Deep foundation pit steel support axial force non-intrusive monitoring device and intelligent monitoring method

Abstract

The invention provides a deep foundation pit steel support axial force non-intrusive monitoring device and an intelligent monitoring method. The monitoring device comprises a piezoelectric ceramic actuator, an MEMS accelerometer, a controller and a self-adaptive bridge type mounting mechanism. Wherein the self-adaptive bridge type mounting mechanism comprises a fixed disc and an adsorption seat, the fixed disc is used for mounting a piezoelectric ceramic actuator or an MEMS accelerometer, and the adsorption seat is used for being in adsorption connection with a steel support; for an existing constructed structure, the monitoring device can be installed at monitoring points where sensors fail or key positions where the monitoring points are not distributed to conduct remedial monitoring or supplementary monitoring. In the steel support dismantling stage, the monitoring device can be dismantled in a lossless mode and used for other follow-up engineering projects, and cyclic reuse is achieved. In addition, the connecting rod is hinged between the fixing disc and the adsorption seats, the span between the two adsorption seats can be changed by adjusting the angle of the connecting rod in the using process, and therefore the device can be attached to the surfaces of steel supports with different curvature radiuses, and universality is good.

Description

Technical Field

[0001] This invention belongs to the field of deep foundation pit monitoring technology, specifically relating to a non-intrusive monitoring device and intelligent monitoring method for the axial force of steel supports in deep foundation pits. Background Technology

[0002] With the continuous expansion of the depth and breadth of urban underground space, deep foundation pit engineering has become a common procedure in urban construction. During deep foundation pit construction, the supporting structure is the core load-bearing system for maintaining pit stability, controlling deformation, and ensuring construction safety; it typically consists of a retaining structure and internal supports. Among these, the axial force of the internal supports is one of the most critical monitoring indicators for judging the stability of the retaining structure, the uniformity of stress distribution, and the existence of potential safety hazards. Currently, the mainstream monitoring method for the axial force of internal supports (including steel and concrete supports) mainly uses invasive sensors, such as vibrating wire sensors, embedded rebar gauges, and welded strain gauges. Meanwhile, at the data analysis level, to achieve intelligent monitoring and risk warning, the industry is increasingly adopting advanced data-driven models, such as BP neural networks, to analyze massive amounts of monitoring data to predict foundation pit deformation or structural conditions.

[0003] While the aforementioned monitoring technologies based on pre-deployment and invasive installation have been recognized in engineering for their accuracy, they still have the following shortcomings in terms of the flexibility, robustness, and economy required for smart monitoring: ① Whether pre-embedded or welded, sensors must be installed before the structure is subjected to stress and become a permanent part of the structure (or its accessories); this installation method strongly interferes with the construction process, and once installed, it is irreversible. ② Sensors are easily damaged in harsh construction environments (such as welding, vibration, and impact), resulting in permanent loss of monitoring data. ③ Existing structures cannot be monitored; for example, if an abnormal stress is found on a support without monitoring points during construction, or if a monitoring point that has been deployed fails, existing technologies cannot provide any remedial monitoring methods. ④ Sensors are for single use only and cannot be disassembled and reused, resulting in high monitoring lifecycle costs. ⑤ Predictive models such as BP neural networks are highly sensitive to the quality of input data. When the performance of pre-embedded sensors drifts due to welding or collisions, producing seemingly correct but actually incorrect data, the model will be "poisoned" and thus output incorrect predictions. When sensors completely fail during construction (such as when cables are cut) or are not deployed due to lack of flexibility, the BP neural network will fall into a state of no data. Due to the lack of input from key nodes, the reliability of the prediction results is low. Summary of the Invention

[0004] This invention provides a flexible, non-invasive monitoring device for the axial force of steel supports in deep foundation pits, which overcomes the aforementioned problems in existing technologies. The monitoring device abandons the traditional "pre-embedded" or "welded" approach, instead employing a non-invasive coupling installation method to achieve an adsorption connection with the steel support. This offers high installation flexibility; after a project is completed, the entire device can be disassembled and reused, reducing overall costs. Furthermore, the monitoring device is applicable to steel support surfaces with different radii of curvature, demonstrating good versatility. Correspondingly, this invention also provides a smart monitoring method using this non-invasive monitoring device for the axial force of steel supports in deep foundation pits. This method actively excites the vibration of the steel support through a piezoelectric ceramic actuator, receives the vibration response signal through a MEMS accelerometer, processes the data with a controller to calculate the corresponding axial force value, and then sends it to a BP neural network prediction model to predict future axial force values. This method performs "refinement" at the source of data generation, ensuring that every data point uploaded to the remote monitoring platform is "clean data" verified by a physical model, thus improving the accuracy of the prediction results.

[0005] The technical solution of this application for the device is as follows: a non-invasive monitoring device for axial force of steel support in deep foundation pits, comprising a piezoelectric ceramic actuator, a MEMS accelerometer, a controller, and an adaptive bridge mounting mechanism; the piezoelectric ceramic actuator and the MEMS accelerometer are connected to the steel support through the adaptive bridge mounting mechanism; the adaptive bridge mounting mechanism includes a mounting plate and two adsorption seats symmetrically arranged on both sides of the mounting plate; a connecting rod is hinged between the mounting plate and the adsorption seats; the piezoelectric ceramic actuator and the MEMS accelerometer are electrically connected to the controller respectively; in use, the adsorption seats are adsorbed and connected to the outer surface of the steel support, and the piezoelectric ceramic actuator and the MEMS accelerometer are fixed on the corresponding mounting plate; the piezoelectric ceramic actuator abuts against the outer surface of the steel support to apply mechanical excitation to the steel support; the MEMS accelerometer contacts the outer surface of the steel support to collect the transient vibration response time-domain signal generated by the steel support under mechanical excitation and send it to the controller.

[0006] Compared with existing technologies, the non-intrusive monitoring device for axial force of steel support in deep foundation pits in this application abandons the traditional "pre-embedded" or "welded" approach and instead adopts a non-intrusive coupling installation method to achieve adsorption connection with the steel support, which has high installation flexibility and low overall cost. Specifically, the monitoring device of this application includes a piezoelectric ceramic actuator, a MEMS accelerometer, a controller, and an adaptive bridge mounting mechanism. The adaptive bridge mounting mechanism includes a fixed plate and an adsorption base. The fixed plate is used to mount the piezoelectric ceramic actuator or MEMS accelerometer, and the adsorption base is used to adsorb and connect with the steel support, thereby temporarily fixing the monitoring device of this application to the surface of the steel support. For existing structures that have already been constructed, the monitoring device of this application can be installed at monitoring points where sensors have failed or at critical locations where no sensors have been installed for remedial or supplementary monitoring. During the steel support removal phase, the monitoring device can be removed without damage and used for subsequent engineering projects, achieving cyclical reuse. Furthermore, a connecting rod is hinged between the fixed plate and the adsorption base. During use, the span between the two adsorption bases can be changed by adjusting the angle of the connecting rod, thereby fitting steel support surfaces with different radii of curvature, demonstrating good versatility.

[0007] As an optimization, in the aforementioned non-intrusive monitoring device for axial force of steel supports in deep foundation pits, the piezoelectric ceramic actuator includes a piezoelectric ceramic element composed of ceramic sheets and electrode stacks; the bottom of the piezoelectric ceramic element is provided with a connecting plate; the connecting plate and the mounting plate are fixed by a threaded connection. This results in a simple structure and convenient installation.

[0008] Furthermore, a heat sink is provided at the top of the piezoelectric ceramic element and at the bottom of the connecting plate; the heat sink has an array of heat dissipation holes. This increases the heat dissipation area, facilitating the rapid dissipation of heat generated by the piezoelectric effect and preventing thermal failure of the piezoelectric ceramic actuator.

[0009] Furthermore, a transmission rod is provided at the center of the bottom of the heat sink; the end of the transmission rod is hemispherical. Thus, when the piezoelectric ceramic actuator operates, the minute displacement of the piezoelectric ceramic element can be amplified and concentrated at a single point on the hemispherical end through the transmission rod, achieving point contact excitation of the steel support surface and improving energy transmission efficiency.

[0010] Furthermore, the heat sink can be made of aluminum alloy. This helps reduce material and manufacturing costs, while also facilitating weight reduction. The heat sink can also be made of other high thermal conductivity metal materials.

[0011] As an optimization, in the aforementioned non-invasive monitoring device for axial force of steel supports in deep foundation pits, the side of the adsorption seat that contacts the steel support is arc-shaped or V-shaped. This allows the adsorption seat to adapt to the curved surface of the steel support, improving the adhesion strength.

[0012] Furthermore, the adsorption base is a switchable magnetic base. The switchable magnetic base controls the on / off state of the magnetic force via a rotary switch, enabling installation and removal, which is simple to operate and easy to implement.

[0013] As an optimization, in the aforementioned non-intrusive monitoring device for axial force of steel supports in deep foundation pits, the controller includes a main control unit, a digital signal processing unit, a power supply unit, and a communication unit; the main control unit is used to control the operation of the piezoelectric ceramic actuator; the digital signal processing unit is used to process the transient vibration response time-domain signal sent by the MEMS accelerometer and calculate the axial force value of the steel support; the communication unit is a wide area network wireless module used to achieve data communication with the remote monitoring platform; and the power supply unit is a rechargeable battery pack.

[0014] Regarding the method, the technical solution of this application is as follows:

[0015] A smart monitoring method for axial force of steel supports in deep foundation pits is proposed. This method is based on the aforementioned non-intrusive monitoring device and BP neural network prediction model for axial force of steel supports in deep foundation pits. Specifically, it includes the following steps: Step 1: During the construction of the deep foundation pit, one or more monitoring devices are installed on the surface of the steel support. The piezoelectric ceramic actuator in the monitoring device applies mechanical excitation to the steel support at a preset cycle. The MEMS accelerometer collects the transient vibration response time-domain signal generated by the steel support under mechanical excitation and sends it to the controller. The controller performs digital filtering and noise reduction on the received time-domain signal, then performs a fast Fourier transform to obtain a frequency-domain signal. The controller then automatically identifies and extracts the current natural frequency of the steel support from the frequency-domain signal. Finally, based on the offset between the pre-stored reference frequency and the current natural frequency, the axial force value of the steel support is calculated. Step 2: The controller sends the calculated axial force value to the BP neural network prediction model in the remote monitoring platform to predict the future axial force value. The remote monitoring platform analyzes and evaluates the prediction results and generates corresponding early warning information.

[0016] Compared with existing technologies, the intelligent monitoring method for axial force of steel supports in deep foundation pits in this application actively excites the vibration of steel supports through piezoelectric ceramic actuators, receives the vibration response signals through MEMS accelerometers, and then processes the data by a controller to calculate the corresponding axial force value before sending it to a BP neural network prediction model to predict future axial force values. This method differs from existing technologies that directly upload raw, massive amounts of vibration waveforms to a remote monitoring platform. By processing the data at its source, it achieves "refinement," ensuring that every uploaded data point is "clean data" verified by a physical model, resulting in high data reliability and improved prediction accuracy. Furthermore, the monitoring device used in this application can be flexibly deployed at monitoring points where sensors have failed or at key locations where no sensors have been deployed. This ensures a continuous supply of high-quality, uninterrupted data to the input layer of the BP neural network prediction model, thus avoiding data loss due to traditional sensor failure and data distortion due to data drift.

[0017] As an optimization, in the aforementioned intelligent monitoring method for axial force of steel supports in deep foundation pits, the BP neural network prediction model is trained using historical data: First, the historical axial force data is divided into a training set and a test set; then, the data in the training set is input into the BP neural network prediction model for iterative training, allowing it to continuously learn the nonlinear law of axial force change until the loss function converges or the predetermined number of training iterations is reached; then, the data in the test set is input into the trained BP neural network prediction model to evaluate its performance, obtain the optimal weights, and complete the training. The reference frequency is obtained in the following way: when the steel support is in a state of zero stress or known initial stress, the monitoring device is installed on the surface of the steel support; a mechanical excitation is applied to the steel support through a piezoelectric ceramic actuator, and a MEMS accelerometer collects the time-domain signal of the transient vibration response generated by the steel support under mechanical excitation and sends it to the controller; the controller performs digital filtering and denoising on the received time-domain signal, then performs a fast Fourier transform to obtain a frequency-domain signal, and then extracts and stores the reference frequency of the steel support from the frequency-domain signal. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the assembly of the non-intrusive monitoring device for axial force of deep foundation pit steel support and the steel support in Embodiment 1 of this application;

[0019] Figure 2 This is a schematic diagram of the piezoelectric ceramic actuator in Example 1;

[0020] Figure 3 This is a schematic diagram of the adaptive bridge mounting mechanism in Example 1;

[0021] Figure 4This is a flowchart of the intelligent monitoring method for axial force of steel supports in deep foundation pits in Example 2;

[0022] Figure 5 This is a flowchart of the prediction process of the BP neural network prediction model in Example 2.

[0023] The labels in the attached diagram are as follows: 1-Piezoelectric ceramic actuator, 11-Piezoelectric ceramic element, 12-Connecting plate, 13-Heat sink, 14-Transmission rod; 2-MEMS accelerometer; 3-Controller; 4-Adaptive bridge mounting mechanism, 41-Mounting plate, 42-Adsorption seat, 43-Connecting rod; 5-Steel support. Detailed Implementation

[0024] The present application will be further described below with reference to the accompanying drawings and embodiments, but this should not be construed as limiting the present application. Contents not described in detail in the following embodiments are all common knowledge in the art.

[0025] Example 1:

[0026] See Figure 1 This embodiment provides a non-invasive monitoring device for the axial force of steel supports in deep foundation pits, including a piezoelectric ceramic actuator 1, a MEMS accelerometer 2, a controller 3, and an adaptive bridge mounting mechanism 4. The piezoelectric ceramic actuator 1 and the MEMS accelerometer 2 are connected to the steel support 5 through the adaptive bridge mounting mechanism 4. The adaptive bridge mounting mechanism 4 includes a mounting plate 41 and two adsorption seats 42 symmetrically arranged on both sides of the mounting plate 41. A connecting rod 43 is hinged between the mounting plate 41 and the adsorption seats 42. The piezoelectric ceramic actuator 1 and the MEMS accelerometer 2 are electrically connected to the controller 3.

[0027] In use, the adsorption seat 42 is adsorbed and connected to the outer surface of the steel support 5, and the piezoelectric ceramic actuator 1 and MEMS accelerometer 2 are fixed on the corresponding mounting plate 41. The piezoelectric ceramic actuator 1 abuts against the outer surface of the steel support 5 to apply mechanical excitation to the steel support 5. The MEMS accelerometer 2 is in contact with the outer surface of the steel support 5 to collect the transient vibration response time-domain signal generated by the steel support 5 under mechanical excitation and send it to the controller 3. The controller is used to process the received time-domain signal, calculate the corresponding axial force value, and send it to the remote monitoring platform.

[0028] In this embodiment, the piezoelectric ceramic actuator 1 is a PZT piezoelectric ceramic actuator, comprising a piezoelectric ceramic element 11 composed of ceramic sheets and electrode stacks; a connecting disk 12 is provided at the bottom of the piezoelectric ceramic element 11; the connecting disk 12 is fixedly connected to the mounting disk 41 by threads. Specifically, the mounting disk 41 is annular, and a set of A threaded holes are spaced apart along the circumference of the mounting disk 41; correspondingly, a set of B threaded holes are provided along the circumference of the connecting disk 12; the lower surface of the connecting disk 12 contacts the upper surface of the mounting disk 41 and is fixedly connected by bolts. The MEMS accelerometer 2 is provided with a connector that mates with the mounting disk 41 and is connected to the mounting disk 41 by threads.

[0029] See Figure 2 In this embodiment, a heat sink 13 is provided on the top of the piezoelectric ceramic element 11 and the bottom of the connecting plate 12; the heat sink 13 has an array of heat dissipation holes. This increases the heat dissipation area, facilitating the rapid dissipation of heat generated by the piezoelectric effect and preventing thermal failure of the piezoelectric ceramic actuator 1. The heat sink 13 is made of aluminum alloy.

[0030] In this embodiment, a transmission rod 14 is provided at the bottom center of the heat sink 13; the end of the transmission rod 14 is hemispherical. Thus, when the piezoelectric ceramic actuator 1 is working, the minute displacement of the piezoelectric ceramic element 11 can be amplified by the transmission rod 14 and concentrated at a point on the hemispherical end, realizing point contact excitation on the surface of the steel support 5 and improving energy transmission efficiency.

[0031] See Figure 3 In this embodiment, the side of the adsorption seat 42 that contacts the steel support 5 is V-shaped. This allows the adsorption seat 42 to adapt to the curved surface of the steel support 5, improving the adsorption strength. Furthermore, the adsorption seat 42 is a switchable magnetic seat (commercially available). The switchable magnetic seat controls the on / off state of the magnetic force via a rotary switch, enabling installation and removal; it is simple to operate and easy to implement. The two ends of the connecting rod 43 are hinged to the mounting plate 41 and the adsorption seat 42, respectively. By adjusting the angle of the connecting rod 43, the span between the two adsorption seats 42 can be changed, thereby fitting the surface of the steel support 5 with different radii of curvature. In the installed state, the two adsorption seats 42 are firmly adsorbed onto the surface of the steel support 5 as 'anchor points', and the piezoelectric ceramic actuator 1 is located on the center line of the two adsorption seats 42 via the mounting plate 41 (the two adsorption seats 42 and the central piezoelectric ceramic actuator 1 form a V-shaped structure); by adjusting the locking angle of the connecting rod 43, a downward preload can be applied to the piezoelectric ceramic actuator 1, forcing the end of the transmission rod 14 to press against the surface of the steel support 5, forming a stable 'three-point' mechanical structure (two magnetic attraction points + one pressing point).

[0032] In this embodiment, the controller 3 includes a main control unit (MCU), a digital signal processing unit (DSP), a power supply unit, and a communication unit. The main control unit is used to control the piezoelectric ceramic actuator 1 to work. The digital signal processing unit is used to process the transient vibration response time-domain signal sent by the MEMS accelerometer 2 and calculate the axial force value of the steel support 4. The communication unit is a wide area network wireless module used to achieve data communication with the remote monitoring platform. The power supply unit is a rechargeable battery pack.

[0033] Example 2:

[0034] See Figure 4 This embodiment provides a smart monitoring method for axial force of steel supports in deep foundation pits. The method is based on the non-intrusive monitoring device for axial force of steel supports in deep foundation pits and the BP neural network prediction model in Embodiment 1, and specifically includes the following steps.

[0035] Step 1, Baseline Calibration: With the steel support 5 under zero stress or a known initial stress state, the monitoring device is installed on the surface of the steel support 5; a mechanical excitation is applied to the steel support 5 through the piezoelectric ceramic actuator 1, and the MEMS accelerometer 2 collects the transient vibration response time-domain signal generated by the steel support 5 under mechanical excitation and sends it to the controller 3; the controller 3 performs digital filtering and noise reduction on the received time-domain signal, and then performs a fast Fourier transform to obtain the frequency-domain signal, and then extracts and stores the reference frequency of the steel support 5 from the frequency-domain signal. .

[0036] Step 2, Monitoring Device Deployment: During the deep foundation pit construction process, one or more monitoring devices will be installed on the corresponding steel support surface 5 according to monitoring requirements. Remedial deployment will be implemented if pre-embedded sensors fail or abnormal stress is detected in areas where no monitoring points are located.

[0037] Step 3: Automated Monitoring and Data Preprocessing: The piezoelectric ceramic actuator 1 in the detection device applies mechanical excitation to the steel support 5 according to a preset cycle. The MEMS accelerometer 2 collects the transient vibration response time-domain signal generated by the steel support 5 under mechanical excitation and sends it to the controller 3. The controller 3 performs digital filtering and noise reduction on the received time-domain signal, and then performs a fast Fourier transform (FFT) to obtain the frequency-domain signal. Then, it automatically identifies and extracts the current natural frequency of the steel support 5 from the frequency-domain signal. Finally, based on the built-in frequency-axial force calibration model, according to the offset between the pre-stored reference frequency and the current natural frequency ( The current axial force value of steel support 5 is calculated.

[0038] For a steel brace simply supported at both ends, under the action of an axial force P, its nth natural frequency (Khz) is approximately: In the formula, L is the effective calculated length of the steel support (m), and E is the elastic modulus of the steel support (N / m). 2 I is the moment of inertia of the steel support section (m) 4 ρ is the density of the steel support (kg / m³), and A is the cross-sectional area of ​​the steel support (m²). 2 P is the axial force (kn; tension is positive, compression is negative; here it specifically refers to the axial force on the steel support), and n is the vibration order (usually we focus on the first-order vibration, i.e., n=1).

[0039] The frequency-axial force calibration model establishment process is as follows: Based on the above theoretical formulas and experimental data, a simplified calibration curve or formula is fitted. The following model is used in this embodiment: Then, by applying stress to a state of zero stress (i.e. A series of known axial forces are applied to the steel support to be tested, and the corresponding frequencies are measured. Then, the coefficients a and b are fitted by the least squares method to obtain the frequency-axial force calibration model.

[0040] Step 4, Axial Force Data Prediction: Controller 3 sends the calculated axial force value to the BP neural network prediction model in the remote monitoring platform to predict the future axial force value;

[0041] The BP neural network prediction model comprises an input layer, a hidden layer, and an output layer. The input layer contains N neurons, corresponding to axial force data from the past N time points (e.g., N=20 days). The hidden layer has one or more neurons used to learn the nonlinear laws governing axial force changes. The output layer contains M neurons, corresponding to predicted axial force values ​​for the next M time points (e.g., M=5 days). The BP neural network prediction model is trained using historical data: First, historical axial force data is divided into a training set and a test set. Then, data from the training set is input into the BP neural network prediction model for iterative training, allowing it to continuously learn the nonlinear laws governing axial force changes until the loss function converges or a predetermined number of training iterations is reached. Next, data from the test set is input into the trained BP neural network prediction model to evaluate its performance, obtain the optimal weights, and complete the training. During monitoring, the latest N days' axial force data (provided by the monitoring device) is used as input, and the trained BP neural network prediction model outputs predicted axial force values ​​for the next M days (see [link to relevant documentation]). Figure 5 When the actual axial force data at time N+1 is uploaded, the new data is added to the input sequence, while the oldest data is removed to form a new input sample, thus achieving rolling prediction.

[0042] Step 5, Assessment and Early Warning: The remote monitoring platform analyzes and assesses the prediction results and generates corresponding early warning information; maintenance personnel take appropriate measures based on the early warning information.

[0043] The foregoing general description of the invention and its specific embodiments should not be construed as a limitation on the technical solution of the invention. Those skilled in the art, based on the disclosure of this application, can add, reduce, or combine the disclosed technical features in the foregoing general description and / or specific embodiments (including examples) without departing from the constituent elements of the invention, to form other technical solutions within the scope of protection of this application.

Claims

1. A non-invasive monitoring device for axial force of steel supports in deep foundation pits, characterized in that: The system includes a piezoelectric ceramic actuator (1), a MEMS accelerometer (2), a controller (3), and an adaptive bridge mounting mechanism (4). The piezoelectric ceramic actuator (1) and the MEMS accelerometer (2) are connected to a steel support (5) via the adaptive bridge mounting mechanism (4). The adaptive bridge mounting mechanism (4) includes a mounting plate (41) and two suction seats (42) symmetrically arranged on both sides of the mounting plate (41). A connecting rod (43) is hinged between the mounting plate (41) and the suction seats (42). The accelerometer (2) is electrically connected to the controller (3); in use, the adsorption seat (42) is adsorbed to the outer surface of the steel support (5), and the piezoelectric ceramic actuator (1) and the MEMS accelerometer (2) are fixed on the corresponding mounting plate (41); the piezoelectric ceramic actuator (1) abuts against the outer surface of the steel support (5) to apply mechanical excitation to the steel support (5); the MEMS accelerometer (2) contacts the outer surface of the steel support (5) to collect the transient vibration response time domain signal generated by the steel support (5) under mechanical excitation and send it to the controller (3).

2. The non-intrusive monitoring device for axial force of steel support in deep foundation pits according to claim 1, characterized in that: The piezoelectric ceramic actuator (1) includes a piezoelectric ceramic element (11) composed of ceramic sheets and electrode stacks; the bottom of the piezoelectric ceramic element (11) is provided with a connecting plate (12); the connecting plate (12) and the mounting plate (41) are fixed by threaded connection.

3. The non-invasive monitoring device for axial force of steel support in deep foundation pits according to claim 2, characterized in that: The top of the piezoelectric ceramic element (11) and the bottom of the connecting plate (12) are respectively provided with a heat sink (13); the heat sink (13) is provided with an array of heat dissipation holes.

4. The non-intrusive monitoring device for axial force of steel support in deep foundation pits according to claim 3, characterized in that: The heat sink (13) has a transmission rod (14) at its bottom center; the end of the transmission rod (14) is hemispherical.

5. The non-intrusive monitoring device for axial force of steel support in deep foundation pits according to claim 1, characterized in that: The side of the adsorption seat (42) that contacts the steel support (5) is arc-shaped or V-shaped.

6. The non-intrusive monitoring device for axial force of steel support in deep foundation pits according to claim 5, characterized in that: The adsorption seat (42) is a switchable magnetic seat.

7. The non-intrusive monitoring device for axial force of steel support in deep foundation pits according to claim 1, characterized in that: The controller (3) includes a main control unit, a digital signal processing unit, a power supply unit, and a communication unit; the main control unit is used to control the piezoelectric ceramic actuator (1) to work; the digital signal processing unit is used to process the transient vibration response time domain signal sent by the MEMS accelerometer (2) and calculate the axial force value of the steel support (4); the communication unit is a wide area network wireless module, used to realize data communication with the remote monitoring platform; the power supply unit is a rechargeable battery pack.

8. A smart monitoring method for axial force of steel supports in deep foundation pits, characterized by: This method is based on the non-intrusive monitoring device for axial force of steel supports in deep foundation pits and the BP neural network prediction model described in claim 7, and specifically includes the following steps: Step 1: During the deep foundation pit construction, one or more monitoring devices are installed on the surface of the steel support (5); the piezoelectric ceramic actuator (1) in the monitoring device applies mechanical excitation to the steel support (5) according to a preset cycle, and the MEMS accelerometer (2) collects the transient vibration response time domain signal generated by the steel support (5) under mechanical excitation and sends it to the controller (3); the controller (3) performs digital filtering and noise reduction on the received time domain signal, and then performs fast Fourier transform to obtain the frequency domain signal, and then automatically identifies and extracts the current natural frequency of the steel support (5) from the frequency domain signal; finally, the axial force value of the steel support (5) is calculated according to the offset between the pre-stored reference frequency and the current natural frequency. Step 2: The controller (3) sends the calculated axial force value to the BP neural network prediction model in the remote monitoring platform to predict the future axial force value; the remote monitoring platform analyzes and evaluates the prediction results and generates corresponding early warning information.

9. The intelligent monitoring method for axial force of steel supports in deep foundation pits according to claim 8, characterized in that, The BP neural network prediction model is trained using historical data: First, the historical axial force data is divided into a training set and a test set; then, the data in the training set is input into the BP neural network prediction model for iterative training, allowing it to continuously learn the nonlinear laws of axial force changes until the loss function converges or the predetermined number of training iterations is reached; finally, the data in the test set is input into the trained BP neural network prediction model to evaluate its performance, obtain the optimal weights, and complete the training.

10. The intelligent monitoring method for axial force of steel supports in deep foundation pits according to claim 8, characterized in that, The reference frequency is obtained in the following way: when the steel support (5) is in a state of zero stress or known initial stress, the monitoring device is installed on the surface of the steel support (5); a mechanical excitation is applied to the steel support (5) by the piezoelectric ceramic actuator (1), the MEMS accelerometer (2) collects the transient vibration response time domain signal generated by the steel support (5) under mechanical excitation, and sends it to the controller (3); the controller (3) performs digital filtering and noise reduction on the received time domain signal, and then performs fast Fourier transform to obtain the frequency domain signal, and then extracts and stores the reference frequency of the steel support (5) from the frequency domain signal.

Images