A system and method for detecting a super-deep stone column composite foundation

Abstract

The application relates to the technical field of gravel pile composite foundation detection, and discloses a super-deep gravel pile composite foundation detection system and method. The system is provided with a distributed optical fiber sensor and a deep excitation source through horizontal directional drilling, avoids the problem of surface excitation energy attenuation, realizes multi-depth continuous detection of a 80m or even deeper gravel pile composite foundation, solves the detection blind area below 30m of a traditional method, significantly improves strain transmission efficiency through three-phase coupling of the optical fiber sensor, grouting and gravel, a multi-feature fusion interpretation method based on deep learning is provided, and signal scattering and multi-solution problems caused by the non-continuity of the gravel pile are effectively overcome.

Description

Technical Field

[0001] This invention relates to the field of deep crushed stone pile composite foundation testing technology, specifically to an ultra-deep crushed stone pile composite foundation testing system and method. Background Technology

[0002] As major national infrastructure construction expands into deeper spaces, ultra-deep (≥80m) gravel pile composite foundations are increasingly widely used in hydropower cofferdams, high embankment roadbeds, and soft soil treatment projects. However, existing detection technologies face severe challenges: (a) Depth limitations and blind zone issues: The effective detection depth of traditional low-strain reflection wave methods is usually no more than 30m, and static load tests are difficult to implement on ultra-deep piles due to the limitations of the reaction system, resulting in the lower part of the pile (30m-80m) becoming a "blind zone" for quality control. (b) Signal distortion caused by the special nature of the medium: Gravel piles are composed of uncemented granular materials with a large number of voids and discontinuous interfaces, resulting in complex propagation paths, severe scattering, and rapid attenuation of acoustic / elastic waves. Conventional geophysical exploration methods (such as cross-hole CT and high-density electrical resistivity tomography) have low resolution and serious multiple solutions problems. (c) Single evaluation index and lack of analysis of synergistic mechanisms: Existing methods mostly focus on pile integrity, making it difficult to quantify and evaluate replacement rate, lateral constraint effect, and pile-soil synergistic working mechanism, and cannot accurately detect compaction. (d) Cost-efficiency contradiction: Core drilling is costly and has poor representativeness, making it difficult to meet the full-coverage testing requirements of large-scale crushed stone pile composite foundation projects. Therefore, there is an urgent need to develop a new testing technology that can overcome depth limitations, adapt to the characteristics of granular media, and achieve multi-parameter fusion evaluation. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an ultra-deep crushed stone pile composite foundation testing system and method, which has advantages such as high testing accuracy and wide coverage, and solves the aforementioned technical problems.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a detection system for ultra-deep crushed stone pile composite foundations, comprising... The data acquisition module is used to collect data at different depths within the crushed stone pile group; Excitation and receiving module, used to set up a seismic source excitation device on the corresponding horizontal layer at each depth; The replacement rate and density evaluation module outputs replacement rate and density cloud maps based on the data collected by the data acquisition module; As a preferred technical solution of the present invention, the data acquisition module uses sensors deployed in the crushed stone pile group, specifically DAS armored optical cables. The deployment of the DAS armored optical cable is specifically as follows: it includes deployment in the depth direction and the horizontal direction. Specifically, in the depth direction, multiple horizontal layers are selected at equal intervals or according to a preset depth based on the design depth of the crushed stone pile group, and a horizontal detection network is deployed on the horizontal layer corresponding to each depth.

[0005] As a preferred technical solution of the present invention, the displacement rate and density evaluation module outputs displacement rate and density cloud maps specifically including the following steps: Step A1: Use tomographic imaging technology to divide the data acquisition module into grids and obtain the spatial distribution cloud map and wave impedance of the P-wave velocity and S-wave velocity of the horizontal layer. Step A2: Evaluate the replacement rate of the crushed stone pile composite foundation based on the deep learning model and obtain the average replacement rate; Step A3: Density contour map based on the density assessment model; Step A4: Output the average displacement rate and density contour maps respectively.

[0006] As a preferred technical solution of the present invention, the deep learning model in step A2 is specifically Attention U-Net, which converts the deep learning model obtained from earthquake inversion into a deep learning model. The velocity distribution map and acoustic impedance property volume at a given location are input into the deep learning model, and the output is a probability map with the same size as the input data. And it was determined to be a target body or foundation of crushed stone piles; Count the total number of pixels with a value of 1 as the nth depth layer Area of ​​the target body ; Get the The displacement rate and average displacement rate of the depth layer are specifically expressed as follows: in, This represents the average replacement rate of the crushed stone pile composite foundation. This indicates the number of layers selected for measurement at different depths. Indicates the first Layer depth replacement rate, Indicates the first depth layer The area of ​​the target body This indicates the area of ​​the foundation being treated.

[0007] As a preferred technical solution of the present invention, step A3 specifically includes: Step A3.1: Obtain the dataset for model learning; Step A3.2: Based on the depth obtained from the seismic inversion in step A1 velocity distribution map at ( , ) and the ratio of longitudinal to transverse waves ( ) Obtain the model's data input ; Step A3.3: Input the model data The input is fed into a fully connected layer for convolution, followed by spatial downsampling, using a stride. Convolutions obtain low-resolution feature maps in space; Step A3.4: Divide the low-resolution feature map in space into multiple residual blocks to obtain an abstract feature map; Step A3.5: Restore the abstract feature map to its original size by transposing the convolution, and decode it to obtain the density cloud map.

[0008] The present invention also provides a method for detecting ultra-deep crushed stone pile composite foundations, which uses the above-mentioned ultra-deep crushed stone pile composite foundation detection system to detect ultra-deep crushed stone pile composite foundations.

[0009] Compared with the prior art, the present invention provides a detection system and method for ultra-deep crushed stone pile composite foundations, which has the following beneficial effects: 1. This invention avoids the problem of surface excitation energy attenuation by deploying distributed optical fiber sensors and deep excitation sources through horizontal directional drilling, and realizes multi-depth continuous detection of crushed stone pile composite foundations at depths of 80m or even deeper, solving the detection blind zone of traditional methods below 30m.

[0010] 2. This invention significantly improves strain transfer efficiency by utilizing the three-phase coupling of fiber optic sensors, grouting, and crushed stone; it also proposes a multi-feature fusion interpretation method based on deep learning, which effectively overcomes the signal scattering and multiple solutions problems caused by the discontinuity of crushed stone piles. Attached Figure Description

[0011] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the equipment layout of the present invention. Detailed Implementation

[0012] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0013] Please see Figure 1 A detection system for ultra-deep crushed stone pile composite foundations, including The data acquisition module is used to collect data within the crushed stone pile group, specifically including: 3D HDD network design: based on the design depth of the gravel pile group ( (≥80m) Multi-layer horizontal directional drilling was carried out in the area of ​​the composite foundation of crushed stone piles to be tested.

[0014] Depth direction: Multiple horizontal layers are selected at equal intervals or within the critical depth range. , , ... ... , (The number of layers selected for measurement at different depths).

[0015] Horizontal direction: Within each depth layer, the drilling trajectory is designed to traverse the soil between the piles in a serpentine pattern, centered on the crushed stone pile group. In areas with complex geology, horizontal boreholes can be designed in a crisscross pattern, at a depth of... A detection network was formed.

[0016] Sensor selection and composite implantation: DAS (Distributed Acoustic Sensing) armored optical cable is used and implanted into the HDD borehole. DAS is used to receive high-frequency seismic wave signals to evaluate the compaction and replacement rate of the crushed stone pile.

[0017] High-fidelity coupling process: After the distributed fiber optic sensors are laid, micro-expansion cement-based slurry is pumped into the borehole through the HDD drill pipe. This eliminates the gap between the fiber optic cable and the borehole wall, enabling good transmission of seismic wave signals and strain fields.

[0018] Excitation and Receiving Module: Source Excitation System Configuration: At each depth layer ( A controllable spark source or electromagnetic source is installed at the end of the horizontal borehole or at a specific node. The broadband seismic wave signal excited by the source propagates through the gravel piles and soil / strata, and is then distributed synchronously acquired by DAS optical cables in the same or adjacent layers. Replacement rate and density assessment module; Step A1: Using tomographic imaging technology, the affected area is meshed based on the seismic signals received at each sampling point. The mesh parameters are then used... This indicates that the corresponding grid is set up for imaging, and the horizontal layer is inverted. The longitudinal wave velocity and transverse wave velocity ( , Spatial distribution cloud map and wave impedance; Step A2: Evaluation of the replacement rate of the gravel pile composite foundation based on the deep learning model (Attention U-Net), specifically including the following steps: Step A2.1: Data Input: Input the depth obtained from seismic inversion. The velocity distribution map and acoustic impedance properties are input to the network.

[0019] Step A2.2: Multi-scale extraction: In deep learning networks, after multiple convolutional layers and non-linear activation functions, the calculated feature volume is an abstract feature of attributes such as coherence volume, variance volume, and gradient volume. This abstract feature is automatically learned by the convolutional layers, and the parameters (weights) are continuously adjusted to ultimately extract the feature volume corresponding to a specific task in the optimal way.

[0020] Step A2.3: The entire intelligent recognition process is divided into shallow networks and deep networks: shallow networks are defined as the first few layers (with outputs) that record the boundaries and shapes of the abnormal objects; deep networks are defined as the entire network (including subsequent layers) that uses convolutional layers in the middle of the network to receive the feature data from the shallow networks and to identify the type of the target object.

[0021] Step A2.3: Attention Allocation: At each layer's jump connection, attention gating automatically calculates the region that contributes most to the final anomaly delineation. For heterogeneous anomalies in seismic data, attention gating significantly amplifies the response at anomaly boundaries.

[0022] Step A2.4: Feature Fusion: The decoder combines the attention-filtered features, i.e., the low-resolution but high-level semantic information of the target object obtained through multiple downsampling, with the upsampled features; Step A2.5: The output layer uses the common Sigmoid activation function, outputting a probability map P with the same size as the input data. The target body is identified as a gravel pile; if It is determined to be the background (foundation or normal strata).

[0023] Step A2.6: Location delineation of crushed stone piles: By performing connected component analysis on the output binarized mask, the geometric center of the target body is calculated, and the center of the anomaly is accurately located in the seismic spatial coordinate system.

[0024] Step A2.7: Evaluation of the size of the crushed stone pile: Count the total number of pixels with a value of 1 in the mask, and combine this with the grid parameters set in the seismic interpretation. The number of grid cells in the target object's projection is calculated, and then the area of ​​the target object is obtained by accumulating these cells. , No. depth layer ).

[0025] Step A2.8: Evaluation model for plane replacement rate and pile range: The proportion of the crushed stone pile area obtained according to A2.7 (No. depth layer ), combined with the treated foundation area Calculate depth The layer replacement rate is: in This represents the average replacement rate of the crushed stone pile composite foundation. This indicates the number of layers selected for measurement at different depths. , , ... ), Indicates the first Layer depth replacement rate, Indicates the first The area of ​​all identified crushed stone piles in the layer. This indicates the area of ​​the foundation being treated.

[0026] Output: Different depths Layer displacement rate and average replacement rate and the ( The range of the crushed stone piles in the ) layer; Step A3: Compaction Assessment Model: Step A3.1: Obtain the dataset for model learning. Based on the laboratory model and a small amount of well logging data, establish a velocity and density label dataset, including density label data for crushed stone piles, foundations, or normal strata (background soil).

[0027] Step A3.2: Data Input for the Model The depth obtained from the seismic inversion in step A1 velocity distribution map at ( , ) and the ratio of longitudinal to transverse waves ( ),Right now First, these three two-dimensional data volumes (each with a size of ( Stacked in three channels, forming ( The initial input tensor enables the model to perceive multiple physical properties of the same space simultaneously.

[0028] Step A3.3: Spatial transformation involves mapping the feature space of the input data. In the fully connected layer, for a two-dimensional data volume, this is achieved through (…). The convolutional form does not change the spatial resolution. For three physical features at a spatial point, the fully connected layer uses ( The convolution of the original data maps to a deeper dimension. ) becomes ( ), The selected dimension. In terms of spatial scale, spatial downsampling is performed to extract global geological topological features, using a step size... Convolution (stride is) )Will( Compressed to ( This yields a low-resolution feature map in space, i.e. ( This allows for the increase of the perceived field of view while removing some noise from the velocity data.

[0029] Step A3.4: Encoding and feature evolution, involving residual convolution model, the feature map is passed through multiple residual blocks, allowing the model to learn the residual increment of compaction relative to the velocity field; nonlinear spatial feature extraction, through the sliding of the convolution kernel, captures the wave velocity gradient anomaly caused by the compaction of the crushed stone pile.

[0030] Step A3.5: Density Convolution: Transposed convolution restores the abstract feature map processed in A3.4 to its original size step by step. The skip connection directly transmits the high-resolution physical details preserved in the early stages of A3.4 to the decoder, ensuring that the pile-soil interface (i.e., the edge of the pile diameter) in the cloud map is not too smooth or blurry.

[0031] Step A3.6: Loss function constraints: Mean square error (MSE) ensures that the density value of each pixel on the cloud image is aligned with the measured value (such as the core sample result); Structural Similarity Index (SSIM) constrains the geometry of the crushed stone pile body to prevent the appearance of scattered spots or broken pile bodies that do not conform to geological logic; Step A4: Output the average displacement rate and density contour maps respectively.

[0032] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A detection system for ultra-deep crushed stone pile composite foundations, characterized in that: include The data acquisition module is used to collect data at different depths within the crushed stone pile group; Excitation and receiving module, used to set up a seismic source excitation device on the corresponding horizontal layer at each depth; The replacement rate and density assessment module outputs replacement rate and density cloud maps based on the data collected by the data acquisition module.

2. The ultra-deep crushed stone pile composite foundation testing system according to claim 1, characterized in that: The data acquisition module uses sensors, specifically DAS armored optical cables, deployed within the pile group. The deployment of the DAS armored optical cable includes both depth and horizontal directions. Specifically, in the depth direction, multiple horizontal layers are selected at equal intervals or according to a preset depth based on the design depth of the crushed stone pile group, and a horizontal detection network is deployed on each corresponding horizontal layer.

3. The ultra-deep crushed stone pile composite foundation testing system according to claim 2, characterized in that: The displacement rate and density assessment module outputs displacement rate and density contour maps, specifically including the following steps: Step A1: Use tomographic imaging technology to divide the data acquisition module into grids and obtain the spatial distribution cloud map and wave impedance of the P-wave velocity and S-wave velocity of the horizontal layer. Step A2: Evaluate the replacement rate of the crushed stone pile composite foundation based on the deep learning model and obtain the average replacement rate; Step A3: Compaction cloud map calculated based on the compaction assessment model; Step A4: Output the average displacement rate and density contour maps respectively.

4. The ultra-deep crushed stone pile composite foundation testing system according to claim 3, characterized in that: The deep learning model in step A2 is specifically Attention U-Net, which uses the deep learning model obtained from earthquake inversion. The velocity distribution map and acoustic impedance property volume at a given location are input into the deep learning model, and the output is a probability map with the same size as the input data. And it was determined to be a target body or foundation of crushed stone piles; Count the total number of pixels with a value of 1 as the nth depth layer Area of ​​the target body ; Get the The displacement rate and average displacement rate of the depth layer are specifically expressed as follows: in, This represents the average replacement rate of the crushed stone pile composite foundation. This indicates the number of layers selected for measurement at different depths. Indicates the first Layer depth replacement rate, Indicates the first depth layer The area of ​​the target body This indicates the area of ​​the foundation being treated.

5. The ultra-deep crushed stone pile composite foundation testing system according to claim 4, characterized in that: Step A3 specifically includes: Step A3.1: Obtain the dataset for model learning; Step A3.2: Based on the depth obtained from the seismic inversion in step A1 velocity distribution map at ( , ) and the ratio of longitudinal to transverse waves ( ) Obtain the model's data input ; Step A3.3: Input the model data The input is fed into a fully connected layer for convolution, followed by spatial downsampling, using a stride. Convolutions obtain low-resolution feature maps in space; Step A3.4: Divide the low-resolution feature map in space into multiple residual blocks to obtain an abstract feature map; Step A3.5: Restore the abstract feature map to its original size by transposing the convolution, and decode it to obtain the density cloud map.

6. A method for testing ultra-deep crushed stone pile composite foundations, characterized in that: The ultra-deep crushed stone pile composite foundation testing system described in any one of claims 1-5 is used for ultra-deep crushed stone pile composite foundation testing.

Images