Overwater anchor rod detection method

Abstract

The invention provides an overwater anchor rod detection method, and belongs to the technical field of overwater anchor rod detection.The overwater anchor rod detection method comprises the steps that a ship type excavator carries detection equipment to construct a platform, firstly, an anchor rod is positioned and fixed, and then original point cloud data is obtained through underwater ultrasonic scanning; meanwhile, real-time shaking parameters caused by water surface fluctuation are measured through an attitude sensor, and a three-dimensional shaking influence matrix is constructed; and performing compensation calculation on the original point cloud based on the shaking data, eliminating water surface fluctuation errors, evaluating the stress state of the anchor rod by applying an anchor rod mechanical analysis equation set, and identifying an abnormal region. The integrity of the internal structure of the anchor rod is analyzed through hammering sound wave detection, sound wave data and a point cloud matrix are fused to generate a comprehensive evaluation model, the health state rating and risk area identification of the anchor rod are analyzed and output through a deep neural network, and the technical problem that the detection precision of the water anchor rod is affected by water surface fluctuation, and consequently the data reliability is low is solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of underwater anchor rod detection, and more particularly, relates to a method for detecting underwater anchor rods. Background Art

[0002] Underwater anchor rods are key load-bearing components in water conservancy projects, offshore platforms, and port construction. Their detection technology is crucial for ensuring project safety. Traditional underwater anchor rod detection mainly uses technical means such as static load tests, ultrasonic detection, and acoustic reflection methods to evaluate their working conditions by measuring the force-deformation characteristics and internal structural integrity of the anchor rods. However, these detection methods usually need to be carried out in a stable environment to obtain accurate and reliable data.

[0003] In practical underwater applications, traditional detection technologies face serious challenges. Due to the continuous shaking of the detection platform caused by water surface fluctuations, the relative position between the detection equipment and the anchor rod continuously changes, thereby introducing systematic errors. At the same time, the vibration interference generated by wave motion will affect the acoustic propagation characteristics and ultrasonic scanning quality, resulting in noise interference and distortion in the collected raw data, greatly reducing the reliability and accuracy of the detection results.

[0004] Existing technologies are difficult to effectively solve the problem of the impact of water surface fluctuations on the detection accuracy of anchor rods. Although some studies have tried to reduce this impact by strengthening the detection platform or optimizing the detection algorithm, due to the lack of systematic analysis of water surface fluctuation characteristics and real-time compensation mechanisms, the reliability of the detection data is still low, unable to meet the high-precision requirements of underwater anchor rod safety assessment. Especially in complex hydrological conditions, the detection results often deviate greatly and are difficult to provide a reliable basis for anchor rod maintenance decisions. That is to say, there is a technical problem in the prior art that the detection accuracy of underwater anchor rods is affected by water surface fluctuations, resulting in low data reliability. Summary of the Invention

[0005] In view of this, the present invention provides a method for detecting underwater anchor rods, which can solve the technical problem in the prior art that the detection accuracy of underwater anchor rods is affected by water surface fluctuations, resulting in low data reliability.

[0006] The present invention is implemented as follows: The present invention provides an anchor rod detection method on water, comprising: using a ship-type excavator equipped with a hollow self-resetting hydraulic cylinder and a jack oil pump to form a detection platform, positioning the anchor rod on water and fixing the position of the hull; scanning the area around the anchor rod by an underwater ultrasonic scanning device to obtain the original anchor rod morphology point cloud matrix; using a posture sensor to measure the real-time shaking data of the detection equipment under water surface fluctuations, and constructing a three-dimensional shaking influence matrix; based on the real-time shaking data, compensating the original anchor rod morphology point cloud matrix to obtain the compensated anchor rod morphology point cloud matrix; applying the anchor rod mechanics analysis equation group to analyze the compensated anchor rod morphology point cloud matrix; performing hammering acoustic wave detection on the anchor rod by a high-speed data acquisition system; fusing the acoustic wave detection data with the anchor rod morphology point cloud matrix to generate a comprehensive evaluation model for anchor rod integrity; using a deep neural network to analyze the comprehensive evaluation model and output an anchor rod health status rating.

[0007] Among them, the three-dimensional sway influence matrix refers to the three-dimensional spatial displacement matrix of the ship-type excavator platform on the water surface caused by wave motion, which includes the displacement in the X, Y, and Z directions and its change function over time.

[0008] The step of constructing a three-dimensional sway influence matrix also includes recording the maximum influencing sway frequency and the influential sway frequency range.

[0009] Among them, the maximum influencing sway frequency refers to the hull sway frequency that has the greatest interference on the detection results; the influential sway frequency range refers to the hull sway frequency range that can have a significant impact on the anchor detection accuracy, and the sway frequency beyond the influential sway frequency range has little effect on the detection results.

[0010] Among them, the maximum impact sway amplitude refers to the critical displacement value at which the hull sway has a significant impact on the detection results. When the sway amplitude exceeds the maximum impact sway amplitude, the reliability of the detection data will significantly decrease.

[0011] Among them, the minimum impact sway amplitude refers to the minimum displacement value of the hull sway that can be recognized and effectively compensated by the detection system. The sway below the minimum impact sway amplitude is regarded as system noise and will be automatically filtered out during data processing.

[0012] Among them, the hammering acoustic wave detection of the anchor rod through the high-speed data acquisition system also includes recording the reflected waveform of the acoustic wave propagating in the anchor rod based on the physical equation of acoustic wave propagation, and analyzing the internal structural integrity of the anchor rod.

[0013] Among them, the physical equation of sound wave propagation is used to calculate the propagation characteristics and reflection characteristics of sound waves in anchor materials. The input includes anchor material density, anchor elastic modulus, sound wave incident angle, sound wave initial intensity, and sound wave frequency. The output is the sound wave propagation speed and reflection waveform characteristic spectrum.

[0014] Among them, the mechanical analysis equations of the anchor rod include the anchor rod stress distribution equation, the anchor rod deformation equation, the anchor rod stability equation, and the anchor rod tensile strength equation.

[0015] Among them, the anchor rod stress distribution equation is used to calculate the stress distribution state at each cross-section position of the anchor rod. The inputs include the axial load of the anchor rod, the cross-sectional area of the anchor rod, the material density of the anchor rod, the burial depth of the anchor rod, and the underwater soil parameters. The output is the stress distribution curve at each position of the anchor rod. The anchor rod deformation equation is used to calculate the deformation of the anchor rod under the stress state. The inputs include the axial load of the anchor rod, the elastic modulus of the anchor rod material, the length of the anchor rod, the moment of inertia of the anchor rod cross-section, and the initial shape of the anchor rod. The output is the deformation displacement field of the anchor rod. The anchor rod stability equation is used to evaluate the overall stability of the anchor rod under the current stress state. The inputs include the critical buckling load of the anchor rod, the actual acting load, the geometric parameters of the anchor rod, the environmental constraint conditions, and the yield strength of the anchor rod material. The output is the safety factor of the anchor rod stability. The anchor rod tensile strength equation is used to evaluate the strength margin of the anchor rod under the tensile state. The inputs include the tensile strength of the anchor rod material, the actual tensile stress, the length of the anchor rod anchorage section, the friction coefficient between the anchor rod and the surrounding medium interface, and the service life of the anchor rod. The output is the safety margin of the anchor rod tensile strength.

[0016] The present invention constructs a detection platform by equipping a ship-type excavator with a variety of sensing devices. Combining real-time attitude monitoring and data compensation technology, the system solves the influence of water surface fluctuations on the detection accuracy of anchor rods. While collecting the original shape point cloud data of the anchor rod, this method real-time monitors the shaking parameters of the platform, constructs a three-dimensional shaking influence matrix, and accordingly precisely compensates the detection data, effectively eliminating the systematic error introduced by water surface fluctuations.

[0017] Compared with the traditional technology, the present invention identifies the maximum influencing shaking frequency and the range of influencing shaking frequencies through shaking frequency analysis, and adopts a differential compensation strategy for different frequency wave interferences, significantly improving the reliability of the detection data. Especially under complex hydrological conditions, this method can accurately distinguish abnormal data that exceeds the maximum influencing shaking amplitude or is lower than the minimum influencing shaking amplitude, avoiding the interference of wrong data on the detection results. At the same time, through a high-speed data acquisition system and deep neural network analysis, a comprehensive evaluation of the internal and external states of the anchor rod is realized.

[0018] The present invention solves the core technical problem that the detection accuracy of underwater anchor rods is affected by water surface fluctuations. By establishing a mathematical relationship model between water surface fluctuations and detection errors, real-time compensation and optimization of detection data are realized, ensuring the accuracy and reliability of the anchor rod detection results in a dynamic water surface environment, and providing a scientific basis for the safety assessment and maintenance decision-making of underwater anchor rods. Brief Description of the Drawings

[0019] Figure 1 It is a flowchart of the method of the present invention.

[0020] Figure 2 This is the overall structural diagram of the water anchor detection system in Example 2.

[0021] Figure 3 This is a partial structural diagram of the hollow self-resetting hydraulic cylinder and the detection device in Example 2.

[0022] Figure 4 This is a structural diagram of the high-speed data acquisition system in Example 2. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0024] like Figure 1 FIG. 1 is a flow chart of a method for detecting an anchor rod on water provided by the present invention, and the method comprises the following steps:

[0025] S01. Use a ship-type excavator equipped with a hollow self-resetting hydraulic cylinder and a jack oil pump to form a testing platform to position the anchor rod above water and fix the position of the hull to ensure that the testing device is perpendicular to the axis of the anchor rod;

[0026] S02, scanning the area around the anchor rod by an underwater ultrasonic scanning device to obtain an original anchor rod shape point cloud matrix in the underwater area;

[0027] S03. Use the attitude sensor carried by the ship-type excavator platform to measure the real-time sway data of the detection equipment under water surface fluctuations, construct a three-dimensional sway influence matrix and record the maximum influencing sway frequency and the influencing sway frequency range;

[0028] S04, performing compensation calculation on the original anchor morphology point cloud matrix based on the real-time sloshing data to eliminate the sloshing error caused by the water surface fluctuation, and obtaining a compensated anchor morphology point cloud matrix;

[0029] S05, applying the anchor mechanics analysis equation group to analyze the compensated anchor morphology point cloud matrix, calculating the anchor stress state, and identifying abnormal areas exceeding the maximum impact sway amplitude or below the minimum impact sway amplitude;

[0030] S06. Perform hammering acoustic wave detection on the anchor bolt through a high-speed data acquisition system, record the reflected waveform of the acoustic wave propagating in the anchor bolt based on the physical equation of acoustic wave propagation, and analyze the internal structural integrity of the anchor bolt;

[0031] S07, fusing the acoustic wave detection data with the anchor morphology point cloud matrix to generate an anchor integrity comprehensive assessment model;

[0032] S08. Analyze the comprehensive evaluation model using a deep neural network and output the bolt health status rating and the identification of potential risk areas;

[0033] S09. Optionally, it further includes formulating a bolt maintenance plan based on the health status rating, including maintenance cycle and suggestions for reinforcement measures.

[0034] Among them, the three-dimensional sway influence matrix specifically refers to the three-dimensional space displacement matrix generated by the wave motion of the barge excavator platform on the water surface, including the displacement amounts in the X, Y, and Z directions and their time-varying functions.

[0035] Among them, the maximum influence sway frequency specifically refers to the hull sway frequency that causes the greatest interference to the detection results, usually between 0.5 and 2.0 Hz. The detection data under the maximum influence sway frequency needs to be specially compensated.

[0036] Among them, the influential sway frequency range specifically refers to the hull sway frequency range that can significantly affect the bolt detection accuracy, generally 0.1 - 5.0 Hz. The sway frequencies outside the influential sway frequency range have less impact on the detection results.

[0037] Among them, the maximum influence sway amplitude specifically refers to the displacement critical value at which the hull sway has a significant impact on the detection results. When the sway amplitude exceeds the maximum influence sway amplitude, the reliability of the detection data drops significantly, and re-detection or additional stabilization measures are required.

[0038] Among them, the minimum influence sway amplitude specifically refers to the minimum displacement value at which the hull sway can be recognized by the detection system and effectively compensated. The sway below the minimum influence sway amplitude is regarded as system noise and will be automatically filtered out in data processing.

[0039] The acoustic wave propagation physical equation is used to calculate the propagation characteristics and reflection characteristics of acoustic waves in the bolt material. The inputs include the bolt material density obtained from the material database, the bolt elastic modulus obtained from the mechanical tester, the acoustic wave incident angle recorded by the hammering device, the initial acoustic wave intensity set by the generator, and the acoustic wave frequency recorded by the acoustic wave generator. The outputs are the acoustic wave propagation speed transmitted to the acoustic wave analysis module and the reflection waveform feature map for use in step S07.

[0040] The bolt mechanical analysis equation set includes the bolt stress distribution equation, the bolt deformation equation, the bolt stability equation, and the bolt tensile strength equation;

[0041] The bolt stress distribution equation is used to calculate the stress distribution state at each cross-section position of the bolt. The inputs include the axial load of the bolt measured by the force sensor, the cross-sectional area of the bolt obtained from the design parameters, the bolt material density obtained from the material database, the bolt embedment depth obtained from the installation record, and the underwater soil parameters obtained from the soil survey. The output is the stress distribution curve at each position of the bolt transmitted to the stress analysis module;

[0042] The bolt deformation equation is used to calculate the deformation of the bolt under the stressed state. The inputs include the axial load of the bolt measured by the force sensor, the elastic modulus of the bolt material obtained from the material database, the length of the bolt obtained from the design parameters, the moment of inertia of the bolt cross-section obtained from the design parameters, and the initial form of the bolt obtained from the initial scan. The output is the bolt deformation displacement field transmitted to the deformation analysis module;

[0043] The bolt stability equation is used to evaluate the overall stability of the bolt under the current stressed state. The inputs include the critical buckling load of the bolt obtained from the mechanical calculation, the actual acting load measured by the force sensor, the geometric parameters of the bolt obtained from the design parameters, the environmental constraint conditions obtained from the environmental monitoring, and the yield strength of the bolt material obtained from the material database. The output is the bolt stability safety factor transmitted to the stability assessment module;

[0044] The bolt tensile strength equation is used to evaluate the strength margin of the bolt under the tensile state. The inputs include the tensile strength of the bolt material obtained from the material database, the actual tensile stress calculated from the stress analysis, the length of the bolt anchorage section obtained from the design parameters, the interface friction coefficient between the bolt and the surrounding medium obtained from the material interface test, and the service life of the bolt obtained from the installation record. The output is the bolt tensile strength safety margin transmitted to the strength assessment module.

[0045] The following describes in detail the specific implementation manners of the above steps.

[0046] The specific implementation of step S01 is to use a standard ship-type excavator as a carrier, and install a hollow self-resetting hydraulic cylinder at the end of its working arm. The inner diameter of this hydraulic cylinder is 107 mm, the stroke is 500 mm, and the stiffness of the self-resetting spring is 15 N / mm. The jack oil pump uses an electric two-way hydraulic pump with a rated pressure of 20 MPa and a flow rate of 2.5 L / min. The detection platform is first roughly positioned through the positioning system of the ship-type excavator, and the detection device is moved to about 0.5 m above the anchor rod. Then, the hull attitude stabilization system is started. This system includes four-corner hydraulic support legs, and each support leg is equipped with a pressure sensor and a displacement sensor. The hull level deviation is maintained less than 0.5° through feedback control. Subsequently, a laser pointer is used to confirm the perpendicularity between the detection device and the axis of the anchor rod. When the angle deviation between the two exceeds 1°, the system will automatically adjust the attitude of the excavator working arm until the angle deviation is less than 0.5°. At this time, the hull position is fixed through the hydraulic locking system. The purpose of this step is to establish a stable detection platform to ensure the accuracy and reliability of subsequent detection data.

[0047] The specific implementation of step S02 is to use a multi-array ultrasonic scanning device to perform an all-round scan of the area around the underwater anchor rod. This device contains 16 ultrasonic probes arranged in a ring, with a working frequency of 500 kHz, a sound wave emission angle of 15°, and a detection depth of up to 20 m. The scanning process starts at 0.2 m above the anchor rod and moves downward at intervals of 0.1 m. At each height position, a 360° rotation scan is performed at a rotation speed of 5° / s. Each probe emits 10 ultrasonic pulses per second and receives the echo signals. The received ultrasonic signals are processed by a signal amplifier, and the envelope detection algorithm is used to extract the echo characteristics. Then, the time-distance conversion algorithm is used to calculate the spatial positions of each point to form point cloud data. Each scan point is measured three times repeatedly, and the average value is taken to improve the accuracy. When the deviation of the three measurement data exceeds 5 mm, the system will mark this point as an uncertain point and perform additional measurements. The final resolution of the original anchor rod morphology point cloud matrix reaches 5 mm, including the three-dimensional coordinate information of about 10 5 spatial points around the anchor rod. The purpose of this step is to obtain the original geometric shape data of the anchor rod underwater and provide a basis for subsequent analysis.

[0048] The specific implementation of step S03 is to install a high-precision attitude sensor system on the barge-type excavator platform. This system includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer. The sampling frequency is 100Hz, the acceleration measurement range is ±16g, and the angular velocity measurement range is ±2000° / s. The sensors fuse multi-source data through the Kalman filtering algorithm to calculate the pitch angle, roll angle, and yaw angle of the platform in real time, with an accuracy better than 0.1°. The system records the displacement changes of the hull under the action of waves to form a three-dimensional sloshing influence matrix, which contains the displacement amounts in the X, Y, and Z directions and their variation functions with time. The sloshing data is subjected to spectral analysis through the fast Fourier transform algorithm to identify the maximum influence sloshing frequency, which is usually between 0.5 and 2.0Hz; at the same time, the range of influential sloshing frequencies is determined, generally 0.1 to 5.0Hz. The system also calculates the sloshing amplitude thresholds, including the maximum influence sloshing amplitude (usually 10mm) and the minimum influence sloshing amplitude (usually 0.5mm). The purpose of this step is to obtain the dynamic characteristic parameters of the detection environment and provide a basis for subsequent sloshing error compensation.

[0049] The specific implementation of step S04 is to perform compensation calculations on the original bolt shape point cloud matrix based on the acquired real-time sloshing data. First, the time synchronization algorithm is used to accurately match the point cloud acquisition time and the sloshing data acquisition time, with a time synchronization accuracy better than 1ms; then, a coordinate transformation model is established according to the principles of rigid body kinematics, which takes into account the influence of the six-degree-of-freedom motion of the hull (three translations and three rotations) on the detection results; the inverse kinematics algorithm is used to calculate the sloshing influence on each point cloud data and construct a compensation matrix; for the sloshing components with frequencies near the maximum influence sloshing frequency (within the range of ±0.1Hz), a weighted compensation coefficient is used for special processing, and the compensation coefficient is 1.2; for the sloshing components with frequencies within the range of influential sloshing frequencies but not near the maximum influence frequency, a standard compensation coefficient is used for processing, and the compensation coefficient is 1.0; for the sloshing components with frequencies outside the range of influential sloshing frequencies, a decay compensation coefficient is used for processing, and the compensation coefficient is 0.5. The compensation matrix is applied to the original point cloud matrix through matrix operations to eliminate the sloshing error caused by water surface fluctuations, and finally the compensated bolt shape point cloud matrix is obtained, with the point cloud accuracy improved to 2mm. The purpose of this step is to eliminate the influence of water surface fluctuations on the measurement results and improve the accuracy of the point cloud data.

[0050] The specific implementation of step S05 is to analyze the compensated bolt morphology point cloud matrix using the bolt mechanics analysis equation set. First, extract the bolt center axis from the point cloud matrix, and use the principal curve fitting algorithm for axis reconstruction with a curve fitting accuracy better than 1 mm. Then, calculate the geometric parameters of each cross-section of the bolt, including diameter, eccentricity, and cross-sectional area. Based on the bolt stress distribution equation, calculate the stress distribution state at each cross-section position, considering input parameters such as axial load, bolt cross-sectional area, bolt material density, bolt embedment depth, and underwater soil parameters, and output the stress distribution curve at each position of the bolt. Use the bolt deformation equation to calculate the deformation of the bolt under the stressed state, considering input parameters such as bolt axial load, bolt material elastic modulus, bolt length, bolt cross-sectional moment of inertia, and bolt initial morphology, and output the bolt deformation displacement field. Use the bolt stability equation to evaluate the overall stability of the bolt under the current stressed state, considering input parameters such as bolt critical buckling load, actual acting load, bolt geometric parameters, environmental constraint conditions, and bolt material yield strength, and output the bolt stability safety factor. Evaluate the strength margin of the bolt under the tensile state through the bolt tensile strength equation, considering input parameters such as bolt material tensile strength, actual tensile stress, bolt anchorage section length, bolt and surrounding medium interface friction coefficient, and bolt service life, and output the bolt tensile strength safety margin. The system also identifies abnormal areas that exceed the maximum influence shaking amplitude (usually 10 mm) or are lower than the minimum influence shaking amplitude (usually 0.5 mm), and specially marks these areas. The purpose of this step is to analyze the mechanical performance state of the bolt and discover potential structural abnormalities.

[0051] The specific implementation of step S06 is to perform impact acoustic wave detection on the bolt through a high-speed data acquisition system. Use a precision impactor to excite the top of the bolt with an impact force of 200 N and an impact duration of 2 ms. Use a piezoelectric sensor array to receive acoustic wave signals, with a sensor sensitivity of 100 mV / g, a frequency response range of 0.5 - 20 kHz, and a sampling frequency of 100 kHz. The dynamic range of the data acquisition system is 120 dB, and the resolution is 24 bit. Conduct 5 impact tests on each bolt, and record 2 s of acoustic wave signals for each test. Convert the time-domain signal to the frequency-domain signal through Fourier transform and analyze the spectral characteristics. Use wavelet transform for time-frequency analysis to identify the propagation mode of acoustic waves in the bolt. Based on the acoustic wave propagation physical equation, establish an acoustic wave propagation model, which considers input parameters such as bolt material density, elastic modulus, acoustic wave incident angle, acoustic wave initial intensity, and acoustic wave frequency, and outputs the acoustic wave propagation speed and the reflection waveform characteristic map. By analyzing the acoustic wave reflection characteristics, identify internal structural discontinuities in the bolt, such as cracks, cavities, or material defects, etc. The detection sensitivity can reach cracks with a diameter of 2 mm or cavities with a diameter of 5 mm. The purpose of this step is to detect the integrity of the bolt internal structure and discover internal defects invisible to the naked eye.

[0052] The specific implementation of step S07 is to fuse the acoustic wave detection data with the bolt morphological point cloud matrix. First, a unified spatial coordinate system is established to register the acoustic wave detection results and the point cloud data in terms of spatial position, with a registration accuracy better than 2 mm. Then, a data fusion algorithm is used to integrate the results of the two detection methods. This algorithm is based on the Bayesian inference framework and comprehensively considers the uncertainties of each measurement data. For the internal defects detected by acoustic wave detection, the influence coefficient on the overall performance of the bolt is calculated according to the defect position and size, and this coefficient is mapped into the point cloud model. For the external deformation or abnormality shown in the point cloud data, the internal structure state is analyzed in combination with the acoustic wave data. A multi-scale analysis method is used to process the fused data, paying attention to both the overall state of the bolt and the local detailed features. Finally, a comprehensive evaluation model of bolt integrity is generated, which contains multi-dimensional information such as the geometric morphology, material properties, defect distribution, stress state, and performance margin of the bolt. The purpose of this step is to comprehensively utilize multi-source data to generate a more comprehensive and accurate bolt state evaluation model.

[0053] The specific implementation of step S08 is to analyze the comprehensive evaluation model using a deep neural network. A deep learning architecture based on a three-dimensional convolutional neural network is adopted. This network includes 5 convolutional layers, 3 pooling layers, and 2 fully connected layers. The input layer receives the multi-dimensional data of the bolt integrity comprehensive evaluation model, and the data dimension is 128×128×64×10, representing a three-dimensional spatial grid and 10 feature channels. A 3×3×3 three-dimensional convolutional kernel is used to extract spatial features. The number of convolutional kernels in the first layer is 32, and it doubles in each subsequent layer. The ReLU activation function and batch normalization technology are used to improve the network performance. 50% random inactivation is introduced in the fully connected layer to prevent overfitting. The output layer includes the bolt health status rating (divided into five grades: A, B, C, D, E) and the identification of potential risk areas (accurate to the specific position and depth of the bolt). The network training adopts a supervised learning method, using a dataset containing 10,000 labeled samples, where the sample proportions of bolts in different grades are 20% for grade A, 30% for grade B, 25% for grade C, 15% for grade D, and 10% for grade E. The cross-entropy loss function and Adam optimizer are used, with an initial learning rate of 0.001, and a cosine annealing scheduling strategy is adopted. During the training process, 5-fold cross-validation is used to evaluate the model performance. The model accuracy reaches 92%, the recall rate reaches 90%, and the F1 score reaches 91%. The purpose of this step is to use artificial intelligence technology to conduct intelligent diagnosis and risk prediction on the bolt state.

[0054] Step S09 is an optional step, and its specific implementation method is to formulate an anchor rod maintenance plan according to the health status rating. For Class A anchor rods (excellent condition, safety factor greater than 2.0), it is recommended that the regular maintenance cycle be 24 months, and the maintenance content includes appearance inspection and anti-corrosion treatment; for Class B anchor rods (good condition, safety factor between 1.5 and 2.0), it is recommended that the maintenance cycle be 12 months, and the maintenance content includes additional strength re-inspection and local protection reinforcement; for Class C anchor rods (average condition, safety factor between 1.2 and 1.5), it is recommended that the maintenance cycle be 6 months, and the maintenance content includes additional load monitoring and comprehensive protection reinforcement; for Class D anchor rods (poor condition, safety factor between 1.0 and 1.2), it is recommended that the maintenance cycle be 3 months, and the maintenance content includes additional emergency reinforcement and preparation of alternative plans; for Class E anchor rods (dangerous condition, safety factor less than 1.0), it is recommended to immediately carry out reinforcement or replacement and formulate an emergency response plan. The maintenance plan also includes specific suggestions for reinforcement measures, such as epoxy resin perfusion (applicable to internal cavities, injection pressure is 0.5 MPa), carbon fiber winding reinforcement (applicable to external corrosion, winding layers are 3 to 5 layers), installation of anchor rod extension sections (applicable to insufficient anchoring force, extension section length is 30% to 50% of the original anchor rod length), etc.

[0055] The following is a detailed description of the mathematical models or calculation processes involved in the present invention.

[0056] In step S03, the mathematical expression of the three-dimensional sloshing influence matrix is specifically as follows:

[0057]

[0058] In the formula, M(t) is the three-dimensional sloshing influence matrix; x(t), y(t), and z(t) are the displacement amounts in the X, Y, and Z directions respectively; θ x (t), θ y (t), θ z (t) are the rotation angles around the X, Y, and Z axes respectively; t is the time variable.

[0059] The relationship between the displacement amount and time can be expressed as:

[0060]

[0061] In the formula, A x , A y , A z are the amplitudes of the main sloshing components, with a range of 5 to 15 mm; f x , f y , f z are the frequencies of the main sloshing components, with a range of 0.5 to 2.0 Hz; φ x , φ y , φ zis the initial phase; a xi 、a yi 、a zi are the amplitudes of the secondary sloshing components; f xi 、f yi 、f zi are the frequencies of the secondary sloshing components; φ xi 、φ yi 、φ zi are the initial phases of the secondary sloshing components; n is the number of sloshing components considered, usually taken as 5 - 10.

[0062] The relationship between the rotation angle and time can be expressed as:

[0063]

[0064] In the formula, B x 、B y 、B z are the amplitudes of the main rotation components, ranging from 0.1 to 1.0°; g x 、g u 、g z are the frequencies of the main rotation components, ranging from 0.3 to 1.5 Hz; ψ x 、ψ y 、ψ z is the initial phase; b xj 、b yj 、b zj are the amplitudes of the secondary rotation components; g xj 、g yj 、g zj are the frequencies of the secondary rotation components; ψ xj 、ψ yj 、ψ zj are the initial phases of the secondary rotation components; m is the number of rotation components considered, usually taken as 3 - 8.

[0065] The reason for choosing the sine function as the basis of the sloshing model is that the water surface fluctuations are usually periodic, and the superposition of multiple sine waves with different frequencies and amplitudes can approximate the actual complex fluctuation situation. Through Fourier transform, the complex waveform can be decomposed into the superposition of a series of simple sine waves, which is in line with the motion characteristics of the structure under the action of waves in actual ocean engineering. Considering multiple frequency components is to more accurately simulate the complexity of the actual water surface fluctuations. The fundamental frequency reflects the main wave influence, and the high-frequency components reflect the secondary influences such as wind waves and ripples.

[0066] In step S04, the point cloud compensation calculation based on the sloshing data is as follows:

[0067] P c = T -1 (t)·P o ;

[0068] Wherein, P c is the compensated point cloud matrix; P o is the original point cloud matrix; T(t) is the time-varying coordinate transformation matrix; t is the point cloud acquisition time.

[0069] The calculation formula of the coordinate transformation matrix T(t) is:

[0070]

[0071] Wherein, R(t) is the rotation matrix; D(t) is the translation vector; 0 1×3 is a 1×3 zero matrix.

[0072] The calculation formula of the rotation matrix R(t) is:

[0073] R(t) = R z (θ z (t))·R y (θ y (t))·R x (θ x (t));

[0074] Wherein, R x (θ x (t)), R y (θ y (t)), R z (θ z (t)) are the rotation matrices about the X, Y, and Z axes respectively.

[0075] The rotation matrix about the X axis is:

[0076]

[0077] The rotation matrix about the Y axis is:

[0078]

[0079] The rotation matrix about the Z axis is:

[0080]

[0081] The calculation formula of the translation vector D(t) is:

[0082]

[0083] The calculation formula of the frequency weight compensation factor is:

[0084]

[0085] where \(W(f)\) is the weight compensation factor corresponding to the frequency \(f\); \(f\) max is the maximum swaying frequency that usually ranges between \(0.5\) and \(2.0\) Hz.

[0086] The coordinate transformation model is established to accurately describe the influence of the six-degree-of-freedom motion of the hull on the point cloud data. The homogeneous coordinate matrix is used to uniformly handle rotation and translation transformations, and the rotation matrix is decomposed into basic rotations around three coordinate axes to simplify calculations and facilitate understanding. The introduction of the frequency weight compensation factor is based on the consideration that the swaying of different frequencies has different degrees of influence on the measurement results. In particular, a larger compensation coefficient is given to the swaying components near the maximum influence frequency, which helps to more accurately eliminate the swaying error.

[0087] In step S05, the mechanical analysis equations of the anchor rod include:

[0088] The axial stress distribution equation of the anchor rod:[[]]

[0089]

[0090] where \(\sigma(z)\) is the axial stress of the anchor rod at a depth of \(z\), with the unit of MPa; \(F\) is the axial load of the anchor rod, with the unit of kN, which is measured by a force sensor; \(A(z)\) is the cross-sectional area of the anchor rod at a depth of \(z\), with the unit of \(mm\) 2 , which is obtained from the design parameters; \(\rho\) is the density of the anchor rod material, with the unit of \(kg / m\) 3 , which is obtained from the material database; \(g\) is the acceleration due to gravity, taking \(9.8m / s\) 2 ; \(\tau(z)\) is the interfacial shear stress between the anchor rod and the surrounding medium at a depth of \(z\), with the unit of MPa, which is obtained from soil exploration; \(d(z)\) is the diameter of the anchor rod at a depth of \(z\), with the unit of \(mm\), which is obtained from the design parameters; \(\sigma\) e (z) is the additional stress term used to consider the stress caused by other factors such as bending and torsion, with the unit of MPa.

[0091] The deformation equation of the anchor rod:[[]]

[0092]

[0093] where \(\delta(z)\) is the axial deformation of the anchor rod at a depth of \(z\), with the unit of \(mm\); \(E(s)\) is the elastic modulus of the anchor rod material at a depth of \(s\), with the unit of GPa, which is obtained from the material database; \(M(u)\) is the bending moment of the anchor rod at a depth of \(u\), with the unit of \(N\cdot m\), which is obtained from mechanical calculations; \(I(u)\) is the moment of inertia of the cross-section of the anchor rod at a depth of \(u\), with the unit of \(mm\) 4 , which is obtained from the design parameters; \(\delta_0(z)\) is the initial deformation, with the unit of \(mm\), which is obtained from the initial scan.

[0094] The stability equation of the anchor rod:[[]]

[0095]

[0096] In the formula, SF is the safety factor of bolt stability, dimensionless; P cr is the critical buckling load of the bolt, in kN, obtained from mechanical calculations; P is the actual acting load, in kN, measured by a force sensor; e max is the maximum eccentricity, in mm, obtained from geometric measurements; e cr is the critical eccentricity, in mm, obtained from theoretical calculations; σ y is the yield strength of the bolt material, in MPa, obtained from the material database; σ max is the maximum actual stress, in MPa, obtained from stress analysis calculations; C e is the environmental impact correction factor, dimensionless, with a value range of 0.8 - 1.0, obtained from environmental monitoring.

[0097] Bolt tensile strength equation:

[0098]

[0099] In the formula, SF t is the safety margin of bolt tensile strength, dimensionless; σ t is the tensile strength of the bolt material, in MPa, obtained from the material database; A min is the minimum cross - sectional area of the bolt, in mm 2 , obtained from cross - section analysis; α is the annual corrosion rate, in year -1 , with a value range of 0.002 - 0.005, determined by material properties and environmental conditions; t is the service life of the bolt, in year, obtained from installation records; γ is the load uncertainty coefficient, dimensionless, with a value range of 1.2 - 1.5; F is the design load, in kN; μ is the friction coefficient at the interface between the bolt and the surrounding medium, dimensionless, with a value range of 0.3 - 0.6, obtained from material interface tests; L a is the anchorage length of the bolt, in m, obtained from design parameters; d avg is the average diameter of the bolt, in mm, obtained from design parameters; C f is the fatigue influence coefficient, dimensionless, with a value range of 0.7 - 0.9, determined by load history analysis.

[0100] The bolt stress distribution equation considers the combined effects of axial load, self - weight, lateral friction, and additional stress, which is in line with the actual stress situation of the bolt. The deformation equation simultaneously considers the superposition of axial deformation and bending deformation, which is more comprehensive than simply considering axial deformation. The stability equation introduces the form of multiplying multiple safety factors, and this conservative design method is widely used in engineering.

[0101] In step S06, the specific expression of the physical equation for acoustic wave propagation is as follows:

[0102] Calculation formula for the propagation speed of acoustic waves in the anchor rod:

[0103]

[0104] In the formula, v is the propagation speed of acoustic waves in the anchor rod material, with the unit of m / s; E is the elastic modulus of the anchor rod material, with the unit of GPa, obtained by a mechanical tester; ρ is the density of the anchor rod material, with the unit of kg / m 3 , obtained from the material database; μ is the Poisson's ratio, dimensionless, with a value range of 0.25 to 0.35, obtained from the material database.

[0105] Calculation formula for the acoustic wave reflection coefficient:

[0106]

[0107] In the formula, R is the acoustic wave reflection coefficient, dimensionless; Z1 is the acoustic impedance of the first medium, with the unit of kg / (m 2 ·s); Z2 is the acoustic impedance of the second medium, with the unit of kg / (m 2 ·s).

[0108] Calculation formula for the acoustic impedance:

[0109] Z = ρ·v;

[0110] In the formula, Z is the acoustic impedance, with the unit of kg / (m 2 ·s); ρ is the material density, with the unit of kg / m 3 ; v is the propagation speed of acoustic waves in this material, with the unit of m / s.

[0111] Calculation formula for the attenuation of acoustic wave intensity:

[0112] I(x) = I0e -αx ;

[0113] In the formula, I(x) is the intensity of the acoustic wave after propagating a distance x, with the unit of W / m 2 ; I0 is the initial acoustic wave intensity, with the unit of W / m 2 , set by the acoustic wave generator; α is the attenuation coefficient, with the unit of m -1 , with a value range of 0.01 to 0.1m -1 ; x is the propagation distance of the acoustic wave, with the unit of m.

[0114] Formula for the relationship between the incident angle and the reflection angle of the acoustic wave:

[0115]

[0116] where θ i is the incident angle of the sound wave, in degrees, recorded by the hammering device; θ r is the reflected angle of the sound wave, in degrees; v1 is the propagation speed of the sound wave in the first medium, in m / s; v2 is the propagation speed of the sound wave in the second medium, in m / s.

[0117] Calculation formula for the time-domain reflection waveform:

[0118]

[0119] where A(t) is the amplitude of the reflection waveform at time t, in Pa; N is the number of reflection interfaces; R i is the reflection coefficient of the i-th interface, dimensionless; A0 is the initial sound wave amplitude, in Pa, recorded by the sound wave generator; α is the attenuation coefficient, in m -1 ; x i is the propagation distance of the sound wave to the i-th interface, in m; f is the sound wave frequency, in Hz, recorded by the sound wave generator; φ i is the phase shift, in radians.

[0120] The physical equation for sound wave propagation adopts the classical wave theory and the acoustic interface reflection theory. The equation considers the influence of the elastic properties of the material on the sound wave propagation speed. The calculation of the interface reflection coefficient is based on the acoustic impedance difference between the two media, which conforms to the physical laws of sound wave propagation. The sound wave intensity attenuation adopts the exponential attenuation model, which is applicable to the energy loss situation in a homogeneous medium. The calculation of the time-domain reflection waveform considers the superposition of multi-interface reflections, distance attenuation, and phase differences, which can accurately simulate the sound wave propagation characteristics in complex structures. The relationship between the incident angle and the reflected angle adopts Snell's law, which is applicable to the wave propagation problem on the interface of different media. The sine function is selected to describe the waveform because the sound wave is essentially a mechanical wave, and its propagation conforms to the characteristics of a sine wave; the exponential attenuation term reflects the energy loss during propagation; the phase term considers the interference effect of the wave.

[0121] In step S07, the mathematical expression of the data fusion processing is as follows:

[0122] Bayesian fusion model:

[0123]

[0124] where P(S|D1, D2) is the posterior probability of the bolt state under the condition of two detection data; P(D1, D2|S) is the likelihood probability of observing two detection data under the condition of the bolt state; P(S) is the prior probability of the bolt state;

[0125] P(D1, D2) is the joint probability of the two detection data.

[0126] Likelihood probability calculation formula:

[0127] P(D1, D2|S) = P(D1|S)P(D2|S|D1);

[0128] In the formula, P(D1|S) is the conditional probability of observing point cloud data under the condition of the given bolt state; P(D2|S|D1) is the conditional probability of observing acoustic wave data under the condition of the given bolt state and point cloud data.

[0129] Point cloud data conditional probability model:

[0130]

[0131] In the formula, M is the dimension of the point cloud data; Indicates that the i-th dimensional point cloud data follows a normal distribution with a mean of μ 1i (S) and a variance of ; d 1i is the i-th dimensional point cloud measurement data; μ 1i (S) is the theoretical mean of the i-th dimensional point cloud data under the given bolt state S; is the variance of the i-th dimensional point cloud data.

[0132] Acoustic wave data conditional probability model:

[0133]

[0134] In the formula, N is the dimension of the acoustic wave data; Indicates that the j-th dimensional acoustic wave data follows a normal distribution with a mean of μ 2j (S, D1) and a variance of ; d 2j is the j-th dimensional acoustic wave measurement data; μ 2j (S, D1) is the theoretical mean of the j-th dimensional acoustic wave data under the given bolt state S and point cloud data D1; is the variance of the j-th dimensional acoustic wave data.

[0135] Prior probability model:

[0136]

[0137] In the formula, S0 is the prior estimated value of the bolt state; is the prior variance of the bolt state.

[0138] State estimation after fusion:

[0139]

[0140] In the formula, is the estimated value of the bolt state after fusion.

[0141] Wavelet Transform in Multiscale Analysis:

[0142]

[0143] Where W f (a, b) is the wavelet transform coefficient of the function f(t); a is the scale parameter, reflecting the fineness of the analysis; b is the translation parameter, reflecting the position of the analysis; ψ(t) is the wavelet mother function; ψ * represents the conjugate complex number of ψ.

[0144] Calculation formula for defect influence coefficient:

[0145]

[0146] Where I d is the defect influence coefficient, dimensionless; V d is the defect volume, with the unit of mm 3 ; V t is the total volume of the anchor rod, with the unit of mm 3 ; d d is the defect depth, with the unit of mm; d t is the total length of the anchor rod, with the unit of mm; r d is the distance from the defect to the central axis of the anchor rod, with the unit of mm; C d is the defect type correction coefficient, dimensionless, taking 1.5 for cracks, 1.0 for cavities, and 0.8 for corrosion.

[0147] Bayesian framework is adopted for data fusion because it can effectively integrate information from different sources with different characteristics and consider the uncertainty of data. The normal distribution assumption is applicable to most measurement error models, which is convenient for mathematical processing. Wavelet transform is used for multiscale analysis because it has good time-frequency localization characteristics and is suitable for analyzing non-stationary signals and discovering local features. The defect influence coefficient formula takes into account the comprehensive effects of defect volume ratio, depth ratio, and radial position. Among them, the volume ratio reflects the size of the defect, the depth ratio reflects the importance of the position, and the inverse square of the radial distance reflects the stress concentration effect. These factors jointly determine the degree of influence of the defect on the overall performance of the anchor rod.

[0148] Specifically, the principle of the present invention is as follows: The core technical principle of the present invention lies in analyzing the characteristics of water surface fluctuations and the real-time compensation mechanism to systematically solve the dynamic error problem in the detection of underwater anchor rods. First, the invention establishes a mapping relationship between the hull sway and the detection error. By constructing a three-dimensional sway influence matrix, it quantitatively describes the influence law of water surface fluctuations in different directions, frequencies, and amplitudes on the detection data. Among them, the identification and analysis of the maximum influence sway frequency (0.5 - 2.0 Hz) and the influential sway frequency range (0.1 - 5.0 Hz) provide a theoretical basis for error compensation.

[0149] The effectiveness of the error compensation mechanism is based on physical principles. When the detection platform sways, the sway parameters are captured in real time through attitude sensors, and the detection data is converted from the moving reference frame to the fixed reference frame to eliminate the position error caused by relative motion. At the same time, the Fourier analysis method is used to decompose the sway components of different frequencies, and compensation algorithms are adopted for different sway frequencies to achieve precise correction of the detection data. Especially for the data under the maximum influence sway frequency, special compensation processing is adopted, significantly improving the data reliability.

[0150] The present invention also establishes a complete set of mechanical analysis equations for the anchor rod, including four aspects: stress distribution, deformation, stability, and tensile strength, forming a comprehensive evaluation system for the stress state of the anchor rod. By analyzing the internal structure of the anchor rod through the physical equation of sound wave propagation and fusing it with the morphological point cloud matrix, the unified evaluation of the internal and external states of the anchor rod is realized. This multi-level and multi-angle comprehensive analysis method conforms to the basic principles of structural mechanics and wave theory, can accurately identify potential risks of the anchor rod, and provides a scientific basis for maintenance decisions.

[0151] The application of the deep neural network further improves the analysis accuracy. By learning a large amount of detection data, a mapping relationship between the anchor rod state and detection features is established, which can automatically identify abnormal patterns and predict potential risks. The entire technical solution forms a closed-loop system, from data collection, error compensation to state evaluation and maintenance recommendations, constructing a complete set of underwater anchor rod detection solutions, effectively solving the technical problem of low detection accuracy of anchor rods in the water surface fluctuation environment.

[0152] Next, a specific embodiment 1 of the present invention is provided. The specific implementation of each step in this embodiment 1 is described in detail as follows.

[0153] The specific implementation of steps S01 - S02 in this embodiment is the same as that described above, and will not be elaborated here in detail.

[0154] In this embodiment, the specific implementation of step S03 is to install a high-precision attitude sensor system on the platform of a barge-mounted excavator. The system includes a three-axis accelerometer, a three-axis gyroscope, and a three-axis magnetometer, with a sampling frequency of 100 Hz, an acceleration measurement range of ±16 g, and an angular velocity measurement range of ±2000° / s. The sensors fuse multi-source data through the Kalman filtering algorithm to calculate the pitch angle, roll angle, and yaw angle of the platform in real time, with an accuracy better than 0.1°. The system records the displacement changes of the hull under the action of waves to form a three-dimensional sloshing influence matrix, and the mathematical expression of this matrix is: In the formula, M(t) is the three-dimensional sloshing influence matrix; x(t), y(t), and z(t) are the displacement amounts in the X, Y, and Z directions respectively; θ x (t), θ y (t), θ z (t) are the rotation angles around the X, Y, and Z axes respectively; t is the time variable. The relationship between the displacement amount and time can be expressed as: In the formula, A x , A y , A z are the amplitudes of the main sloshing components, with a range of 5 - 15 mm; f x , f y , f z are the frequencies of the main sloshing components, with a range of 0.5 - 2.0 Hz; φ x , φ y , φ z are the initial phases; a xi , a yi , a zi are the amplitudes of the secondary sloshing components; f xi , f yi , f zi are the frequencies of the secondary sloshing components; φ xi , φ yi , φ zi are the initial phases of the secondary sloshing components; n is the number of sloshing components considered, usually taking 5 - 10. The relationship between the rotation angle and time can be expressed as: In the formula, B x , B y , B z are the amplitudes of the main rotation components, with a range of 0.1 - 1.0°; g x , g y , g z are the frequencies of the main rotation components, with a range of 0.3 - 1.5 Hz; ψ x , ψ y , ψ z are the initial phases; bxj , b yj , b zj is the amplitude of the secondary rotation component; g xj , g yj , g zj is the frequency of the secondary rotation component; ψ xj , ψ yj , ψ zj is the initial phase of the secondary rotation component; m is the number of rotation components considered, usually taking values from 3 to 8. The sloshing data is subjected to spectral analysis through the fast Fourier transform algorithm to identify the maximum sloshing frequency, which is usually between 0.5 and 2.0 Hz; at the same time, the range of influential sloshing frequencies is determined, generally 0.1 to 5.0 Hz. The system also calculates the sloshing amplitude thresholds, including the maximum influential sloshing amplitude (usually 10 mm) and the minimum influential sloshing amplitude (usually 0.5 mm). The purpose of this step is to obtain the dynamic characteristic parameters of the detection environment and provide a basis for subsequent sloshing error compensation.

[0155] The specific implementation of step S04 is to perform compensation calculations on the original anchor bolt shape point cloud matrix based on the acquired real-time sloshing data. First, the time synchronization algorithm is used to accurately match the point cloud acquisition time with the sloshing data acquisition time, and the time synchronization accuracy is better than 1 ms; then, according to the rigid body kinematics principle, a coordinate transformation model is established, and the point cloud compensation calculation formula is: P c = T -1 (t)·P o ; where P c is the compensated point cloud matrix; P o is the original point cloud matrix; T(t) is the time-varying coordinate transformation matrix; t is the point cloud acquisition time. The calculation formula for the coordinate transformation matrix T(t) is: where R(t) is the rotation matrix; D(t) is the translation vector; 0 1×3 is a 1×3 zero matrix. The calculation formula for the rotation matrix R(t) is: R(t) = R z (θ z (t))·R y (θ y (t))·R x (θ x (t)); where R x (θ x (t)), R y (θ y (t)), R z (θ z (t)) are the rotation matrices about the X, Y, and Z axes respectively. The rotation matrix about the X axis is: The rotation matrix about the Y axis is: The rotation matrix about the Z-axis is as follows: The calculation formula for the translation vector D(t) is: Use the inverse kinematics algorithm to calculate the influence of shaking on each point cloud data, and construct a compensation matrix. The calculation formula for the frequency weight compensation factor is: In the formula, W(f) is the weight compensation factor corresponding to the frequency f; f max is the maximum influencing shaking frequency, usually between 0.5 and 2.0 Hz. For the shaking components with frequencies near the maximum influencing shaking frequency (within the range of ±0.1 Hz), a weighted compensation coefficient is used for special processing, and the compensation coefficient is 1.2; for the shaking components with frequencies within the range of the influencing shaking frequencies but not near the maximum influencing frequency, a standard compensation coefficient is used for processing, and the compensation coefficient is 1.0; for the shaking components with frequencies exceeding the range of the influencing shaking frequencies, an attenuation compensation coefficient is used for processing, and the compensation coefficient is 0.5. Apply the compensation matrix to the original point cloud matrix through matrix operations to eliminate the shaking error caused by water surface fluctuations, and finally obtain the compensated anchor rod shape point cloud matrix, with the point cloud accuracy improved to 2 mm. The purpose of this step is to eliminate the influence of water surface fluctuations on the measurement results and improve the accuracy of the point cloud data.

[0157] The specific implementation of step S05 is to analyze the compensated anchor rod shape point cloud matrix using the anchor rod mechanical analysis equation set. First, extract the center axis of the anchor rod from the point cloud matrix, and use the principal curve fitting algorithm to reconstruct the axis, with the curve fitting accuracy better than 1 mm; then, calculate the geometric parameters of each cross-section of the anchor rod, including the diameter, eccentricity, and cross-sectional area; calculate the stress distribution state at each cross-section position based on the anchor rod stress distribution equation. The specific formula is: In the formula, σ(z) is the axial stress of the anchor rod at a depth of z, with the unit of MPa; F is the axial load of the anchor rod, with the unit of kN, which is measured by a force sensor; A(z) is the cross-sectional area of the anchor rod at a depth of z, with the unit of mm 2 , which is obtained from the design parameters; ρ is the density of the anchor rod material, with the unit of kg / m 3 , which is obtained from the material database; g is the acceleration due to gravity, taking 9.8 m / s 2 ; τ(z) is the interfacial shear stress between the anchor rod and the surrounding medium at a depth of z, with the unit of MPa, which is obtained from soil exploration; d(z) is the diameter of the anchor rod at a depth of z, with the unit of mm, which is obtained from the design parameters; σ e (z) is an additional stress term used to consider the stress caused by other factors such as bending and torsion, with the unit of MPa. Calculate the deformation of the anchor rod under the stressed state using the anchor rod deformation equation. The specific formula is: In the formula, δ(z) is the axial deformation of the anchor rod at a depth of z, with the unit of mm; E(s) is the elastic modulus of the anchor rod material at a depth of s, with the unit of GPa, which is obtained from the material database; M(u) is the bending moment of the anchor rod at a depth of u, with the unit of N·m, which is obtained from mechanical calculations; I(u) is the moment of inertia of the cross-section of the anchor rod at a depth of u, with the unit of mm 4 , which is obtained from the design parameters; δ0(z) is the initial deformation, with the unit of mm, which is obtained from the initial scan. The overall stability of the anchor rod under the current stress state is evaluated using the anchor rod stability equation, and the specific formula is: In the formula, SF is the safety factor of the anchor rod stability, dimensionless; P cr is the critical buckling load of the anchor rod, with the unit of kN, which is obtained from mechanical calculations; P is the actual applied load, with the unit of kN, which is measured by a force sensor; e max is the maximum eccentricity, with the unit of mm, which is obtained from geometric measurements; e cr is the critical eccentricity, with the unit of mm, which is obtained from theoretical calculations; σ y is the yield strength of the anchor rod material, with the unit of MPa, which is obtained from the material database; σ max is the maximum actual stress, with the unit of MPa, which is obtained from stress analysis calculations; C e is the environmental impact correction factor, dimensionless, with a value range of 0.8 - 1.0, which is obtained from environmental monitoring. The strength margin of the anchor rod under the tensile state is evaluated using the anchor rod tensile strength equation, and the specific formula is: C f ; In the formula, SF t is the safety margin of the anchor rod tensile strength, dimensionless; σ t is the tensile strength of the anchor rod material, with the unit of MPa, which is obtained from the material database; A min is the minimum cross-sectional area of the anchor rod, with the unit of mm 2 , which is obtained from cross-section analysis; α is the annual corrosion rate, with the unit of year -1 , with a value range of 0.002 - 0.005, which is determined by the material properties and environmental conditions; t is the service life of the anchor rod, with the unit of year, which is obtained from the installation records; γ is the load uncertainty coefficient, dimensionless, with a value range of 1.2 - 1.5; F is the design load, with the unit of kN; μ is the friction coefficient at the interface between the anchor rod and the surrounding medium, dimensionless, with a value range of 0.3 - 0.6, which is obtained from material interface tests; L a is the length of the anchored section of the anchor rod, with the unit of m, which is obtained from the design parameters; d avg is the average diameter of the anchor rod, with the unit of mm, which is obtained from the design parameters; C fis the fatigue influence coefficient, dimensionless, with a value range of 0.7 to 0.9, determined by load history analysis. The system also identifies abnormal areas that exceed the maximum influence sway amplitude (usually 10 mm) or are lower than the minimum influence sway amplitude (usually 0.5 mm), and special marks are made on these areas. The purpose of this step is to analyze the mechanical performance state of the anchor bolts and discover potential structural abnormalities.

[0158] The specific implementation of step S06 is to conduct impact acoustic wave detection on the anchor bolts through a high-speed data acquisition system. A precision impactor is used to excite the top of the anchor bolt, with an impact force of 200 N and an impact duration of 2 ms; a piezoelectric sensor array is used to receive the acoustic wave signals, with a sensor sensitivity of 100 mV / g, a frequency response range of 0.5 to 20 kHz, and a sampling frequency of 100 kHz; the dynamic range of the data acquisition system is 120 dB and the resolution is 24 bit. Five impact tests are conducted on each anchor bolt, and 2 s of acoustic wave signals are recorded for each test; the time-domain signals are converted into frequency-domain signals through Fourier transform to analyze the spectral characteristics; wavelet transform is used for time-frequency analysis to identify the propagation mode of the acoustic waves in the anchor bolts; an acoustic wave propagation model is established based on the physical equation of acoustic wave propagation, and the acoustic wave propagation speed calculation formula is: In the formula, v is the propagation speed of the acoustic wave in the anchor bolt material, with the unit of m / s; E is the elastic modulus of the anchor bolt material, with the unit of GPa, obtained by a mechanical tester; ρ is the density of the anchor bolt material, with the unit of kg / m 3 , obtained from the material database; μ is the Poisson's ratio, dimensionless, with a value range of 0.25 to 0.35, obtained from the material database. The acoustic wave reflection coefficient calculation formula is: In the formula, R is the acoustic wave reflection coefficient, dimensionless; Z1 is the acoustic impedance of the first medium, with the unit of kg / (m 2 ·s); Z2 is the acoustic impedance of the second medium, with the unit of kg / (m 2 ·s). The acoustic impedance calculation formula is: Z = ρ·v; in the formula, Z is the acoustic impedance, with the unit of kg / (m 2 ·s); ρ is the material density, with the unit of kg / m 3 ; v is the propagation speed of the acoustic wave in this material, with the unit of m / s. The acoustic wave intensity attenuation calculation formula is: I(x) = I0e -ax ; in the formula, I(x) is the intensity of the acoustic wave after propagating a distance x, with the unit of W / m 2 ; I0 is the initial acoustic wave intensity, with the unit of W / m 2 , set by the acoustic wave generator; α is the attenuation coefficient, with the unit of m -1 , with a value range of 0.01 to 0.1 m -1 ; x is the acoustic wave propagation distance, with the unit of m. The relationship formula between the acoustic wave incident angle and the reflection angle is: In the formula, θi is the incident angle of the sound wave, in degrees, recorded by the hammering device; θ r is the reflected angle of the sound wave, in degrees; v1 is the propagation speed of the sound wave in the first medium, in m / s; v2 is the propagation speed of the sound wave in the second medium, in m / s. The time-domain reflection waveform calculation formula is: In the formula, A(t) is the amplitude of the reflection waveform at time t, in Pa; N is the number of reflection interfaces; R i is the reflection coefficient of the i-th interface, dimensionless; A0 is the initial sound wave amplitude, in Pa, recorded by the sound wave generator; α is the attenuation coefficient, in m -1 ; x i is the propagation distance of the sound wave to the i-th interface, in m; f is the sound wave frequency, in Hz, recorded by the sound wave generator; φ i is the phase shift, in radians. By analyzing the sound wave reflection characteristics, discontinuity surfaces inside the anchor rod, such as cracks, cavities or material defects, etc., can be identified, and the detection sensitivity can reach cracks with a diameter of 2 mm or cavities with a diameter of 5 mm. The purpose of this step is to detect the integrity of the internal structure of the anchor rod and discover internal defects that are invisible to the naked eye.

[0159] The specific implementation of step S07 is to fuse and process the sound wave detection data with the anchor rod morphology point cloud matrix. First, establish a unified spatial coordinate system, register the sound wave detection results and the point cloud data in terms of spatial position, and the registration accuracy is better than 2 mm; then, use a data fusion algorithm to integrate the results of the two detection methods. This algorithm is based on the Bayesian inference framework and comprehensively considers the uncertainties of each measurement data. The specific formula of the Bayesian fusion model is: In the formula, P(S|D1, D2) is the posterior probability of the anchor rod state under the condition of the two detection data;

[0160] P(D1, D2|S) is the likelihood probability of observing the two detection data under the condition of the anchor rod state; P(S) is the prior probability of the anchor rod state; P(D1, D2) is the joint probability of the two detection data. The likelihood probability calculation formula is: P(D1, D2|S) = P(D1|S)P(D2|S|D1); in the formula, P(D1|S) is the conditional probability of observing the point cloud data under the condition of the anchor rod state; P(D2|S|D1) is the conditional probability of observing the sound wave data under the condition of the anchor rod state and the point cloud data. The conditional probability model of the point cloud data is: In the formula, M is the dimension of the point cloud data; indicates that the i-th dimension of the point cloud data follows a normal distribution with a mean of μ 1i (S) and a variance of ; d 1iis the point cloud measurement data of the i-th dimension; μ 1i (S) is the theoretical mean of the i-th dimension point cloud data under the given bolt state S; is the variance of the i-th dimension point cloud data. The conditional probability model of acoustic wave data is: In the formula, N is the dimension of acoustic wave data; indicates that the j-th dimension acoustic wave data follows a normal distribution with a mean of μ 2j (S, D1) and a variance of ; d 2j is the j-th dimension acoustic wave measurement data; μ 2j (S, D1) is the theoretical mean of the j-th dimension acoustic wave data under the given bolt state S and point cloud data D1; is the variance of the j-th dimension acoustic wave data. The multi-scale analysis method is used to process the fused data. The wavelet transform formula in multi-scale analysis is: In the formula, W f (a, b) is the wavelet transform coefficient of the function f(t); a is the scale parameter, reflecting the fineness of analysis; b is the translation parameter, reflecting the position of analysis; ψ(t) is the wavelet mother function; ψ * represents the conjugate complex number of ψ. For the internal defects found by acoustic wave detection, the influence coefficient on the overall performance of the bolt is calculated according to the defect position and size. The formula for the defect influence coefficient is: In the formula, I d is the defect influence coefficient, dimensionless; V d is the defect volume, with the unit of mm 3 ; V t is the total volume of the bolt, with the unit of mm 3 ; d d is the defect depth, with the unit of mm; d t is the total length of the bolt, with the unit of mm; r d is the distance from the defect to the central axis of the bolt, with the unit of mm; C d is the defect type correction coefficient, dimensionless. Take 1.5 for cracks, 1.0 for cavities, and 0.8 for corrosion. The prior probability model is: In the formula, S0 is the prior estimated value of the bolt state; is the prior variance of the bolt state. The state estimate after fusion is: In the formula, It is the estimated value of the state of the bolt after fusion. For the external deformation or anomalies shown in the point cloud data, analyze its internal structural state in combination with the acoustic wave data; adopt a multi-scale analysis method to process the fused data, paying attention to both the overall state of the bolt and the local detailed features; finally, generate a comprehensive evaluation model for the integrity of the bolt, which includes multi-dimensional information such as the geometric shape, material properties, defect distribution, stress state, and performance margin of the bolt. The purpose of this step is to comprehensively utilize multi-source data to generate a more comprehensive and accurate bolt state evaluation model.

[0161] The specific implementation of step S08 is to analyze the comprehensive evaluation model using a deep neural network. A deep learning architecture based on a three-dimensional convolutional neural network is adopted. This network consists of 5 convolutional layers, 3 pooling layers, and 2 fully connected layers. First, a three-dimensional convolutional neural network model is established. The input layer receives multi-dimensional data from the comprehensive evaluation model of bolt integrity. The data dimension is 128×128×64×10, representing a three-dimensional spatial grid and 10 feature channels. Among them, the 10 feature channels respectively represent: geometric shape features (channels 1-3), material property features (channels 4-5), defect distribution features (channels 6-7), and stress state features (channels 8-10). The convolutional layer uses a 3×3×3 three-dimensional convolutional kernel to extract spatial features. The number of convolutional kernels in the first layer is 32, and it doubles in each subsequent layer; the ReLU activation function and batch normalization technology are used to improve the network performance; 50% dropout is introduced in the fully connected layer to prevent overfitting; the output layer includes the bolt health status rating (divided into five grades: A, B, C, D, E) and the identification of potential risk areas (accurate to the specific location and depth of the bolt). The establishment of the training dataset includes the following steps: First, historical detection data is collected, including point cloud data and acoustic detection data of bolts in different states, totaling 15,000 samples; then, these data are preprocessed, including data cleaning, normalization, and augmentation, to generate training samples in a standard format; then, professional engineers label the samples, classify them into five grades: A, B, C, D, E according to the actual state of the bolts, and mark the location and scope of the potential risk areas; finally, the labeled samples are divided into a training set (10,000 samples), a validation set (2,500 samples), and a test set (2,500 samples). The training process includes the following steps: First, the network parameters are initialized. The weights are initialized using the He initialization method, and the bias initial value is set to 0; then, the training parameters are set, including a batch size of 64, an initial learning rate of 0.001, and a cosine annealing scheduling strategy; then, the cross-entropy loss function is used to evaluate the performance of the classification task, and the mean absolute error is used to evaluate the performance of the regression task; then, the Adam optimizer is used for parameter optimization, and the early stopping strategy is used during the training process. When the performance of the validation set does not improve for 10 consecutive epochs, the training stops; finally, 5-fold cross-validation is used to evaluate the model performance, and the model with the best performance on the validation set is selected. The performance metrics of the final model on the test set are: classification accuracy of 92%, recall rate of 90%, F1 score of 91%, and average error of risk area localization of 2.5 mm. The purpose of this step is to use artificial intelligence technology to perform intelligent diagnosis and risk prediction on the bolt state.

[0162] The specific implementation of step S09 is the same as the foregoing and will not be elaborated in detail here.

[0163] Specifically, the Kalman filter algorithm adopted in step S03 is used to fuse the data of the accelerometer, gyroscope, and magnetometer. Its state equation and observation equation are as follows: x k = Ax k-1 + Bu k + w k ; z k = Hx k + v k ; In the formula, x k is the state vector at time k, including the position, velocity, and attitude of the platform; A is the state transition matrix; u k is the control input; B is the control matrix; w k is the process noise, which follows a normal distribution with a mean of 0 and a covariance of Q; z k is the observation vector at time k, including the measurement values of each sensor; H is the observation matrix; v k is the observation noise, which follows a normal distribution with a mean of 0 and a covariance of R. The prediction and update processes of the Kalman filter are divided into two steps: Prediction step: Update step: K k = P k|k-1 H T (HP k|k-1 H T + R) -1 ; P k|k = (I - K k H)P k|k-1 ; In the formula, is the prior state estimate at time k; P k|k-1 is the prior estimate error covariance; K k is the Kalman gain; is the posterior state estimate at time k; P k|k is the posterior estimate error covariance; I is the identity matrix. This algorithm can effectively process the noise and errors in the sensor data, provide an accurate estimate of the platform attitude, and provide a reliable basis for subsequent jitter compensation.

[0164] In Embodiment 1, the positioning and detection of the anchor rod are realized through the detection platform carried by the barge-mounted excavator. The original shape data of the anchor rod is obtained by using an ultrasonic scanning device. An attitude sensor is used to measure the platform shaking and perform compensation calculations. The mechanical analysis equations and acoustic detection technology are applied to evaluate the state of the anchor rod. Through data fusion and deep neural network analysis, a comprehensive evaluation result is generated, and finally a targeted maintenance plan is formulated. This method fully considers the particularity of the underwater detection environment, solves the problem of shaking error caused by water surface fluctuation, combines mechanical analysis and acoustic detection to achieve a comprehensive detection of the external shape and internal structure of the anchor rod, and improves the accuracy and reliability of the detection results through artificial intelligence technology, providing technical support for the safety management of underwater anchor rods.

[0165] To better understand and implement the present invention, the following provides Embodiment 2 of a specific application scenario of the present invention: Researchers carried out the practical application of the underwater anchor rod detection method in a certain offshore wind farm. There are 25 wind turbines in this wind farm, and each wind turbine foundation is fixed by 12 anchor rods with a diameter of 90 mm. The material of the anchor rod is Q345 steel, the tensile strength is 540 MPa, the total length is 15 m, and the underwater burial depth is 12 m. Since it was found that some wind turbines showed slight inclination after 5 years of operation of the wind farm, it was suspected that the anchor rods might be damaged, so a comprehensive detection was carried out.

[0166] As Figure 2 shown, the detection team used a modified barge-mounted excavator, and a hollow self-resetting hydraulic cylinder with an inner diameter of 110 mm was installed at the end of the working arm, equipped with a jack oil pump with a rated pressure of 25 MPa. A high-precision attitude sensor system was installed on the excavator platform, including a three-axis accelerometer, a three-axis gyroscope and a three-axis magnetometer, with a sampling frequency of 120 Hz and measurement ranges of ±16 g and ±2000° / s respectively. Figure 3 shows the partial structure of the hollow self-resetting hydraulic cylinder and the detection device. The hollow self-resetting hydraulic cylinder is the core execution component of this detection method. Its unique design lies in the centrally penetrating hollow channel, which allows the anchor rod to pass through the hydraulic cylinder, realizing the precise positioning and force control detection of the anchor rod. Self-resetting springs are provided at the upper and lower parts of the hydraulic cylinder, which can ensure that the piston returns to the initial position when the hydraulic system fails, improving the safety of the system. The hydraulic oil is controlled through the oil inlet and return port on the left to form a complete hydraulic circuit. The hydraulic piston is in the middle of the cylinder body, and it also has a hollow structure to accommodate the anchor rod to pass through. The ultrasonic scanning device is located at the bottom, rotating 360° around the anchor rod to scan and obtain the surface morphology data of the underwater anchor rod. The hammering detection device is located on the right, used to generate a standardized impact force and record the acoustic wave propagation characteristics. The attitude sensor is installed in the upper right corner, real-time monitoring the three-dimensional attitude changes of the whole device, providing shaking compensation parameters for subsequent data processing. The whole structural design fully considers the particularity of the underwater operation environment, provides a stable supporting force through the hydraulic system, and at the same time, each sensor works together to achieve a comprehensive detection of the anchor rod.

[0167] The researchers recorded the three-dimensional sway influence matrix data of the hull under the action of waves, as shown in Table 1:

[0168] Table 1 Main parameter table of the three-dimensional sway influence matrix

[0169]

[0170]

[0171] Through fast Fourier transform analysis, it is obtained that the maximum influence sway frequency is 0.92 Hz, the range of influential sway frequencies is 0.2 - 4.8 Hz, the maximum influence sway amplitude is 12 mm, and the minimum influence sway amplitude is 0.6 mm.

[0172] The detection team used 16 ultrasonic probes arranged in a ring, with a working frequency of 580 kHz, an angle of 15°, and a detection depth of 18 m, to conduct an all-round scan of each anchor rod. A 360° rotational scan was performed at each height position to obtain the original point cloud data with a resolution of 4.5 mm. Then, based on the sway data, compensation calculations were performed on the point cloud, and finally, the accuracy of the compensated point cloud was improved to 1.8 mm.

[0173] The compensated point cloud data was analyzed using the anchor rod mechanics analysis equations. The measurement results of the mechanical parameters of some anchor rods are shown in Table 2:

[0174] Table 2 Measurement results table of the mechanical parameters of some anchor rods

[0175]

[0176] The anchor rods were subjected to hammering acoustic wave detection through a high-speed data acquisition system. A precision hammer was used to excite the top of the anchor rod, with a hammering force of 250 N and a duration of 1.8 ms. A piezoelectric sensor array with a sensitivity of 120 mV / g was used to receive the acoustic wave signals, and the sampling frequency was 110 kHz. As Figure 4As shown, in this embodiment, the high-speed data acquisition system is a key component of this underwater anchor rod detection method, mainly used for the hammering acoustic wave detection link. This system adopts a multi-channel synchronous acquisition architecture, including a front-end signal conditioning module, a high-speed ADC conversion module, an FPGA real-time processing module, and a data storage module. The system sampling rate can reach 500 kHz, ensuring the complete capture of the high-frequency acoustic wave signals in the anchor rod. The front-end signal conditioning module uses a programmable gain amplifier to automatically adjust the gain coefficient according to the real-time environmental noise, improving the signal-to-noise ratio. The system arranges 8 high-sensitivity piezoelectric sensors on the anchor rod, evenly distributed around the anchor rod to form an acoustic wave acquisition array. When the hammering device applies a standardized impact on the anchor rod, the acoustic wave propagates inside the anchor rod, and the piezoelectric sensors convert mechanical vibrations into electrical signals, which are sent to the ADC conversion after signal conditioning. The FPGA processing module implements an adaptive waveform extraction algorithm, which can effectively separate the target signal from the environmental noise under the condition of hull shaking. The system is also equipped with a GPS positioning module to associate the collected data with the spatial position, facilitating the fusion with the point cloud matrix during subsequent data processing. The calculated acoustic wave propagation speed is 5150 m / s, and some acoustic wave detection results are shown in Table 3:

[0177] Table 3 Partial Acoustic Wave Detection Results of Anchor Rods

[0178]

[0179] The acoustic wave detection data and the point cloud data are fused and processed to establish a unified spatial coordinate system, and the registration accuracy reaches 1.5 mm. A data fusion algorithm using a Bayesian inference framework comprehensively considers the uncertainty of each measurement data and generates a comprehensive evaluation model for the integrity of the anchor rod. For the detected internal defects, the defect influence coefficient is calculated and mapped into the point cloud model.

[0180] The comprehensive evaluation model is analyzed based on a three-dimensional convolutional neural network, and the neural network parameters used are shown in Table 4:

[0181] Table 4 Three-Dimensional Convolutional Neural Network Parameter Table

[0182]

[0183] 12,500 labeled samples are used for model training, and the sample proportions of anchor rods of grades A, B, C, D, and E are 24%, 32%, 23%, 14%, and 7% respectively. The training uses an Adam optimizer with a batch size of 64 and an initial learning rate of 0.001. After 250 epochs of training, the final classification accuracy on the test set reaches 94.2%, and the average error of risk area positioning is 2.1 mm.

[0184] According to the health status rating, a maintenance plan is formulated for each anchor rod, as shown in Table 5:

[0185] Table 5 Maintenance Plan Table for Anchor Bolts of Different Grades

[0186]

[0187] For Class E anchor bolts, replacements were immediately carried out. For Class D anchor bolts, carbon fiber winding reinforcement (4 layers of winding) was adopted. For Class C anchor bolts, epoxy resin perfusion (injection pressure of 0.55 MPa) and local reinforcement were used. Through this comprehensive inspection and targeted maintenance, the overall safety of the anchor bolts in the wind farm has been effectively improved.

[0188] Traditional underwater anchor bolt detection mainly uses manual diving detection or simple ultrasonic detection, which has the following problems: First, the underwater environment is complex, and manual detection has low efficiency and safety hazards; second, the fluctuation of the water surface causes the measurement platform to shake, affecting the detection accuracy; third, a single detection method cannot simultaneously obtain the external shape and internal structure state of the anchor bolt; fourth, there is a lack of systematic data analysis methods, making it difficult to accurately evaluate the health status of the anchor bolt. The underwater anchor bolt detection method of the present invention solves the stability problem through the detection platform carried by a ship-type excavator, realizes the comprehensive detection of the anchor bolt by using ultrasonic scanning and acoustic detection, eliminates the influence of water surface fluctuation by using a real-time shaking compensation algorithm, improves the accuracy of the detection result through data fusion and deep learning technology, and finally realizes the intelligent evaluation of the anchor bolt state and the scientific formulation of the maintenance plan. Compared with the traditional method, the detection efficiency of the present invention is increased by about 3 times, the detection accuracy is improved by 2.5 times, and the risk prediction accuracy rate is increased by 35%, significantly reducing the detection cost and maintenance risk, and providing effective technical support for the safety management of underwater anchor bolts.

[0189] It should be noted that the detailed explanations of the variables involved in the present invention are shown in Tables 6, 7, and 8 below.

[0190] Table 6 Variable Explanation Table (First Part)

[0191]

[0192] Table 7 Variable Explanation Table (Second Part)

[0193]

[0194]

[0195] Table 8 Variable Explanation Table (Third Part)

[0196]

[0197]

[0198] The above are only 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 can easily think of changes or substitutions, which should all be covered within the protection scope of the present invention.

Claims

1. An underwater anchor rod detection method, comprising: A ship-type excavator equipped with a hollow self-resetting hydraulic cylinder and a jack oil pump is used to form a detection platform to locate the anchor rod above water and fix the position of the hull; the area around the anchor rod is scanned by an underwater ultrasonic scanning device to obtain the original anchor rod morphology point cloud matrix; the attitude sensor is used to measure the real-time sway data of the detection equipment under water surface fluctuations to construct a three-dimensional sway influence matrix; Based on the real-time shaking data, the original anchor shape point cloud matrix is ​​compensated and calculated to obtain the compensated anchor shape point cloud matrix; the anchor mechanics analysis equation group is used to analyze the compensated anchor shape point cloud matrix; The anchor bolts are tested by hammering sonic wave through a high-speed data acquisition system; The acoustic wave detection data is fused with the anchor morphology point cloud matrix to generate a comprehensive assessment model for anchor integrity. The comprehensive assessment model is analyzed using a deep neural network to output the anchor health status rating.

2. The underwater anchor rod detection method according to claim 1, wherein, The three-dimensional sway influence matrix refers to the three-dimensional spatial displacement matrix of the ship-type excavator platform on the water surface caused by wave motion, which includes the displacement in the X, Y, and Z directions and its change function over time.

3. The underwater anchor rod detection method according to claim 2, wherein The step of constructing a three-dimensional sway influence matrix also includes recording the maximum influencing sway frequency and the influencing sway frequency range.

4. The underwater anchor rod detection method according to claim 3, wherein The maximum influencing sway frequency refers to the hull sway frequency that has the greatest interference with the detection results; the influential sway frequency range refers to the hull sway frequency range that can have a significant impact on the anchor detection accuracy, and the sway frequency beyond the influential sway frequency range has little effect on the detection results.

5. The underwater anchor rod detection method according to claim 4, wherein The maximum impact sway amplitude refers to the critical displacement value at which the hull sway has a significant impact on the test results. When the sway amplitude exceeds the maximum impact sway amplitude, the reliability of the test data will significantly decrease.

6. The underwater anchor rod detection method according to claim 5, wherein, The minimum impact sway amplitude refers to the minimum displacement value of the hull sway that can be recognized and effectively compensated by the detection system. Sway below the minimum impact sway amplitude is regarded as system noise and will be automatically filtered out during data processing.

7. The underwater anchor rod detection method according to claim 6, wherein, The hammering acoustic wave detection of the anchor bolt by the high-speed data acquisition system also includes recording the reflected waveform of the acoustic wave propagating in the anchor bolt based on the physical equation of acoustic wave propagation, and analyzing the internal structural integrity of the anchor bolt.

8. The underwater anchor rod detection method according to claim 7, wherein The physical equation of sound wave propagation is used to calculate the propagation characteristics and reflection characteristics of sound waves in anchor materials. The input includes anchor material density, anchor elastic modulus, sound wave incident angle, sound wave initial intensity, and sound wave frequency. The output is the sound wave propagation speed and reflection waveform characteristic spectrum.

9. The underwater anchor rod detection method according to claim 8, wherein, The anchor bolt mechanical analysis equation group includes the anchor bolt stress distribution equation, anchor bolt deformation equation, anchor bolt stability equation and anchor bolt tensile strength equation.

10. The underwater anchor rod detection method according to claim 9, wherein The bolt stress distribution equation is used to calculate the stress distribution state at each cross-section position of the bolt. The inputs include the axial load of the bolt, the cross-sectional area of the bolt, the material density of the bolt, the embedment depth of the bolt, and the underwater soil parameters. The output is the stress distribution curve at each position of the bolt. The bolt deformation equation is used to calculate the deformation of the bolt under the stress state. The inputs include the axial load of the bolt, the elastic modulus of the bolt material, the length of the bolt, the moment of inertia of the bolt cross-section, and the initial shape of the bolt. The output is the bolt deformation displacement field. The bolt stability equation is used to evaluate the overall stability of the bolt under the current stress state. The inputs include the critical buckling load of the bolt, the actual acting load, the geometric parameters of the bolt, the environmental constraint conditions, and the yield strength of the bolt material. The output is the bolt stability safety factor. The bolt tensile strength equation is used to evaluate the strength margin of the bolt under the tensile stress state. The inputs include the tensile strength of the bolt material, the actual tensile stress, the length of the bolt anchorage section, the friction coefficient at the interface between the bolt and the surrounding medium, and the service life of the bolt. The output is the bolt tensile strength safety margin.

Images