Pile foundation compactness real-time detection method

Abstract

The invention belongs to the technical field of pile foundation detection, and particularly relates to a pile foundation compactness real-time detection method which comprises the following steps: step S1, a multi-mode sensor array is pre-buried in a pile foundation construction stage, and the multi-mode sensor array comprises a piezoelectric-capacitance composite sensor, a nano conductive film coating and an MEMS (Micro Electro Mechanical System) fiber grating chain; and S2, synchronously acquiring multi-physical field data through the vibration wave propagation velocity, the dielectric constant gradient, the strain field distribution and the electrical impedance change. According to the real-time detection method for the compactness of the pile foundation, by arranging the pre-embedded multi-mode sensor array, multi-dimensional monitoring on the density of the concrete of the pile foundation is achieved. And real-time data acquisition is carried out on different physical quantities and in the pile foundation concrete pouring process through various sensors, so that the comprehensiveness and accuracy of detection are enhanced, the effect of accurately evaluating the overall compactness of the pile foundation in real time is achieved, and the problem that detection is not timely in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of pile foundation testing technology, and in particular to a method for real-time testing of pile foundation compaction. Background Technology

[0002] The main methods for pile foundation testing include static load testing, core drilling, low-strain method, high-strain method, and sonic logging. Existing pile foundation compaction testing technologies have the following limitations:

[0003] The acoustic transmission method requires the pre-embedded acoustic logging pipes, which increases construction costs and is easily affected by mud.

[0004] Distributed fiber optic sensing requires high installation precision and cannot detect local microscopic defects.

[0005] Traditional AI algorithms rely on cloud computing, which makes it difficult to meet real-time requirements.

[0006] Furthermore, in actual testing, existing traditional testing methods all test the compaction of the pile foundation after the pile foundation is formed, and cannot test the compaction of the concrete during the concrete pouring process. This results in untimely testing, so a real-time pile foundation compaction testing method is needed. Summary of the Invention

[0007] Based on existing technical problems, this invention proposes a method for real-time detection of pile foundation compaction.

[0008] The present invention proposes a method for real-time detection of pile foundation compaction, comprising the following steps:

[0009] Step S1: Pre-embed a multi-modal sensor array during the pile foundation construction stage, including a piezoelectric-capacitive composite sensor, a nano-conductive thin film coating, and a MEMS fiber optic grating chain.

[0010] Step S2: Simultaneously acquire multi-physics data by measuring vibration wave propagation velocity, dielectric constant gradient, strain field distribution, and electrical impedance changes.

[0011] Step S3: Use the edge computing terminal to perform spatiotemporal alignment preprocessing on the original data and extract joint time-frequency domain features.

[0012] Step S4: Construct a compaction mapping model based on a dynamic convolutional neural network (DynCNN) and output a three-dimensional defect probability distribution map of the pile body.

[0013] Step S5: Transmit key data to the cloud via the UWB-NB-IoT hybrid communication network, and optimize global model parameters by combining the federated learning mechanism.

[0014] Preferably, in step S1, some of the piezoelectric-capacitive composite sensors are fixedly mounted on a chain support, the top of which is provided with a hoisting and positioning mechanism. The piezoelectric-capacitive composite sensor adopts a silicon carbide encapsulation structure and has a size ≤3cm. 3 The frequency response range is 1Hz-10kHz, the withstand voltage is ≥50MPa, and the sensitivity ratio of the capacitance detection module to the piezoelectric module is 1:(0.2-0.5).

[0015] Preferably, the chain support includes a mounting plate, the upper surface of which is fixedly connected to a fixing conduit, and the lower surface of which is fixedly connected to a conical buoyancy airbag.

[0016] Preferably, a hinged mounting plate is fixedly connected to the surface of the mounting plate, and a plurality of hinged mounting plates are arranged in a circular array with the axis of the mounting plate as the center. The upper surface of each of the plurality of hinged mounting plates is fixedly connected with a hinged chain strip by bolts.

[0017] The hinged chain plate is composed of multiple first hinge plates, multiple second hinge plates with one end of the first hinge plates hinged by a pin, and multiple bucket-shaped buoyancy airbags fixedly connected to the lower surface of the second hinge plates.

[0018] The adjacent first hinge plate and the second hinge plate are fixedly connected by bolts.

[0019] Preferably, the hoisting and positioning mechanism includes a hoisting fixing frame, and the upper surface of the hoisting fixing frame is fixedly connected with three lifting lugs arranged in a circular array with the axis of the hoisting fixing frame as the center.

[0020] Preferably, the surface of the hoisting frame is rotatably connected to a connecting rod column via a pin, and the three connecting rod columns are arranged in a circular array around the axis of the hoisting frame. One end of each connecting rod column is rotatably connected to a positioning support column via a pin, and the surface of the positioning support column is in the shape of a character. A movable frame is provided below the hoisting frame, and the surfaces of the three positioning support columns are rotatably connected to the surface of the movable frame via pins.

[0021] One end of the positioning support column is fixedly connected to an arc-shaped positioning slide, which is made of polytetrafluoroethylene.

[0022] Preferably, the upper surface of the hoisting frame is rotatably connected to an adjusting screw tube via a bearing, one end of the adjusting screw tube passing through and extending to the lower surface of the movable frame, and the surface of the adjusting screw tube being threadedly connected to the surface of the movable frame.

[0023] Preferably, an adjusting handwheel is fixedly sleeved at the other end of the adjusting screw, and the lower surface of the adjusting handwheel is slidably connected to the upper surface of the hoisting frame.

[0024] The inner wall of the adjusting solenoid is rotatably connected to the surface of the fixed conduit via a bearing.

[0025] Preferably, the dynamic convolutional neural network in step S4 includes deformable convolutional kernels, whose deformation offset Δ is calculated by the following formula:

[0026] Δ = Softmax(W) d *X+b d )

[0027] Where X is the input feature map, W d and b d As learnable parameters, the sub-pixel positioning accuracy of the output defect boundary is ≤5mm.

[0028] Preferably, the federated learning mechanism in step S5 employs differential privacy protection, and the noise addition satisfies the following:

[0029] M(D)=f(D)+N(0,σ 2 S 2 I)

[0030] The sensitivity SS ≤ 0.3, the privacy budget ∈ ≤ 1.0, and the model aggregation cycle is 6-24 hours.

[0031] Step S6: Trigger graded early warning based on the density threshold and generate a visual repair plan on the BIM platform.

[0032] The beneficial effects of this invention are as follows:

[0033] 1. By setting up a pre-embedded multimodal sensor array, multi-dimensional monitoring of the density of pile foundation concrete is achieved. Multiple sensors collect real-time data on different physical quantities and during the pile foundation concrete pouring process, enhancing the comprehensiveness and accuracy of the detection. This achieves accurate and real-time assessment of the overall compactness of the pile foundation, solving the problem of untimely detection in existing technologies.

[0034] 2. By setting up a dynamic convolutional neural network and using deformable convolutional kernels to perform deep learning analysis on pile foundation compaction, it has higher accuracy and reliability compared with traditional detection methods.

[0035] 3. By setting up a hoisting and positioning mechanism, the testing equipment can move stably and accurately within the pile foundation through precise positioning control, ensuring the stability and measurement accuracy of the multi-modal sensor array. Meanwhile, the chain support provides strong support and dynamic adjustment capabilities, allowing the testing equipment to adapt to height changes during different pile foundation concrete pouring processes, making the pile foundation compaction testing more accurate and efficient. Attached Figure Description

[0036] Figure 1 This is a schematic diagram of a chain support structure for a real-time pile compaction detection method proposed in this invention;

[0037] Figure 2 This is a three-dimensional view of the chain support structure of the real-time pile compaction detection method proposed in this invention;

[0038] Figure 3 This invention proposes a real-time detection method for pile foundation compaction. Figure 2 Enlarged view of the structure at point A in the middle;

[0039] Figure 4 This is a three-dimensional view of the hinged chain strip structure of the real-time pile compaction detection method proposed in this invention.

[0040] In the diagram: 1. Chain bracket; 101. Mounting plate; 102. Fixed conduit; 103. Conical buoyancy airbag; 104. Hinge mounting plate; 105. Hinge chain strip; 1051. First hinge plate; 1052. Second hinge plate; 1053. Bucket-shaped buoyancy airbag; 2. Lifting and positioning mechanism; 201. Lifting fixing frame; 202. Lifting lug; 203. Connecting rod column; 204. Positioning support column; 205. Moving frame; 206. Arc-shaped positioning slide; 207. Adjusting screw; 208. Adjusting handwheel. Detailed Implementation

[0041] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0042] A method for real-time detection of pile foundation compaction includes the following steps:

[0043] Step S1: Pre-embed a multi-modal sensor array during the pile foundation construction stage, including a piezoelectric-capacitive composite sensor, a nano-conductive thin film coating, and a MEMS fiber optic grating chain.

[0044] Furthermore, in this embodiment, an optical fiber pressure sensor can be used instead of a piezoelectric-capacitive composite sensor. The optical fiber pressure sensor uses the refraction or reflection of optical fibers to detect changes in the pressure of the pile foundation concrete, thereby enabling the monitoring of the compactness of the pile foundation concrete.

[0045] The density of pile foundations can be assessed by monitoring pressure changes in the concrete using fiber optic pressure sensors. Dense pile foundations withstand greater pressure, while loose pile foundations exhibit a smaller pressure response.

[0046] Reference Figures 1-4In step S1, several piezoelectric-capacitive composite sensors are fixedly mounted on a chain bracket 1. A hoisting and positioning mechanism 2 is provided at the top of the chain bracket 1. The piezoelectric-capacitive composite sensors adopt a silicon carbide encapsulation structure and have a size ≤3cm. 3 The frequency response range is 1Hz-10kHz, the withstand voltage is ≥50MPa, and the sensitivity ratio of the capacitance detection module to the piezoelectric module is 1:(0.2-0.5).

[0047] The chain support 1 includes a mounting plate 101, with a fixed cable conduit 102 fixedly connected to the upper surface of the mounting plate 101 and a conical buoyancy airbag 103 fixedly connected to the lower surface of the mounting plate 101.

[0048] Furthermore, the conical buoyancy airbag 103 is used to maintain the buoyancy of the mounting plate 101 in the concrete, preventing the mounting plate 101 from sinking too deep into the concrete and causing a falling pull on the hinge chain strip 105.

[0049] In use, the fixed conduit 102 is used to connect the chain bracket 1 to the hoisting and positioning mechanism 2, and to thread and protect the cable of the piezoelectric-capacitive composite sensor installed on the chain bracket 1.

[0050] A hinged mounting plate 104 is fixedly connected to the surface of the mounting plate 101. Multiple hinged mounting plates 104 are arranged in a ring array with the axis of the mounting plate 101 as the center. The upper surfaces of the multiple hinged mounting plates 104 are all fixedly connected with hinged chain strips 105 by bolts.

[0051] The hinged chain plate 105 is composed of multiple first hinge plates 1051, multiple second hinge plates 1052 with one end of the first hinge plates 1051 hinged by a pin, and a bucket-shaped buoyancy airbag 1053 fixedly connected to the lower surface of multiple second hinge plates 1052.

[0052] Furthermore, the buoyancy calculation formulas for the conical buoyancy airbag 103 and the bucket-shaped buoyancy airbag 1053 are as follows:

[0053] F b =ρ concrete ·V 气囊 ·gm 传感器 ·g

[0054] Where ρ concrete V is the density of concrete. 气囊 The airbag volume can be adjusted from 0-50cm in depth by adjusting the inflation volume.

[0055] The adjacent first hinge plate 1051 and second hinge plate 1052 are fixedly connected by bolts.

[0056] In use, the hinged chain plate is fixedly connected to the hinged chain plate by a first hinged plate 1051 near one end of the hinged mounting plate 104 using fixing bolts. Then, two adjacent first hinged plates 1051 and second hinged plates 1052 are fixedly connected by bolts to form the hinged chain plate strip 105. The piezoelectric-capacitive composite sensor is installed on the second hinged plate 1052 at one end of the hinged chain plate strip 105 away from the mounting plate 101.

[0057] During the process of testing the compactness of the pile foundation concrete, multiple hinged chain plates 105 are first unfolded and hoisted into the bottom of the pile foundation, and the bucket-shaped buoyancy airbags 1053 under multiple second hinged plates 1052 are selectively inflated according to the diameter of the pile foundation.

[0058] Specifically, the inflated bucket-shaped buoyancy airbag 1053 will generate buoyancy on the second hinge plate 1052 in the concrete, causing it to float or partially float in the concrete. The uninflated bucket-shaped buoyancy airbag 1053 will not generate buoyancy, causing the second hinge plate 1052 to sink deep into the concrete under its own weight, gravity and the pressure of the concrete. By adjusting the depth of the different hinge chain strips 105 sinking into the concrete and the distance from the axis of the mounting plate 101, the compactness of the pile foundation concrete at different locations and depths can be detected.

[0059] Furthermore, during use, multiple inflation tubes connected to the inflation pump can be passed through the fixed conduit 102 and then connected to the inflation pump via a solenoid valve. This allows for dynamic adjustment of the position and depth of the detection sensor during the testing process, thereby achieving better results in detecting the density of pile foundation concrete.

[0060] The hoisting and positioning mechanism 2 includes a hoisting fixing frame 201, and three lifting lugs 202 are fixedly connected to the upper surface of the hoisting fixing frame 201 in a circular array centered on the axis of the hoisting fixing frame 201.

[0061] The surface of the hoisting frame 201 is rotatably connected to the connecting rod column 203 via a pin. The three connecting rod columns 203 are arranged in a circular array with the axis of the hoisting frame 201 as the center. One end of the connecting rod column 203 is rotatably connected to the positioning support column 204 via a pin. The surface of the positioning support column 204 is in the shape of a "7". A movable frame 205 is provided below the hoisting frame 201. The surfaces of the three positioning support columns 204 are rotatably connected to the surface of the movable frame 205 via pins.

[0062] One end of the positioning support column 204 is fixedly connected to an arc-shaped positioning slide 206, which is made of polytetrafluoroethylene.

[0063] Furthermore, polytetrafluoroethylene has an extremely low coefficient of friction and is characterized by high temperature resistance, corrosion resistance, wear resistance, and good stability.

[0064] In use, the arc-shaped positioning slide plate 206 contacts the inner wall of the pile foundation casing to provide sliding positioning support for the positioning support column 204 and the hoisting fixing frame 201.

[0065] The upper surface of the hoisting frame 201 is rotatably connected to the adjusting screw tube 207 via a bearing. One end of the adjusting screw tube 207 passes through and extends to the lower surface of the movable frame 205. The surface of the adjusting screw tube 207 is threadedly connected to the surface of the movable frame 205.

[0066] The other end of the adjusting screw tube 207 is fixedly sleeved with an adjusting handwheel 208, and the lower surface of the adjusting handwheel 208 is slidably connected to the upper surface of the hoisting fixing frame 201.

[0067] The inner wall of the adjusting screw tube 207 is rotatably connected to the surface of the fixed conduit tube 102 via a bearing.

[0068] In use, first use a crane or other lifting equipment to lift the lifting and positioning mechanism 2 and the chain bracket 1. Then move the chain bracket 1 and the lifting and positioning mechanism 2 into the pile foundation casing. Adjust the diameter of the three positioning support columns 204 after unfolding according to the inner diameter of the pile foundation to ensure that the arc-shaped positioning slide plate 206 slides and connects with the inner wall of the pile foundation casing after the three positioning support columns 204 are opened.

[0069] During adjustment, by rotating the adjusting handwheel 208, the adjusting handwheel 208 drives the adjusting screw tube 207 to rotate. The rotation of the adjusting screw tube 207 drives the moving frame 205 to move, adjusting the distance between the moving frame 205 and the hoisting fixed frame 201. By adjusting the distance between the moving frame 205 and the hoisting fixed frame 201, under the limiting action of the connecting rod column 203, the positioning support column 204 is driven to expand outward or retract inward around the axis of the adjusting screw tube 207, adjusting the distance between the three positioning support columns 204.

[0070] Then, based on the actual inner diameter of the pile foundation, multiple bucket-shaped buoyancy airbags 1053 on the chain support 1 are inflated. By controlling the inflation of multiple bucket-shaped buoyancy airbags 1053 on a single hinged chain strip 105, the length of the single hinged chain strip 105 that floats inside the pile foundation is controlled, thereby enabling the detection of concrete density at different locations and depths inside the pile foundation. During the detection, a piezoelectric-capacitive composite sensor is fixedly installed at the end of the hinged chain strip 105.

[0071] Then, the hoisting positioning mechanism 2 and the chain support 1 are placed into the pile foundation using hoisting equipment. During the descent, the hoisting positioning mechanism 2 positions the chain support 1 to ensure that the chain support 1 is concentric with the axis of the pile foundation. During the descent, the chain support 1 is placed into the bottom wall of the pile foundation in an extended state using auxiliary supports or auxiliary slings. During the concrete pouring process, the buoyancy generated by the concrete on the bucket-shaped buoyancy airbag 1053 and the upward lifting force of the hoisting equipment cause the chain support 1 to drive the piezoelectric-capacitive composite sensor to be distributed into different positions and depths within the pile foundation to perform multi-directional detection of the compactness of the pile foundation concrete.

[0072] By setting up a hoisting and positioning mechanism 2, the testing equipment can move stably and accurately within the pile foundation through precise positioning control, ensuring the stability and measurement accuracy of the multimodal sensor array. Meanwhile, the chain support 1 provides strong support and dynamic adjustment capabilities, allowing the testing equipment to adapt to height changes during different pile foundation concrete pouring processes, making the pile foundation compaction testing more accurate and efficient.

[0073] Furthermore, the piezoelectric-capacitive composite sensor incorporates a piezoelectric material (PZT lead zirconate titanate or PVDF polymer). When the concrete is subjected to external vibration or internal stress changes, the piezoelectric material deforms, causing the positive and negative charge centers within it to separate, generating a surface charge proportional to the stress (positive piezoelectric effect). The charge Q satisfies: Q = d ij ·σ·A, where d ij Where C is the piezoelectric constant (C / N), σ is the applied stress (Pa), and A is the electrode area (m²). 2 The wave velocity v is calculated by measuring the time difference Δt between the transmitting and receiving ends of the vibration wave. p =L / Δt, reflecting the elastic modulus and density of concrete, where L is the sensor spacing.

[0074] By analyzing the signal amplitude attenuation rate α = 20log(A) in / A out (dB / m) to determine internal defects (such as voids and cracks).

[0075] The sensor comprises a pair of parallel electrode plates, with concrete filling the space between them as the dielectric material. The capacitance value CC is determined by the following formula:

[0076]

[0077] Where ε r Let ε0 be the relative permittivity of concrete, ε0 be the vacuum permittivity, A be the area of ​​the electrodes, and d be the distance between the electrodes. Changes in the density of concrete will alter its porosity, leading to changes in ε0. r Changes (density ↑ → porosity ↓ → ε) rBy measuring the dynamic fluctuations of capacitance (accuracy up to 0.1pF), the gradient distribution of dielectric constant is inverted to locate segregation or abnormal moisture content areas.

[0078] Specifically, the piezoelectric element and capacitor electrode are designed to be coplanar, with the piezoelectric material serving as a component of the capacitor dielectric. For example, a metal layer plated on the surface of PZT piezoelectric ceramic acts as both a piezoelectric electrode and a capacitor plate. High-frequency vibration signals (>1kHz) are acquired through the piezoelectric path, while low-frequency dielectric changes (DC-100Hz) are detected through the capacitive path, avoiding spectral interference. The data fusion algorithm includes spatiotemporal alignment and a joint calibration model.

[0079] Among them, spatiotemporal alignment, for the vibration wave propagation time series v p Align C(t) with the capacitance change curve C(t) using wavelet transform to eliminate the difference in sensor response delay.

[0080] Among them, the joint calibration model establishes a two-parameter mapping equation for the density ρ:

[0081]

[0082] Where a, b, and c are experimental calibration coefficients, C0 is the initial capacitance value, and the cross term v p • ΔC is used to correct for temperature and humidity coupling errors.

[0083] Alternatively, a Kistler 8763A or PCB Piezotronics 352C33 can be used as a piezoelectric sensor, and a MicroStrain 3DM-GX5-25 or Honeywell C2080 as a capacitive sensor, and then these two sensor components can be integrated using high-precision silicon carbide packaging technology.

[0084] Furthermore, after the nano-conductive film is sprayed onto the surface of the reinforcing steel, its electrical impedance change is closely related to the concrete porosity and interfacial bond strength. When the concrete debonds from the reinforcing steel due to insufficient compaction or segregation, the coating resistance changes significantly, thereby locating the defect area.

[0085] Referring to the crack-resistant principle of protein / polysaccharide composite nanofilms, the nano-conductive coating reduces the generation of microcracks through a dense molecular structure, maintaining stable conductivity when the pile foundation is bent or vibrated, thereby improving the reliability of sensor signals.

[0086] Furthermore, MEMS fiber optic grating chains are embedded in the pile body at a density of 5 measuring points per meter, enabling real-time measurement of the concrete strain field distribution and temperature gradient via Bragg wavelength shift. Its advantages include:

[0087] High resolution: strain resolution up to 0.1 με, temperature accuracy ±0.5℃.

[0088] Interference resistance: The wire drawing tower grating chain with ORMOCER coating has a mechanical strength 5 times that of traditional optical fiber and a temperature range of -180℃ to +200℃.

[0089] Step S2: Simultaneously acquire multi-physics data by measuring vibration wave propagation velocity, dielectric constant gradient, strain field distribution, and electrical impedance changes.

[0090] Step S3: Use the edge computing terminal to perform spatiotemporal alignment preprocessing on the original data and extract joint time-frequency domain features.

[0091] Step S4: Construct a compaction mapping model based on a dynamic convolutional neural network (DynCNN) and output a three-dimensional defect probability distribution map of the pile body.

[0092] In step S4, the dynamic convolutional neural network contains deformable convolutional kernels, and their deformation offset Δ is calculated using the following formula:

[0093] Δ = Softmax(W) d *X+b d )

[0094] Where X is the input feature map, W d and b d As learnable parameters, the sub-pixel positioning accuracy of the output defect boundary is ≤5mm.

[0095] Step S5: Transmit key data to the cloud via the UWB-NB-IoT hybrid communication network, and optimize global model parameters by combining the federated learning mechanism.

[0096] In step S5, the federated learning mechanism employs differential privacy protection, and the noise addition satisfies the following conditions:

[0097] M(D)=f(D)+N(0,σ 2 S 2 I)

[0098] The sensitivity SS ≤ 0.3, the privacy budget ∈ ≤ 1.0, and the model aggregation cycle is 6-24 hours.

[0099] Step S6: Trigger graded early warning based on the density threshold and generate a visual repair plan on the BIM platform.

[0100] In this embodiment, another part involves the deployment of multiple sensors: for example, in a 2m diameter cast-in-place pile, a ring-shaped sensor array (6 layers in total) is arranged every 2m along the longitudinal direction.

[0101] Data acquisition: Full pile scan every 10 seconds during the pouring stage, and sampling every minute during the hardening stage.

[0102] Anomaly Handling: If the density at 3m below the pile bottom is detected to be <85%, the system will automatically recommend a grouting pressure of 12MPa and a grouting volume of 0.8m³. 3 .

[0103] By setting up a dynamic convolutional neural network (DynCNN) and using deformable convolutional kernels to perform deep learning analysis on pile foundation compaction, it has higher accuracy and reliability compared with traditional detection methods.

[0104] By setting up a pre-embedded multimodal sensor array, multi-dimensional monitoring of the density of pile foundation concrete was achieved. Furthermore, by using multiple sensors to collect real-time data on different physical quantities and during the pile foundation concrete pouring process, the comprehensiveness and accuracy of the detection were enhanced. This enabled accurate and real-time assessment of the overall compactness of the pile foundation, solving the problem of untimely detection in existing technologies.

[0105] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A method for real-time detection of pile foundation compaction, characterized in that, Includes the following steps: Step S1: Pre-embed a multi-modal sensor array during the pile foundation construction stage, including a piezoelectric-capacitive composite sensor, a nano-conductive thin film coating, and a MEMS fiber optic grating chain; Step S2: Simultaneously acquire multi-physics data by measuring vibration wave propagation velocity, dielectric constant gradient, strain field distribution, and electrical impedance changes; Step S3: Use an edge computing terminal to perform spatiotemporal alignment preprocessing on the raw data and extract joint time-frequency domain features; Step S4: Construct a compaction mapping model based on a dynamic convolutional neural network (DynCNN) and output a three-dimensional defect probability distribution map of the pile body; Step S5: Transmit key data to the cloud via the UWB-NB-IoT hybrid communication network, and optimize global model parameters by combining the federated learning mechanism.

2. The method for real-time detection of pile foundation compaction according to claim 1, characterized in that: In step S1, some of the piezoelectric-capacitive composite sensors are fixedly mounted on a chain bracket (1). The top of the chain bracket (1) is equipped with a hoisting and positioning mechanism (2). The piezoelectric-capacitive composite sensor adopts a silicon carbide encapsulation structure and has a size ≤3cm. 3 The frequency response range is 1Hz-10kHz, the withstand voltage is ≥50MPa, and the sensitivity ratio of the capacitance detection module to the piezoelectric module is 1:(0.2-0.5).

3. The method for real-time detection of pile foundation compaction according to claim 2, characterized in that: The chain support (1) includes a mounting plate (101), a fixing conduit (102) is fixedly connected to the upper surface of the mounting plate (101), and a conical buoyancy airbag (103) is fixedly connected to the lower surface of the mounting plate (101).

4. The method for real-time detection of pile foundation compaction according to claim 3, characterized in that: The mounting plate (101) is fixedly connected to a hinged mounting plate (104). Multiple hinged mounting plates (104) are arranged in a ring array with the axis of the mounting plate (101) as the center. The upper surfaces of multiple hinged mounting plates (104) are fixedly connected to hinged chain strips (105) by bolts. The hinge chain strip (105) is composed of multiple first hinge plates (1051), multiple second hinge plates (1052) with one end of the first hinge plates (1051) hinged by a pin, and a bucket-shaped buoyancy airbag (1053) fixedly connected to the lower surface of multiple second hinge plates (1052). The first hinge plate (1051) and the second hinge plate (1052) are fixedly connected by bolts.

5. The method for real-time detection of pile foundation compaction according to claim 4, characterized in that: The hoisting and positioning mechanism (2) includes a hoisting fixing frame (201), and three lifting lugs (202) are fixedly connected to the upper surface of the hoisting fixing frame (201) in a circular array centered on the axis of the hoisting fixing frame (201).

6. The method for real-time detection of pile foundation compaction according to claim 5, characterized in that: The surface of the hoisting frame (201) is rotatably connected to a connecting rod column (203) via a pin. The three connecting rod columns (203) are arranged in a circular array with the axis of the hoisting frame (201) as the center. One end of the connecting rod column (203) is rotatably connected to a positioning support column (204) via a pin. The surface of the positioning support column (204) is shaped like the number 7. A movable frame (205) is provided below the hoisting frame (201). The surfaces of the three positioning support columns (204) are rotatably connected to the surface of the movable frame (205) via pins. One end of the positioning support column (204) is fixedly connected to an arc-shaped positioning slide (206), which is made of polytetrafluoroethylene.

7. The method for real-time detection of pile foundation compaction according to claim 6, characterized in that: The upper surface of the hoisting frame (201) is rotatably connected to an adjusting screw tube (207) via a bearing. One end of the adjusting screw tube (207) passes through and extends to the lower surface of the movable frame (205). The surface of the adjusting screw tube (207) is threadedly connected to the surface of the movable frame (205).

8. The method for real-time detection of pile foundation compaction according to claim 7, characterized in that: The other end of the adjusting screw tube (207) is fixedly sleeved with an adjusting handwheel (208), and the lower surface of the adjusting handwheel (208) is slidably connected to the upper surface of the hoisting fixing frame (201). The inner wall of the adjusting solenoid (207) is rotatably connected to the surface of the fixed conduit (102) via a bearing.

9. The method for real-time detection of pile foundation compaction according to claim 1, characterized in that: The dynamic convolutional neural network in step S4 includes deformable convolutional kernels, whose deformation offset Δ is calculated by the following formula: Δ=Softmax(W d *X+b d ) Where X is the input feature map, W d and b d As learnable parameters, the sub-pixel positioning accuracy of the output defect boundary is ≤5mm.

10. The method for real-time detection of pile foundation compaction according to claim 1, characterized in that: The federated learning mechanism in step S5 employs differential privacy protection, and the noise addition satisfies the following: M(D)=f(D)+N(0,σ 2 S 2 I) The sensitivity SS ≤ 0.3, the privacy budget ∈ ≤ 1.0, and the model aggregation cycle is 6-24 hours; Step S6: Trigger graded early warning based on the density threshold and generate a visual repair plan on the BIM platform.

Images