An online detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silt layers

Abstract

The present invention belongs to the technical field of on-line detection of pile foundation anti-collapse holes, and in particular to an on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers, comprising the following steps: starting the operation of the drill pipe machine to drive the drill pipe to drill downward, and while the drill pipe is drilling, transmitting sound waves through a sound wave emission probe installed on the drill pipe. This on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers can, through the on-line detection method, monitor the state of the hole wall during the drilling process in real time, timely detect potential collapse risks, and thus take preventive measures to avoid the occurrence of collapse accidents; through sound wave detection and analysis, it can accurately identify cracks and cavities in the hole wall, evaluate the stability of the hole wall, and ensure that the quality of the drilling meets the engineering requirements; by using a sound wave transmitter, a receiver, a data acquisition module and a processing unit, the automation and intelligence of the detection process are realized, manual intervention is reduced, and the detection efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of on-line detection of pile foundation anti-collapse holes, and particularly to an on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers. Background Art

[0002] With the acceleration of the urbanization process and the vigorous development of infrastructure construction, pile foundation engineering, as a key supporting part of buildings, its construction quality and safety have been increasingly emphasized. However, when carrying out pile foundation drilling construction under geological conditions such as alluvial plains rich in silty sand layers, serious problems such as hole collapse, hole wall cracks and cavities are often faced. The existence of these problems not only affects the quality of the drilling hole, resulting in a decrease in the bearing capacity of the pile foundation, but also may trigger construction safety accidents, causing economic losses and casualties.

[0003] Traditional methods are all through manual detection or detection methods relying on construction experience. These methods have many drawbacks, such as low detection efficiency, poor accuracy, inability to monitor in real time, etc., and are difficult to meet the high standards of modern construction. In addition, due to the complexity of the geological conditions of the silty sand layer, conventional detection means are often difficult to effectively identify the fine cracks and hidden cavities on the hole wall, resulting in hidden dangers not being dealt with in time. Therefore, an on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers is needed. Summary of the Invention

[0004] Based on the existing technical problems, the present invention proposes an on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers.

[0005] An on-line detection method for pile foundation anti-collapse holes in alluvial plains rich in silty sand layers proposed by the present invention includes the following steps: Step 1: Start the drill rod machine to drive the drill rod to drill downward. While the drill rod is drilling, the acoustic wave transmitter installed on the drill rod emits acoustic waves, and the acoustic wave receivers arranged at different positions on the arc surface of the drill rod receive the acoustic waves reflected at different positions. Moreover, a plurality of acoustic wave receivers are arranged in a spiral equidistant arrangement on the arc surface of the drill rod, and a plurality of acoustic wave receivers are also symmetrically distributed with respect to the vertical direction center of the drill rod. The data of the acoustic wave emission is collected through the data acquisition module, and the time relay is used to record the time difference between the acoustic wave emission and reception, and the data is transmitted to the processing unit.

[0006] Step 2: First perform noise reduction processing on the collected acoustic wave signals. The noise reduction processing is carried out by performing convolution operation on the Gaussian blur filter and the acoustic wave signals.

[0007] Step 3: Reconstruct the image of the hole wall using the processed data; when sound waves encounter media with different acoustic impedances, energy reflection occurs; by analyzing the intensity and time of the reflected waves, the properties and positions of the media are inferred; use a sound wave emission probe to emit a sound wave pulse and use a receiver to record the time when the sound wave arrives and the time when the reflected wave returns; this time difference is the total time for the sound wave to travel back and forth between the emission point and the reflection interface.

[0008] Step 4: Analyze the image, apply the trained CNN model to the test set to detect cracks and voids; identify cracks and voids in the hole wall and evaluate the stability of the hole wall based on the attenuation and reflection characteristics of the sound wave energy.

[0009] Step 5: According to the CNN model applied to the test set, evaluate and make a comparison judgment. When it is unqualified, drive the concrete pouring equipment to fill the inside of the detected drilling hole.

[0010] Preferably, the data collected in Step 1 are the sound wave velocity, sound wave attenuation coefficient, and reflection coefficient:

[0011] Calculation of sound wave velocity: ; where is the sound wave velocity, is the distance traveled by the sound wave, is the time taken for the sound wave to travel.

[0012] Sound wave attenuation coefficient: ; where is the attenuation coefficient, is the initial sound intensity, is the propagation distance and the sound intensity after that.

[0013] Reflection coefficient: ; where is the reflection coefficient, is the reflected sound intensity, is the incident sound intensity.

[0014] Preferably, the formula for Gaussian blur filtering in Step 2 is: ; where represents the value of the Gaussian function, which is a function of two variables and , is the coordinate relative to the center point within the kernel, is the standard deviation of the Gaussian distribution, represents pi; represents the base of the natural logarithm, represents the point and the square of the Euclidean distance from the origin.

[0015] To generate a Gaussian kernel, calculate the weights of each point within the kernel: ; where is the unnormalized Gaussian kernel weight, denotes the summation over all points within the kernel.

[0016] For the acoustic wave signal The process of applying Gaussian filtering is expressed as: ; where is the filtered signal, is the kernel radius, is the value of the original signal at position .

[0017] Preferably, the steps for calculating the position of the reflection interface in step three are as follows: a1. Record the time difference; record the time when the acoustic wave is emitted ; record the time when the acoustic wave is reflected back and captured by the receiver ; calculate the time difference: ; a2. Calculate the one-way travel time; since the recorded time is for the round trip of the acoustic wave, the time difference needs to be divided by 2 to obtain the one-way travel time : ; a3. Calculate the distance; use the acoustic wave velocity and the one-way travel time to calculate the distance from the reflection interface to the transmitter : .

[0018] Based on the calculated distance positions of the reflection interfaces, mark each reflection interface on the borehole section. When the borehole wall is layered, each reflection interface represents a layer with different acoustic properties; since the initial borehole wall forms a layered structure, the thickness and acoustic properties of each layer are initially estimated; use the reflection interfaces as the boundaries between different layers, and based on the intensity and waveform characteristics of the reflection signals, initially estimate the acoustic properties of each layer; assign initially estimated acoustic parameters to each layer, and the parameters include sound velocity, density, and attenuation coefficient; use geological modeling software to visualize the parameter information as an initial borehole wall model; use the initial borehole wall model to perform forward modeling of acoustic wave propagation to predict the reflection and propagation of acoustic waves at different positions.

[0019] Preferably, in step four, when analyzing the image, use a trained CNN model to analyze the image of the borehole inner wall;

[0020] b1. Image normalization: Scale the image pixel values to a fixed range, select ;

[0021] ; where ​​​is the original image, and are the minimum and maximum pixel values of the image, respectively.

[0022] b2. Feature extraction: Use filters of different sizes in the convolutional layer to capture features at different scales: ; where is the scaled image, is the scaling ratio, is a function name representing the operation of performing image scaling; Extract features through the convolutional layer: ; is the feature map, is the convolutional kernel, is the bias term, represents the convolutional operation.

[0023] c3. Model training: First, construct the CNN architecture: Design a network containing convolutional layers, pooling layers, activation functions, and fully connected layers: ; is the output, is the input image, is the network weight, represents the convolutional neural network; Then, select a loss function to evaluate the performance of the model, using: ; where represents the true label, represents the predicted probability, represents the loss; Finally, perform the optimization algorithm and use the SGD optimization algorithm to update the network weight: ; where, represents the learning rate, represents the gradient of the loss function with respect to the weight.

[0024] c4. Model evaluation: Evaluate the performance of the model on an independent validation set, using metrics such as accuracy, recall, and F1-score: ; ; ; where, is the number of samples correctly identified as cracks or voids, is the number of samples that are incorrectly identified as cracks or voids when they are not, is the number of samples that fail to correctly identify actual cracks or voids, is the precision, representing the proportion of actual positive samples among all samples predicted as positive, is the recall, representing the proportion of samples correctly predicted as positive among all actual positive samples, represents the harmonic mean of precision and recall, used to measure the accuracy and comprehensiveness of the model.

[0025] Preferably, based on the high accuracy and high recall values measured in Step 4, a CNN model applied to the test set is used to determine whether the cracks and cavities on the inner wall of the drill hole are qualified; if the detected cracks and cavities are lower than or meet the preset standards in all aspects, it is determined that the inner wall of the drill hole is qualified, otherwise it is determined as unqualified. When it is unqualified, the concrete pouring equipment is driven to pour and fill the detected drill hole.

[0026] Preferably, in Step 1, a circumferential motion launching mechanism and a vertical motion launching mechanism are respectively arranged on the surface of the drill pipe of the drill pipe machine. The circumferential motion launching mechanism includes a turntable bearing. A driving groove is formed on the arc surface of the drill pipe. The top arc surface and the bottom arc surface of the driving groove are fixedly installed with the inner ring of the turntable bearing, and the outer ring of the turntable bearing is fixedly installed with a driving rotating ring.

[0027] Preferably, a toothed ring is further fixedly installed on the arc surface at the bottom of the driving groove. The toothed ring is located at the top of the lower turntable bearing. Vertical arc plates are fixedly installed on the opposite surfaces of the two driving rotating rings. A rotating motor is fixedly installed on the bottom arc surface of the vertical arc plate. The output end of the rotating motor is fixedly installed with a rotating gear through a coupling. The teeth of the rotating gear are engaged with the tooth grooves of the toothed ring.

[0028] Preferably, the vertical motion launching mechanism includes fixing plates installed on the top surface and the bottom surface of the vertical arc plate. The opposite surfaces of the two fixing plates are rotatably connected with threaded rods through bearings. A vertical driving motor is further fixedly installed at the bottom of the bottom fixing plate. The output end of the vertical driving motor is fixedly installed with the bottom of the threaded rod through a coupling.

[0029] Preferably, a slider is threadedly connected to the arc surface of the threaded rod. A vertical sliding groove is formed on the surface of the vertical arc plate. The inner wall of the vertical sliding groove is slidably inserted with the surface of the slider. An installation plate is fixedly installed on the top of the slider. The surface of the installation plate is fixedly installed with the installation surface of the acoustic wave transmitting probe.

[0030] The beneficial effects of the present invention are as follows:

[0031] 1. Through the circumferential motion emission mechanism, sound waves can conduct all-round scanning along the arc surface of the drill pipe, ensuring that every part of the inner wall of the borehole is covered by the detection without leaving any blind spots; the vertical motion emission mechanism can ensure accurate detection of sound waves in the vertical direction of the drill pipe, thereby improving the recognition accuracy of wall cracks and cavities; by combining circumferential motion and vertical motion, multi-dimensional sound wave detection can be achieved, providing a more comprehensive understanding of the state of the borehole wall; it can not only improve the detection coverage and accuracy, but also enhance the detection efficiency, reduce equipment wear, enhance equipment stability, facilitate maintenance and replacement, adapt to different geological conditions, improve data acquisition quality, and ultimately enhance the safety of construction.

[0032] 2. Through the online detection method, the state of the borehole wall during the drilling process can be monitored in real time, potential borehole collapse risks can be detected in a timely manner, and preventive measures can be taken to avoid the occurrence of borehole collapse accidents; through sound wave detection and analysis, cracks and cavities in the borehole wall can be accurately identified, the stability of the borehole wall can be evaluated, and the quality of the borehole can be ensured to meet the engineering requirements; by using sound wave transmitters, receivers, data acquisition modules and processing units, the automation and intelligence of the detection process have been realized, reducing manual intervention and improving the detection efficiency; through Gaussian fuzzy filtering for noise reduction and CNN model analysis, the accuracy of sound wave signal processing and the precision of image recognition have been improved, providing reliable data support for the evaluation of the borehole wall state; when an unqualified borehole wall is detected, the system can quickly drive the concrete pouring equipment to fill it, handle the problem in a timely manner, ensuring the continuity and safety of construction; this method is applicable to pile foundation construction in alluvial plains rich in silty sand layers, has strong environmental adaptability, and can be popularized and applied to borehole construction in similar geological conditions; by establishing a borehole wall model and visual display, the state of the borehole wall becomes more intuitive, facilitating analysis and decision-making, and at the same time providing traceable data records for subsequent construction and quality control; through real-time monitoring, automated processing, precise analysis and rapid response, the quality, safety and efficiency of borehole construction have been effectively improved, and the construction cost has been reduced, with significant beneficial effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 It is a schematic structural diagram of an online detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty sand layers;

[0034] Figure 2 It is a three-dimensional diagram of the drill pipe structure of an online detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty sand layers;

[0035] Figure 3 It is a three-dimensional diagram of the driving groove structure of an online detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty sand layers;

[0036] Figure 4 It is an online detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty sand layers Figure 3Three-dimensional view of the structure at location A;

[0037] Figure 5 For an on-line detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty layers Figure 2 Three-dimensional view of the structure at location B;

[0038] Figure 6 Three-dimensional view of the driving swivel structure for an on-line detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty layers.

[0039] In the figure: 1, drill rod machine; 2, drill rod; 3, acoustic wave emission probe; 4, acoustic wave receiver; 5, circumferential motion emission mechanism; 51, turntable bearing; 52, driving groove; 53, driving swivel; 54, toothed ring; 55, vertical arc plate; 56, rotating motor; 57, rotating gear; 6, vertical motion emission mechanism; 61, fixed plate; 62, threaded rod; 63, vertical driving motor; 64, slider; 65, vertical sliding groove; 66, mounting plate. Specific implementation method

[0040] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments.

[0041] Refer to Figures 1-6 , an on-line detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silty layers, including the following steps: Step 1, start the drill rod machine 1 to drive the drill rod 2 to drill downward. While the drill rod 2 is drilling, the acoustic wave emission probe 3 installed on the drill rod 2 emits acoustic waves, and the acoustic wave receivers 4 arranged at different positions on the arc surface of the drill rod 2 receive the acoustic waves reflected from different positions. Moreover, multiple acoustic wave receivers 4 are arranged in a spiral equidistant manner on the arc surface of the drill rod 2, and multiple acoustic wave receivers 4 are also symmetrically distributed with respect to the vertical direction center of the drill rod 2. The data of the acoustic wave emission is collected through the data acquisition module, and the time difference between the acoustic wave emission and reception is recorded by using a time relay, and the data is transmitted to the processing unit.

[0042] The data collected in Step 1 are acoustic wave velocity, acoustic wave attenuation coefficient, and reflection coefficient:

[0043] Calculation of acoustic wave velocity: ; where is the acoustic wave velocity, is the distance traveled by the acoustic wave, is the time taken for the acoustic wave to travel;

[0044] Acoustic wave attenuation coefficient: ; where is the attenuation coefficient, is the initial sound intensity, is the propagation distance The sound intensity after

[0045] Reflection coefficient: ; where is the reflection coefficient, is the reflected sound intensity, is the incident sound intensity.

[0046] In step one, a circumferential motion transmitting mechanism 5 and a vertical motion transmitting mechanism 6 are respectively arranged on the surface of the drill pipe 2 of the drill pipe machine 1. The circumferential motion transmitting mechanism 5 includes a turntable bearing 51. A driving groove 52 is formed on the arc surface of the drill pipe 2. The top arc surface and the bottom arc surface of the driving groove 52 are fixedly installed with the inner ring of the turntable bearing 51. The outer ring of the turntable bearing 51 is fixedly installed with a driving ring 53.

[0047] Specifically implemented in this way, through the circumferential motion transmitting mechanism 5, sound waves can perform an all-round scan along the arc surface of the drill pipe 2 to ensure that every part of the inner wall of the drill hole is covered by the detection without leaving dead corners; the vertical motion transmitting mechanism 6 can ensure that sound waves can also perform accurate detection in the vertical direction of the drill pipe 2, thereby improving the recognition accuracy of cracks and cavities in the hole wall; combining circumferential motion and vertical motion can achieve multi-dimensional sound wave detection and more comprehensively understand the state of the hole wall; it can not only improve the detection coverage and accuracy, but also improve the detection efficiency, reduce equipment wear, enhance equipment stability, facilitate maintenance and replacement, adapt to different geological conditions, improve the quality of data collection, and ultimately enhance the safety of construction.

[0048] The bottom arc surface of the driving groove 52 is also fixedly installed with a toothed ring 54. The toothed ring 54 is located on the top of the lower turntable bearing 51. The opposite surfaces of the two driving rings 53 are fixedly installed with vertical arc plates 55. The bottom arc surface of the vertical arc plate 55 is fixedly installed with a rotary motor 56. The output end of the rotary motor 56 is fixedly installed with a rotary gear 57 through a coupling. The teeth of the rotary gear 57 are engaged with the tooth grooves of the toothed ring 54.

[0049] Specifically implemented in this way, the meshing design of the toothed ring 54 and the rotary gear 57 ensures the precise control of the rotary motion, enabling the sound wave transmitting mechanism to rotate according to the preset trajectory and speed, thereby improving the accuracy and consistency of the detection.

[0050] The vertical motion transmitting mechanism 6 includes fixing plates 61 installed on the top surface and the bottom surface of the vertical arc plate 55. The opposite surfaces of the two fixing plates 61 are rotatably connected with a threaded rod 62 through bearings. The bottom of the bottom fixing plate 61 is also fixedly installed with a vertical driving motor 63. The output end of the vertical driving motor 63 is fixedly installed with the bottom of the threaded rod 62 through a coupling.

[0051] Specifically, through the rotational connection between the threaded rod 62 and the fixed plate 61 and the drive of the vertical drive motor 63, precise control of the vertical movement can be achieved, ensuring the accurate position of the acoustic wave emission mechanism in the vertical direction; the design of components such as the fixed plate 61, threaded rod 62, vertical drive motor 63, and coupling in the vertical movement emission mechanism 6 not only realizes precise, stable, and smooth vertical movement control, but also improves the adjustment efficiency, reduces manual intervention, enhances the consistency and repeatability of the equipment, adapts to different depth requirements, improves energy utilization efficiency, facilitates maintenance and replacement, enhances the safety of the equipment, and realizes automated and intelligent control, ultimately improving the accuracy of data collection.

[0052] The arc surface of the threaded rod 62 is threadedly connected with a slider 64. A vertical chute 65 is provided on the surface of the vertical arc plate 55. The inner wall of the vertical chute 65 is slidably inserted into the surface of the slider 64. The top of the slider 64 is fixedly installed with a mounting plate 66. The surface of the mounting plate 66 is fixedly installed with the mounting surface of the acoustic wave emission probe 3.

[0053] Specifically, the threaded connection between the threaded rod 62 and the slider 64 allows for precise adjustment of the position of the slider 64 by rotating the threaded rod 62, thereby realizing the precise positioning and adjustment of the acoustic wave emission probe 3; the sliding insertion design of the slider 64 and the vertical chute 65 reduces friction and resistance during movement, making the movement of the acoustic wave emission probe 3 smoother; by rotating the threaded rod 62, the height and position of the acoustic wave emission probe 3 can be easily adjusted to adapt to different detection requirements; it realizes the precise, stable, and flexible adjustment of the acoustic wave emission probe 3, improves the adaptability, detection efficiency, and maintenance convenience of the equipment, and at the same time ensures the consistency and safety of detection, providing guarantee for high-quality data collection.

[0054] During operation, the rotation motor 56 works to drive the rotation gear 57 to rotate. Then, the rotation gear 57 rotates in a circular motion on the surface of the toothed ring 54, thereby driving the vertical arc plate 55 to perform a circular motion. During the circular motion of the vertical arc plate 55, the acoustic wave emission probe 3 provided on the surface is driven to perform a circular motion to emit acoustic waves. At the same time, the vertical drive motor 63 can also be controlled to work, driving the threaded rod 62 to rotate. Then, under the sliding cooperation of the vertical chute 65, the slider 64 slides in the vertical direction, driving the acoustic wave emission probe 3 to perform a vertical movement to emit acoustic waves.

[0055] Step 2: First, perform denoising processing on the collected acoustic wave signals. The denoising processing is carried out by performing a convolution operation on the Gaussian blur filter and the acoustic wave signals.

[0056] The formula for the Gaussian blur filter in Step 2 is: ; where represents the value of the Gaussian function, which is a function of two variables and function, is the coordinate within the nucleus relative to the center point, is the standard deviation of the Gaussian distribution, is represented as pi; is represented as the base of the natural logarithm, represents the point to the square of the Euclidean distance from the origin;

[0057] To generate the Gaussian kernel, calculate the weight of each point within the kernel: ; where, is the unnormalized Gaussian kernel weight, represents the sum over all points within the kernel;

[0058] For the acoustic wave signal The process of applying Gaussian filtering is expressed as: ; where, is the filtered signal, is the kernel radius, is the original signal at the position value.

[0059] Step 3. Through the processed data, use an algorithm to reconstruct the image of the hole wall; when an acoustic wave encounters a medium with different acoustic impedances, energy reflection will occur; by analyzing the intensity and time of the reflected wave, infer the properties and position of the medium; use the acoustic wave emission probe 3 to emit an acoustic wave pulse, and use the receiver to record the time when the acoustic wave arrives and the time when the reflected wave returns; this time difference is the total time for the acoustic wave to travel back and forth between the emission point and the reflection interface.

[0060] The steps to calculate the position of the reflection interface in Step 3: a1. Record the time difference; record the time when the acoustic wave is emitted ; record the time when the acoustic wave is reflected back and captured by the receiver ; calculate the time difference: ; a2. Calculate the one-way time; since the recorded time is for the round trip of the acoustic wave, the time difference needs to be divided by 2 to obtain the one-way time : ; a3. Calculate the distance; use the acoustic wave velocity and the one-way time to calculate the distance from the reflection interface to the transmitter : ;

[0061] Mark each reflection interface on the borehole profile based on the calculated distance position of the reflection interface. When the borehole wall is layered, each reflection interface represents a layer with different acoustic properties. Since the initial borehole wall forms a layered structure, the thickness and acoustic properties of each layer are initially estimated. Use the reflection interface as the boundary between different layers and preliminarily estimate the acoustic properties of each layer according to the intensity and waveform characteristics of the reflection signal. Assign preliminarily estimated acoustic parameters to each layer, including sound velocity, density, and attenuation coefficient. Use geological modeling software to visualize the parameter information as an initial borehole wall model. Use the initial borehole wall model to perform forward simulation of acoustic wave propagation and predict the reflection and propagation of acoustic waves at different positions.

[0062] Step 4: Analyze the image, apply the trained CNN model to the test set for crack and void detection; identify cracks and voids in the borehole wall and evaluate the stability of the borehole wall based on the attenuation and reflection characteristics of acoustic waves.

[0063] In Step 4, when analyzing the image, use the trained CNN model to analyze the image of the inner wall of the borehole.

[0064] b1. Image normalization: Scale the pixel values of the image to a fixed range, and select ; where

[0065] ; among them, is the original image, and are the minimum and maximum pixel values of the image respectively;

[0066] b2. Feature extraction: Use filters of different sizes in the convolutional layer to capture features of different scales: ; where is the scaled image, is the scaling ratio, is a function name representing the operation of performing image scaling; extract features through the convolutional layer: ; is the feature map, is the convolutional kernel, is the bias term, represents the convolutional operation;

[0067] c3. Model training: First, construct the CNN architecture: Design a network including convolutional layers, pooling layers, activation functions, and fully connected layers: ; is the output, is the input image, is the network weight, represents the convolutional neural network; Select a loss function to evaluate the performance of the model, and adopt: ; where represents the true label, represents the predicted probability, represents the loss; finally, an optimization algorithm is performed, using the SGD optimization algorithm to update the network weights: ; where, represents the learning rate, represents the gradient of the loss function with respect to the weights;

[0068] c4. Model evaluation: Evaluate the performance of the model on an independent validation set, using metrics such as accuracy, recall, and F1-score: ; ; ; where, is the number of samples correctly identified as cracks or voids, is the number of samples that are incorrectly identified as cracks or voids when they are not, is the number of samples that incorrectly fail to identify actual cracks or voids, is the precision, which represents the proportion of samples that are actually positive among all samples predicted as positive, is the recall, which represents the proportion of samples that are correctly predicted as positive among all samples that are actually positive, represents the harmonic mean of precision and recall, which is used to measure the accuracy and comprehensiveness of the model.

[0069] Step Five, based on the CNN model applied to the test set, conduct a comparative judgment. When it fails, drive the concrete pouring equipment to fill the inside of the detected borehole; through the high accuracy and high recall values measured in Step Four, use the CNN model applied to the test set to determine whether the cracks and voids on the inner wall of the borehole are qualified; if the detected cracks and voids are lower than or meet the preset standards in all aspects, it is determined that the inner wall of the borehole is qualified, otherwise it is determined as unqualified. When it is unqualified, drive the concrete pouring equipment to fill the inside of the detected borehole.

[0070] Through the online detection method, the wall state of the borehole can be monitored in real time, potential borehole collapse risks can be detected in a timely manner, and preventive measures can be taken to avoid the occurrence of borehole collapse accidents; through acoustic wave detection and analysis, cracks and cavities in the borehole wall can be accurately identified, the stability of the borehole wall can be evaluated, and the quality of the borehole can be ensured to meet the engineering requirements; by using acoustic wave transmitters, receivers, data acquisition modules and processing units, the automation and intelligence of the detection process are realized, manual intervention is reduced, and the detection efficiency is improved; through Gaussian fuzzy filtering denoising and CNN model analysis, the accuracy of acoustic wave signal processing and the precision of image recognition are improved, providing reliable data support for the evaluation of the borehole wall state; when an unqualified borehole wall is detected, the system can quickly drive the concrete pouring equipment to fill it, handle the problem in a timely manner, and ensure the continuity and safety of the construction; this method is applicable to the pile foundation construction in alluvial plains rich in silty sand layers, has strong environmental adaptability, and can be popularized and applied to borehole construction in similar geological conditions; by establishing a borehole wall model and visual display, the borehole wall state becomes more intuitive, facilitating analysis and decision-making, and at the same time providing traceable data records for subsequent construction and quality control; through real-time monitoring, automated processing, precise analysis and rapid response, the quality, safety and efficiency of borehole construction are effectively improved, the construction cost is reduced, and significant beneficial effects are achieved.

[0071] The above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent substitutions or changes, and should be covered by the protection scope of the present invention.

Claims

1. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers, characterized in that: It includes the following steps: Step 1, start the drill pipe machine (1) to work and drive the drill pipe (2) to drill downward. While the drill pipe (2) is drilling, sound waves are emitted through the sound wave emission probe (3) installed on the drill pipe (2). The surface of the drill pipe (2) of the drill pipe machine (1) is respectively provided with a circumferential motion emission mechanism (5) and a vertical motion emission mechanism (6). The circumferential motion emission mechanism (5) drives the sound wave emission probe (3) to move in the circumferential direction, and the vertical motion emission mechanism (6) drives the sound wave emission probe (3) to move in the vertical direction; the sound wave receivers (4) arranged at different positions on the arc surface of the drill pipe (2) are used to receive the sound waves reflected from different positions, and multiple sound wave receivers (4) are arranged in a spiral equidistant arrangement on the arc surface of the drill pipe (2), and multiple sound wave receivers (4) are also symmetrically distributed with respect to the vertical direction center of the drill pipe (2). The data of sound wave emission is collected through the data acquisition module, and the time difference between sound wave emission and reception is recorded by using a time relay, and the data is transmitted to the processing unit; Step 2, perform denoising processing on the collected sound wave signals. The denoising processing is carried out by performing convolution operation on the Gaussian blur filter and the sound wave signals; Step 3: Reconstruct the image of the hole wall using the processed data; when sound waves encounter media with different acoustic impedances, energy reflection occurs; by analyzing the intensity and time of the reflected waves, infer the properties and positions of the media; use the acoustic wave emission probe (3) to emit an acoustic wave pulse, and use the receiver to record the time when the acoustic wave arrives and the time when the reflected wave returns; this time difference is the total time for the sound wave to travel back and forth between the emission point and the reflection interface; Step 4, analyze the image. Apply the trained CNN model to the test set to detect cracks and cavities; identify the cracks and cavities on the hole wall, and evaluate the stability of the hole wall according to the attenuation and reflection characteristics of the sound wave energy; Step 5, make a comparison and judgment according to the CNN model applied to the test set. When it is unqualified, drive the concrete pouring equipment to pour and fill the inside of the detected drill hole.

2. The on-line detection method for preventing hole collapse of pile foundation in alluvial plain rich in silt layer according to claim 1, wherein: The data collected in Step 1 is the sound wave velocity, sound wave attenuation coefficient and reflection coefficient: Calculation of sound wave velocity: ; where is the sound wave velocity, is the distance traveled by the sound wave, is the time taken for the sound wave to travel; Sound attenuation coefficient: ; wherein, is the attenuation coefficient, is the initial sound intensity, is the propagation distance and is the sound intensity after propagation; Reflection coefficient: ; where is the reflection coefficient, is the reflected sound intensity, is the incident sound intensity.

3. An on-line detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silt layers according to claim 1, characterized in that: The formula for Gaussian blur filtering in the second step is as follows: ; where represents the value of the Gaussian function, which is a function of two variables and , is the coordinate relative to the center point within the kernel, is the standard deviation of the Gaussian distribution, represents pi; represents the base of the natural logarithm, represents the point the square of the Euclidean distance to the origin; To generate a Gaussian kernel, calculate the weight of each point within the kernel: ; where is the unnormalized Gaussian kernel weight, denotes the summation over all points within the kernel; For the acoustic wave signal The process of applying Gaussian filtering is expressed as: ; where is the filtered signal, is the kernel radius, is the value of the original signal at position ​ 4. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers according to claim 1, characterized in that: The steps for calculating the position of the reflection interface in step 3: a1. Record the time difference; record the time when the sound wave is emitted ; record the time when the sound wave is reflected back and captured by the receiver ; calculate the time difference: ; a2. Calculate the one-way time; since the recorded time is for the round trip of the sound wave, the time difference needs to be divided by 2 to obtain the one-way time : ; a3. Calculate the distance; use the acoustic wave velocity and the one-way travel time to calculate the distance from the reflection interface to the transmitter : ; Mark each reflection interface on the drill hole section through the calculated distance position of the reflection interface. When the hole wall is layered, each reflection interface represents a layer with different acoustic characteristics; because the initial hole wall forms a layered structure, the thickness and acoustic characteristics of each layer are initially estimated; use the reflection interface as the boundary between different layers, and initially estimate the acoustic characteristics of each layer according to the intensity and waveform characteristics of the reflection signal; assign the initially estimated acoustic parameters to each layer, and the parameters include sound velocity, density and attenuation coefficient; use geological modeling software to visualize the parameter information as an initial hole wall model; use the initial hole wall model to perform forward simulation of sound wave propagation to predict the reflection and propagation of sound waves at different positions.

5. The on-line detection method for preventing hole collapse of pile foundation in alluvial plain rich in silt layer according to claim 1, characterized in that: In Step 4, the trained CNN model is used to analyze the image of the inner wall of the drill hole when analyzing the image; b1. Image normalization: Scale the pixel values of the image to a fixed range, select ; ; wherein, is the original image, and are the minimum and maximum pixel values of the image, respectively; b2. Feature extraction: Use filters of different sizes in the convolutional layer to capture features at different scales: ; where is the scaled image, is the scaling ratio, is a function name representing the operation of performing image scaling; Extract features through the convolutional layer: ; is the feature map, is the convolutional kernel, is the bias term, represents the convolution operation; c3. Model Training: First, construct a CNN architecture: Design a network that includes convolutional layers, pooling layers, activation functions, and fully connected layers: ; is the output, is the input image, is the network weight, represents the convolutional neural network; Then, select a loss function to evaluate the performance of the model, using: ; where represents the true label, represents the predicted probability, represents the loss; Finally, perform an optimization algorithm. Use the SGD optimization algorithm to update the network weights: ; where, represents the learning rate, represents the gradient of the loss function with respect to the weights; c4. Model Evaluation: Evaluate the performance of the model on an independent validation set, using metrics such as accuracy, recall, and F1-score: ; ; ; where is the number of samples correctly identified as cracks or voids, is the number of samples that are incorrectly identified as cracks or voids when they are not, is the number of samples that incorrectly fail to identify actual cracks or voids, is the precision, which represents the proportion of samples that are actually positive among all samples predicted as positive, is the recall, which represents the proportion of samples that are correctly predicted as positive among all samples that are actually positive, represents the harmonic mean of precision and recall, and is used to measure the accuracy and comprehensiveness of the model.

6. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers according to claim 1, characterized in that: Based on the high accuracy rate and high recall rate values measured in Step 4, use the CNN model applied to the test set to judge whether the cracks and cavities on the inner wall of the drill hole are qualified; if the detected cracks and cavities are lower than or meet the preset standards in all aspects, it is judged that the inner wall of the drill hole is qualified, otherwise it is judged as unqualified. When it is unqualified, drive the concrete pouring equipment to pour and fill the inside of the detected drill hole.

7. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers according to claim 1, characterized in that: The circumferential motion launching mechanism (5) includes a turntable bearing (51). A driving groove (52) is formed on the arc surface of the drill pipe (2). The top arc surface and the bottom arc surface of the driving groove (52) are fixedly installed with the inner ring of the turntable bearing (51), and an outer ring of the turntable bearing (51) is fixedly installed with a driving rotating ring (53).

8. An on-line detection method for preventing borehole collapse of pile foundations in alluvial plains rich in silt layers according to claim 7, characterized in that: A toothed ring (54) is further fixedly installed on the arc surface at the bottom of the driving groove (52). The toothed ring (54) is located at the top of the lower turntable bearing (51). Vertical arc plates (55) are fixedly installed on the opposite surfaces of the two driving rotating rings (53). A rotating motor (56) is fixedly installed on the bottom arc surface of the vertical arc plate (55). An output end of the rotating motor (56) is fixedly installed with a rotating gear (57) through a coupling. Teeth of the rotating gear (57) are engaged with tooth grooves of the toothed ring (54).

9. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers according to claim 8, characterized in that: The vertical motion launching mechanism (6) includes fixing plates (61) installed on the top surface and the bottom surface of the vertical arc plate (55). Threaded rods (62) are rotatably connected to the opposite surfaces of the two fixing plates (61) through bearings. A vertical driving motor (63) is further fixedly installed at the bottom of the bottom fixing plate (61). An output end of the vertical driving motor (63) is fixedly installed with the bottom of the threaded rod (62) through a coupling.

10. An on-line detection method for preventing hole collapse of pile foundations in alluvial plains rich in silt layers according to claim 9, characterized in that: A slider (64) is threadedly connected to the arc surface of the threaded rod (62). A vertical sliding groove (65) is formed on the surface of the vertical arc plate (55). An inner wall of the vertical sliding groove (65) is slidably inserted with the surface of the slider (64). An installation plate (66) is fixedly installed on the top of the slider (64). The surface of the installation plate (66) is fixedly installed with the installation surface of the acoustic wave transmitting probe (3).

Images