Quality detection method for wharf armor block

Abstract

The invention discloses a wharf armor block quality detection method, and belongs to the technical field of concrete structure detection.The method comprises the steps that a surface model is constructed through a movable three-dimensional laser scanning device, and a high-risk defect area is recognized; performing encryption scanning and impact echo detection in the high-risk area by adopting a frequency-adjustable dual-mode flaw detection device to generate a three-dimensional defect distribution diagram; a core sample is drilled in the defect area for compressive strength testing, and the material performance is analyzed in combination with a non-contact infrared spectrum; and establishing a dynamic evaluation model, calculating a safety coefficient by integrating surface defects, internal defects and material discreteness data, and generating a three-dimensional visual report. The problems that a traditional detection method is low in efficiency, insufficient in internal defect positioning precision, one-sided in material performance evaluation and the like are solved, full-dimensional detection from the surface to the interior and from the macroscopic level to the microscopic level is achieved through a multi-source data fusion and closed-loop feedback mechanism, and the reliability of structural safety evaluation and the scientificity of maintenance decision are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of concrete structure detection, and particularly to a method for detecting the quality of quay apron blocks. Background Art

[0002] In port and coastal engineering, quay apron blocks, as the core protective structures against wave impact, are subjected to complex marine environment loads for a long time, and are prone to damage such as surface cracking, internal cavities, and material aging. These defects will significantly reduce the structural durability and even cause local collapse accidents. Therefore, regular quality inspection of apron blocks is crucial for ensuring the safety of port operation.

[0003] In the prior art, the detection of apron blocks mainly relies on the following methods: Manual visual inspection: Technicians observe surface cracks by climbing or approaching by boat. This method is limited by the experience of inspectors, unable to quantify the crack depth, and has high risks in high-altitude / waterfront operations. For example, a certain port once suffered a dam break of apron blocks due to the omission of underwater cracks during manual inspection. Single-point ultrasonic testing: A handheld flaw detector is used to locally scan the suspected area. Due to the high surface roughness of apron blocks (Ra>12.5μm), ultrasonic waves are prone to scattering, resulting in a detection rate of internal cavities of less than 60%, and unable to obtain three-dimensional defect distribution information. Destructive core sampling test: Core samples are randomly drilled for laboratory strength testing. This method has a low sampling rate (usually <0.1%), damages the structural integrity, and cannot reflect the spatial heterogeneity of material properties. A certain study shows that traditional core sampling schemes may miss more than 80% of the micro-defects in the interfacial transition zone.

[0004] The prior art has the following key problems: Contradiction between detection efficiency and coverage: High-precision detection (such as laser scanning) requires long-time static operation and is difficult to adapt to the detection window limitations caused by tidal changes; Lack of collaborative analysis of multi-scale defects: Surface morphology, internal defects, and material property data are isolated, lacking a fusion evaluation model; Insufficient suppression of environmental interference: Temperature and humidity fluctuations, mechanical vibrations, etc. in the marine environment are likely to cause detection data distortion. For example, the signal-to-noise ratio of ultrasonic echo signals is often lower than 20dB. In addition, most of the existing detection results are presented in two-dimensional reports, which cannot intuitively reflect the three-dimensional spatial relationship of defects, resulting in the formulation of maintenance plans relying on empirical judgments and having the risks of over-repair or under-repair. For example, a certain project wasted 30% of the reinforcement materials due to the misjudgment of crack directions in two-dimensional drawings.

[0005] The above problems indicate that there is an urgent need for a full-life cycle detection technology for apron blocks that is efficient, accurate, and adaptable to complex marine environments to meet the needs of modern port intelligent operation and maintenance. Summary of the Invention

[0006] The present invention overcomes the problems of low efficiency, insufficient accuracy in locating internal defects, and one-sided evaluation of material properties in traditional detection methods, and provides a method for detecting the quality of quay apron blocks. Through multi-source data fusion and a closed-loop feedback mechanism, it realizes full-dimensional detection from the surface to the interior and from the macroscopic to the microscopic, significantly improving the reliability of structural safety assessment and the scientific nature of maintenance decisions.

[0007] To achieve the above object, the present invention adopts the following scheme: A method for detecting the quality of quay apron blocks, comprising the following steps: S1: Use a mobile three-dimensional laser scanning device to scan the apron blocks to obtain laser point cloud data, construct a three-dimensional surface model through a real-time point cloud registration algorithm, and extract surface geometric features based on a deep convolutional neural network, including the depth of surface cracks and the spalling area. Mark the areas where the depth of surface cracks exceeds a preset depth threshold or the spalling area is greater than a preset area threshold as high-risk areas, and generate a heat map of high-risk areas; S2: Based on the heat map of high-risk areas, use an adjustable frequency dual-mode flaw detection device to detect cavities and cracks inside the apron blocks: perform encrypted scanning with ultrasonic probes in high-risk areas; arrange multiple impact points along the stress diffusion path in the projection area of the surface crack extension direction, and use an array impact echo device to collect vibration signals at the impact points; Fuse the ultrasonic data and vibration signal data through a multi-sensor data fusion algorithm to establish a three-dimensional defect probability distribution map, and mark the areas where the internal cavity diameter or crack length is greater than the corresponding upper limit threshold as severely defective areas; S3: Drill multiple groups of core samples within a predetermined distance range around the severely defective areas and conduct compressive strength tests and acoustic emission damage detections; conduct an infrared spectral composition census on non-severely defective areas to detect the degree of concrete carbonation and chloride ion erosion index; in the defect-matrix interface transition zone, arrange micro-area hardness test points according to the material property gradient and conduct hardness tests; S4: Establish a dynamic evaluation model including the surface defect distribution entropy value, the proportion of internal defect volume, and the material property dispersion coefficient; calculate the comprehensive safety factor S through a multi-objective optimization algorithm, and compare it with the upper safety threshold S u and the lower limit S d for comparison to generate a three-dimensional visual inspection report.

[0008] Preferably, in the step S1, the robotic arm of the mobile three-dimensional laser scanning device is integrated with a dual-mode lidar and an inertial measurement unit, and performs the following operations: Based on continuous three - point cloud slice data, the differential geometry algorithm is used to calculate the local curvature change rate of the surface of the armor block in real - time. The curvature change rate is quantified by the ratio of the Gaussian curvature difference between adjacent measurement points to the arc - length spacing; when the local minimum curvature radius is less than 2 meters, the surface roughness index exceeds the standard - defined threshold of 0.5 mm, and ambient vibration in the range of 0.1 - 5 Hz is detected, the high - frame - rate lidar mode is activated, and the scanning frame rate is increased to not less than 200 Hz. The adaptive Kalman filter is used to eliminate the low - frequency vibration noise caused by waves. By fusing the state equations of the robotic arm's pitch angle, angular velocity, and linear acceleration with the observation data of the inertial measurement unit, the covariance matrices of the process noise and the observation noise are updated in real - time to suppress wave interference. In the high - frame - rate lidar mode, the moving speed of the end of the robotic arm is dynamically adjusted according to the local minimum curvature radius. When the minimum curvature radius is less than 1 meter, a low - speed scan of 0.2 m / s is used; when the curvature radius is in the range of 1 - 2 meters, a medium - speed scan of 0.5 m / s is used. At the same time, the laser line density is increased to 800 lines per meter, and the point - cloud overlap rate of adjacent scan paths is ensured to be not less than 40%.

[0009] Preferably, in step S1, the real - time point - cloud registration algorithm specifically includes the following steps: Coarse registration: The improved iterative closest point algorithm is used, and the weighted sum of the geometric distance and the curvature difference is used as the optimization target. The curvature difference is quantified by the mixed difference between the Gaussian curvature and the mean curvature of the corresponding points, and the Gaussian curvature weight factor is set to 0.5; when screening the matching point pairs, only the point pairs with the ratio of the principal curvature radii greater than 0.7 and the cosine value of the normal - vector included angle exceeding 0.9 are retained, and the mismatches with inconsistent surface geometric features are excluded. Fine registration: Based on the field - programmable gate array - accelerated Gauss - Newton iterative algorithm, the Jacobian matrix containing the rotation component and the translation component is calculated in parallel, and the optimal transformation parameters are solved by the damped least - squares method. The damping factor is set to 0.0001 to prevent matrix singularity; combined with the three - dimensional feature descriptor matching, the local surface is divided into 32 azimuth intervals and the height distribution histogram is statistically calculated. The chi - square distance is used to calculate the feature similarity, and the regions with matching scores lower than the threshold are marked as low - confidence regions. Dynamic evaluation and re - planning: The registration error of the local area is calculated in real - time. When the root - mean - square error of the area exceeds 0.8 mm, it is determined as a registration - failure area; the scan - path re - planning is triggered. On the basis of the original scan path, a deflection angle of ±15 degrees is added, the point - cloud line density is increased to 1200 lines per meter, the overlap rate is increased to 60%, and the low - confidence regions are subjected to supplementary scanning operations.

[0010] Preferably, in step S2, the frequency-tunable dual-mode flaw detection device includes an ultrasonic probe and an array impact echo device. The ultrasonic probe adopts the following layered waveguide structure and dynamic control strategy: The waveguide layer of the ultrasonic probe is composed of a piezoelectric composite substrate and five layers of gradient impedance matching layers. The impedance values of each layer gradually change according to an exponential law from the substrate to the couplant interface. The total thickness of the matching layer is 2 mm to suppress the acoustic energy scattering caused by surface roughness. When the arithmetic mean deviation of the surface roughness is detected to exceed 12.5 μm, the wide-band frequency-sweeping detection mode is started, and continuous frequency sweeping is performed at a rate of 50 kHz per second within the emission frequency range of 1.5 MHz to 2.5 MHz, and the emission pulse width is dynamically adjusted according to the frequency-sweeping range; Based on the longitudinal wave velocity of the material and the current emission frequency, the effective focusing depth is calculated. When the fluctuation of the focusing depth exceeds ±10% of the nominal value, the calibration mode is triggered: emit a 1 MHz reference pulse, and reverse correct the actual wave velocity of the material through the echo time difference to eliminate the sound velocity error caused by concrete aging; A non-linear relationship model between the contact pressure and the surface roughness is established, and the probe height is adjusted in real time through a piezoelectric actuator to keep the couplant pressure stable in the range of 0.2 to 0.5 MPa; a proportional-integral-derivative controller is used to achieve dynamic balance, with its proportional coefficient set to 0.8, the integral time set to 0.1 s, the derivative time set to 0.05 s, and the sampling frequency of the pressure sensor is 1 kHz.

[0011] Preferably, in step S2, the array impact echo device is used for internal defect location by using the following method: The impact energy is controlled in five levels by adjusting the impact speed of the array impact echo device: 0.5 J, 1 J, 2 J, 3.5 J, 5 J; the impact frequency is dynamically selected according to the surface Vickers hardness: high-frequency impact at 10 Hz is used when the hardness is lower than 300 MPa, medium-frequency impact at 5 Hz is used when the hardness is in the range of 300 - 500 MPa, and low-frequency impact at 2 Hz is used when the hardness exceeds 500 MPa; The surface wave velocity is calculated according to the elastic modulus, density and Poisson's ratio of the material. The surface wave wavelength is calculated based on the center analysis frequency and the surface wave velocity. Piezoelectric sensors are arranged at intervals of one-eighth of the surface wave wavelength, and several piezoelectric sensors are arranged in a spiral-grid hybrid layout to form a sensor array covering three-dimensional space; The vibration signals collected by the echo device are processed by the fusion of short-time Fourier transform and wavelet packet decomposition; among them, the short-time Fourier transform uses a Gaussian-modulated cosine window function, and the window width is dynamically adjusted according to the impact frequency; the wavelet packet decomposition selects the 10th-order Daubechies wavelet basis, and according to the sampling frequency f sample and the center frequency f central the decomposition layer number is set to ; Adopt a high-low frequency dual-channel analysis method to calculate the signal energy integral value for the low-frequency channel of 0.5 - 10 kHz, and calculate the kurtosis index for the high-frequency channel of 10 - 50 kHz; Based on the data of the sensor array, use the time reversal algorithm to reconstruct the three-dimensional wave field, and set the observation time window length to 50 milliseconds; when the signal energy integral value of the low-frequency channel exceeds 0.8 times the reference value of healthy concrete, the kurtosis index value of the high-frequency channel exceeds the threshold, and the diameter of the three-dimensional wave field focusing area is greater than 10 cm, the redundant area is determined as an effective internal defect.

[0012] Preferably, in the step S3, the core sampling and detection specifically include the following steps: Based on the fused data of the laser displacement sensor and the inertial measurement unit, calculate the offset angle of the drill bit axis in real time: measure the lateral offset through the laser sensor, and compensate for the mechanical vibration error by combining the gyroscope angular velocity data, with the compensation coefficient set to 0.05 s / radian; use a fuzzy proportional-integral-derivative controller to adjust the stroke of the four groups of hydraulic cylinders, aiming to minimize the weighted sum of the offset angle and the change rate of the axial force, with the maximum deflection angle limited to ±15 degrees, and trigger an alarm when the limit is exceeded; Calculate the dimensionless working condition index in real time, which is defined as the product of the axial drilling force and the drill bit radius divided by the product of the feed rate and the hydraulic pressure. When this index exceeds the set threshold, implement three-level protection: adjust the retraction speed to 2 times the current feed rate plus 50 mm / s; reduce the drill bit speed to 300 revolutions per minute; spray coolant at a flow rate of 0.2 times the axial force; The following enhancement methods are adopted for the core sample compressive strength test: Control the loading rate in three stages: use a slow loading rate of 0.5 MPa / s when the estimated strength is below 30%, increase the speed to 1.0 MPa / s in the range of 30% - 70%, and reduce the speed to 0.2 MPa / s after exceeding 70%; Synchronously collect acoustic emission signals, calculate the damage evolution b value by counting the number of events with amplitudes exceeding the range of 45 dB to 95 dB. When the change rate of the b value exceeds the set critical threshold, it is determined as a precursor to critical failure; Establish a linear positive correlation model between the ratio of axial force to feed rate and the compressive strength, and a negative correlation model between the time integral of the dimensionless working condition index and the strength, and comprehensively calculate the actual strength of the core sample.

[0013] Preferably, the infrared spectral composition census of the non-serious defect area in the step S3 specifically includes the following steps: For the short-wave infrared channel of 2.5 - 5 microns, a quantum cascade laser with a central frequency of 120 terahertz is used as the light source to scan the characteristic peaks of silicate crystallization at a resolution of 4 cm⁻¹; for the long-wave infrared channel of 8 - 12 microns, the characteristic peaks of carbonate and chloride ions are scanned with a step size of 0.5 microns, and the single integration time is set to 50 milliseconds; based on the signal-to-noise ratio of the two bands, the fusion weights are dynamically allocated, and the weight coefficient is set to the normalized result of the square value of the signal-to-noise ratio of each band. Calibrate the influence of the ambient temperature by referring to the blackbody radiation source at a specified temperature, and establish a wavelength drift model including the steady-state thermal expansion coefficient and the transient response coefficient; perform non-linear correction on the original spectrum, and use the Planck blackbody radiation formula to eliminate the interference of ambient thermal radiation. An improved convolutional recurrent neural network is used for analysis. The input layer of the network receives a 256×256-dimensional dual-band spectral splicing matrix; the convolutional module uses a dilated convolutional kernel with a dilation rate of 2 to extract multi-scale features, and the recurrent module uses a bidirectional long short-term memory network to capture the spectral sequence dependence, and the hidden layer dimension is set to 128; when training the model, a weighted cross-entropy loss function is used, and the weights are inversely allocated according to the material aging level, and ±2% spectral perturbation is added to generate adversarial samples to improve robustness. When calculating the aging index, the output probability of the neural network is fused and a weight of 60% is assigned to it, the carbonation index is calculated and a weight of 30% is assigned to it, and the chloride ion erosion index is calculated and a weight of 10% is assigned to it; the final aging index is calculated by weighted summation. When the aging index exceeds 0.6, it is determined to be severely aged, and a maintenance warning is generated.

[0014] Preferably, in step S4, the method for the dynamic evaluation model to perform a comprehensive structural safety assessment is as follows: An improved weighted Shannon entropy model is used to calculate the disorder degree of the surface defect distribution, and the coupling effect of the crack depth and area is used as the weight factor; a fractal dimension correction term is introduced. When the fractal dimension exceeds 1.2, the total entropy value is increased by 30% of the fractal dimension deviation amount to enhance the sensitivity to complex crack networks. Calculate the effective stress concentration coefficient, set the reference volume to 1000 cubic centimeters, and the void volume is accumulated according to the equivalent sphere radius; when the crack length exceeds 30 centimeters, the stress concentration coefficient increases an additional correction term, and the correction amplitude is proportional to the length excess value; when continuous surface cracks are detected, the internal defect weight factor is increased by 40%. Based on the spatial autocorrelation Moran's I index, evaluate the uniformity of the material property distribution. When calculating the index, an exponential decay function is used to define the spatial weight matrix, and the decay coefficient is set to 0.5 meters; when the Moran's I index exceeds 0.3, it is determined that the material properties are clustered, and the dispersion coefficient is amplified according to the exponential difference ratio, and the amplification factor is set to 0.5 at most. Dynamically synthesize the safety factor: Calculate the comprehensive safety factor by using the reverse weight formula with the surface entropy value, the corrected internal defect coefficient, and the dispersion coefficient in a weight ratio of 0.5:0.3:0.2; generate a red warning when the safety factor is lower than 1.2, and mark the high-risk area in the 3D model.

[0015] Preferably, in step S4, an IoT module is set to optimize the evaluation process. The operation process of the IoT module includes the following steps: Perform wavelet packet decomposition and dynamic quantization compression on the detected data stream. Use the Daubechies 9 / 7 wavelet basis function to decompose the data into multiple sub-bands, and dynamically adjust the quantization step size according to the signal-to-noise ratio of each sub-band; use a compression ratio of 10:1 in the normal state, and switch to a 5:1 high-fidelity compression mode when high-risk defects are detected, giving priority to retaining high-frequency detail features; Construct a multi-physical field compensation matrix for temperature, humidity, and pressure to quantify the influence of environmental factors on the measured values. Set the sensitivity of the strain measurement value to temperature to be 0.3 microstrain / °C, and the sensitivity of the stress measurement value to humidity to be 0.05 MPa / % relative humidity; collect environmental data in real time and substitute it into the compensation matrix to correct the strain and stress detection results; Construct a time series prediction model based on the long short-term memory network. The input feature vector includes the surface defect entropy value, the internal defect volume ratio, the material dispersion coefficient, and environmental parameters; set the dimension of the network hidden layer to 64, with the time step covering 24 hours of continuous data, and output the risk probability value for the next 30 days; divide the risk level into three levels: a probability value lower than 0.3 is low risk, 0.3 - 0.6 is medium risk, and exceeding 0.6 is high risk; identify the key parameters through the Fisher information matrix, set an elastic coefficient of 0.8 to limit the update amplitude of the parameters, and retain the historical data feature distribution.

[0016] Preferably, the IoT module includes a visualization terminal interface unit. The 3D visualization detection report uses an augmented reality interface, including the following method steps: Construct an environmental point cloud map based on the simultaneous localization and mapping technology to achieve the six-degree-of-freedom space alignment of the detection data coordinate system and the real scene, and control the real-time update error of the pose matrix to be less than 0.3 mm; the surface topography layer is superimposed and displayed at a resolution of 0.5 mm / pixel, and the defect distribution layer is colored with a red - orange gradient, and the coloring intensity increases linearly with the crack depth. When the depth exceeds 3 mm, the transparency drops to 0.2; Identify gesture actions based on a 3D convolutional neural network. Input a sequence of 16 consecutive depth images, perform gesture space trajectory fitting, and output operation instructions; perform voxel cutting on any specified plane to generate a material property profile, extract the mechanical parameter distribution of the profile through the Gaussian kernel interpolation algorithm to generate a profile stress distribution map, and set the interpolation bandwidth to 5 mm fixed to suppress noise interference; Calculate the repair volume of the defective area, use the Heaviside function to determine the cavity boundary, and set the integration step size to 1 mm; aiming at maximizing stiffness and minimizing material usage, the volume of the reinforcing material does not exceed 30% of the original structure, set the optimization density threshold to 0.3, and perform topology optimization to design the reinforcing structure; the visualization terminal interface unit displays the von Mises stress nephogram of the structure after reinforcement in real time; use gesture interaction to modify the reinforcement scheme, dynamically update the mechanical simulation results, and the optimization iteration period is less than 2 seconds.

[0017] The present invention has at least the following beneficial effects: (1) By integrating 3D laser scanning technology and intelligent recognition algorithms, automatic and rapid detection of surface defects of armor blocks is achieved. Compared with the traditional manual inspection method, this method can comprehensively cover the surface of the structure, effectively identify micro-cracks and local spalling areas, and significantly improve the accuracy of defect location. At the same time, the multi-angle scanning and real-time data fusion mechanism overcomes the limitations of single-view detection, ensuring no missed detection in areas with complex geometric features. Through a standardized data processing process, human judgment errors are greatly reduced, providing reliable basic data support for subsequent grading evaluation. (2) The adaptive scanning strategy designed for the complex curved surface characteristics of armor blocks effectively improves the quality of point cloud acquisition in high-curvature areas by dynamically adjusting scanning parameters. Combining anti-interference algorithms and closed-loop control mechanisms significantly suppresses measurement deviations caused by factors such as wave vibration and temperature and humidity changes in the marine environment. The optimized point cloud registration process not only ensures millimeter-level accuracy but also shortens the overall operation time, making large-scale structure detection feasible in engineering. This method is particularly suitable for the maintenance scenarios of special-shaped armor blocks with complex surface textures and variable geometric shapes. (3) The multi-physical-field coupling detection technology gives full play to the complementary advantages of ultrasonic flaw detection and impact echo methods. High-frequency ultrasonic scanning provides fine imaging of near-surface defects, while the impact echo technology enhances the detection ability of deep defects. Through signal fusion and 3D reconstruction algorithms, the spatial distribution characteristics of defects are completely revealed, avoiding the vision limitations of traditional single-modal detection. This solution shows good adaptability in the identification of typical defects such as internal cavities and crack networks in concrete, significantly reducing the risk of missed detection and providing multi-dimensional data basis for structural integrity assessment. (4) The hierarchical detection strategy based on material property gradients realizes the collaborative evaluation of macroscopic mechanical properties and microscopic composition changes. The non-contact spectroscopic analysis technology breaks through the spatial limitations of traditional sampling detection and can quickly survey the carbonation depth and erosion conditions in a large area. Combining local high-precision hardness testing, the performance mutation characteristics at the defect interface transition zone are accurately captured. While reducing structural damage, this method significantly improves the spatial resolution of material property data, providing a refined data basis for durability degradation analysis. (5) The constructed multi-source data dynamic evaluation model objectively reflects the overall and regional characteristics of the structural safety state through quantitative index fusion and weight self-adaptive mechanism. The augmented reality visualization system transforms abstract detection data into an interactive model in 3D space, intuitively presenting the distribution law and development trend of defects. Combining the repair plan generated by predictive maintenance algorithms effectively guides the accurate allocation of maintenance resources. This technology system significantly shortens the response cycle from detection to decision-making, improves the pertinence and economy of maintenance work, and provides intelligent decision-making support for project management. Brief Description of the Drawings

[0018] Figure 1 It is a principle flow chart of the method for detecting the quality of the quay armor block provided by the present invention. Specific implementation mode

[0019] The following further elaborates on the present invention in conjunction with the accompanying drawings, so that those skilled in the art can implement it with reference to the text of the specification.

[0020] As Figure 1 shown, the method for quality inspection of quay armor blocks includes the following steps: S1: Use a mobile three-dimensional laser scanning device to scan the armor blocks to obtain laser point cloud data, construct a three-dimensional surface model through a real-time point cloud registration algorithm, extract surface geometric features based on a deep convolutional neural network, including surface crack depth and spalling area. Mark the areas where the surface crack depth exceeds the preset depth threshold or the spalling area is greater than the preset area threshold as high-risk areas, and generate a heat map of high-risk areas; S2: Based on the heat map of high-risk areas, use a frequency-tunable dual-mode flaw detection device to detect voids and cracks inside the armor blocks: perform encrypted scanning with ultrasonic probes in high-risk areas; arrange multiple impact points along the stress diffusion path in the projection area of the surface crack extension direction, and use an array impact echo device to collect vibration signals at the impact points; Fuse the ultrasonic data and vibration signal data through a multi-sensor data fusion algorithm to establish a three-dimensional defect probability distribution map, and mark the areas where the internal void diameter or crack length is greater than the corresponding upper limit threshold as severely defective areas; S3: Drill multiple groups of core samples within a predetermined distance range around the severely defective areas and conduct compressive strength tests and acoustic emission damage detections; conduct infrared spectroscopy composition surveys on non-severely defective areas to detect the carbonation degree and chloride ion erosion index of the concrete; in the defect-matrix interface transition zone, arrange micro-area hardness test points according to the material property gradient and conduct hardness tests; S4: Establish a dynamic evaluation model including surface defect distribution entropy value, internal defect volume ratio, and material property dispersion coefficient; calculate the comprehensive safety factor S through a multi-objective optimization algorithm, and compare it with the upper safety threshold Su and the lower safety threshold Sd to generate a three-dimensional visual inspection report.

[0021] The scanning accuracy of the three-dimensional laser scanning device can be set from 0.5 mm to 2 mm, the crack depth determination threshold is set to 3 mm or 5 mm, and the spalling area threshold can be set to 50 square centimeters or 80 square centimeters. Use a FARO Focus series laser scanner, equipped with a six-degree-of-freedom robotic arm of the UR10e model, and install a scanning probe at the end of the robotic arm. The scanning probe is installed within a range of 0.3 m to 1.5 m from the surface of the armor block. The robotic arm moves and scans according to a preset path, stitches the point cloud data in real time, and automatically marks the defect areas through a deep convolutional neural network (such as ResNet-50), which can greatly improve the surface crack recognition accuracy and effectively shorten the single scan time.

[0022] The frequency of the ultrasonic probe can be selected as 1.5 MHz, 2 MHz or 2.5 MHz, and the scanning grid spacing can be set to 5 cm × 5 cm or 10 cm × 10 cm. The impact energy of impact echo detection can be divided into three levels: 0.5 J, 2 J, and 5 J, and the vibration signal acquisition duration can be set to 40 ms, 50 ms or 60 ms. Olympus Epoch series ultrasonic flaw detectors and Proceq Pundit PL-200 impact echo meters can be used. The ultrasonic probe can be installed at the center position of the surface of the detection area, and the impact hammer can be arranged at intervals of 10 cm to 30 cm along the crack extension direction. After high-frequency ultrasonic scanning, switch to the impact echo mode, and fuse the two types of data to generate a three-dimensional defect map, which can effectively reduce the positioning error of voids inside the block and the measurement error of crack length.

[0023] The core drilling diameter can be selected as 50 mm or 75 mm, and the compressive strength loading rate can be set to 0.3 MPa / s, 0.5 MPa / s or 1.0 MPa / s. The spacing of the hardness test points can be set to 2 mm, 5 mm or 10 mm. A Hilti DD120 core drill is used, and an Instron 5982 universal testing machine is used for strength testing. The core drill can be installed 10 cm to 20 cm away from the defect edge, and the infrared spectroscopy probe can cover the entire surface of the non-defect area. After drilling the core sample, it is pressurized in stages until failure, and the peak frequency of the acoustic emission signal is recorded synchronously, which can reduce the repeatability error of the compressive strength test and control the detection accuracy of chloride ion content to 0.01%.

[0024] The safety factor threshold can be set to 1.2, 1.5 or 1.8, and the weight adjustment range can be selected as 20%, 40% or 60%. A Dell Precision workstation can be used to run the evaluation algorithm, and the calculation is accelerated by an NVIDIA Quadro GPU. The data processing server can be deployed in the dock control center and connected to the detection equipment through a 5G network. After importing the surface defect distribution, internal defect volume and material performance data, the safety level is automatically calculated and a PDF / 3D model dual-format report is generated, which has a very short warning response time for high-risk areas, and the provided data analysis results can effectively improve the accuracy of the maintenance plan generation.

[0025] In another technical solution, in the step S1, the robotic arm of the mobile three-dimensional laser scanning device is integrated with a dual-mode lidar and an inertial measurement unit, and the following operations are performed: Based on continuous three - point cloud slice data, a differential geometry algorithm is used to calculate the local curvature change rate on the surface of the armor block in real - time. The curvature change rate is quantified by the ratio of the Gaussian curvature difference between adjacent measurement points to the arc - length spacing. When the local minimum curvature radius is less than 2 meters, the surface roughness index exceeds the 0.5 - millimeter threshold defined by the standard, and ambient vibrations in the range of 0.1 - 5 Hz are detected, the high - frame - rate lidar mode is activated, and the scanning frame rate is increased to no less than 200 Hz. An adaptive Kalman filter is used to eliminate the low - frequency vibration noise caused by waves. By fusing the state equations of the robotic arm's pitch angle, angular velocity, and linear acceleration with the observation data of the inertial measurement unit, the process noise and observation noise covariance matrices are updated in real - time to suppress wave interference. In the high - frame - rate lidar mode, the moving speed of the robotic arm end is dynamically adjusted according to the local minimum curvature radius. When the minimum curvature radius is less than 1 meter, a low - speed scan of 0.2 m / s is adopted; when the curvature radius is in the range of 1 - 2 meters, a medium - speed scan of 0.5 m / s is adopted. At the same time, the laser line density is increased to 800 lines per meter, and the point - cloud overlap rate of adjacent scan paths is ensured to be no less than 40%.

[0026] The number of joints of the robotic arm can be selected as 6 - axis or 7 - axis, and the end - load capacity can be set to 5 kg or 10 kg. The curvature - radius judgment threshold can be set to 2 meters, 1.5 meters, or 1 meter. A six - degree - of - freedom robotic arm of the Universal Robots UR10e model can be used, and the material of the robotic - arm body can be selected as aluminum alloy or carbon - fiber composite material. The robotic - arm base can be installed at the central position on the top of the mobile scanning platform and rigidly connected to the scanning device. The robotic arm is controlled to move along the normal direction of the armor - block surface through a preset program, and the scanning speed can be set to 0.2 m / s or 0.5 m / s. In the efficient scanning mode, it can maintain a high coverage rate for complex surfaces and reduce the robotic - arm positioning error to 0.1 mm.

[0027] The time interval for calculating the curvature change rate can be set to 0.1 s or 0.2 s, and the surface - roughness threshold can be set to 0.3 mm, 0.5 mm, or 0.8 mm. The lidar can be of the FARO Focus S 350 model, and the scanning frame rate can be switched to 100 Hz, 200 Hz, or 300 Hz. The lidar can be installed 30 cm in front of the flange at the end of the robotic arm, and the scanning angle can be adjusted to ±45 degrees. In the specific implementation process, the curvature difference between adjacent three - point cloud slices is calculated in real - time. When the curvature radius is less than the threshold and the roughness exceeds the standard, the scanning frame rate is automatically increased to more than 200 Hz, the point - cloud density in the high - curvature area can be increased to 1200 points per square meter, and the data - acquisition efficiency is increased by 40%.

[0028] The wave vibration filtering bandwidth can be set to 0.1 - 5 Hz or 0.5 - 3 Hz, and the scan line density can be set to 500 lines / m, 800 lines / m, or 1000 lines / m. The inertial measurement unit can be of the Xsens MTi-670 model, and the acceleration measurement range can be set to ±8g or ±16g. The inertial measurement unit can be installed 10 cm behind the lidar and connected to the radar data interface through the RS485 bus. The manipulator attitude data and the IMU vibration signal are fused through a Kalman filter to correct the scan path coordinate offset in real time. By filtering and denoising the data, the point cloud distortion rate caused by wave interference is reduced to less than 5%, and the registration error in the overlapping area is less than 0.8 mm.

[0029] In step S1, the real-time point cloud registration algorithm specifically includes the following steps: Perform coarse registration: Use the improved iterative closest point algorithm, with the weighted sum of the geometric distance and the curvature difference as the optimization target, where the curvature difference is quantified by the mixed difference between the Gaussian curvature and the mean curvature of the corresponding points, and the Gaussian curvature weight factor is set to 0.5; when screening matching point pairs, only retain the point pairs with a principal curvature radius ratio greater than 0.7 and a cosine value of the normal vector angle exceeding 0.9, and exclude the false matches with inconsistent surface geometric features.

[0030] The curvature consistency weight factor can be set to 0.3, 0.5, or 0.7, the principal curvature radius ratio threshold can be set to 0.6, 0.7, or 0.8, and the cosine value threshold of the normal vector angle can be set to 0.85, 0.9, or 0.95. The coarse registration algorithm can be implemented using an Intel Arria 10 series FPGA chip, and the matching point pair screening module can use an off-the-shelf point cloud processing library such as PCL (PointCloud Library). The FPGA module can be installed in the PCIe slot of the scan device control cabinet and connected to the main control computer through Gigabit Ethernet. The original point cloud is sliced at 5 mm intervals, the curvature difference between adjacent slices is calculated, and the false matching points that do not meet the curvature ratio and normal vector conditions are removed. Under the condition of a low false matching rate, the coarse registration has a very fast matching speed.

[0031] Perform fine registration: Based on the field-programmable gate array (FPGA) to accelerate the Gauss-Newton iterative algorithm, parallelly calculate the Jacobian matrix including the rotation component and the translation component, and solve the optimal transformation parameters through the damped least squares method. The damping factor is set to 0.0001 to prevent matrix singularity; combined with three-dimensional feature descriptor matching, divide the local surface into 32 azimuth intervals and statistically calculate the height distribution histogram, calculate the feature similarity using the chi-square distance, and mark the areas with matching scores lower than the threshold as low-confidence areas; the Gauss-Newton iterative damping factor can be set to 0.0001, 0.001, or 0.01, the azimuth interval of the feature descriptor can be divided into 16, 32, or 64 partitions, and the histogram resolution can be set to 0.3 mm, 0.5 mm, or 1.0 mm. The Xilinx Kintex UltraScale FPGA development board can be used to accelerate matrix operations, and the feature matching unit can be implemented using the Open3D open-source framework. The FPGA acceleration card can be integrated into the data processing workstation and communicate with the CPU through the PCIe 3.0 interface. Based on the rough registration result, parallelly calculate the Jacobian matrix and iteratively solve the optimal transformation parameters, and at the same time match the SHOT feature descriptor. The fine registration accuracy can reach 0.5 mm and it takes very little single-frame processing time.

[0032] Perform dynamic evaluation and replanning: Calculate the registration error of the local area in real time. When the root mean square error of the area exceeds 0.8 mm, it is determined as a registration failure area; trigger the rescan path replanning, add a deflection angle of ±15 degrees on the original scan path, increase the point cloud line density to 1200 lines per meter, increase the overlap rate to 60%, and perform a supplementary scan operation on the low-confidence areas.

[0033] The local registration error threshold can be set to 0.5 mm, 0.8 mm, or 1.0 mm, the scan path deflection angle can be set to ±10 degrees, ±15 degrees, or ±20 degrees, and the supplementary scan line density can be set to 1000 lines / m, 1200 lines / m, or 1500 lines / m. The Velodyne VLP-16 lidar can be used for supplementary scanning, and the path planning algorithm can be selected from the ROS (Robot Operating System) navigation package. The supplementary scanning device can be installed 50 cm to the side of the original scanning device and synchronized with the main control system through the CAN bus. Calculate the registration error of each area in real time. When the error exceeds the threshold, generate a spiral progressive supplementary scan path, automatically adjust the scan angle and line density, which can effectively reduce the time-consuming of repeated scanning in low-confidence areas and ensure a very high integrity of the overall point cloud.

[0034] In another technical solution, in the step S2, the frequency-tunable dual-mode flaw detection device includes an ultrasonic probe and an array impact echo device, and the ultrasonic probe adopts the following hierarchical waveguide structure and dynamic control strategy: The waveguide layer of the ultrasonic probe is composed of a piezoelectric composite substrate and five layers of gradient impedance matching layers. The impedance values of each layer gradually change according to an exponential law from the substrate to the couplant interface. The total thickness of the matching layer is 2 mm to suppress the acoustic energy scattering caused by surface roughness. When the arithmetic mean deviation of the surface roughness is detected to exceed 12.5 μm, the wide-band frequency-sweeping detection mode is started. In the emission frequency range of 1.5 MHz to 2.5 MHz, continuous frequency sweeping is performed at a rate of 50 kHz per second, and the emission pulse width is dynamically adjusted according to the frequency-sweeping range. The total number of layers of the waveguide layer can be set to 3 layers, 5 layers or 7 layers, and the frequency-sweeping range can be set to 1 to 2.5 MHz or 1.5 to 3 MHz. The ultrasonic probe can be selected as the Olympus V358-SU model, and the piezoelectric composite material can be selected as PZT-5H or PMN-PT material. The waveguide layer of the probe can be installed between the transducer wafer and the protective film, and the total thickness is controlled within the range of 2 to 3 mm. When the surface roughness exceeds 12.5 μm, the 1.5 to 2.5 MHz frequency-sweeping mode is automatically started. Ultrasonic waves with an emission pulse width of 3 to 5 μs are emitted at each frequency point. The signal-to-noise ratio of the rough surface echo can be increased to more than 30 dB, and the defect detection rate exceeds 90%.

[0035] Calculate the effective focus depth based on the longitudinal wave velocity of the material and the current emission frequency. When the fluctuation of the focus depth exceeds ±10% of the nominal value, trigger the calibration mode: emit a 1 MHz reference pulse, and inversely correct the actual wave velocity of the material through the echo time difference to eliminate the acoustic velocity error caused by concrete aging. The wave velocity correction coefficient can be set to 2.3, 2.5 or 2.8, and the roughness correction exponent can be selected as 0.25, 0.3 or 0.35. The focus calibration pulse frequency can be set to 1 MHz or 0.5 MHz, and the emission pulse width can be set to 2 μs or 3 μs. The calibration signal emission module can be integrated inside the probe handle and connected to the main control unit through a coaxial cable. Calibration pulses are automatically emitted every 10 minutes, the calculated value of the wave velocity is corrected according to the echo time difference, and the material parameters in the depth formula are updated, with a very small focus depth control error, reducing the detection deviation caused by concrete aging to within 3%.

[0036] Establish a non-linear relationship model between contact pressure and surface roughness, and adjust the probe height in real time through a piezoelectric actuator to stabilize the coupling agent pressure in the range of 0.2 to 0.5 MPa; use a proportional-integral-derivative controller to achieve dynamic balance, with the proportional coefficient set to 0.8, the integral time set to 0.1 s, the derivative time set to 0.05 s, and the sampling frequency of the pressure sensor to 1 kHz. The contact pressure range can be set from 0.15 to 0.5 MPa, and the PID control parameters can be set to a proportional coefficient of 0.6, an integral time of 0.2 s, and a derivative time of 0.1 s. The piezoelectric actuator can be selected as the P-623.1CD model of PI company, and the pressure sensor can be selected from the Honeywell FMA series. The actuator can be installed on the top of the probe housing and arranged in parallel with the pressure sensor at an interval of 5 cm. Adjust the downward pressing distance of the probe in real time through closed-loop feedback to maintain the coupling agent film thickness between 0.1 and 0.3 mm, control the pressure fluctuation amplitude within ±0.05 MPa, and reduce the signal loss rate caused by poor contact.

[0037] In the step S2, the array impact echo device is used for internal defect location by the following method: Control the impact energy in five levels by adjusting the impact speed of the array impact echo device: 0.5 J, 1 J, 2 J, 3.5 J, 5 J; dynamically select the impact frequency according to the surface Vickers hardness: use a high-frequency impact of 10 Hz when the hardness is less than 300 MPa, use a medium-frequency impact of 5 Hz in the range of 300 - 500 MPa, and use a low-frequency impact of 2 Hz when the hardness exceeds 500 MPa; the mass of the impact hammer can be selected as 30 g, 50 g or 80 g, and the impact speed can be divided into three levels of 4.47 m / s, 6.32 m / s, 8.94 m / s corresponding to the energy of 0.5 to 5 joules. The impact device can be selected as the Proceq Pundit PL-300 type, and the hammer head material can be selected as tungsten carbide or zirconia ceramic. The impact hammer can be installed at a height of 2 to 5 cm from the detection surface and driven to fall vertically through an electromagnetic coil. Automatically select the impact gear according to the surface hardness value, use the high-frequency mode of 10 Hz when the hardness is below 300 MPa, and use the low-frequency mode of 2 Hz above, fully suppressing the incidence of signal overload and ensuring a high effective data acquisition rate.

[0038] Calculate the surface wave velocity according to the elastic modulus, density and Poisson's ratio of the material, calculate the surface wave wavelength according to the center analysis frequency and the surface wave velocity, arrange the piezoelectric sensors at an interval of one-eighth of the surface wave wavelength, and arrange several piezoelectric sensors in a spiral-grid hybrid layout to form a sensor array covering three-dimensional space; Perform short-time Fourier transform and wavelet packet decomposition fusion processing on the vibration signals collected by the echo device; among them, the short-time Fourier transform uses a Gaussian-modulated cosine window function, and the window width is dynamically adjusted according to the impact frequency; the wavelet packet decomposition selects the 10th-order Daubechies wavelet basis, according to the sampling frequency fsample With the center frequency f central Set the decomposition level to ; Adopt a high-low frequency dual-channel analysis method. Calculate the signal energy integral value for the low-frequency channel of 0.5 - 10 kHz, and calculate the kurtosis index for the high-frequency channel of 10 - 50 kHz; Based on the data of the sensor array, use the time reversal algorithm to reconstruct the three-dimensional wave field, and set the observation time window length to 50 milliseconds; when the signal energy integral value of the low-frequency channel exceeds 0.8 times the reference value of healthy concrete, the kurtosis index value of the high-frequency channel exceeds the threshold, and the diameter of the three-dimensional wave field focusing area is greater than 10 cm, the redundant area is determined as an effective internal defect.

[0039] Surface wave wavelength calculation formula: , surface wave velocity , where v is the Poisson's ratio, and the center frequency f central can take 25 kHz, the elastic modulus E can take 30 to 50 GPa, and the material density ρ is set to 2300 to 2500 kg / m 3 . The piezoelectric sensor can be selected as the PCB Piezotronics 352C03 model, and the array substrate material can be selected as FR-4 epoxy board or aluminum alloy. The sensors can be arranged at intervals of 10 to 15 cm along the crack direction and installed symmetrically offset by 20 to 30 degrees on both sides. Excite surface waves through a 25 kHz center frequency, collect vibration signals at intervals of 1 / 8 wavelength, and the detection accuracy rate of the crack extension direction exceeds 90%, and the blind area range is reduced to within 5 cm.

[0040] The width of the short-time Fourier transform window function can be set to 5 milliseconds, 10 milliseconds, or 20 milliseconds, and the wavelet packet decomposition level can be selected as 3 layers, 5 layers, or 7 layers. The signal processor can be selected as the NI PXIe-8840 controller, equipped with a 14-bit high-speed acquisition card. The data analysis module can be deployed in an industrial computer and directly connected to the acquisition card through a PCIe interface. Integrate the energy integral value of the 0.5 to 10 kHz low-frequency band and the kurtosis index of the 10 to 50 kHz high-frequency band to generate a heat map of the defect probability distribution. The detection error of the internal cavity diameter is controlled to be less than 1 cm, and the detection time can be shortened from 45 minutes to 20 minutes.

[0041] In another technical solution, in step S3, the core sampling and detection specifically include the following steps: Based on the fused data of the laser displacement sensor and the inertial measurement unit, the offset angle of the drill bit axis is calculated in real time: the lateral offset is measured by the laser sensor, and the mechanical vibration error is compensated by combining the angular velocity data of the gyroscope, with the compensation coefficient set to 0.05 seconds / radian; a fuzzy proportional-integral-derivative controller is used to adjust the stroke of the four hydraulic cylinders, aiming to minimize the weighted sum of the offset angle and the change rate of the axial force, with the maximum deflection angle limited to ±15 degrees, and an alarm is triggered when the limit is exceeded; The dimensionless working condition index is calculated in real time, which is defined as the product of the axial drilling force and the drill bit radius divided by the product of the feed rate and the hydraulic pressure. When this index exceeds the set threshold, three-level protection is executed: the retraction speed is adjusted to 2 times the current feed rate plus 50 mm / s; the drill bit rotation speed is reduced to 300 revolutions per minute; coolant is sprayed at a flow rate of 0.2 times the axial force; The following strengthening methods are adopted for the core sample compressive strength test: The loading rate is controlled in three stages: slow loading at 0.5 MPa / s when the estimated strength is below 30%, the speed is increased to 1.0 MPa / s in the range of 30%-70%, and the speed is reduced to 0.2 MPa / s after exceeding 70%; The acoustic emission signal is collected synchronously. By counting the number of events with amplitudes exceeding the range of 45 dB to 95 dB, the damage evolution b value is calculated. When the change rate of the b value exceeds the set critical threshold, it is determined as a precursor to critical failure; A linear positive correlation model between the ratio of axial force to feed rate and the compressive strength, and a negative correlation model between the time integral of the dimensionless working condition index and the strength are established, and the actual strength of the core sample is calculated comprehensively.

[0042] The stroke adjustment accuracy of the hydraulic cylinder of the core sampling equipment can be set to 0.1 mm or 0.2 mm, and the threshold of the ratio of drilling torque to feed rate can be set to 120 N·s / mm or 150 N·s / mm. The hydraulic servo valve can select the 4WRPE series of BoschRexroth, and the servo motor can select the Yaskawa SGM7G model. The hydraulic cylinders can be installed symmetrically on both sides of the drill rig column and connected to the main shaft through a rigid coupling. The offset of the drill bit is monitored in real time by the laser displacement sensor. When the offset angle exceeds the preset threshold, the hydraulic system automatically adjusts the pressure difference of the four cylinders to improve the drilling perpendicularity accuracy and reduce the risk of core sample breakage.

[0043] The loading rates for the compressive strength test can be divided into three levels: 0.3 MPa / s, 0.5 MPa / s, or 0.8 MPa / s, and the acoustic emission signal acquisition frequency can be set to 100 kHz or 200 kHz. An Instron 3367 universal testing machine can be selected for the testing machine, and a PAC Micro80 model can be selected for the acoustic emission sensor. The core sample fixture can be installed between the upper and lower platen of the testing machine, and the sensor is attached to the middle of the side of the core sample. The pressure is loaded in stages until the specimen fails, and the cumulative number of acoustic emission events and the amplitude distribution are recorded synchronously to accurately identify the critical point of material damage and avoid sudden fracture.

[0044] The linear correlation coefficient of the data fusion model can be set to 0.1 to 0.2, and the integration time range can be set to 30 seconds or 60 seconds. An NI USB-6366 type data acquisition card can be selected, and the control software can be developed based on the LabVIEW platform. The processing module can be integrated into the industrial control computer and communicate with the drill controller through Ethernet. Calculate the correlation between drilling parameters and compressive strength in real time, generate a material property mapping curve, establish a quantitative relationship between the drilling process and material strength, and support in-situ quality assessment.

[0045] In another solution, the implementation of the infrared spectral composition census for non-severe defect areas in step S3 specifically includes the following steps: For the short-wave infrared channel of 2.5 - 5 μm, a quantum cascade laser with a central frequency of 120 THz is used as the light source to scan the characteristic peaks of silicate crystallization at a resolution of 4 cm⁻¹; for the long-wave infrared channel of 8 - 12 μm, scan the characteristic peaks of carbonate and chloride ions with a step size of 0.5 μm, and the single integration time is set to 50 ms; based on the signal-to-noise ratio of the two bands, dynamically allocate the fusion weight, and the weight coefficient is set to the normalized result of the square value of the signal-to-noise ratio of each band; Calibrate the influence of the ambient temperature by referring to the blackbody radiation source at a specified temperature, and establish a wavelength drift model including the steady-state thermal expansion coefficient and the transient response coefficient; perform non-linear correction on the original spectrum and use the Planck blackbody radiation formula to eliminate the interference of ambient thermal radiation; Use an improved convolutional recurrent neural network for analysis. The input layer of the network receives a 256×256-dimensional dual-band spectral splicing matrix; the convolutional module uses a dilated convolutional kernel with a dilation rate of 2 to extract multi-scale features, and the recurrent module uses a bidirectional long short-term memory network to capture the spectral sequence dependence relationship, and the hidden layer dimension is set to 128; when training the model, use a weighted cross-entropy loss function, and the weights are allocated inversely according to the material aging level, and add ±2% spectral perturbation to generate adversarial samples to improve robustness; When calculating the aging index, fuse the output probability of the neural network and assign it a weight of 60%, calculate the carbonation index and assign it a weight of 30%, calculate the chloride ion erosion index and assign it a weight of 10%; calculate the final aging index by weighted summation. When the aging index exceeds 0.6, it is determined as severe aging, and a maintenance warning is generated.

[0046] The spectral resolution of the short-wave infrared channel can be set to 4 cm⁻¹ or 8 cm⁻¹, and the integration time of the long-wave infrared can be set to 40 ms or 60 ms. The quantum cascade laser can select the LaserTune series of Block Engineering, and the infrared detector can select the FLIR A67 model. The laser emission head can be installed 20 cm above the scanning bracket, and the detector is kept 50 cm away from the sample surface. Alternate scanning in two bands, separate signals of different wavelengths through an optical beam splitter, and capture the characteristic absorption peaks of hydration products and erosion products simultaneously.

[0047] The temperature of the blackbody radiation source for temperature drift compensation can be set to 30 °C or 40 °C, and the thermal expansion coefficient of the wavelength correction model can be set to 0.015 nm / °C or 0.025 nm / °C. The reference blackbody can select the LandcalR1500TI model, and the temperature control module can select the Omega CNi16 series. The blackbody source can be placed 10 cm beside the sample to be measured and coupled to the spectrometer entrance through an optical fiber. Collect the blackbody reference spectrum every 5 minutes, dynamically correct the wavelength shift caused by the ambient temperature, and suppress the spectral distortion caused by temperature fluctuations.

[0048] The number of hidden layer nodes of the convolutional recurrent neural network can be set to 64 or 128, and the sample size of the training data set can be set to 1500 groups or 2000 groups. The processor can select the NVIDIA Jetson AGX Xavier, and the framework can select TensorFlow Lite. The computing unit can be embedded inside the spectrometer host and connected to the main control board through a PCIe interface. Stitch the two-band spectral data into a two-dimensional matrix and input it into the network to output the material aging probability value, and automatically identify the signs of early carbonation and chloride ion erosion.

[0049] In another technical solution, in the step S4, the method for the dynamic evaluation model to perform a comprehensive structural safety evaluation is as follows: Use an improved weighted Shannon entropy model to calculate the disorder degree of the surface defect distribution, and use the coupling effect of crack depth and area as the weight factor; introduce a fractal dimension correction term. When the fractal dimension exceeds 1.2, increase the total entropy value by 30% of the fractal dimension deviation amount to enhance the sensitivity to complex crack networks; Calculate the effective stress concentration factor, set the reference volume to 1000 cubic centimeters, and accumulate the void volume according to the equivalent sphere radius; when the crack length exceeds 30 centimeters, an additional correction term is added to the stress concentration factor, and the correction amplitude is proportional to the excess length value; when continuous surface cracks are detected, the internal defect weight factor is increased by 40%. Evaluate the uniformity of material property distribution based on the spatial autocorrelation Moran's I index. When calculating the index, use an exponential decay function to define the spatial weight matrix, and set the decay coefficient to 0.5 meters; when the Moran's I index exceeds 0.3, it is determined that the material properties are clustered, and the coefficient of variation is amplified according to the exponential difference ratio, with the maximum amplification factor set to 0.5. Dynamically synthesize the safety factor: Calculate the comprehensive safety factor using the reverse weight formula with the surface entropy value, the corrected internal defect coefficient, and the coefficient of variation in a weight ratio of 0.5:0.3:0.2; when the safety factor is lower than 1.2, generate a red warning and mark the high-risk area in the 3D model.

[0050] The fractal dimension correction coefficient of the surface defect distribution entropy value can be set to 0.2 or 0.3, and the reference volume in the formula for calculating the effective stress concentration factor can be set to 800 cubic centimeters or 1200 cubic centimeters. The computing server can choose an HP Z8G4 workstation, and the optimization algorithm can be implemented based on the MATLAB global optimization toolbox. The data processing terminal can be deployed in the detection site control room and connected to the scanning device through optical fiber. After importing the 3D defect model, automatically calculate the entropy value and stress concentration factor of each area to quantify the impact degree of defects on the overall stability of the structure.

[0051] The decay coefficient of the weight matrix for spatial autocorrelation analysis can be set to 0.3 meters or 0.5 meters, and the correction amplitude of the coefficient of variation can be set to 0.4 or 0.6. The geographic information system software can choose ArcGIS Pro, and the statistical analysis module can be developed based on the Python Scipy library. The spatial analysis unit can run on a cloud server and interact with the local system through an API interface. The working process is as follows: perform spatial interpolation on the hardness test point data, calculate the Moran's I index, and correct the coefficient of variation. The technical effect is to identify abnormal aggregation areas of material properties.

[0052] The weight distribution ratio of the safety factor synthesis formula can be set to 5:3:2 or 6:2:2. When set to 5:3:2, the reverse weight formula for calculating the safety factor S is as follows: , where Ht is the comprehensive entropy index, α is the defect volume weight factor, is the effective stress concentration factor, is the coefficient of variation, and the decision threshold can be set to 1.1 or 1.3. The visualization software can be ParaView or CloudCompare, and the report generation module can be based on the JasperReports tool. The display terminal can be configured with a touch screen operation interface and installed on the monitoring console of the dock scheduling center. The comprehensive evaluation results automatically generate a color-coded 3D model and a text report, providing hierarchical early warning and maintenance priority suggestions.

[0053] In another technical solution, in step S4, an Internet of Things module is set to optimize the evaluation process. The operation process of the Internet of Things module includes the following steps: Perform wavelet packet decomposition and dynamic quantization compression on the detection data stream. Use the Daubechies 9 / 7 wavelet basis function to decompose the data into multiple sub-bands, and dynamically adjust the quantization step according to the signal-to-noise ratio of each sub-band; in the normal state, use a compression ratio of 10:1, and switch to a 5:1 high-fidelity compression mode when detecting high-risk defects, giving priority to retaining high-frequency detail features; Construct a multi-physical field compensation matrix for temperature, humidity, and pressure to quantify the influence of environmental factors on the measured values. Set the sensitivity of the strain measurement value to temperature to be 0.3 microstrain / degree Celsius, and the humidity sensitivity of the stress measurement value to be 0.05 MPa / percent relative humidity; collect environmental data in real time and substitute it into the compensation matrix to correct the strain and stress detection results; Construct a time series prediction model based on a long short-term memory network. The input feature vector includes the surface defect entropy value, the internal defect volume ratio, the material coefficient of variation, and environmental parameters; the dimension of the network hidden layer is set to 64, the time step covers 24 hours of continuous data, and the output is the risk probability value for the next 30 days; the risk level is divided into three levels: a probability value lower than 0.3 is low risk, 0.3 - 0.6 is medium risk, and higher than 0.6 is high risk; identify key parameters through the Fisher information matrix, set an elastic coefficient of 0.8 to limit the parameter update amplitude, and retain the historical data feature distribution.

[0054] The data acquisition frequency of the Internet of Things module can be set to 1 time / minute or 1 time / 5 minutes. The environmental temperature compensation range can be set to -10 to 50 degrees Celsius, and the humidity compensation range can be set to 20% to 95%RH. The edge computing device can be the Huawei Atlas 500 intelligent small station, and the wireless transmission module can be the Quectel EC25 series 4G module. The edge node can be installed in the dock distribution box and connected to the detection device through the RS485 bus. Package the surface defect coordinates, material hardness data, and environmental parameters in JSON format, encrypt them with AES, and upload them to the Alibaba Cloud IoT platform. Through edge-cloud collaborative processing, remote monitoring and long-term storage of detection data are realized, the local storage pressure is reduced, and the real-time early warning ability for abnormal states is improved.

[0055] The time step of the LSTM network can be set to 24 hours or 48 hours, the number of neurons in the hidden layer can be set to 64 or 128, and the risk level division threshold can be set to 0.3 or 0.5. The training server can be selected as Dell PowerEdge R750xa, equipped with NVIDIA A100 GPU acceleration card. The prediction model can be deployed on the cloud Kubernetes cluster, providing query services through REST API. Calculate the risk probability for the next 30 days every 6 hours, and the results are overlaid on the 3D model in the form of a heat map. Through time series data analysis, the prediction of the structural performance degradation trend is realized, the incidence of sudden failures is reduced, and the operability of the preventive maintenance plan is improved.

[0056] The elastic coefficient of the online learning algorithm can be set to 0.5 or 1.0, and the historical data retention period can be set to 1 year or 2 years. The incremental learning framework can be selected as TensorFlow Federated, and the storage server can be selected as Inspur AS5500G5. The data warehouse can be deployed in the OSS bucket of the cloud service provider, and is interconnected with the training cluster through a dedicated line. Automatically screen samples with a confidence level higher than 90% in the newly added data every week, update the model weights and retain the protection of key parameters. Through the continuous learning mechanism, the adaptive optimization of the model to the new environmental conditions is realized, and the compatibility of the detection schemes for different docks is improved.

[0057] The Internet of Things module includes a visualization terminal interface unit, and the three-dimensional visualization detection report adopts an augmented reality interface, including the following method steps: Construct an environmental point cloud map based on the simultaneous localization and mapping technology, realize the six-degree-of-freedom space alignment of the detection data coordinate system and the real scene, and control the real-time update error of the pose matrix to be less than 0.3 mm; the surface topography layer is superimposed and displayed at a resolution of 0.5 mm / pixel, and the defect distribution layer is colored with a red-orange gradient, and the coloring intensity increases linearly with the crack depth. When the depth exceeds 3 mm, the transparency drops to 0.2; Identify gesture actions based on a three-dimensional convolutional neural network, input a sequence of 16 consecutive depth images, perform gesture space trajectory fitting, and output operation instructions; perform voxel cutting on any specified plane to generate a material performance profile, extract the distribution of the profile mechanical parameters through the Gaussian kernel interpolation algorithm to generate a profile stress distribution map, and the interpolation bandwidth is fixed at 5 mm to suppress noise interference; Calculate the repair volume of the defect area, use the Heaviside function to determine the cavity boundary, and set the integration step size to 1 mm; aiming at maximizing stiffness and minimizing material usage, the volume of the reinforcement material does not exceed 30% of the original structure, set the optimization density threshold to 0.3, and perform topology optimization to design the reinforcement structure; the visualization terminal interface unit displays the von Mises stress nephogram of the reinforced structure in real time; use gesture interaction to modify the reinforcement scheme, dynamically update the mechanical simulation results, and the optimization iteration period is less than 2 seconds.

[0058] The hierarchical transparency of the augmented reality display can be set to 0.5 or 0.7, the profile interpolation bandwidth can be set to 3 mm or 8 mm, and the gesture recognition delay can be controlled within 50 ms or 100 ms. The AR glasses can be selected as Microsoft HoloLens2, and the gesture recognition sensor can be selected as the Intel RealSense D455 depth camera. The display terminal can be installed at the front end of the safety helmet and communicate with the data processing terminal via Wi-Fi 6. Align the three-dimensional defect model with the real scene video stream in space, and the operator can switch the display layer through the pinch gesture. Through virtual-real fusion visualization, the three-dimensional presentation of hidden defects is realized, improving the on-site decision-making efficiency.

[0059] The volume calculation accuracy of the intelligent repair scheme can be set to 1% or 3%, and the number of optimization iterations of the reinforcement structure can be set to 100 or 200. The CAD software can be selected as Autodesk Revit, and the finite element analysis engine can be selected as ANSYS Mechanical. The design module can run on a Dell Precision 7760 mobile workstation and interact with the AR system through the STEP format. After selecting the defect area, a three-dimensional mesh model of the repair body is automatically generated, and the variable density method topology optimization is performed and the BOM list is exported. Through the automated design process, the rapid generation of the repair plan is realized, reducing the manual modeling time and improving the mechanical rationality of the reinforcement structure.

[0060] The refresh rate of the stress nephogram in the virtual-real simulation can be set to 25 frames per second or 60 frames per second, and the interaction response delay can be controlled within 0.5 seconds or 1 second. The mechanical solver can be selected as COMSOL Multiphysics, and the rendering engine can be selected as Unity3D. The simulation server can be deployed in the local computer room and connected to a 4K display through an HDMI 2.1 interface. The operator adjusts the parameters of the reinforcement scheme through gestures, updates the stress distribution in real time and displays the maximum deformation. Through interactive simulation verification, the visualization optimization of the design scheme is realized, improving the economy of the repair plan.

[0061] It should be noted that although the above steps are described in a specific order, it does not mean that the steps must be executed in the above specific order. In fact, some of these steps can be executed concurrently or even in a different order, as long as the required functions can be achieved. The number of devices and the processing scale described here are used to simplify the description of the present invention, and the application, modification, and variation of the present invention are obvious to those skilled in the art.

[0062] Although the embodiments of the present invention have been disclosed above, it is not limited to the applications listed in the specification and embodiments. It can be fully applied to various fields suitable for the present invention. For those familiar with the art, additional modifications can be easily achieved. Therefore, without departing from the general concept defined by the claims and the equivalent scope, the present invention is not limited to the specific details and the illustrated examples herein.

Claims

1. Method for quality inspection of quay apron block, characterized in that, It includes the following steps: S1: Use a mobile three-dimensional laser scanning device to scan the armor blocks to obtain laser point cloud data, construct a three-dimensional surface model through a real-time point cloud registration algorithm, and extract surface geometric features based on a deep convolutional neural network, including surface crack depth and spalling area. Mark the areas where the surface crack depth exceeds the preset depth threshold or the spalling area is greater than the preset area threshold as high-risk areas, and generate a high-risk area heat map; S2: Based on the high-risk area heat map, use an adjustable frequency dual-mode flaw detection device to detect cavities and cracks inside the armor blocks: perform encrypted scanning with ultrasonic probes in high-risk areas; arrange multiple impact points along the stress diffusion path in the projection area of the surface crack extension direction, and use an array impact echo device to collect vibration signals at the impact points; Fuse the ultrasonic data and vibration signal data through a multi-sensor data fusion algorithm to establish a three-dimensional defect probability distribution map, and mark the areas where the internal cavity diameter or crack length is greater than the corresponding upper limit threshold as severely defective areas; S3: Drill multiple groups of core samples within a predetermined distance range around the severely defective areas and conduct compressive strength tests and acoustic emission damage detections; conduct infrared spectral composition surveys on non-severely defective areas to detect the concrete carbonation degree and chloride ion erosion index; In the defect-matrix interface transition zone, arrange micro-area hardness test points according to the material property gradient and conduct hardness tests; S4: Establish a dynamic evaluation model including the surface defect distribution entropy value, the internal defect volume ratio, and the material property dispersion coefficient; calculate the comprehensive safety factor S through a multi-objective optimization algorithm, and compare it with the upper safety threshold S u and the lower safety threshold S d to generate a three-dimensional visualization detection report.

2. The quality inspection method for the quay apron block according to claim 1, wherein, In step S1, the robotic arm of the mobile three-dimensional laser scanning device is integrated with a dual-mode lidar and an inertial measurement unit, and performs the following operations: Based on continuous three-point cloud slice data, use a differential geometry algorithm to calculate the local curvature change rate of the surface of the armor blocks in real time, and quantify the curvature change rate through the ratio of the Gaussian curvature difference between adjacent measurement points to the arc length spacing; when the local minimum curvature radius is less than 2 meters, the surface roughness index exceeds the standard-defined 0.5 mm threshold, and environmental vibrations in the range of 0.1 - 5 Hz are detected, activate the high-frame-rate lidar mode and increase the scanning frame rate to not less than 200 Hz; Use an adaptive Kalman filter to eliminate low-frequency vibration noise caused by waves, fuse the state equations of the robotic arm pitch angle, angular velocity, and linear acceleration with the observation data of the inertial measurement unit, and update the process noise and observation noise covariance matrices in real time to suppress wave interference; In the high-frame-rate lidar mode, dynamically adjust the moving speed of the robotic arm end according to the local minimum curvature radius. When the minimum curvature radius is less than 1 meter, use a low speed of 0.2 m / s for scanning. When the curvature radius is in the range of 1 - 2 meters, use a medium speed of 0.5 m / s for scanning. At the same time, increase the laser line density to 800 lines per meter and ensure that the point cloud overlap rate of adjacent scanning paths is not less than 40%.

3. The method for detecting the quality of the quay apron block according to claim 2, characterized in that, In step S1, the real-time point cloud registration algorithm specifically includes the following steps: Perform coarse registration: The improved iterative closest point algorithm is adopted, and the weighted sum of geometric distance and curvature difference is used as the optimization objective. The curvature difference is quantified by the mixed difference between the Gaussian curvature and the mean curvature of the corresponding points, and the Gaussian curvature weight factor is set to 0.

5. When screening the matching point pairs, only the point pairs with the principal curvature radius ratio greater than 0.7 and the cosine value of the normal vector included angle exceeding 0.9 are retained, and the false matches with inconsistent surface geometric features are excluded. Perform fine registration: Based on the field programmable gate array to accelerate the Gauss-Newton iterative algorithm, the Jacobian matrix including the rotation component and the translation component is calculated in parallel, and the optimal transformation parameters are solved by the damped least squares method. The damping factor is set to 0.0001 to prevent matrix singularity. Combining the three-dimensional feature descriptor matching, the local surface is divided into 32 azimuth intervals and the height distribution histogram is statistically calculated. The chi-square distance is used to calculate the feature similarity, and the regions with the matching score lower than the threshold are marked as low-confidence regions. Perform dynamic evaluation and replanning: Calculate the registration error of the local region in real time. When the root mean square error of the region exceeds 0.8 mm, it is determined as a registration failure region. Trigger the scanning path replanning, increase the deflection angle by ±15 degrees on the basis of the original scanning path, increase the point cloud line density to 1200 lines per meter, increase the overlap rate to 60%, and perform supplementary scanning operations on the low-confidence regions.

4. The method for detecting the quality of quay apron blocks according to claim 1, characterized in that, In step S2, the tunable frequency dual-mode flaw detection device includes an ultrasonic probe and an array impact echo device, and the ultrasonic probe adopts the following hierarchical waveguide structure and dynamic control strategy: The waveguide layer of the ultrasonic probe is composed of a piezoelectric composite substrate and five layers of gradient impedance matching layers. The impedance values of each layer gradually change according to the exponential law from the substrate to the couplant interface. The total thickness of the matching layer is 2 mm to suppress the acoustic energy scattering caused by surface roughness. When the arithmetic mean deviation of the surface roughness is detected to exceed 12.5 μm, the wideband frequency sweep detection mode is started, and continuous frequency sweeping is performed at a rate of 50 kHz per second in the emission frequency range of 1.5 MHz to 2.5 MHz, and the emission pulse width is dynamically adjusted according to the frequency sweep range. Calculate the effective focusing depth based on the longitudinal wave velocity of the material and the current emission frequency. When the focusing depth fluctuation exceeds ±10% of the nominal value, trigger the calibration mode: emit a 1 MHz reference pulse, and reverse correct the actual wave velocity of the material through the echo time difference to eliminate the sound velocity error caused by concrete aging. Establish a non-linear relationship model between the contact pressure and the surface roughness, and adjust the probe height in real time through the piezoelectric actuator to make the couplant pressure stable in the range of 0.2 to 0.5 MPa. Adopt a proportional-integral-differential controller to achieve dynamic balance, with the proportional coefficient set to 0.8, the integral time set to 0.1 s, the differential time set to 0.05 s, and the sampling frequency of the pressure sensor set to 1 kHz.

5. The method for detecting the quality of quay apron block according to claim 4, wherein In step S2, the array impact echo device performs internal defect location using the following method: The impact energy is controlled in five levels by adjusting the impact speed of the array impact echo device: 0.5 J, 1 J, 2 J, 3.5 J, 5 J; the impact frequency is dynamically selected according to the surface Vickers hardness: high-frequency impact at 10 Hz is used when the hardness is below 300 MPa, medium-frequency impact at 5 Hz is used when the hardness is in the range of 300 - 500 MPa, and low-frequency impact at 2 Hz is used when the hardness exceeds 500 MPa; The surface wave velocity is calculated based on the elastic modulus, density and Poisson's ratio of the material, the surface wave wavelength is calculated based on the central analysis frequency and the surface wave velocity, and piezoelectric sensors are arranged at intervals of one-eighth of the surface wave wavelength. A number of piezoelectric sensors are arranged in a spiral-grid hybrid layout to form a sensor array with three-dimensional space coverage; Perform short-time Fourier transform and wavelet packet decomposition fusion processing on the vibration signals collected by the echo device; among them, the short-time Fourier transform uses a Gaussian-modulated cosine window function, and the window width is dynamically adjusted according to the impact frequency; the wavelet packet decomposition selects the 10th-order Daubechies wavelet basis, and according to the sampling frequency f sample and the center frequency f central Set the decomposition level to ; The high-low frequency dual-channel analysis method is adopted to calculate the signal energy integral value for the low-frequency channel of 0.5 - 10 kHz and the kurtosis index for the high-frequency channel of 10 - 50 kHz; Based on the data of the sensor array, the three-dimensional wave field is reconstructed using the time reversal algorithm, and the observation time window length is set to 50 ms; when the signal energy integral value of the low-frequency channel exceeds 0.8 times the reference value of healthy concrete, the kurtosis index value of the high-frequency channel exceeds the threshold, and the diameter of the focusing area of the three-dimensional wave field is greater than 10 cm, the redundant area is determined as an effective internal defect.

6. The quality inspection method for the quay apron block according to claim 1, characterized in that In step S3, the core sample drilling and testing specifically include the following steps: Based on the fusion data of the laser displacement sensor and the inertial measurement unit, the offset angle of the drill bit axis is calculated in real time: the lateral offset is measured by the laser sensor, and the mechanical vibration error is compensated by combining the gyroscope angular velocity data, and the compensation coefficient is set to 0.05 s / rad; a fuzzy proportional-integral-derivative controller is used to adjust the stroke of four hydraulic cylinders, with the goal of minimizing the weighted sum of the offset angle and the axial force change rate, and the maximum deflection angle is limited to ±15 degrees, and an alarm is triggered when the limit is exceeded; The dimensionless working condition index is calculated in real time, which is defined as the product of the axial drilling force and the drill bit radius divided by the product of the feed rate and the hydraulic pressure. When this index exceeds the set threshold, three-level protection is implemented: the retraction speed is adjusted to 2 times the current feed rate plus 50 mm / s; the drill bit speed is reduced to 300 rpm; coolant is sprayed at a flow rate of 0.2 times the axial force; The following enhancement methods are used for the core sample compressive strength test: The loading rate is controlled in three stages: slow loading at 0.5 MPa / s is used when the estimated strength is below 30%, the speed is increased to 1.0 MPa / s in the range of 30% - 70%, and the speed is reduced to 0.2 MPa / s after exceeding 70%; The acoustic emission signals are collected synchronously, and the damage evolution b value is calculated by counting the number of events with amplitudes exceeding the range of 45 dB to 95 dB. When the change rate of the b value exceeds the set critical threshold, it is determined as a precursor of critical failure; A linear positive correlation model between the axial force-feed rate ratio and the compressive strength, and a negative correlation model between the time integral of the dimensionless working condition index and the strength are established, and the actual strength of the core sample is calculated comprehensively.

7. The quality inspection method for the quay apron block according to claim 1, characterized in that In step S3, the infrared spectral composition census of the non-serious defect area specifically includes the following steps: For the short-wave infrared channel of 2.5 - 5 microns, a quantum cascade laser with a central frequency of 120 terahertz is used as the light source to scan the characteristic peaks of silicate crystallization at a resolution of 4 cm⁻¹; for the long-wave infrared channel of 8 - 12 microns, the characteristic peaks of carbonate and chloride ions are scanned with a step size of 0.5 microns, and the single integration time is set to 50 milliseconds; based on the dynamic distribution of the signal-to-noise ratio of the two bands, the fusion weight is assigned, and the weight coefficient is set as the normalized result of the square value of the signal-to-noise ratio of each band. Calibrate the influence of the ambient temperature by referring to the blackbody radiation source at a specified temperature, and establish a wavelength drift model including the steady-state thermal expansion coefficient and the transient response coefficient; perform nonlinear correction on the original spectrum, and use the Planck blackbody radiation formula to eliminate the interference of ambient thermal radiation. An improved convolutional recurrent neural network is used for analysis. The input layer of the network receives a 256×256-dimensional dual-band spectral splicing matrix; the convolutional module uses a dilated convolutional kernel with a dilation rate of 2 to extract multi-scale features, and the recurrent module uses a bidirectional long short-term memory network to capture the spectral sequence dependence, and the hidden layer dimension is set to 128; when training the model, a weighted cross-entropy loss function is used, and the weights are assigned reversely according to the material aging level, and spectral perturbations of ±2% are added to generate adversarial samples to improve robustness. When calculating the aging index, fuse the output probability of the neural network and assign it a weight of 60%, calculate the carbonation index and assign it a weight of 30%, calculate the chloride ion erosion index and assign it a weight of 10%; calculate the final aging index by weighted summation. When the aging index exceeds 0.6, it is determined as severe aging, and a maintenance warning is generated.

8. The method for detecting the quality of the quay apron block as claimed in claim 1, wherein In step S4, the method for the dynamic evaluation model to conduct a comprehensive structural safety assessment is as follows: An improved weighted Shannon entropy model is used to calculate the disorder degree of surface defect distribution, and the coupling effect of crack depth and area is used as the weight factor; a fractal dimension correction term is introduced. When the fractal dimension exceeds 1.2, the total entropy value is increased by 30% of the fractal dimension deviation amount to enhance the sensitivity to complex crack networks. Calculate the effective stress concentration factor, set the reference volume to 1000 cubic centimeters, and the void volume is accumulated according to the equivalent sphere radius. When the crack length exceeds 30 centimeters, an additional correction term is added to the stress concentration factor, and the correction amplitude is proportional to the excess length value; when continuous surface cracks are detected, the internal defect weight factor is increased by 40%. Based on the spatial autocorrelation Moran's I index, evaluate the uniformity of material property distribution. When calculating the index, an exponential decay function is used to define the spatial weight matrix, and the decay coefficient is set to 0.5 meters; when the Moran's I index exceeds 0.3, it is determined as a clustered distribution of material properties, and the dispersion coefficient is amplified according to the exponential difference ratio, and the amplification factor is set to 0.5 at most. Perform dynamic synthesis on the safety factor: calculate the comprehensive safety factor by using the reverse weight formula with the surface entropy value, the corrected internal defect coefficient, and the dispersion coefficient in a weight ratio of 0.5:0.3:0.2; when the safety factor is lower than 1.2, a red warning is generated, and high-risk areas are marked in the 3D model.

9. The quality inspection method for the quay armor units according to claim 1, characterized in that, In step S4, an Internet of Things module is set to optimize the evaluation process. The operation process of the Internet of Things module includes the following steps: Perform wavelet packet decomposition and dynamic quantization compression on the detected data stream. Use the Daubechies 9 / 7 wavelet basis function to decompose the data into multiple sub-bands, and dynamically adjust the quantization step size according to the signal-to-noise ratio of each sub-band. When in the normal state, a compression ratio of 10:1 is adopted, and when a high-risk defect is detected, it switches to a 5:1 high-fidelity compression mode, giving priority to retaining high-frequency detail features; Construct a multi-physical field compensation matrix for temperature, humidity, and pressure to quantify the influence of environmental factors on the measured values. Set the sensitivity of the strain measurement value to temperature to be 0.3 microstrain / degree Celsius, and the humidity sensitivity of the stress measurement value to be 0.05 MPa / percent relative humidity. Real-time collect environmental data and substitute it into the compensation matrix to correct the strain and stress detection results; Construct a time series prediction model based on a long short-term memory network. The input feature vector includes the surface defect entropy value, the internal defect volume ratio, the material dispersion coefficient, and environmental parameters. The dimension of the network hidden layer is set to 64, and the time step covers 24 hours of continuous data. Output the risk probability value for the next 30 days. Divide the risk level into three levels: a probability value lower than 0.3 is low risk, 0.3 - 0.6 is medium risk, and more than 0.6 is high risk. Identify key parameters through the Fisher information matrix, set the elastic coefficient limit parameter update amplitude to 0.8, and retain the historical data feature distribution.

10. The method for detecting the quality of quay apron blocks according to claim 9, characterized in that, The Internet of Things module includes a visualization terminal interface unit. The three-dimensional visualization detection report uses an augmented reality interface, including the following method steps: Construct an environmental point cloud map based on simultaneous localization and mapping technology to achieve six-degree-of-freedom spatial alignment between the detection data coordinate system and the real scene, and control the real-time update error of the pose matrix to be less than 0.3 mm. The surface topography layer is superimposed and displayed at a resolution of 0.5 mm / pixel. The defect distribution layer is colored with a red - orange gradient, and the coloring intensity linearly increases with the crack depth. When the depth exceeds 3 mm, the transparency drops to 0.2; Identify gesture actions based on a three-dimensional convolutional neural network. Input a sequence of 16 consecutive depth images, perform gesture space trajectory fitting, and output operation instructions. Perform voxel cutting on any specified plane to generate a material property profile, and extract the profile mechanical parameter distribution through the Gaussian kernel interpolation algorithm to generate a profile stress distribution map. The interpolation bandwidth is fixed at 5 mm to suppress noise interference; Calculate the repair volume of the defect area, use the Heaviside function to determine the cavity boundary, and set the integration step size to 1 mm. With the goal of maximizing stiffness and minimizing material usage, the volume of the reinforcing material does not exceed 30% of the original structure, the optimization density threshold is set to 0.3, and perform topology optimization to design the reinforcing structure. The visualization terminal interface unit real-time displays the von Mises stress cloud map of the reinforced structure; Modify the reinforcement plan using gesture interaction, dynamically update the mechanical simulation results, and the optimization iteration period is less than 2 seconds.

Images