Pile body section change-considered defect positioning accurate identification method

Abstract

The invention provides a defect positioning accurate identification method considering the change of the section of a pile body, and the method comprises the steps: carrying out the simulation comparison of a constructed sound wave signal feature vector and a pre-constructed sound wave feature vector of a defect-free pile foundation, calculating the feature difference degree of the two, setting a difference degree threshold value, and if the feature difference degree exceeds the threshold value, carrying out the recognition of the defect-free pile foundation. If yes, preliminarily determining that the pile body has defects; on the basis of the adjusted sound wave detection, sound wave signals are collected again, signal processing and feature extraction are conducted through a self-adaptive path tracking algorithm, and the three-dimensional coordinates of the defect in the pile body are calculated according to the time delay and the propagation speed of reflected waves by combining the established multi-reflection model and an inversion sound wave propagation path; and according to the calculated three-dimensional coordinates of the defect reflection points, the influence degree of the defects on the bearing capacity of the pile foundation is calculated by combining the defect size estimated through the amplitude of reflected waves, the defect type recognized through waveform features and the strength parameter of a pile body material through a finite element method.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for accurately identifying defect locations by taking into account changes in pile body cross-sections. Background Art

[0002] Defect detection in composite variable-section piles, such as steel-concrete-filled tube piles and prestressed pipe piles, is a complex and critical technical challenge in deep subway excavation construction. Fluctuations in pile material properties, such as concrete strength, steel tube wall thickness, and interfacial bonding quality, as well as external environmental factors such as changes in surrounding soil properties and groundwater levels, significantly impact the effectiveness of acoustic detection methods. While acoustic detection methods based on multiple reflection models and adaptive path tracing can initially identify pile defects, these methods require further optimization to adapt to complex working conditions in practice. First, changes in concrete strength alter the propagation velocity of acoustic waves in the pile, directly affecting defect location accuracy. Second, fluctuations in steel tube wall thickness can alter the reflection characteristics of acoustic waves at the steel-concrete interface, further complicating defect identification. Furthermore, uneven interfacial bonding quality can cause the acoustic wave propagation path at the interface to shift, making defect location even more difficult. External environmental factors, such as changes in surrounding soil properties and groundwater levels, can alter the propagation characteristics of acoustic waves between the pile and the soil, further complicating detection. Therefore, how to achieve accurate parameter setting of the acoustic wave detection method and optimization of the defect location strategy under dynamically changing material properties and external environment has become a technical problem that needs to be solved urgently. Summary of the Invention

[0003] The present invention provides a method for accurately identifying defect locations that takes into account changes in pile cross-sections, which mainly includes:

[0004] Obtain the different properties corresponding to different pile materials, use finite element analysis or ray tracing methods to establish the propagation speed, attenuation coefficient and reflection coefficient of sound waves in different pile materials, construct a multi-reflection sound wave propagation model with reflection and refraction effects, and simulate the complete propagation path of sound waves in the pile body;

[0005] The acoustic wave signal of a specific frequency and waveform is emitted to the pile body, and the reflected acoustic wave signal is collected. The acoustic wave signal is analyzed in the time domain and frequency domain using an adaptive path tracing algorithm. The acoustic wave signal is decomposed into different path components, and the characteristic parameters of each component are extracted to construct the acoustic wave signal feature vector.

[0006] The constructed acoustic wave signal feature vector is simulated and compared with the acoustic wave feature vector of a pre-constructed defect-free pile foundation, and the difference between the two features is calculated. At the same time, a difference threshold is set. If the feature difference exceeds the threshold, it is preliminarily determined that the pile body has defects.

[0007] If a defect is initially determined to exist, a multi-reflection model is pre-built, and the properties of the surrounding soil layers and groundwater level data are input to simulate the changes in the sound wave propagation path under different soil layers and water levels. The transmission frequency of the sound wave detection and the position of the transmitting / receiving probes are adjusted according to the simulation results;

[0008] Based on the adjusted acoustic wave detection, the acoustic wave signal is re-collected and an adaptive path tracing algorithm is used for signal processing and feature extraction. Combined with the established multiple reflection model, the acoustic wave propagation path is inverted. Based on the time delay and propagation speed of the reflected wave, the three-dimensional coordinates of the defect within the pile body are calculated.

[0009] Based on the calculated three-dimensional coordinates of the defect reflection point, combined with the defect size estimated by the reflection wave amplitude, the defect type identified by the waveform characteristics, and the strength parameters of the pile material, the finite element method is used to calculate the impact of the defect on the bearing capacity of the pile foundation.

[0010] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0011] The present invention discloses a method for accurately locating and identifying defects that takes into account changes in the cross-section of the pile body. This method establishes a propagation model of sound waves in different pile body materials, combined with the surrounding soil layer and groundwater level data, to achieve accurate positioning and evaluation of pile foundation defects. First, a ray tracing method is used to construct a propagation model of sound waves in the pile body material to obtain the sound wave propagation characteristics. Then, a specific sound wave signal is emitted and the reflected signal is collected, and the characteristic vector is extracted through an adaptive path tracing algorithm and compared with the characteristics of a defect-free pile foundation. If a potential defect is found, the detection parameters are adjusted according to the multiple reflection model, the signal is re-collected and the sound wave propagation path is inverted, and the three-dimensional coordinates of the defect are calculated. Finally, the finite element method is used to evaluate the impact of the defect on the bearing capacity of the pile foundation in combination with the defect characteristics and the pile body material parameters. The present invention can accurately detect and evaluate pile foundation defects and improve the accuracy and efficiency of pile foundation quality inspection. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 This is a flow chart of a method for accurately identifying defect locations that takes into account changes in pile cross-sections of the present invention.

[0013] Figure 2 Schematic diagram of a defect location and accurate identification method taking into account changes in pile cross-section according to the present invention.

[0014] Figure 3 This is another schematic diagram of a defect location and accurate identification method that takes into account changes in the cross-section of the pile body according to the present invention. DETAILED DESCRIPTION

[0015] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0016] like Figure 1-3 In this embodiment, a method for accurately identifying defect locations that takes into account changes in pile cross-sections may specifically include:

[0017] Step S101, obtain different properties corresponding to different pile body materials, use finite element analysis or ray tracing method to establish the propagation speed, attenuation coefficient and reflection coefficient of sound waves in different pile body materials, construct a multiple reflection sound wave propagation model with reflection and refraction effects, and simulate the complete propagation path of sound waves in the pile body.

[0018] Acquire the acoustic wave propagation velocity value and elastic modulus value of the pile body material, calculate the acoustic wave attenuation coefficient matrix according to the acoustic wave propagation velocity value and elastic modulus value, and use the acoustic wave attenuation coefficient matrix to obtain a single reflection coefficient set; perform three-dimensional grid division on the pile body material according to the single reflection coefficient set, use the three-dimensional grid division data to calculate the relationship parameter between the acoustic wave refraction angle data set and the material thickness in the grid unit, and obtain acoustic wave propagation path data through the acoustic wave refraction angle data set; use the acoustic wave propagation path data to calculate the waveform distortion coefficient in the grid unit, construct a convolutional neural network model according to the waveform distortion coefficient, and obtain an acoustic wave propagation eigenvector through the convolutional neural network model; calculate the grid unit normalized reflection coefficient data set for the acoustic wave propagation eigenvector, and construct a complete propagation path of the acoustic wave in the pile body material according to the normalized reflection coefficient data set.

[0019] For example, the initial value of the propagation velocity and elastic modulus of the sound wave in the pile body material is obtained, and the attenuation coefficient matrix of the sound wave under different density parameters is calculated using the sound wave propagation attenuation coefficient matrix operation method. The single reflection coefficient set is calculated based on the stress distribution law of the pile body and the sound wave incident angle. A three-dimensional grid partitioning method is used to generate a grid unit data structure based on the pile body material components. The sound wave refraction angle data set and the material thickness dimension relationship parameters are iteratively calculated in each grid unit using the sound wave propagation distance weight. The pile body sound wave propagation path data is calculated based on the single reflection intensity set. A mathematical model of sound wave reflection intensity attenuation is constructed. Based on the sound wave propagation path data, the waveform distortion coefficient of the sound wave in each grid unit is calculated. A convolutional neural network is used to fit the waveform distortion coefficient and the sound wave propagation path data to obtain the sound wave propagation eigenvector. The sound wave reflection coefficient in the grid unit is normalized according to the sound wave propagation eigenvector to generate a normalized reflection coefficient data set. The multi-reflection sound wave propagation data is calculated using the normalized reflection coefficient data set. Based on the multi-reflection acoustic wave propagation data, a propagation path mapping matrix was constructed. Singular value decomposition was performed on the propagation path mapping matrix, and the eigenvalue set of the multi-reflection propagation of the pile acoustic wave was obtained from the singular value decomposition results. The complete propagation path vector of the acoustic wave in the pile material was constructed based on the multi-reflection propagation eigenvalue set. The propagation law parameters of the reflection and refraction effects of the acoustic wave in the pile material were obtained from the complete propagation path vector. The propagation characteristics of the acoustic wave in the pile material are closely related to the elastic modulus and density of the material. The acoustic wave propagation velocity can be calculated from the elastic modulus and density parameters. When the acoustic wave propagation velocity is initially 3000 meters per second, the elastic modulus of the pile concrete is 30 gigapascals, and the density is 2400 kilograms per cubic meter. When the acoustic wave propagates in the pile concrete, it will attenuate. The attenuation coefficients at different locations in the attenuation coefficient matrix are 0.3, 0.5, and 0.8, respectively. The attenuation coefficient increases with propagation distance. When the acoustic wave enters the pile material at a 45-degree angle, the single reflection coefficient is calculated to be 0.6 based on the stress distribution law. A three-dimensional meshing method was used to divide the pile body into cubic grid cells with a side length of 10 cm. Within each grid cell, the sound wave refraction angle was iteratively calculated based on the weight of the sound wave propagation distance. When the sound wave incident angle was 45 degrees, the refraction angle was 30 degrees. When the material thickness was 50 cm, the reflection intensity was 70% of the original intensity. The sound wave propagation path within the pile body was distributed in a zigzag pattern, and the propagation path data recorded the sound wave propagation trajectory within each grid cell. Based on the propagation law of sound waves within the pile body, a mathematical model for the attenuation of sound wave reflection intensity was established. After five reflections, the reflection intensity decayed to 25% of the original intensity, at which point the waveform was distorted with a waveform distortion coefficient of 0.35. A convolutional neural network was used to fit the waveform distortion coefficient to obtain a 12-dimensional sound wave propagation feature vector.The reflection coefficient of the sound wave within each grid cell is normalized to a value between 0 and 1. This normalized reflection coefficient dataset captures the reflection characteristics of the sound wave during multiple reflections within the pile. The propagation path mapping matrix describes the complete propagation path of the sound wave in the pile material. The matrix dimensions are 100×100. Singular value decomposition of this matrix yields a set of eigenvalues, with the largest eigenvalue being 2.5 and the second largest eigenvalue being 1.8. This set of eigenvalues ​​reflects the propagation characteristics of the sound wave in the pile material. The complete propagation path vector constructed based on the propagation eigenvalue set has a dimension of 50. Each element in the vector corresponds to a node on the sound wave propagation path, with a node spacing of 5 cm. The complete propagation path vector captures the reflection and refraction patterns of the sound wave in the pile concrete. After continuous reflection, the reflection angle and intensity gradually decrease, and the reflection intensity essentially disappears after a propagation distance exceeding 2 meters.

[0020] The pile body materials include concrete, steel pipe and bonding layer. The properties of the pile body materials include concrete strength grade, steel pipe wall thickness and interface bonding quality.

[0021] An ultrasonic detector is used to obtain the initial value of concrete compressive strength and concrete elastic modulus data to form a concrete strength characteristic vector. The concrete strength values ​​are graded using a support vector machine based on the concrete strength characteristic vector to obtain a pile body concrete strength classification dataset. The Gaussian mixture clustering method is used to group the steel pipe characteristic data matrix based on the pile body concrete strength classification dataset to obtain a steel pipe wall thickness distribution parameter set. A comprehensive property matrix of the pile body material is constructed based on the steel pipe wall thickness distribution parameter set and the cementing layer interface stress distribution dataset.

[0022] For example, an ultrasonic detector is used to obtain the initial value of concrete compressive strength and concrete elastic modulus data, a stress and strain sensor is used to measure the tensile strength and yield strength of steel pipes, and a basic data set of bond layer shear strength and solidification degree is obtained using an acoustic wave reflection detection device. A concrete strength feature vector is constructed based on the concrete density parameters and mix ratio data. The concrete strength values ​​are graded using a support vector machine based on the concrete strength feature vector, and a pile body concrete strength grading data set is obtained based on the grading results. A steel pipe feature data matrix is ​​constructed based on the steel pipe surface roughness and wall thickness values. The steel pipe feature data matrix is ​​grouped using a Gaussian mixture clustering method, and a steel pipe wall thickness distribution parameter set is obtained from the grouping results. A bond layer feature vector is constructed based on the bond layer thickness values ​​and bonding performance data. The bond layer interface stress distribution parameters are calculated based on the bond layer feature vector, and a bond layer interface stress distribution data set is obtained. A comprehensive pile body material attribute matrix is ​​constructed based on the concrete strength grading data set, the steel pipe wall thickness distribution parameter set, and the bond layer interface stress distribution data set. The material interface bonding strength index is calculated using the comprehensive pile body material attribute matrix. A pile material property evaluation vector was constructed based on the material interface bonding strength index. This vector was used to obtain a comprehensive assessment of the concrete strength grade, steel pipe wall thickness, and the bond quality of the adhesive layer interface. The pile concrete strength grade was tested using ultrasonic testing. The ultrasonic detector emitted ultrasonic waves at a frequency of 50 kHz. The compressive strength of the concrete was calculated by measuring the propagation velocity of the ultrasonic waves in the concrete. When the ultrasonic wave propagation velocity was 4000 meters per second, the corresponding compressive strength of the concrete was 35 MPa, and the concrete elastic modulus was 32 GPa. The mechanical properties of the steel pipe were measured using stress-strain sensors. Strain gauges were placed on the surface of the steel pipe to measure the strain of the steel pipe under load. The stress corresponding to the strain reaching 0.2% was the yield strength of the steel pipe. The tensile strength of the steel pipe was measured to be 400 MPa, and the yield strength was 235 MPa. The concrete strength feature vector includes parameters such as concrete density, water-cement ratio, and aggregate particle size distribution. The concrete density is 2400 kg per cubic meter, the water-cement ratio is 0.45, and the coarse aggregate particle size distribution ranges from 5 mm to 25 mm. A support vector machine (SVM) is used to grade concrete strength values, dividing it into 10 grades from C15 to C60. Each grade corresponds to a different strength range. For example, the C30 strength grade corresponds to a compressive strength range of 29.6 MPa to 38.5 MPa. The surface roughness of the steel pipe is measured using a laser 3D scanner, with surface roughness values ​​ranging from 10 microns to 50 microns. The wall thickness of the steel pipe is measured using an ultrasonic thickness gauge, with wall thickness values ​​ranging from 8 mm to 16 mm. A Gaussian mixture clustering method divides the steel pipe feature data matrix into three categories, each representing a different steel pipe performance level. The clustering results yield a set of steel pipe wall thickness distribution parameters.The thickness of the bond layer was measured using an ultrasonic thickness gauge and was between 2 mm and 5 mm. The bonding properties of the bond layer were determined using a pull-out test, with the interfacial bond strength being no less than 2 MPa. The bond layer characteristic vector includes parameters such as thickness uniformity, interfacial bond strength, and degree of solidification. These parameters are used to calculate the bond layer interface stress distribution dataset. The pile body material comprehensive attribute matrix integrates the concrete strength grading dataset, the steel pipe wall thickness distribution parameter set, and the bond layer interface stress distribution dataset. The matrix dimension is 100×100, and the singular value decomposition method is used to extract the material interface bonding strength characteristic values. The pile body material attribute evaluation vector dimension is 50, with each element in the vector corresponding to a material performance indicator. The comprehensive evaluation results show that the concrete strength grade is C35, the steel pipe wall thickness is 12 mm, and the bond layer interface bonding quality meets the design requirements.

[0023] Obtain a set of physical property parameters of the pile body material. Based on the physical property parameters of different pile body materials, use the ray tracing method to set the propagation path of the sound wave in the material, obtain the change curves of the attenuation coefficient and reflection coefficient, establish a physical model of the sound wave in the pile body material, and output the propagation speed, attenuation coefficient and reflection coefficient data set of the sound wave in different pile body materials.

[0024] The elastic modulus distribution data of the pile body material and the initial frequency parameter of the acoustic wave are obtained, a ray grid data structure is constructed based on the elastic modulus distribution data of the pile body material, a recursive neural network is used to predict the acoustic wave propagation path, and an acoustic wave refraction path vector is obtained; a spectral clustering algorithm is used to perform segmented calculations on the acoustic wave propagation path based on the acoustic wave refraction path vector, and the initial acoustic wave attenuation coefficient and the acoustic wave refractive index are obtained from the segmented calculations to generate an acoustic wave propagation characteristic vector; an acoustic wave attenuation function is constructed based on the acoustic wave propagation characteristic vector, and the acoustic wave propagation velocity and reflection coefficient of the acoustic wave in different pile body materials are calculated from the acoustic wave attenuation function to obtain an acoustic wave reflection coefficient curve data set.

[0025] For example, an ultrasonic detector is used to obtain initial data on the elastic modulus and density distribution of the pile material, an optical sensor is used to measure the surface roughness of the material, and the acoustic detector records the basic parameter set of the initial frequency and incident angle of the acoustic wave. A ray grid data structure is constructed based on the initial data on the elastic modulus and density distribution of the pile material. A recursive neural network is used within the ray grid data structure to predict the acoustic wave propagation path, and the acoustic wave refraction path vector is obtained from the prediction results. A spectral clustering algorithm is used to segment the acoustic wave propagation path based on the acoustic wave refraction path vector and the propagation distance data. The initial acoustic wave attenuation coefficient and the initial acoustic wave refractive index are calculated based on the segmentation results to obtain the acoustic wave propagation eigenvector. Based on the acoustic wave propagation eigenvector, an acoustic wave attenuation function is constructed, and the acoustic wave intensity attenuation values ​​at different propagation distances are calculated from the acoustic wave attenuation function to obtain an acoustic wave attenuation coefficient curve dataset. An acoustic wave physical propagation function is established based on the acoustic wave attenuation coefficient curve dataset, and the acoustic wave propagation velocity values ​​in different pile materials are calculated using the acoustic wave physical propagation function to obtain an acoustic wave propagation velocity dataset. An acoustic reflection calculation function was constructed based on a dataset of acoustic wave propagation velocity and material surface roughness values. The acoustic reflection coefficients at different pile material interfaces were obtained from this function, resulting in a dataset of acoustic reflection coefficient curves. The propagation characteristics of acoustic waves in pile materials depend on the material's physical parameters. The ultrasonic detector emitted acoustic waves at a frequency of 50 kHz. The measured elastic modulus of the pile concrete material was 32 GPa, with a density ranging from 2300 to 2500 kg / m³. The surface roughness of the material was 25 microns, and the acoustic wave incident angle was set at 45 degrees. The ray mesh data structure constructed based on the material's physical parameters uses a three-dimensional spatial partitioning method with a 5 mm grid cell size. Each grid cell contains the material's elastic modulus and density values. The recursive neural network input features include the physical parameters of the grid cell and the acoustic wave propagation direction. 100 hidden layer nodes are used to predict the acoustic wave propagation direction in the next grid cell. The prediction results form an acoustic wave refraction path vector with a dimension of 1000. As sound waves propagate, their intensity gradually decays. When the propagation distance is 500 mm, a spectral clustering algorithm divides the propagation path into five segments, each 100 mm long. The initial sound wave attenuation coefficient is calculated to be 0.8, and the initial sound wave refractive index is 1.5. The sound wave propagation feature vector has a dimension of 50 and captures the attenuation and refraction characteristics of the sound wave during propagation. The sound wave attenuation function describes how the sound wave intensity changes with propagation distance. When the propagation distance is 100 mm, the sound wave intensity attenuates by 0.9, at 200 mm, by 0.8, and at 300 mm, by 0.7. This creates a sound wave attenuation coefficient curve with 300 data points. The sound wave physical propagation function, based on the Huygens principle, calculates the sound wave propagation velocity in concrete to be 4000 meters per second and in steel to be 5100 meters per second.Sound waves reflect at material interfaces, and the reflection coefficient is closely related to the surface roughness. When the surface roughness is 25 microns, the acoustic reflection calculation function calculates a reflection coefficient of 0.95 for the concrete-air interface and 0.35 for the concrete-steel interface. The acoustic reflection coefficient curve dataset documents the reflection characteristics of sound waves at different incident angles. As the incident angle increases from 0 to 90 degrees, the reflection coefficient exhibits a nonlinear change, reaching a maximum of 0.98 at an angle of 45 degrees. Correlation analysis between acoustic wave propagation velocity and reflection coefficient reveals that greater surface roughness is associated with a smaller reflection coefficient and slower propagation velocity. As the surface roughness increases from 10 to 50 microns, the reflection coefficient decreases from 0.98 to 0.85, and the propagation velocity decreases from 4200 to 3800 meters per second. This correlation demonstrates the strong sensitivity of acoustic wave propagation characteristics to material surface conditions.

[0026] Step S102: emit an acoustic wave signal of a specific frequency and waveform to the pile body, collect the reflected acoustic wave signal, use an adaptive path tracing algorithm to perform time domain analysis and frequency domain analysis on the acoustic wave signal, decompose the acoustic wave signal into different path components, extract the characteristic parameters of each component, and construct an acoustic wave signal feature vector.

[0027] An ultrasonic transmitter is used to transmit a sound wave signal of a specific frequency to the pile body, and a sound wave receiver is used to collect time-domain signal data of reflected echo corresponding to the sound wave signal; median filtering and wavelet threshold processing are performed on the reflected echo time-domain signal data to obtain filtered sound wave signal data; an adaptive path tracing algorithm is used to calculate the sound wave propagation path length of the filtered sound wave signal data, and a reflection wave time delay value sequence is obtained from the propagation path length; wavelet transform is used to perform multi-scale decomposition on the filtered sound wave signal data according to the reflection wave time delay value sequence to obtain sound wave components corresponding to different propagation paths; a convolutional neural network is used to extract waveform distortion characteristics and amplitude attenuation characteristics of the sound wave components of different propagation paths to obtain sound wave signal feature vectors.

[0028] Exemplarily, an ultrasonic transmitter emits sound waves of a specific frequency into the pile body, and a sound receiver collects reflected echo signals. The original sound wave time domain signal data is recorded based on the number of sound wave sampling points and the sampling interval. The original sound wave time domain signal data is subjected to denoising preprocessing, using a median filter to eliminate impulse noise and a wavelet threshold method to eliminate Gaussian noise, thereby obtaining filtered sound wave signal data. Time domain feature extraction is performed on the filtered sound wave signal data, and an adaptive path tracing algorithm is used to calculate the sound wave propagation path length. A sequence of reflected wave delay values ​​is calculated from the propagation path length. A fast Fourier transform is performed on the filtered sound wave signal data to obtain the sound wave signal's frequency domain characteristics. Spectral analysis is then used to determine the frequency component distribution and frequency resolution parameters. Based on the reflected wave delay value sequence and frequency domain feature data, a wavelet transform is used to perform multi-scale decomposition of the sound wave signal to obtain sound wave components corresponding to different propagation paths. A convolutional neural network is used to extract waveform distortion and amplitude attenuation characteristics based on the sound wave components along different propagation paths, and a sound wave signal feature vector is constructed through feature combination. An ultrasonic transmitter transmits a 50 kHz acoustic signal into the pile shaft. The sampling frequency is set to 1 MHz, the number of sampling points is 10,000, and the sampling interval is 1 microsecond. The recorded raw acoustic time-domain signal consists of the transmitted waveform and multiple reflected echoes. The raw signal contains both impulse noise and Gaussian noise, with a signal-to-noise ratio of approximately 15 decibels. A median filter is used to remove impulse noise with an amplitude exceeding 2 volts, and a wavelet thresholding method is used to reduce the Gaussian noise to below 0.1 volt. The filtered acoustic signal contains multiple reflected waveforms. The time-domain waveforms show that the first reflected wave arrives at 100 microseconds, the second at 200 microseconds, and the third at 300 microseconds. An adaptive path tracing algorithm calculates the propagation path length based on the acoustic wave propagation velocity. When the acoustic wave propagates at a speed of 4000 meters per second in concrete, the corresponding path length for the first reflected wave is 400 mm, the second at 800 mm, and the third at 1200 mm. The acoustic signal undergoes a fast Fourier transform (FFT) to obtain its frequency domain characteristics. Spectral analysis reveals a dominant frequency component between 45 and 55 kHz, with a frequency resolution of 100 Hz. Sideband frequency components exist around the dominant frequency component, with a frequency difference of 5 kHz between the sideband frequency and the dominant frequency. This spectral characteristic reflects the frequency modulation effect of the acoustic wave during propagation. Wavelet transforms are used to decompose the acoustic signal into five layers, with different scales corresponding to acoustic components along different propagation paths. The first layer of decomposition corresponds to high-frequency components, reflecting the rapidly changing characteristics of the acoustic wave, while the fifth layer of decomposition corresponds to low-frequency components, reflecting the slowly changing characteristics of the acoustic wave.The acoustic wave components corresponding to different propagation paths exhibit different waveform characteristics. The first reflected wave has relatively good waveform integrity, with a waveform distortion of 0.1 and an amplitude attenuation of 0.2. The second reflected wave exhibits some waveform distortion, with a waveform distortion of 0.3 and an amplitude attenuation of 0.5. The third reflected wave exhibits significant waveform distortion, with a waveform distortion of 0.6 and an amplitude attenuation of 0.8. A convolutional neural network extracts waveform features through five convolutional layers. The first layer extracts basic waveform features, such as peak and trough positions, while the fifth layer extracts advanced waveform features, such as waveform symmetry and periodicity. The feature vector has a dimension of 100 and includes characteristic parameters such as waveform distortion, amplitude attenuation, frequency modulation, and phase modulation. These characteristic parameters comprehensively describe the propagation characteristics of acoustic waves in the pile material.

[0029] In step S103, the constructed acoustic wave signal feature vector is simulated and compared with the pre-constructed acoustic wave feature vector of a defect-free pile foundation, and the feature difference between the two is calculated. A difference threshold is set at the same time. If the feature difference exceeds the threshold, it is preliminarily determined that there is a defect in the pile body.

[0030] For the pre-constructed acoustic wave feature vector of the defect-free pile foundation and the acoustic wave feature vector of the pile foundation to be tested, the Euclidean distance is used to calculate the distance value between the feature vectors to obtain the acoustic wave feature space difference data; based on the acoustic wave feature space difference data, the waveform correlation and phase difference value are calculated using the cross-correlation algorithm to obtain the acoustic wave time domain feature difference parameters; for the acoustic wave time domain feature difference parameters, the frequency distribution difference and the amplitude attenuation ratio are calculated through spectrum analysis to obtain the acoustic wave frequency domain feature difference parameters; based on the acoustic wave time domain feature difference parameters and the acoustic wave frequency domain feature difference parameters, the feature differences are classified using a support vector machine and normalized using a multi-layer perceptron to obtain a standardized difference value; if the standardized difference value exceeds the upper limit of the preset threshold range, it is determined that the pile body has defects.

[0031] For example, the Euclidean distance is used to calculate the distance between the pre-constructed acoustic wave feature vector of a defect-free pile foundation and the acoustic wave feature vector of the pile foundation to be tested, thereby obtaining acoustic wave feature space difference data. A time-domain comparison function is constructed based on the acoustic wave feature space difference data, and a cross-correlation algorithm is used to calculate waveform correlation and phase difference values ​​to obtain acoustic wave time-domain feature difference parameters. Frequency domain transformation is performed on the acoustic wave time-domain feature difference parameters, and frequency distribution difference and amplitude attenuation ratio are calculated through spectrum analysis to obtain acoustic wave frequency-domain feature difference parameters. A simulation comparison matrix is ​​constructed based on the acoustic wave time-domain feature difference parameters and the frequency-domain feature difference parameters. Feature classification of the simulation comparison matrix is ​​performed using a support vector machine to obtain an acoustic wave feature difference classification vector. A feature difference comprehensive score is calculated based on the acoustic wave feature difference classification vector, and the feature difference comprehensive score is normalized using a multi-layer perceptron to obtain a standardized difference value. A threshold range is set for the standardized difference value. If the standardized difference value falls within the threshold range, the pile body is determined to be normal. If the standardized difference value exceeds the upper limit of the threshold range, the pile body is determined to be defective, and a defect determination indicator is obtained. The pre-constructed acoustic wave feature vector for a defect-free pile foundation has a dimension of 100 and consists of both time-domain and frequency-domain features. Time-domain features include parameters such as waveform amplitude, phase, and duration, while frequency-domain features include parameters such as frequency distribution and energy distribution. The acoustic wave feature vector of the pile foundation under test has the same dimension as the feature vector of a defect-free pile foundation. The distance between the two feature vectors is calculated using Euclidean distance. A distance value less than 0.1 indicates high similarity, while a distance value greater than 0.5 indicates significant differences. A time-domain contrast function performs point-by-point correlation analysis on the waveforms, and a cross-correlation algorithm is used to calculate the waveform correlation. The correlation value ranges from -1 to 1, with a correlation of 1 indicating identical waveforms and a correlation of -1 indicating completely opposite waveforms. Phase difference is calculated using a Hilbert transform. The phase difference for a defect-free pile foundation is typically less than 30 degrees, but increases significantly when defects are present. To calculate the frequency-domain feature difference, a fast Fourier transform is used to obtain the spectrum. The frequency distribution difference reflects the difference in the frequency components of the two signals. A dominant frequency offset exceeding 5 kHz indicates an anomaly. The amplitude attenuation ratio reflects the attenuation of signal energy. The amplitude attenuation ratio of a normal pile foundation ranges from 0.7 to 0.9, while that of a defective pile foundation is typically less than 0.5. The simulation comparison matrix has a dimension of 10×10, with matrix elements containing parameters for time-domain and frequency-domain feature differences. A support vector machine uses a radial basis kernel function to classify features, resulting in a feature difference classification vector with a dimension of 20. A comprehensive feature difference score is calculated using a weighted summation, with a weight of 0.6 for time-domain features and 0.4 for frequency-domain features. A multilayer perceptron (MLP) contains three hidden layers, with 64, 32, and 16 nodes per layer, respectively. A sigmoid activation function is used to perform nonlinear mapping on the comprehensive feature difference score, resulting in a standardized difference value between 0 and 1.The judgment threshold range is set between 0.2 and 0.8. When the normalized difference value is less than 0.2, the pile body is in good condition, while when it is greater than 0.8, it indicates a defect. In a specific application, a pile foundation was inspected and the calculated Euclidean distance value was 0.7, the time domain waveform correlation was 0.4, the phase difference value was 60 degrees, the frequency offset was 8 kHz, and the amplitude attenuation ratio was 0.3. After feature extraction and normalization of these parameters, the normalized difference value was 0.85, exceeding the upper threshold limit of 0.8, and the pile foundation was ultimately determined to be defective.

[0032] In step S104, if it is preliminarily determined that there are defects, a multiple reflection model is pre-built, and the properties of the surrounding soil layers and groundwater level data are input to simulate the changes in the sound wave propagation path under different soil layers and water level conditions. The transmission frequency of the sound wave detection and the position of the transmitting / receiving probe are adjusted according to the simulation results.

[0033] The porosity and density data of the soil layer are obtained, a soil layer data structure is constructed, and a layered soil acoustic characteristic data set is obtained; a multi-reflection acoustic wave propagation function is established for the layered soil acoustic characteristic data set, and the soil layer interface reflection and transmission coefficients are calculated through a deep neural network to obtain acoustic wave propagation path prediction data; an acoustic wave frequency response function is constructed based on the acoustic wave propagation path prediction data, and the acoustic wave emission frequency range is determined through a frequency scanning method to obtain an acoustic wave excitation parameter set; a probe layout optimization function is established for the acoustic wave excitation parameter set, and the spatial layout parameters of the transmitting probe and the receiving probe are calculated through a genetic algorithm to obtain the probe layout coordinates.

[0034] For example, a geological radar is used to obtain initial soil porosity and density data. A water level meter is used to record the groundwater level depth and soil moisture content. A soil permeability detector is used to obtain basic permeability parameters for soil layers at different depths. A soil layer data structure is constructed based on the basic physical parameters of the soil layer. A layered interpolation algorithm is used to calculate the acoustic impedance parameters of each soil layer, resulting in a layered soil acoustic characteristic dataset. A multi-reflection acoustic wave propagation function is established for the layered soil acoustic characteristic dataset. A deep neural network is used to calculate the reflection and transmission coefficients of the sound wave at the soil layer interface, resulting in predicted acoustic wave propagation path data. Based on the predicted acoustic wave propagation path data, an acoustic wave frequency response function is constructed. A frequency sweep method is used to determine the optimal acoustic wave emission frequency range, resulting in a set of acoustic wave excitation parameters. Based on the set of acoustic wave excitation parameters, a probe placement optimization function is established. Within this optimization function, a genetic algorithm is used to calculate the spatial placement parameters of the transmitting and receiving probes, resulting in a probe placement plan. A simulation of the acoustic wave propagation path is performed for the probe placement plan. The integrity of the multi-reflection waveforms is verified using a backtracking algorithm, resulting in the probe position coordinates and emission frequency parameters. Geological radar, which uses electromagnetic waves at a frequency of 200 MHz to detect soil structure, measured soil porosity ranging from 15% to 35%. Soil density gradually increases from top to bottom, reaching 85% at the surface and 95% at deeper layers. A water level gauge indicated a groundwater depth of 3 meters. Soil moisture content exhibits a stepped distribution with depth, from 20% at the surface to 38% below the waterline, where saturation is achieved. Soil permeability decays exponentially from the surface to a depth of 5 meters, from 5.6 × 10-4 cm / s at the surface to 1.2 × 10-4 cm / s at the bottom. The soil layer data structure divides the strata into five layers, each 1 meter thick. Cubic spline interpolation was used to calculate the acoustic impedance of the soil layers. The surface impedance is 3.2×10^6 Pascals per second per meter. The impedance increases with depth, with a sudden change at the groundwater level, reaching 4.8×10^6 Pascals per second per meter below the water level. A multi-reflection acoustic wave propagation function is developed based on the laws of reflection and refraction of sound waves at interfaces. The deep neural network input features include parameters such as soil acoustic impedance, moisture content, and compaction. The reflection coefficient of each interface is calculated using four hidden layers. The water-soil interface has the highest reflection coefficient, reaching 0.45, while the reflection coefficients of other soil interfaces range from 0.15 to 0.25. The acoustic frequency response function uses a frequency sweep method to calculate the acoustic wave propagation characteristics over the range of 20 kHz to 100 kHz. Low frequencies result in insufficient penetration, while high frequencies attenuate rapidly. The optimal transmission frequency was determined to be 50 kHz. At this frequency, the speed of sound waves in saturated soil with a moisture content of 38% is 1600 meters per second, and the speed of sound waves in unsaturated soil with a moisture content of 20% is 1800 meters per second.Probe placement optimization uses a genetic algorithm with a population size of 100 and 50 generations of evolution. The fitness function is used to optimize the spacing and depth between the transmitting and receiving probes. The transmitting probe was buried at a depth of 0.5 meters, the receiving probe at a depth of 1.5 meters, and the horizontal spacing between the two probes was 2 meters. Under this placement scheme, the sound wave maintained a signal strength of over 25% after three reflections. A reverse tracking algorithm was used to verify the sound wave propagation path, inferring the sound wave propagation trajectory from the receiving probe position. The verification results showed that the sound wave was strongly reflected at the water-soil interface, with the reflected wave energy accounting for 45% of the incident wave. The waveform integrity was maintained well, the main frequency of the signal spectrum remained near 50 kHz, and the bandwidth was controlled within 10 kHz.

[0035] Obtain soil layer data and water level data of the target area, combine with the preset sound wave velocity model, calculate the propagation path of the sound wave under different conditions, generate a propagation path diagram based on the change in the propagation path, and analyze the propagation characteristics of the sound wave under different soil layer properties and water level conditions. If the path diagram shows that there is a change in the sound wave propagation, adjust the transmission frequency, and determine the layout plan of the probe position based on the propagation path diagram and the adjusted transmission frequency.

[0036] A soil layer structure model is constructed based on the permeability and acoustic impedance values ​​of the soil layers, and a layered acoustic wave velocity dataset is obtained by calculation using physical parameters. A sound wave propagation path prediction function is constructed based on the layered acoustic wave velocity dataset, and the reflection and refraction characteristics of the sound waves at the soil layer interface are calculated using a recursive neural network to obtain a path change sequence. The path change sequence is used to construct a probe layout optimization function, and the horizontal spacing and burial depth of the probes are calculated using a grid search method. If the integrity of the detection data meets the preset benchmark value, the probe spatial layout plan is obtained.

[0037] For example, initial soil permeability and porosity data are acquired using geological radar, water level depth is recorded using a water level gauge, and basic soil impedance and propagation attenuation data are measured using acoustic sensors. A layered soil structure is constructed based on a preset acoustic velocity model. The acoustic propagation velocity of each layer is calculated using soil physical parameters to generate a layered acoustic velocity dataset. A sound wave propagation path prediction function is constructed for the layered acoustic velocity dataset. A recursive neural network is used to calculate the reflection and refraction characteristics of sound waves at soil interfaces, generating a path change sequence. A sound wave propagation trajectory is plotted based on the path change sequence, and the coordinates of characteristic points along the propagation path are extracted using image processing methods to generate the spatial distribution data of the sound wave path. A frequency response function is established for the spatial distribution data of the sound wave path, and a genetic algorithm is used to optimize the acoustic wave emission frequency. If the displacement of a path characteristic point exceeds a preset baseline value, the emission frequency is adjusted to the optimized range. A probe placement optimization function is constructed based on the adjusted emission frequency. Within this optimization function, a grid search method is used to calculate the horizontal spacing and burial depth of the probes. The acoustic wave propagation coverage is calculated based on the probe placement parameters. The integrity of the probe data is verified using a reverse verification method, resulting in a proposed probe spatial layout plan. Geological radar uses electromagnetic waves at a frequency of 200 MHz to detect soil structure. The measured surface soil permeability is 5.6 × 10^-4 cm / s, the porosity is 32%, and the water level is 3.5 meters deep. The acoustic impedance of the soil increases from 3.2 × 10^6 Pascals / s / m at the surface to 4.8 × 10^6 Pascals / s / m at the bottom, with propagation attenuation increasing by 0.2 decibels per meter with depth. A pre-defined acoustic velocity model divides the strata into five layers, each 1 meter thick. Based on soil physical parameters, the surface acoustic velocity is calculated to be 1500 meters / s. This velocity increases gradually with depth, reaching a sudden change at the water level, reaching 1800 meters / s in the water-saturated layer. The recursive neural network input features include acoustic velocity, acoustic impedance, and water level depth. Three hidden layers are used to calculate the reflection and refraction angles of sound waves at the soil interface. The maximum reflection coefficient occurs at the soil-water interface, reaching 0.42. The acoustic wave propagation trajectory is plotted using a rectangular coordinate system, with horizontal distance on the horizontal axis and depth on the vertical axis. Feature point extraction reveals significant refraction of the sound wave at the water level, causing the propagation path to deflect. The displacement of the path feature points reached 0.8 meters, exceeding the preset baseline value of 0.5 meters. A genetic algorithm optimized the transmission frequency, initially set at 50 kHz. After 50 generations of evolution, the frequency was adjusted to 35 kHz, achieving the optimal balance between acoustic penetration depth and resolution. Probe placement optimization employed a 20×20 grid search space with a grid spacing of 0.1 meters. Calculations revealed an optimal horizontal probe spacing of 2 meters, with a buried depth of 0.5 meters for the transmitting probe and 1.5 meters for the receiving probe.Under this deployment scheme, the acoustic wave propagation coverage area is a rectangular area with a depth of 5 meters and a horizontal width of 4 meters. Reverse verification shows that the acoustic wave signal attenuation within this area does not exceed 30 decibels, the waveform distortion is less than 0.3, and the detection data integrity is good. Analysis of different soil conditions shows that when the soil layer has high permeability, the acoustic wave propagation speed changes significantly, and the permeability changes from 10 to 10. -4 Increase to 10 -3 When the velocity of sound waves decreases by 200 meters per second, the speed of sound waves decreases. Increased porosity leads to increased sound wave attenuation: for every 10% increase in porosity, the attenuation increases by 0.15 decibels per meter. Water level fluctuations have the greatest impact on the sound wave propagation path: for every 1-meter increase in water level, the sound wave refraction angle changes by 5 degrees, and the propagation path deflects by 0.3 meters. These patterns provide an important basis for optimizing probe placement. By adjusting the transmission frequency and probe position, we can effectively respond to the changing sound wave propagation characteristics in different formation conditions.

[0038] In step S105, based on the adjusted acoustic wave detection, the acoustic wave signal is recollected, and the adaptive path tracing algorithm is used for signal processing and feature extraction. Combined with the established multiple reflection model, the acoustic wave propagation path is inverted, and the three-dimensional coordinates of the defect in the pile body are calculated based on the time delay and propagation speed of the reflected wave.

[0039] Acoustic wave signals are collected according to the adjusted detection parameters, and the environmental noise is eliminated by a digital filter to obtain a filtered signal. An adaptive path tracing algorithm is used to extract the waveform delay value and the reflection wave amplitude parameters of the filtered signal; a reflection wave separation function is constructed based on the waveform delay value and the reflection wave amplitude parameters, and the acoustic wave propagation delay sequence and the amplitude attenuation sequence are extracted from the reflected waveform through a time-frequency analysis method; a multiple reflection path matrix is ​​established based on the acoustic wave propagation delay sequence and the amplitude attenuation sequence, and a deep neural network is used to calculate the acoustic wave propagation velocity and the spatial scattering angle to obtain the propagation path inversion data; a defect location equation group is constructed based on the propagation path inversion data, and an iterative solution method is used to calculate the depth coordinates and azimuth angle of the defect position to obtain the defect spatial position data.

[0040] For example, acoustic wave signals are collected based on the adjusted acoustic wave detection parameters, and environmental noise is eliminated through a digital filter. An adaptive path tracing algorithm is used to extract waveform delay values ​​and reflection wave amplitude parameters from the filtered signal to obtain an initial acoustic wave feature data set. A reflection wave separation function is constructed based on a multiple reflection acoustic wave model, and multiple reflection waveforms are identified through a time-frequency analysis method. The acoustic wave propagation delay sequence and amplitude attenuation sequence are extracted from the reflection waveforms. A multiple reflection path matrix is ​​established for the acoustic wave propagation delay sequence and amplitude attenuation sequence, and a deep neural network is used to calculate the acoustic wave propagation velocity and spatial scattering angle to obtain acoustic wave propagation path inversion data. A three-dimensional spatial positioning function is constructed based on the acoustic wave propagation path inversion data, and the propagation trajectory of each reflection wave on the pile cross section is calculated using the least squares method to obtain a spatial distribution map of the acoustic wave propagation path. A set of defect location equations is established based on the acoustic wave propagation path spatial distribution map, and an iterative solution method is used to calculate the depth coordinates and azimuth angle of the defect position to obtain the defect spatial location data. A coordinate accuracy verification function was constructed for the spatial location data of the defect, and the confidence interval of the three-dimensional coordinates of the defect was estimated using the bootstrap resampling method to obtain the defect location coordinate value and positioning accuracy evaluation results. The acoustic wave detection signal acquisition adopted a transmission frequency of 50 kHz, a sampling frequency of 1 MHz, and a sampling duration of 10 milliseconds. The collected original signal contained multiple reflection waveforms. A bandpass filter was used to eliminate 50 Hz power frequency interference and high-frequency noise. The waveform delay value extracted by the adaptive path tracking algorithm showed that the arrival time of the first reflection wave was 100 microseconds and the arrival time of the second reflection wave was 200 microseconds. The reflection wave amplitudes were 80% and 50% of the initial amplitude, respectively. The multiple reflection acoustic wave model separated the reflected waveforms. The time-frequency analysis results showed that the main frequency component of the first reflection wave was concentrated between 45 and 55 kHz, and the second reflection wave had a wider frequency band of 40 to 60 kHz due to propagation attenuation. The delay sequence of the acoustic wave during propagation increases linearly, with the delay interval between adjacent reflected waves being 100 microseconds. The amplitude decay sequence decays exponentially with a coefficient of 0.6. The deep neural network consists of five hidden layers, with 128, 64, 32, 16, and 8 nodes in each layer, respectively. Input features include parameters such as delay, amplitude, and frequency. The output acoustic wave propagation velocity is 4000 meters per second, with a spatial scattering angle of 30 degrees. Inversion results of the acoustic wave propagation path within the pile body show that after multiple reflections, the sound wave exhibits a zigzag propagation characteristic. The three-dimensional spatial positioning function uses a cylindrical coordinate system to describe the acoustic wave propagation trajectory. The least squares method calculates the propagation trajectory, resulting in a spiral projection on the cross section. The defect location equations include a depth equation, an azimuth equation, and a radial distance equation. The iterative solution yields the defect location coordinates as 2.5 meters in depth, 135 degrees in azimuth, and a radius of 0.3 meters from the pile center.The bootstrap resampling method repeatedly sampled the defect coordinates 1000 times, resulting in a 95% confidence interval for the depth coordinate of 2.45 to 2.55 meters, a 95% confidence interval for the azimuth of 132 to 138 degrees, and a 95% confidence interval for the radial distance of 0.28 to 0.32 meters. Positioning accuracy assessments showed that the depth positioning error was less than 5 centimeters, the azimuth positioning error was less than 3 degrees, and the radial distance positioning error was less than 2 centimeters. The propagation characteristics of sound waves in different materials significantly affect defect location accuracy. When sound waves propagate from concrete to rebar, the propagation velocity increases from 4000 to 5100 meters per second, and the scattering angle decreases from 30 degrees to 20 degrees. The sudden change in propagation velocity at the waterline causes a change in the sound wave refraction angle, resulting in a 5 to 10 centimeter deviation in the positioning of the defect depth coordinates. Through comprehensive analysis and correction of multiple reflection waveforms, the positioning error was ultimately controlled to within 5 centimeters.

[0041] Step S106, based on the calculated three-dimensional coordinates of the defect reflection point, combined with the defect size estimated by the reflection wave amplitude, the defect type identified by the waveform characteristics, and the strength parameters of the pile body material, the finite element method is used to calculate the degree of influence of the defect on the bearing capacity of the pile foundation.

[0042] The defect area value and crack width value are calculated based on the pile body reflection amplitude data to obtain a defect geometric parameter set; the defect geometric parameter set is used to construct a stress-strain relationship matrix, and the mechanical property data of the pile body material is obtained from the material strength value and elastic modulus value; the convolutional neural network model is trained using the pile body material mechanical property data, and the defect type discrimination data is determined from the defect morphological characteristics and defect inclination angle values; if the defect type discrimination data meets the stress analysis conditions, the strain energy density distribution value is calculated, and the stress concentration factor and the stress influencing parameter set are determined using the Gaussian integral method.

[0043] For example, a tetrahedral meshing function for the pile body is constructed based on the three-dimensional coordinates of the defect reflection point, boundary constraints are set in the meshing function, and the defect area value and crack width value are calculated by the amplitude reflectivity to obtain a set of defect geometric parameters. A constitutive equation of the pile body material is established for the defect geometric parameter set, and a stress-strain relationship matrix is ​​constructed from the material strength value and the elastic modulus value to obtain the mechanical property data of the pile body material. A defect recognition function is constructed based on the material mechanical property data, and a convolutional neural network is used to extract the defect morphological characteristics and defect inclination values. The defect spatial distribution characteristics are calculated through morphological analysis to obtain defect type discrimination data. A stress field calculation function is established based on the defect type discrimination data, and the node displacement equation group is solved using the finite element method. The unit stress distribution is calculated by strain field interpolation to obtain the pile body stress distribution data. A strain energy density calculation function is constructed for the pile body stress distribution data, and the unit strain energy is calculated using the Gaussian integral method. The stress concentration coefficient is determined from the strain energy density distribution to obtain a set of stress influencing parameters. A bearing capacity calculation function was established based on a set of stress-influencing parameters. The ultimate bearing capacity of the pile body was calculated using the strength verification criterion. The degree of bearing capacity influence was determined through the load distribution cloud diagram, and the bearing capacity assessment result was obtained. The tetrahedral meshing of the pile body was divided using adaptive meshing technology. The mesh size in the defect area was set to 20 mm, and the mesh size away from the defect area was set to 50 mm. The boundary constraints were set as free at the top of the pile and fixed at the bottom of the pile. The reflected wave amplitude was calculated to be 300 square centimeters, the crack width was 2 mm, and the defect depth was located at 2.5 meters in the pile body. The constitutive equation of the pile body material adopted the elastic-plastic constitutive model. The elastic modulus of the concrete material was 30 GPa, the Poisson's ratio was 0.2, the compressive strength was 30 MPa, and the tensile strength was 2.5 MPa. The stress-strain relationship matrix was in incremental form. In the elastic stage, the stress and strain were linearly related. After exceeding the yield stress, it entered the plastic stage, and the stress-strain curve showed a nonlinear change. The convolutional neural network consists of five convolutional layers and three fully connected layers. Input features include the defect's geometric dimensions, location coordinates, and waveform characteristics. Morphological analysis determined the defect to be a horizontal through-hole crack with an inclination of 85 degrees, an elliptical spatial distribution, and a major-to-minor axis ratio of 3:1. Finite element stress calculations used a 20-node quadratic element with a stress singular element at the defect site. The results showed a stress concentration factor of 2.8 at the defect tip, a maximum principal stress of 65 MPa, a minimum principal stress of 5 MPa, and a shear stress of 15 MPa. The stress distribution exhibited a butterfly-shaped expansion, with a stress concentration zone at the defect tip. Strain energy density was calculated using a 27-point Gaussian integral. The element strain energy decayed exponentially with increasing distance from the defect, reaching 0.15 joules per cubic centimeter at the defect tip and decreasing to 0.02 joules per cubic centimeter in the far field. The stress concentration factor increased with defect size, increasing by 40% when the defect area doubled.The bearing capacity calculation employed strength verification criteria, comprehensively evaluating the pile's bearing capacity using both principal stress and maximum shear stress criteria. The results showed that the defect caused a 25% decrease in the pile's bearing capacity. The load distribution contours revealed a stress redistribution around the defect, with the bearing capacity above the defect decreasing by 35%, the bearing capacity below by 20%, and the lateral bearing capacity by 15%. This stress redistribution indicates that the defect significantly impacted the pile's overall bearing capacity, with the impact increasing with increasing load.

[0044] The above disclosure is only a preferred embodiment of the present invention, and it is certainly not intended to limit the scope of the present invention. A person skilled in the art can understand that all or part of the processes of the above embodiment and equivalent changes made in accordance with the claims of the present invention are still within the scope of the invention.

Claims

1. A method for accurately identifying defect locations taking into account changes in pile cross-section, characterized in that: The method comprises: Obtain the different properties corresponding to different pile materials, use finite element analysis or ray tracing methods to establish the propagation speed, attenuation coefficient and reflection coefficient of sound waves in different pile materials, construct a multi-reflection sound wave propagation model with reflection and refraction effects, and simulate the complete propagation path of sound waves in the pile body; The acoustic wave signal of a specific frequency and waveform is emitted to the pile body, and the reflected acoustic wave signal is collected. The acoustic wave signal is analyzed in the time domain and frequency domain using an adaptive path tracing algorithm. The acoustic wave signal is decomposed into different path components, and the characteristic parameters of each component are extracted to construct the acoustic wave signal feature vector. The constructed acoustic wave signal feature vector is simulated and compared with the acoustic wave feature vector of a pre-constructed defect-free pile foundation, and the difference between the two features is calculated. At the same time, a difference threshold is set. If the feature difference exceeds the threshold, it is preliminarily determined that the pile body has defects. If a defect is initially determined to exist, a multi-reflection model is pre-built, and the properties of the surrounding soil layers and groundwater level data are input to simulate the changes in the sound wave propagation path under different soil layers and water levels. The transmission frequency of the sound wave detection and the position of the transmitting / receiving probes are adjusted according to the simulation results; Based on the adjusted acoustic wave detection, the acoustic wave signal is re-collected and an adaptive path tracing algorithm is used for signal processing and feature extraction. Combined with the established multiple reflection model, the acoustic wave propagation path is inverted. Based on the time delay and propagation speed of the reflected wave, the three-dimensional coordinates of the defect within the pile body are calculated. Based on the calculated three-dimensional coordinates of the defect reflection point, combined with the defect size estimated by the reflection wave amplitude, the defect type identified by the waveform characteristics, and the strength parameters of the pile material, the finite element method is used to calculate the impact of the defect on the bearing capacity of the pile foundation.

2. The method according to claim 1, characterized in that The method of obtaining different properties corresponding to different pile materials, using finite element analysis or ray tracing methods to establish the propagation speed, attenuation coefficient and reflection coefficient of sound waves in different pile materials, constructing a multi-reflection sound wave propagation model with reflection and refraction effects, and simulating the complete propagation path of sound waves in the pile body includes: Acquiring the acoustic wave propagation velocity and elastic modulus of the pile material, calculating an acoustic wave attenuation coefficient matrix based on the acoustic wave propagation velocity and elastic modulus, and obtaining a single reflection coefficient set using the acoustic wave attenuation coefficient matrix; Performing three-dimensional grid division on the pile body material according to the single reflection coefficient set, using the three-dimensional grid division data to calculate the relationship parameter between the acoustic wave refraction angle data set and the material thickness in the grid unit, and obtaining the acoustic wave propagation path data through the acoustic wave refraction angle data set; The sound wave propagation path data is used to calculate the waveform distortion coefficient within the grid unit, a convolutional neural network model is constructed according to the waveform distortion coefficient, and a sound wave propagation feature vector is obtained through the convolutional neural network model; A grid unit normalized reflection coefficient data set is calculated for the acoustic wave propagation characteristic vector, and a complete propagation path of the acoustic wave in the pile body material is constructed based on the normalized reflection coefficient data set.

3. The method according to claim 2, characterized in that The pile body materials include concrete, steel pipe and bonding layer. The properties of the pile body materials include concrete strength grade, steel pipe wall thickness and interface bonding quality.

4. The method according to claim 2, characterized in that Also includes: Obtain a set of physical property parameters of the pile body material. Based on the physical property parameters of different pile body materials, use the ray tracing method to set the propagation path of the sound wave in the material, obtain the change curves of the attenuation coefficient and reflection coefficient, establish a physical model of the sound wave in the pile body material, and output the propagation speed, attenuation coefficient and reflection coefficient data set of the sound wave in different pile body materials. Specifically, it includes: Obtaining the elastic modulus distribution data of the pile body material and the initial frequency parameter of the acoustic wave, constructing a ray grid data structure based on the elastic modulus distribution data of the pile body material, using a recursive neural network to predict the acoustic wave propagation path, and obtaining the acoustic wave refraction path vector; A spectral clustering algorithm is used to segmentally calculate the acoustic wave propagation path based on the acoustic wave refraction path vector, and an initial acoustic wave attenuation coefficient and an acoustic wave refractive index are obtained from the segmented calculation to generate an acoustic wave propagation feature vector; An acoustic wave attenuation function is constructed according to the acoustic wave propagation characteristic vector, and the acoustic wave propagation velocity and reflection coefficient in different pile body materials are calculated from the acoustic wave attenuation function to obtain an acoustic wave reflection coefficient curve data set.

5. The method according to claim 1, wherein The method includes: emitting an acoustic wave signal of a specific frequency and waveform toward the pile body, collecting the reflected acoustic wave signal, performing time domain analysis and frequency domain analysis on the acoustic wave signal using an adaptive path tracing algorithm, decomposing the acoustic wave signal into different path components, extracting characteristic parameters of each component, and constructing an acoustic wave signal characteristic vector, including: An ultrasonic transmitter is used to transmit a sound wave signal of a specific frequency to the pile body, and a sound wave receiver is used to collect the time domain signal data of the reflected echo corresponding to the sound wave signal; Performing median filtering and wavelet threshold processing on the reflected echo time domain signal data to obtain filtered sound wave signal data; Calculating the acoustic wave propagation path length using an adaptive path tracing algorithm for the filtered acoustic wave signal data, and obtaining a reflected wave delay value sequence from the propagation path length; Performing multi-scale decomposition on the filtered acoustic wave signal data using wavelet transform according to the reflected wave time delay value sequence to obtain acoustic wave components corresponding to different propagation paths; A convolutional neural network is used to extract waveform distortion features and amplitude attenuation features of the acoustic wave components of different propagation paths to obtain acoustic wave signal feature vectors.

6. The method according to claim 1, characterized in that The constructed acoustic wave signal characteristic vector is simulated and compared with the acoustic wave characteristic vector of the pre-constructed defect-free pile foundation, the characteristic difference between the two is calculated, and a difference threshold is set at the same time. If the characteristic difference exceeds the threshold, it is preliminarily determined that the pile body has defects, including: For the pre-constructed acoustic wave feature vector of the defect-free pile foundation and the acoustic wave feature vector of the pile foundation to be tested, the Euclidean distance is used to calculate the distance between the feature vectors to obtain the acoustic wave feature space difference data; According to the acoustic wave feature spatial difference data, a cross-correlation algorithm is used to calculate the waveform correlation and phase difference value to obtain the acoustic wave time domain feature difference parameter; For the acoustic wave time domain characteristic difference parameter, the frequency distribution difference and the amplitude attenuation ratio are calculated by spectrum analysis to obtain the acoustic wave frequency domain characteristic difference parameter; According to the acoustic wave time domain characteristic difference parameter and the acoustic wave frequency domain characteristic difference parameter, a support vector machine is used to classify the characteristic difference and a multi-layer perceptron is used to perform normalization processing to obtain a standardized difference value; If the standardized difference value exceeds the upper limit of the preset threshold range, it is determined that the pile body has defects.

7. The method according to claim 1, characterized in that If it is preliminarily determined that there is a defect, a multiple reflection model is pre-built, and the properties of the surrounding soil layers and groundwater level data are input to simulate the changes in the sound wave propagation path under different soil layers and water levels. The transmission frequency of the sound wave detection and the position of the transmitting / receiving probe are adjusted according to the simulation results, including: Obtain soil porosity and density data, construct soil layer data structure, and obtain layered soil acoustic characteristic data set; A multi-reflection sound wave propagation function is established for the layered soil acoustic characteristic data set, and the soil interface reflection and transmission coefficients are calculated through a deep neural network to obtain the sound wave propagation path prediction data; Constructing an acoustic wave frequency response function based on the acoustic wave propagation path prediction data, determining the acoustic wave emission frequency interval by a frequency scanning method, and obtaining an acoustic wave excitation parameter set; A probe layout optimization function is established for the acoustic wave excitation parameter set, and the spatial layout parameters of the transmitting probe and the receiving probe are calculated by a genetic algorithm to obtain the probe layout coordinates.

8. The method according to claim 7, characterized in that Also includes: Obtain soil layer data and water level data for the target area, combine with the preset sound wave velocity model, calculate the propagation path of the sound wave under different conditions, generate a propagation path diagram based on the change in the propagation path, and analyze the propagation characteristics of the sound wave under different soil layer properties and water level conditions. If the path diagram shows that there is a change in sound wave propagation, adjust the transmission frequency. Based on the propagation path diagram and the adjusted transmission frequency, determine the layout plan of the probe position, which specifically includes: A soil layer structure model was constructed based on the soil permeability and acoustic impedance values, and a layered acoustic velocity data set was obtained using physical parameters. A sound wave propagation path prediction function is constructed for the layered sound wave velocity data set, and the reflection and refraction characteristics of the sound wave at the soil layer interface are calculated through a recursive neural network to obtain a path change sequence; The path change sequence is used to construct a probe layout optimization function, and the horizontal spacing and burial depth of the probes are calculated through a grid search method. If the integrity of the detection data meets the preset benchmark value, the probe spatial layout plan is obtained.

9. The method according to claim 1, characterized in that The adjusted acoustic wave detection method recollects acoustic wave signals, performs signal processing and feature extraction using an adaptive path tracing algorithm, and inverts the acoustic wave propagation path based on the established multiple reflection model. The three-dimensional coordinates of the defect within the pile body are calculated based on the time delay and propagation speed of the reflected wave, including: Acquiring an acoustic wave signal according to the adjusted detection parameters, eliminating environmental noise through a digital filter to obtain a filtered signal, and extracting waveform delay values ​​and reflection wave amplitude parameters from the filtered signal using an adaptive path tracing algorithm; Constructing a reflection wave separation function based on the waveform delay value and the reflection wave amplitude parameters, and extracting the sound wave propagation delay sequence and amplitude attenuation sequence from the reflection waveform through a time-frequency analysis method; A multiple reflection path matrix is ​​established based on the acoustic wave propagation delay sequence and the amplitude attenuation sequence, and a deep neural network is used to calculate the acoustic wave propagation velocity and spatial scattering angle to obtain propagation path inversion data; A defect location equation group is constructed based on the propagation path inversion data, and an iterative solution method is used to calculate the depth coordinates and azimuth angle of the defect position to obtain the defect spatial position data.

10. The method according to claim 1, characterized in that The method of calculating the influence of the defect on the bearing capacity of the pile foundation by using the finite element method based on the calculated three-dimensional coordinates of the defect reflection point, the defect size estimated by the reflected wave amplitude, the defect type identified by the waveform characteristics, and the strength parameters of the pile body material includes: Calculate the defect area and crack width based on the pile body reflection amplitude data to obtain the defect geometric parameter set; The defect geometric parameter set is used to construct a stress-strain relationship matrix, and the mechanical property data of the pile body material is obtained from the material strength value and elastic modulus value; The convolutional neural network model is trained by the mechanical property data of the pile body material, and the defect type discrimination data is determined from the defect morphological characteristics and defect inclination values; If the defect type discrimination data meets the stress analysis conditions, the strain energy density distribution value is calculated, and the stress concentration factor and the stress influencing parameter set are determined by the Gaussian integral method.

Images