failure, and greatly improving the reliability and accuracy of excavation pit heave monitoring.

[0029] In one embodiment, the protective pipe 4 is detachably connected along the axial direction by at least one collection pipe section and at least one extension pipe section, with adjacent pipe sections adopting a variable diameter nested structure and locked by fasteners. Specifically, the extension pipe section is adapted to be connected to the upper end and / or lower end of the collection pipe section; the protective pipe 4 is provided on the outer side of both the collection pipe section and the extension pipe section, and the end diameters of the collection pipe section and the extension pipe section are complementary and nested with each other. A locking nut is provided on the outside of the connection to ensure that the protective pipe 4 on the collection pipe section and the extension pipe section can undergo overall axial displacement when driven by the soil. The protective tube 4 of the acquisition section is externally fitted with the magnetic ring 6, and internally integrates the magnetostrictive displacement sensor 7, which is fixed to the internal measuring rod 5. The magnetostrictive displacement sensor 7 is connected to the intelligent acquisition module 9 via the flexible cable 8, and is used to sense the displacement of the magnetic ring 6 and process the data. It should be noted that the internal measuring rod 5 can also be extended by splicing.

[0030] The acquisition section is the part of the protective pipe 4 that contains the magnetostrictive displacement sensor 7, and its main function is to acquire displacement data. Extension sections are auxiliary sections used to increase the overall length of the protective pipe 4. Their number and length can be flexibly configured according to the actual burial depth requirements, and they can be connected to the acquisition section or other extension sections to achieve the required total length. Detachable connections between sections can be achieved through various mechanical structures, such as threaded connections, snap-fit ​​connections, flange connections, or pin connections. This makes the assembly, disassembly, and length adjustment of the protective pipe 4 convenient, greatly improving the device's on-site adaptability and maintenance efficiency. The sections are connected axially, ensuring that the connected protective pipe 4 maintains good straightness, thus not affecting the normal operation of the internal measuring rod 5 and the smooth sliding of the magnetic ring 6. Adjacent sections adopt a variable-diameter nested structure. That is, the outer diameter of one section's end is slightly smaller than the inner diameter of another section's end, allowing one section to be inserted into the interior of another. This nested structure provides excellent alignment, ensuring precise alignment of the pipe sections' axes after connection, effectively reducing radial sway and misalignment risks at the connection point, thereby significantly improving the overall structural stability and rigidity of the protective pipe 4. Through the above technical solution, the protective pipe 4 is designed to consist of a detachably connected acquisition pipe section and an extension pipe section, and uses a variable-diameter nested structure and fasteners for locking. This application effectively solves the problem of device adaptability caused by varying pit depths. It should be noted that the extension pipe section exposed above the pit surface can be destroyed.

[0031] In a preferred embodiment, the protective tube 4 is sleeved on the outside of the cylindrical guide rail 3, maintaining a radial gap A of 0.5mm-20mm, preferably 10mm. The radial gap A is filled with lubricating oil and covered with a waterproof membrane material to reduce sliding resistance and achieve a waterproof seal. The sleeved relationship between the protective tube 4 and the cylindrical guide rail 3 ensures that the measuring unit can perform axial displacement under the support of the anchoring unit. Maintaining a radial gap A of 0.5mm-20mm provides necessary space for the free sliding of the protective tube 4 relative to the cylindrical guide rail 3, avoiding jamming caused by direct contact or excessive tightness. This gap allows for certain manufacturing tolerances and installation deviations, while also providing space for subsequent filling with lubricating material, ensuring smooth sliding. The range of the radial gap A balances sliding freedom and structural stability; too small a gap can easily cause jamming, while too large a gap may cause the measuring unit to wobble, affecting accuracy.

[0032] Lubricating oil is filled into the radial gap A between the protective pipe 4 and the cylindrical guide rail 3. Its main function is to significantly reduce the coefficient of friction when the two slide relative to each other. The lubricating oil forms an oil film on the sliding surface, separating the metal surfaces, thereby reducing wear and heat generated by direct contact, and ensuring that the protective pipe 4 can respond to the small heave displacement of the soil with extremely low resistance. Lubricating oils with good water resistance, oxidation resistance, and viscosity-temperature properties, such as silicone oil, lithium-based grease, or synthetic lubricating oils, can be selected to adapt to the humid and temperature-changing environment inside the pit. The lubricating oil can be injected during installation or replenished periodically through the reserved oil filling hole to maintain its lubrication effect. A water-resistant membrane material is covered on the outside of the radial gap A filled with lubricating oil to form an effective physical barrier to prevent external moisture, mud, corrosive substances, etc. from penetrating into the sliding mechanism. The water-resistant membrane material can protect the lubricating oil from dilution, emulsification, or contamination, thereby extending the service life of the lubricating oil and maintaining its lubricating performance. Meanwhile, the waterproof membrane material also prevents hard particles such as silt from entering the sliding gap, thus preventing them from acting as abrasives to accelerate component wear or cause jamming. The waterproof membrane material can be made of flexible, corrosion-resistant, and wear-resistant polymer materials, such as nitrile rubber, fluororubber, or TPU (thermoplastic polyurethane), forming sealing rings, corrugated pipes, or flexible sleeves, which are fixed to the connection between the protective pipe 4 and the cylindrical guide rail 3 by compression, bonding, or snap-fitting. This design ensures that the measuring unit can slide smoothly and accurately relative to the anchoring unit during soil heave, effectively eliminating measurement errors caused by friction and environmental erosion, and guaranteeing the authenticity and validity of the pit heave monitoring data.

[0033] In this application, the magnetostrictive displacement sensor 7 and the intelligent acquisition module 9 are connected by a flexible cable 8. The flexible cable 8 is a TPU or PVC-coated spring wire to accommodate the sliding displacement between the protective tube 4 and the cylindrical guide rail 3. For example, the flexible cable 8 adopts a spring wire structure. A spring wire, also known as a spiral cable or coiled wire, has its internal conductor and external insulation layer spirally wound. This unique structure gives the wire excellent elasticity and resilience, allowing it to effectively extend when stretched and automatically return to its original length after the external force is released. This enables it to adapt to large dynamic displacements and effectively solves the technical problem that traditional cables are easily damaged by repeated stretching and bending when there is relative sliding displacement between the protective tube 4 and the cylindrical guide rail 3. The inherent elasticity of the spring wire allows it to freely adapt to the axial sliding of the protective tube 4 relative to the cylindrical guide rail 3 during soil heave, avoiding excessive stress concentration or breakage of the cable due to excessive displacement.

[0034] Example 2 Combination Figure 6As shown, this embodiment discloses a method for monitoring bottom heave during the entire excavation process of a foundation pit. It utilizes the aforementioned device for monitoring bottom heave and instability early warning during the entire excavation process of a foundation pit, and works collaboratively with a remote monitoring platform communicatively connected to the intelligent acquisition module 9. The method includes the following steps: First, before excavation of the foundation pit, boreholes are drilled into stable soil layers unaffected by pit heave; these boreholes are used to insert the cone 1 into the stable soil layer. This step aims to provide a stable mounting foundation for the monitoring device. Before excavation of the foundation pit or after excavation to a predetermined depth, the location of the stable soil layer below the bottom of the foundation pit is determined through geological survey. Subsequently, boreholes with the required diameter and depth are drilled at the predetermined locations using specialized drilling equipment.

[0035] Then, lubricating oil is injected at the connection between the cylindrical guide rail 3 and the protective pipe 4 to fill the connection tightly with lubricating oil, and a water-proof membrane material is installed; bagged cement is wrapped around the bottom of the cone head 1 of the anchoring unit, in preparation for the anchoring unit and the bagged cement to be lowered to the bottom of the buried hole simultaneously. Install the bottom cone 1 to the cylindrical guide rail 3, then inject lubricating oil at the connection between the cylindrical guide rail 3 and the protective pipe 4 to fill the connection tightly with lubricating oil, and then install a water-proof membrane material such as a sealing ring to prevent mud and sand from flowing in; connect and lower the anchoring unit and measuring unit in sequence; by controlling the lowering depth, make the magnetic ring 6 accurately located at the preset excavation depth of the foundation pit as the initial monitoring reference point, and make the uppermost end of the measuring unit protrude more than 0.5m above the initial ground surface before the foundation pit is excavated; for example, the installation sequence can be anchoring unit - extension pipe section - acquisition pipe section - extension pipe section, and make the uppermost acquisition pipe section protrude more than 0.5m above the initial ground surface before the foundation pit is excavated.

[0036] Next, the end of the cone 1 is reinforced by grouting through an in-hole grouting pipe, which improves end stability after solidification. The intelligent acquisition module 9 controls the measurement unit to read the initial value and perform zero calibration, and collects the displacement output of the magnetostrictive displacement sensor 7 during excavation. The axial displacement of the magnetic ring 6 relative to the measuring rod 5 is calculated to obtain the pit bottom heave. While collecting the displacement signal, the ambient temperature and construction vibration parameters are acquired simultaneously. The built-in error compensation model is used to correct the original heave data in real time to eliminate measurement deviations caused by environmental interference. The intelligent acquisition module 9 uploads the compensated heave monitoring data to a remote monitoring platform for trend prediction and early warning.

[0037] In this embodiment, temperature and / or vibration parameters are also collected, and an error compensation model is used to compensate for the displacement results. This step aims to improve the accuracy of the uplift measurement data and eliminate the interference of environmental factors (such as temperature changes and construction vibrations) on the sensor output. Simultaneously with displacement data acquisition, the intelligent acquisition module 9 or its built-in sensing unit synchronously collects ambient temperature and / or construction vibration parameters. Temperature changes may cause thermal expansion and contraction of sensor materials, affecting measurement accuracy; construction vibrations may cause instantaneous fluctuations in sensor signals or structural resonance. To correct these errors, the system utilizes a pre-established error compensation model. This model can be built based on empirical data, physical principles, or machine learning algorithms. It takes the real-time collected temperature and / or vibration parameters as input, calculates correction values ​​for the displacement measurement results, and thus compensates for the original displacement data to obtain a more accurate value for the pit bottom uplift.

[0038] Specifically, the intelligent acquisition module 9 is used to receive sensor data, control waveguide wire pulse emission and response acquisition, process magnetic ring 6 displacement data, and transmit it to a remote monitoring platform via a network. Simultaneously, the intelligent acquisition module 9 integrates a temperature sensing unit, a vibration sensing unit, and a processor unit. An error compensation model is embedded in the processor unit to correct measurement errors caused by ambient temperature and construction vibration in real time. The temperature sensing unit in the intelligent acquisition module 9 is used to acquire real-time temperature information of the environment in which the device is located; for example, it can use a thermistor, thermocouple, or integrated temperature sensor. The vibration sensing unit is used to monitor the vibration at the construction site in real time; for example, it can use an accelerometer or piezoelectric sensor to capture vibration intensity, frequency, and other characteristics. The processor unit is the core computing component of the intelligent acquisition module 9; for example, it can use a high-performance microcontroller (MCU) or embedded processor to perform tasks such as data acquisition, signal processing, model calculation, and communication protocol management. These built-in units work together to provide the necessary multi-source real-time data and computing power for subsequent environmental interference calibration.

[0039] The core of the error compensation model embedded in the processor unit lies in its ability to comprehensively consider the combined effects of multiple environmental factors on the measurement results, rather than simply compensating for a single factor independently. Its goal is to establish an accurate mapping relationship that correlates the original measurement values ​​with environmental parameters (such as temperature and vibration), thereby predicting and eliminating measurement deviations caused by these environmental factors.

[0040] The model building process includes: S81. Multi-source sample data acquisition: Before the device is buried in the foundation pit, an environmental interference sample library needs to be constructed by simulating the environment. The intelligent acquisition module performs time-frequency domain feature extraction on the acquired raw signals. First, the original acceleration sequence obtained by the vibration sensing unit is used. The process involves calculating the effective value of vibration that reflects the overall background energy of the environment. ,in The total number of sample points within a single sampling window, and the peak factor reflecting the mechanical impact characteristics. ,in The maximum absolute peak value of the vibration signal within the sampling period; simultaneously, for the time-domain signal Perform a Fast Fourier Transform to obtain the spectrum function, as shown in the formula: ; in For time variables, To meet The imaginary unit, Angular frequency; Extract the angular frequency corresponding to the maximum amplitude of the spectral function. Calculate the main frequency To identify and locate frequency interference points that may cause resonance in magnetostrictive waveguide wires; Finally, the known calibration standard displacement in the laboratory will be determined. Sensor original displacement value Real-time ambient temperature The extracted vibration feature sets are then fused to construct the input feature vector. And based on the calculated measured deviation value An offline training database was established to serve as the prediction label. S82. Model Construction and Optimization Based on SVR: The SVR algorithm is used to model the nonlinear error law and solve the measurement deviation problem caused by the coupling of temperature drift and vibration noise. Using the radial basis function kernel to transform the low-dimensional input feature vector Mapping to a high-dimensional feature space to handle complex environmental interference, the formula is: ; in Parameters for controlling the mapping range of kernel functions, Support vector samples determined during the offline training phase of the model; Furthermore, by combining grid search Cross-validation, for the penalty factor Insensitive loss coefficient and kernel function parameters The formula for joint optimization is: ; in , For Lagrange multipliers, The total number of support vectors selected. This is the bias term; the model establishes the relationship between environmental characteristics and prediction bias. The precise mapping relationship between them; S83. Real-time compensation for uplift displacement: The trained and optimized model parameters are embedded in the embedded processor of the intelligent acquisition module; during the monitoring of the foundation pit excavation, the intelligent acquisition module synchronously reads the original displacement signals from the sensors. Real-time temperature and the vibration characteristic group after processing by the processor unit Input the real-time feature vector into the SVR decision function. The comprehensive prediction deviation under the current environmental conditions is calculated using a formula. The corrected output yields the final high-precision result of the pit bottom heave displacement. .

[0041] Through the above technical solution, this application effectively solves the problem of nonlinear interference caused by changes in ambient temperature and construction vibration during the excavation of the foundation pit, affecting the measurement accuracy of the magnetostrictive displacement sensor 7. The temperature sensing unit and vibration sensing unit built into the intelligent acquisition module 9 can capture multi-source environmental information in real time, which is then preliminarily processed by the processor unit. Based on this, by embedding a multi-source environmental interference calibration model based on SVR in the processor unit, this model can establish a precise nonlinear mapping relationship between environmental characteristics and measurement deviations using training samples obtained through pre-calibration experiments. During actual monitoring, the intelligent acquisition module 9 can acquire the original displacement signal, ambient temperature, and vibration characteristics in real time and input them into the SVR model to quickly calculate the comprehensive prediction deviation under the current environment, thereby providing real-time, high-precision compensation for the original displacement data. This method avoids the limitations of traditional linear compensation models, especially under the coupling effect of temperature drift and vibration noise, and can significantly improve the accuracy and reliability of pit bottom heave monitoring data, providing more accurate data support for the safe construction of foundation pit projects.

[0042] In another preferred embodiment, the intelligent acquisition module 9 uploads the compensated uplift monitoring data to a remote monitoring platform for trend prediction and early warning. This step enables remote and intelligent management and application of monitoring data. The high-precision uplift monitoring data, after error compensation processing, is transmitted to the remote monitoring platform in real time or at regular intervals via the communication unit built into the intelligent acquisition module 9 (e.g., GPRS, 4G / 5G, LoRa, etc. wireless communication modules). Upon receiving the data, the remote monitoring platform stores and visualizes it, and uses its integrated analysis algorithms to perform trend analysis on historical data to predict future uplift development trends. When the uplift amount or its rate of change exceeds a preset threshold, the platform automatically triggers an early warning mechanism, promptly issuing warning information to relevant management personnel via SMS, email, APP notifications, etc., thereby providing timely and effective decision support for the safety management of the foundation pit project.

[0043] Specifically, the remote monitoring platform is used to adjust and manage the sampling frequency, data content, and communication status of the measurement unit in real time. Simultaneously, the remote monitoring platform integrates a finite element model-driven foundation pit stability proxy model to predict and warn of instability risks during the foundation pit excavation process. The model establishment process includes: S91. Establishment of the Finite Element Model: Based on the actual geological report and support design, a three-dimensional finite element model reflecting the unloading characteristics of the foundation pit is established; considering the temporal nature of the foundation pit excavation, it is divided according to the construction conditions. The excavation sequence is as follows: Static soil parameters affecting the stability of the foundation pit are selected, including the soil elastic modulus. Cohesion internal friction angle The overall excavation sequence characterizing the project scale Current construction excavation sequence And the amount of bottom heave of the foundation pit corresponding to the current step sequence. Together as model input variables Assume random variables Following a specific probability distribution, it is generated using Latin hypercube sampling. A combination of static soil parameters was used to extract the state variables corresponding to each step of the process through finite element simulation. Together, we construct the input sample set ; each group of samples Input the finite element model to simulate the entire excavation process and extract the sequence of each excavation step. Deformation of key nodes that induce foundation pit instability This forms the corresponding output response sequence. This leads to the construction of a training sample database containing a "parameter-state-response" mapping relationship. ; S92. Establishing a spatiotemporal surrogate model for a CNN-BiLSTM hybrid neural network: Establishing the input variables The deformation time sequence of the entire foundation pit output The model uses a nonlinear mapping relationship between the input layer, a CNN feature extraction layer, a BiLSTM temporal prediction layer, and a fully connected output layer to replace the time-consuming finite element calculation. This represents the input variable under the t-th excavation step; Using size convolution kernel For the input vector Perform sliding convolution to extract feature maps ; As the convolution kernel traverses the entire The generated complete feature sequence ,in For input vectors The feature dimension is set to 6 here; convolutional feature maps Downsampling is performed: Max pooling is used as the specific implementation for feature dimensionality reduction to retain the most significant mechanical feature information and generate the output feature vector. and will As a BiLSTM unit in the Input at any moment; Forward propagation follows the excavation steps from arrive Calculate and capture the cumulative deformation effect of the foundation pit excavation, i.e. Specifically, it can be broken down as follows: ; in, Represents the Gate of Oblivion Represents the input gate. Represents the state of the candidate unit. Represents cell state update. Represents the output gate. Represents a hidden state. It is the Sigmoid activation function. The hyperbolic tangent activation function is used. , , , and , , , These are the corresponding weight matrix and bias vector, respectively; Represents the Hadamard product (element-wise multiplication of a matrix / vector).

[0044] Backpropagation from excavation steps Calculation up to 1 captures the feedback effect of subsequent excavation and unloading on the current state, i.e. The formula expansion structure is the same as above, only the time direction is reversed; BiLSTM in step sequence The final output is from the forward hidden state. and backward hidden state It is pieced together, that is Finally, the predicted output is generated by mapping the deformation value to the fully connected layer. ; S93, Using Mean Squared Error (MSE) as the Loss Function To measure the predicted deformation value The actual value calculated by finite element method The difference between them is minimized by using the loss function to drive the update of network weights using the backpropagation algorithm; where, Represents the number of samples; S94. Reliability Risk Calculation: In the remote monitoring platform, a large-scale stochastic simulation is performed using a trained CNN-BiLSTM surrogate model to calculate the failure probability of the foundation pit under the current excavation conditions; assuming the foundation pit is in the [missing information]th ... The deformation safety warning threshold under the excavation sequence is Define function ,when This indicates that the foundation pit is in a reliable state, when This indicates that the foundation pit is in a failed state; generation A set of random parameter samples that conform to the statistical characteristics of the soil in the field. ,in Inputting the CNN-BiLSTM surrogate model allows for rapid prediction of the corresponding deformation response. The statistical prediction results satisfy Number of failure samples Calculate the failure probability ,in ( ) is an indicator function that takes the value 1 when the condition is met and 0 otherwise. S95. Current Risk Level Warning: Based on the calculated failure probability. Combined with reliability indicators Real-time early warning of the current excavation sequence of the foundation pit. It is the inverse function of the standard normal distribution function; set the target reliability threshold. ,like The remote monitoring platform automatically triggers an early warning signal, prompting the site to immediately stop excavation and take reinforcement measures; if If so, it indicates that the foundation pit is in a safe state; S96. Risk Classification and Early Warning for the Next Construction Step: After completing the risk classification and early warning for the current construction step, the remote monitoring platform will automatically predict the risk situation of the next construction step; known geological parameters include soil elastic modulus E, cohesion c, and internal friction angle. Total Excavation Sequence The time step sequence is updated to , before Step bulge sequence [ , , ], obtained through polynomial fitting extrapolation Excavation sequence uplift volume ; Obtain the input variables under the (t+1)th excavation step sequence After reliability risk calculation, the following results were obtained. And in the risk classification and early warning of the next construction step, it is obtained and with reliability threshold Make judgments on early warning information.

[0045] The aforementioned monitoring methods first ensured the reliable anchoring of the monitoring device in the stable soil layer at the bottom of the foundation pit, providing a stable benchmark for subsequent accurate measurements. During excavation, the system continuously and in real-time collected data on the heave at the bottom of the pit. Through an error compensation model integrating environmental parameters (such as temperature and vibration), the system effectively eliminated the impact of external interference on measurement accuracy, significantly improving the accuracy and reliability of the heave data. Finally, the intelligent acquisition module 9 uploaded the compensated high-precision data to the remote monitoring platform, enabling prediction and risk warning of the foundation pit heave trend. This transformed the single device's measurement capability into a complete, intelligent, and high-precision full-process foundation pit safety monitoring and early warning system, providing a scientific basis and decision support for the safe construction of foundation pit projects.

[0046] Through the aforementioned technical solutions, the remote monitoring platform not only enables flexible remote management of the sampling frequency, data content, and communication status of the measurement units, optimizing data acquisition efficiency and system operational stability, but more importantly, by integrating a CNN-BiLSTM hybrid neural network surrogate model driven by a finite element model, it achieves forward-looking prediction and graded early warning of instability risks throughout the entire foundation pit excavation process. This surrogate model can efficiently simulate the deformation response of the foundation pit under different soil parameters and excavation steps, overcoming the drawback of long computation time in traditional finite element calculations, making large-scale stochastic simulation possible. Through real-time calculation of failure probability and reliability indicators, the platform can promptly detect potential instability risks and automatically trigger early warnings when the risk reaches a preset threshold, guiding timely preventative reinforcement measures on-site, thereby effectively avoiding foundation pit engineering accidents and significantly improving the safety and intelligent management level of foundation pit engineering. This forward-looking risk prediction capability, combined with the real-time compensation for environmental interference by the intelligent acquisition module 9, constitutes a more comprehensive and reliable foundation pit heave monitoring and risk management system.

[0047] It should be understood that the above are merely preferred embodiments of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions that fall within the scope of the present invention are within the scope of protection of the present invention.

[0048] The accompanying drawings used in the above description of the embodiments only illustrate certain embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

Claims

1. A device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, characterized in that, include: An anchoring unit and a measuring unit are connected axially front to back; wherein... The anchoring unit includes a cone head, a radially expandable umbrella-type support assembly, and a cylindrical guide rail. The cone head is located at the front end of the umbrella-type support assembly and is used to insert into the stable soil layer below the bottom of the foundation pit for positioning. The umbrella-type support assembly is located at the front end of the cylindrical guide rail and is connected to the cone head. It can be radially expanded after being inserted into the stable soil layer to increase the contact area with the surrounding soil. The measuring unit includes a protective tube movably connected to the other end of the cylindrical guide rail, a measuring rod disposed inside the protective tube and fixedly connected to the cylindrical guide rail, a magnetic ring movably sleeved on the outer wall of the protective tube, a magnetostrictive displacement sensor located inside the protective tube for sensing changes in the position of the magnetic ring, a flexible cable electrically connected to the magnetostrictive displacement sensor, and an intelligent acquisition module; wherein, the protective tube is configured to have a torque-counteracting travel amount that slides along the axial direction of the cylindrical guide rail, so that the protective tube can slide relative to the cylindrical guide rail during soil heave, and eliminate the upward torque exerted on the measuring unit by the heave of the foundation pit, avoiding errors caused by the overall upward movement of the measuring unit; The magnetostrictive displacement sensor is adapted to detect and output the axial displacement of the magnetic ring relative to the measuring rod as the amount of pit bottom heave. The intelligent acquisition module is used to control the data acquisition of the magnetostrictive displacement sensor, process the soil displacement data, and realize data communication with the remote monitoring platform.

2. The device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, as described in claim 1, is characterized in that... The umbrella-type support assembly is connected to the cylindrical guide rail via a spring, so that the umbrella-type support assembly is constrained and closed by external force during insertion, and automatically expands radially under the action of the spring after reaching a predetermined depth.

3. The device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, as described in claim 2, is characterized in that... The umbrella-type support assembly includes an umbrella-type support plate, which adopts a sheet-like, arc-shaped, or hollowed-out reinforced structure, and has anti-slip ridges or rough surfaces on its outer surface to increase the frictional resistance with the soil and enhance the device's pull-out and shear stability.

4. The device for monitoring and instability early warning of pit bottom heave throughout the entire excavation process as described in claim 1, characterized in that, The measurement unit includes a data acquisition section and at least one set of extension sections, the extension sections being adapted to be connected to the upper and / or lower ends of the data acquisition section; The protective pipe is provided on the outside of both the acquisition pipe section and the extension pipe section. The end diameters of the acquisition pipe section and the extension pipe section are complementary and nested together. A locking nut is provided on the outside of the connection to ensure that the protective pipe on the acquisition pipe section and the extension pipe section can be axially displaced as a whole when driven by the soil. The magnetic ring is provided on the outside of the protective tube of the acquisition tube section, and the magnetostrictive displacement sensor is integrated inside. The magnetostrictive displacement sensor is connected to the intelligent acquisition module through the flexible cable and is used to sense the displacement of the magnetic ring and process the data.

5. The device for monitoring and instability early warning of pit bottom heave throughout the entire excavation process as described in claim 1, characterized in that, The protective tube is sleeved on the outside of the cylindrical guide rail and maintains a radial gap of 0.5mm-20mm. The radial gap is filled with lubricating oil and covered with a water-proof membrane material to reduce sliding resistance and achieve waterproof sealing.

6. The device for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, as described in claim 1, is characterized in that... The flexible cable is a spring conductor to accommodate the sliding displacement of the protective tube and the cylindrical guide rail.

7. A method for monitoring bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, characterized in that, The device for monitoring and instability early warning of pit bottom heave during the entire excavation process as described in any one of claims 1 to 6, and the remote monitoring platform communicatively connected to the intelligent acquisition module, comprises the following steps: S1. Drill holes before the foundation pit is excavated to ensure that the drilling depth reaches the deep stable soil layer that is not affected by the excavation of the foundation pit. S2. Inject lubricating oil at the connection between the cylindrical guide rail and the protective tube to fill the connection tightly with lubricating oil, and install a water-proof membrane material. S3. Wrap the bottom of the cone of the anchoring unit with a bag of cement, and prepare to lower the anchoring unit and the bag of cement to the bottom of the hole at the same time. S4. Connect the anchoring unit and the measuring unit in sequence and lower them synchronously. By controlling the lowering depth, make the magnetic ring accurately located at the preset depth of the foundation pit excavation surface, as the initial monitoring reference point. S5. Make the uppermost end of the measuring unit protrude more than 0.5m above the initial ground surface before the foundation pit is excavated; S6. After the cone is lowered into place, grout is injected into the bottom of the hole through the grouting pipe pre-installed in the hole, so that the cone end and the bagged cement are fixed with the stable soil layer, and the bottom anchor end is locked. S7. The intelligent acquisition module controls the measurement unit to read the initial value and perform zero calibration, and acquires the displacement output of the magnetostrictive displacement sensor during the excavation process, and calculates the axial displacement of the magnetic ring relative to the measuring rod to obtain the amount of pit bottom heave. S8. While acquiring displacement signals, the ambient temperature and construction vibration parameters are acquired simultaneously. The built-in error compensation model is used to correct the original uplift data in real time, eliminating measurement deviations caused by environmental interference. S9. The intelligent acquisition module uploads the compensated uplift monitoring data to the remote monitoring platform for trend prediction and early warning.

8. The method for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, as described in claim 7, is characterized in that... The intelligent acquisition module is used to receive sensor data, control waveguide wire pulse emission and response acquisition, process magnetic ring displacement data, and transmit it to a remote monitoring platform via network. Simultaneously, the intelligent acquisition module incorporates a temperature sensing unit, a vibration sensing unit, and a processor unit. Furthermore, an error compensation model is embedded in the processor unit to correct measurement errors caused by ambient temperature and construction vibration in real time. The model establishment process includes: S81. Multi-source sample data acquisition: Before the device is buried in the foundation pit, an environmental interference sample library needs to be constructed by simulating the environment. The intelligent acquisition module performs time-frequency domain feature extraction on the acquired raw signals. First, the original acceleration sequence obtained by the vibration sensing unit is used. The process involves calculating the effective value of vibration that reflects the overall background energy of the environment. ,in The total number of sample points within a single sampling window, and the peak factor reflecting the mechanical impact characteristics. ,in The maximum absolute peak value of the vibration signal within the sampling period; simultaneously, for the time-domain signal Perform a Fast Fourier Transform to obtain the spectrum function, as shown in the formula: ； in For time variables, To meet The imaginary unit, Angular frequency; Extract the angular frequency corresponding to the maximum amplitude of the spectral function. Calculate the main frequency To identify and locate frequency interference points that may cause resonance in magnetostrictive waveguide wires; Finally, the known calibration standard displacement in the laboratory will be determined. Sensor original displacement value Real-time ambient temperature The extracted vibration feature sets are then fused to construct the input feature vector. And based on the calculated measured deviation value An offline training database was established to serve as the prediction label. S82. Model Construction and Optimization Based on SVR: The SVR algorithm is used to model the nonlinear error law and solve the measurement deviation problem caused by the coupling of temperature drift and vibration noise. Using the radial basis function kernel to transform the low-dimensional input feature vector Mapping to a high-dimensional feature space to handle complex environmental interference, the formula is: ; in Parameters for controlling the mapping range of kernel functions, Support vector samples determined during the offline training phase of the model; Furthermore, by combining grid search Cross-validation, for the penalty factor Insensitive loss coefficient and kernel function parameters The formula for joint optimization is: ; in , For Lagrange multipliers, The total number of support vectors selected. This is the bias term; the model establishes the relationship between environmental characteristics and prediction bias. The precise mapping relationship between them; S83. Real-time compensation for uplift displacement: The trained and optimized model parameters are embedded in the embedded processor of the intelligent acquisition module; during the monitoring of the foundation pit excavation, the intelligent acquisition module synchronously reads the original displacement signals from the sensors. Real-time temperature and the vibration characteristic group after processing by the processor unit Input the real-time feature vector into the SVR decision function. The comprehensive prediction deviation under the current environmental conditions is calculated using a formula. The corrected output yields the final high-precision result of the pit bottom heave displacement. .

9. The method for monitoring pit bottom heave and providing early warning of instability throughout the entire excavation process of a foundation pit, as described in claim 7, is characterized in that... The remote monitoring platform is used to adjust and manage the sampling frequency, data content, and communication status of the measurement units in real time. Simultaneously, the remote monitoring platform integrates a finite element model-driven foundation pit stability proxy model to predict and warn of instability risks during the foundation pit excavation process. The model establishment process includes: S91. Establishment of the Finite Element Model: Based on the actual geological report and support design, a three-dimensional finite element model reflecting the unloading characteristics of the foundation pit is established; considering the temporal nature of the foundation pit excavation, it is divided according to the construction conditions. The excavation sequence is as follows: Static soil parameters affecting the stability of the foundation pit are selected, including the soil elastic modulus. Cohesion internal friction angle The overall excavation sequence characterizing the project scale Current construction excavation sequence And the amount of bottom heave of the foundation pit corresponding to the current step sequence. Together as model input variables Assume random variables Following a specific probability distribution, it is generated using Latin hypercube sampling. A combination of static soil parameters was used to extract the state variables corresponding to each step of the process through finite element simulation. Together, we construct the input sample set ; each group of samples Input the finite element model to simulate the entire excavation process and extract the sequence of each excavation step. Deformation of key nodes that induce foundation pit instability This forms the corresponding output response sequence. This leads to the construction of a training sample database containing a "parameter-state-response" mapping relationship. ; S92. Establishing a spatiotemporal surrogate model for a CNN-BiLSTM hybrid neural network: Establishing the input variables The deformation time sequence of the entire foundation pit output The model uses a nonlinear mapping relationship between the input layer, a CNN feature extraction layer, a BiLSTM temporal prediction layer, and a fully connected output layer to replace the time-consuming finite element calculation. This represents the input variable under the t-th excavation step; Using size convolution kernel For the input vector Perform sliding convolution to extract feature maps ; As the convolution kernel traverses the entire The generated complete feature sequence ,in For input vectors The feature dimension is set to 6 here; convolutional feature maps Downsampling is performed: Max pooling is used as the specific implementation for feature dimensionality reduction to retain the most significant mechanical feature information and generate the output feature vector. and will As a BiLSTM unit in the Input at any moment; Forward propagation follows the excavation steps from arrive Calculate and capture the cumulative deformation effect of the foundation pit excavation, i.e. ; Backpropagation from excavation steps Calculation up to 1 captures the feedback effect of subsequent excavation and unloading on the current state, i.e. The formula expansion structure is the same as above, only the time direction is reversed; BiLSTM in step sequence The final output is from the forward hidden state. and backward hidden state It is pieced together, that is Finally, the predicted output is generated by mapping the deformation value to the fully connected layer. ; S93. Using mean squared error as a loss function to measure predicted deformation values. The actual value calculated by finite element method The difference between them is minimized by using the backpropagation algorithm to drive the update of network weights; S94. Reliability Risk Calculation: In the remote monitoring platform, a large-scale stochastic simulation is performed using a trained CNN-BiLSTM surrogate model to calculate the failure probability of the foundation pit under the current excavation conditions; assuming the foundation pit is in the [missing information]th ... The deformation safety warning threshold under the excavation sequence is Define function ,when This indicates that the foundation pit is in a reliable state, when This indicates that the foundation pit is in a failed state; generation A set of random parameter samples that conform to the statistical characteristics of the soil in the field. ,in Inputting the CNN-BiLSTM surrogate model allows for rapid prediction of the corresponding deformation response. The statistical prediction results satisfy Number of failure samples Calculate the failure probability ; S95. Current Risk Level Warning: Based on the calculated failure probability. Combined with reliability indicators Real-time early warning of the current excavation sequence of the foundation pit. It is the inverse function of the standard normal distribution function; set the target reliability threshold. ,like The remote monitoring platform automatically triggers an early warning signal, prompting the site to immediately stop excavation and take reinforcement measures; if If so, it indicates that the foundation pit is in a safe state; S96. Risk Classification and Early Warning for the Next Construction Step: After completing the risk classification and early warning for the current construction step, the remote monitoring platform will automatically predict the risk situation of the next construction step; known geological parameters include soil elastic modulus E, cohesion c, and internal friction angle. Total Excavation Sequence The time step sequence is updated to , before Step bulge sequence [ , , ], obtained through polynomial fitting extrapolation Excavation sequence uplift volume ; Obtain the input variables under the (t+1)th excavation step sequence After reliability risk calculation, the following results were obtained. And in the risk classification and early warning of the next construction step, it is obtained and with reliability threshold Make judgments on early warning information.

Images