A negative pressure adsorption, radar scanning, AI identification and three-dimensional mapping four-in-one internal steel bar corrosion detection wall climbing robot and system

Abstract

The application discloses a four-in-one internal steel bar corrosion detection wall-climbing robot and system with negative pressure adsorption, radar scanning, AI identification and three-dimensional mapping, and the wall-climbing robot comprises a robot main body, an adsorption mechanism and a corrosion detection mechanism. The adsorption mechanism is arranged at the bottom center of a robot chassis and is used for being adsorbed on the surface of a vertical or inclined steel reinforced concrete structure. The corrosion detection mechanism is arranged at the tail of the robot chassis and is opposite to the surface of the steel reinforced concrete structure. The corrosion detection mechanism comprises a protective shell, a ground penetrating radar arranged in the protective shell and an auxiliary sensor. The ground penetrating radar scans and detects the steel reinforced concrete structure along with the movement of the wall-climbing robot. The complete process from automatic grid scanning, multi-modal data acquisition, signal preprocessing, feature extraction to deep learning classification and detection result structure mapping is realized through an on-board computing unit, the problems of high risk, low efficiency and non-standard detection in traditional manual detection are avoided, and the steel bar corrosion detection efficiency and operation safety are significantly improved.

Description

Technical Field

[0001] This invention relates to the field of non-destructive testing and health monitoring technology for civil engineering structures, specifically to a method combining negative pressure adsorption, radar scanning, AI recognition, and 3D modeling. Figure 4 An integrated wall-climbing robot and system for detecting internal steel reinforcement corrosion. Background Technology

[0002] Reinforced concrete structures are widely used in various engineering fields such as construction, bridges, and municipal works due to their high strength, durability, and moderate cost. During long-term service, the reinforcing steel is susceptible to corrosion due to factors such as environmental temperature and humidity, carbon dioxide erosion, chloride ion penetration, and concrete carbonation. Corrosion of the reinforcing steel leads to a reduction in cross-sectional area and a decline in mechanical properties. Furthermore, the volume expansion of corrosion products can cause cracking and spalling of the concrete cover, severely affecting the structure's load-bearing capacity and durability, and even causing safety accidents.

[0003] Existing methods for detecting rebar corrosion mainly fall into two categories: manual inspection and non-destructive testing (NDT). Manual inspection often employs localized destructive methods, such as drilling cores to directly observe the corrosion state or weighing to measure the degree of corrosion. While this method is highly accurate, it damages the structure, has a limited detection range, and is inefficient, making it unsuitable for large-area structural inspections. NDT methods primarily include half-cell potential testing, ultrasonic testing, and ground-penetrating radar (GPR). Half-cell potential testing can preliminarily determine the likelihood of corrosion but cannot quantitatively assess the degree of corrosion. Ultrasonic testing is sensitive to internal defects in concrete but is significantly affected by the arrangement of rebar and the quality of the concrete, resulting in limited accuracy. GPR, by emitting electromagnetic waves and receiving reflected signals, can acquire information on the location of rebar, the thickness of the protective layer, and internal structure. It offers advantages such as non-contact, high efficiency, and large-area detection. However, current GPR inspections largely rely on manual handheld operation, which is difficult and poses high safety risks for high-rise building facades, bridge piers, and other high-altitude or complex structures. Furthermore, manual operation can lead to non-standard inspection paths and unstable data acquisition accuracy, affecting the accuracy of corrosion identification. Summary of the Invention

[0004] This invention aims to provide a method for negative pressure adsorption, radar scanning, AI recognition, and 3D modeling. Figure 4 The integrated internal steel reinforcement corrosion detection wall-climbing robot and system integrates motion mechanism, adsorption mechanism, corrosion detection mechanism and control and communication mechanism on a single wall-climbing robot platform. Through the onboard computing unit, it realizes the complete process from automatic grid scanning, multimodal data acquisition, signal preprocessing, feature extraction to deep learning classification and detection result structure mapping. It avoids the high risk, low efficiency and non-standard detection problems of traditional manual inspection, and significantly improves the efficiency and safety of steel reinforcement corrosion detection.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0006] A wall-climbing robot for detecting internal corrosion of steel bars includes a robot body, an adsorption mechanism, and a corrosion detection mechanism;

[0007] The robot body includes a robot chassis and motion mechanisms located on both sides of the robot chassis. The motion mechanisms include drive motors and drive wheels connected to the drive motors.

[0008] The adsorption mechanism is located at the bottom center of the robot chassis and is used to adsorb onto the surface of vertical or inclined reinforced concrete structures.

[0009] The corrosion detection mechanism is located at the rear of the robot chassis, opposite the surface of the reinforced concrete structure; the corrosion detection mechanism includes a protective shell, a ground-penetrating radar and auxiliary sensors housed within the protective shell;

[0010] Ground-penetrating radar scans and detects reinforced concrete structures as the wall-climbing robot moves.

[0011] The protective shell has an internal electromagnetic shielding layer, and an antenna detection opening is provided at the bottom of the protective shell, which is matched with the antenna transmitting surface of the ground penetrating radar.

[0012] The ground-penetrating radar antenna emitter maintains a constant distance from the surface of the reinforced concrete structure.

[0013] Preferably, auxiliary sensors include temperature sensors and humidity sensors, used to collect ambient temperature and humidity data to provide a basis for signal correction.

[0014] Preferably, the inner wall of the protective housing is provided with a shock-absorbing buffer layer to protect the ground-penetrating radar antenna and auxiliary sensors from impacts and vibrations.

[0015] Preferably, it also includes a control and communication mechanism, which is electrically connected to the ground-penetrating radar, the auxiliary sensor and the drive motor respectively;

[0016] The control and communication mechanism includes an onboard computing unit, a drive unit, a power supply, and a wireless communication unit;

[0017] The airborne computing unit includes a signal preprocessing module, a feature extraction module, and a corrosion classification network module.

[0018] A method for detecting internal corrosion of reinforcing bars includes the following steps:

[0019] S1. Adsorption: Adsorb the wall-climbing robot onto the surface of the reinforced concrete structure, plan the climbing path and detection area, and control the wall-climbing robot to move at a constant speed along the preset path;

[0020] S2. Radar Scan: Perform B-Scan profile scanning on the reinforced concrete structure according to the set sampling frequency, measurement point spacing and scanning speed; preprocess the acquired raw radar signals to generate B-Scan grayscale or pseudo-color images.

[0021] S3. Feature Extraction: Extract the physical features of the reinforcing bars from the preprocessed signal; extract the image features of the reinforcing bars from the image using a convolutional neural network;

[0022] S4. Corrosion Classification and Result Judgment: A convolutional neural network is used to construct a corrosion classification network. The physical features and image features of each detection point are input into the corrosion classification network, and the classification result of the steel corrosion state of the detection point and its confidence level are output.

[0023] When the confidence level is lower than the preset threshold, the detection point is marked as "suspected corrosion", prompting manual review or controlling the wall-climbing robot to scan repeatedly;

[0024] S5. Result Recording and Map Mapping: Based on the pose, movement trajectory, and ground-penetrating radar antenna parameters of the wall-climbing robot, determine the spatial coordinates of each detection point, associate them with the classification results of the steel reinforcement corrosion status, and map them to the three-dimensional structural map and its corresponding building structure diagram to realize the marking, statistics, and location of the corrosion area.

[0025] As a preferred option, it also includes S6 and damage verification.

[0026] Representative points were selected in each preset area, and core samples were obtained by drilling. The thickness of the protective layer and the actual degree of corrosion of the steel bars were measured and compared with the radar detection results to correct the detection model parameters.

[0027] Preferably, the sampling frequency is no less than 6 times the center frequency of the bottom-penetrating radar antenna, and the scanning speed is matched with the radar triggering frequency;

[0028] The robot's crawling speed is matched with the radar triggering frequency; for each preset detection area, the wall-climbing robot is controlled to perform at least two reciprocating scans.

[0029] Preferably, the physical characteristics of the reinforcing bars include the position of the apex of the hyperbola reflecting the reinforcing bars, the amplitude intensity, and the waveform characteristics;

[0030] The corrosion classification network can be a deep residual network, an efficient convolutional neural network, or a lightweight convolutional neural network.

[0031] Preferably, step S3 also includes polarization feature extraction and correction of the dual-polarization radar signal data, specifically...

[0032] The azimuth angle of the reinforcing bar is solved, and the corrected fully polarized waveform of the reinforcing bar at the 0° azimuth angle is obtained by Alford rotational linear transformation. Multidimensional polarization features are extracted from the waveform, including polarization coherence matrix, polarization covariance matrix, Pauli decomposition coefficients and polarization similarity attributes. The multidimensional polarization features are fused with physical features and image features at the feature level and then input into the corrosion classification network.

[0033] Preferably, S4 also includes a domain adaptation module to address the differences in feature distribution between different concrete structures, including a domain discriminator, a fast adaptation mechanism, or an amplitude threshold adaptation mechanism.

[0034] The present invention has the following beneficial effects:

[0035] 1. Integrate motion mechanism, adsorption mechanism, corrosion detection mechanism and control and communication mechanism on a single wall-climbing robot platform. Through the onboard computing unit, realize the complete process from automatic grid scanning, multimodal data acquisition, signal preprocessing, feature extraction to deep learning classification and detection result structure mapping. Avoid the high risk, low efficiency and non-standard detection problems of traditional manual inspection, and significantly improve the efficiency of steel bar corrosion detection and operation safety.

[0036] 2. The corrosion detection mechanism uses a rigid fixed bracket to lock the ground penetrating radar antenna and auxiliary sensors inside the robot. The antenna transmitting surface maintains a constant distance from the concrete surface and moves at a constant speed with the robot body to form a continuous scanning line. By precisely controlling the walking speed of the motion module and the radar trigger frequency, high-density grid-like signal acquisition is achieved. This improves long-term operational reliability and is suitable for rapid inspection of long-distance, large-span structures, balancing detection efficiency and system durability.

[0037] 3. By employing a high-frequency shielded ground-penetrating radar antenna, coupled with temperature and humidity auxiliary sensors, and combining a multi-step signal preprocessing process, environmental noise and interference are effectively suppressed, the signal-to-noise ratio is improved, and a high-quality data foundation is provided for subsequent feature extraction and corrosion classification.

[0038] 4. Deploy a feature extraction module and a corrosion classification network on the airborne computing unit. By extracting physical and image features and combining them with deep learning algorithms, the system can accurately classify the corrosion level of steel bars. Compared with traditional signal processing and manual judgment methods, this method can better explore the subtle differences in electromagnetic wave reflection characteristics of different corrosion levels, thereby improving the accuracy and robustness of corrosion identification.

[0039] 5. Optional damage verification and model calibration steps can continuously optimize the detection model parameters through actual measurement data, thereby continuously improving the accuracy and reliability of the detection system and meeting the detection needs of different engineering scenarios. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall structure of the wall-climbing robot of the present invention.

[0041] Figure 2 This is a schematic diagram of the internal structure of the motion mechanism and the adsorption mechanism of the present invention.

[0042] Figure 3 This is a schematic diagram of the bottom view structure of the present invention.

[0043] Figure 4 This is a schematic diagram of the corrosion detection mechanism of the present invention.

[0044] Figure 5 This is a flowchart of the detection process of the present invention.

[0045] The components include: 1. Robot body; 2. Adsorption mechanism; 3. Corrosion detection mechanism.

[0046] 11. Robot chassis; 12. Motion mechanism; 121. Drive motor; 122. Drive wheel;

[0047] 21. Ducted fan; 22. Negative pressure sealing chamber;

[0048] 31. Protective outer shell; 32. Ground penetrating radar; 33. Shock-absorbing buffer layer; 34. Antenna detection opening. Detailed Implementation

[0049] The present invention will now be described in further detail with reference to the accompanying drawings and specific preferred embodiments.

[0050] In the description of this invention, it should be understood that the terms "left side," "right side," "upper part," "lower part," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. "First," "second," etc., do not indicate the importance of the components, and therefore should not be construed as a limitation of this invention. The specific dimensions used in this embodiment are only for illustrating the technical solution and do not limit the scope of protection of this invention.

[0051] like Figures 1-5 As shown, a method combining negative pressure adsorption, radar scanning, AI recognition, and 3D modeling... Figure 4 The integrated internal steel reinforcement corrosion detection wall-climbing robot includes a robot body 1, an adsorption mechanism 2, a corrosion detection mechanism 3, and a control and communication mechanism.

[0052] The robot body 1 includes a robot chassis 11 and motion mechanisms 12 disposed on both sides of the robot chassis 11. The motion mechanism 12 includes a drive motor 121 and drive wheels 122 connected to the drive motor 121, which are used to drive the wall-climbing robot to crawl at a constant speed along a preset path on the surface of a reinforced concrete structure. The drive motor 121 is a DC geared motor, and the outer surface of the drive wheel 122 is covered with a high-friction rubber layer to enhance the friction with the structural surface and ensure crawling stability.

[0053] The adsorption mechanism 2 is located at the bottom center of the robot chassis 11 and includes a ducted fan 21 and a negative pressure sealing chamber 22. When the ducted fan 21 is working, it draws air from inside the negative pressure sealing chamber 22, making the air pressure inside the chamber lower than the external atmospheric pressure. This generates a stable negative pressure adsorption force between the robot chassis 11 and the reinforced concrete structure surface, allowing the wall-climbing robot to reliably adhere to vertical or inclined concrete walls. By adjusting the rotation speed of the ducted fan 21, the adsorption force can be flexibly adjusted to adapt to concrete structure surfaces with different inclination angles. The bottom of the negative pressure sealing chamber 22 is also equipped with a flexible sealing structure made of elastic materials such as silicone, rubber, or polyurethane to compensate for unevenness on the reinforced concrete surface and improve the negative pressure sealing effect.

[0054] The corrosion detection mechanism 3 is located at the rear of the robot chassis 11, opposite the surface of the reinforced concrete structure. The corrosion detection mechanism 3 includes a high-frequency shielded ground-penetrating radar 32 and auxiliary sensors fixed to the robot chassis 11, and a protective shell 31 covering the ground-penetrating radar 32 and auxiliary sensors. The high-frequency shielded ground-penetrating radar 32 continuously receives scanning information as the wall-climbing robot moves as a whole, performing high-density grid scanning detection on the area within its coverage range. The corrosion detection mechanism 3 uses a rigid fixing bracket to lock the ground-penetrating radar 32 antenna and auxiliary sensors inside the robot, maintaining a constant distance between the antenna's transmitting surface and the concrete surface.

[0055] The high-frequency shielded ground-penetrating radar 32 antenna has a center frequency of no less than 1.6 GHz and is fixed inside the robot chassis 11. The antenna's transmitting surface faces the concrete structure surface. As the robot moves, it continuously emits electromagnetic waves and receives reflected signals to perform real-time data acquisition and complete a high-density grid scan of a detection zone, obtaining information on the internal structure of reinforced concrete (reinforcing bar location, protective layer thickness, and corrosion status). Auxiliary sensors, including temperature and humidity sensors, are fixed near the ground-penetrating radar 32 antenna to collect real-time environmental temperature and humidity data of the detection area, providing a basis for radar signal correction and improving detection accuracy.

[0056] The protective housing 31 is made of hard and wear-resistant material, and a shock-absorbing buffer layer 33 is provided on the inside to protect the ground penetrating radar 32 antenna and auxiliary sensors from impact and vibration. The protective housing 31 has an antenna detection opening 34 on the side facing the structural surface. The antenna detection opening 34 matches the antenna transmitting surface of the ground penetrating radar 32 to ensure normal transmission of radar signals.

[0057] This embodiment is equipped with a 1.8GHz high-frequency shielded radar antenna with a detection depth of 0-100cm, suitable for detecting the thickness of the concrete structure's reinforcing steel protective layer. The protective shell 31 is integrally injection molded from ABS engineering plastic, with dimensions of 200mm×120mm×80mm, and has an IP65 dustproof and waterproof rating. It also has an internal electromagnetic shielding layer to shield against external electromagnetic interference. The shock-absorbing buffer layer 33 is made of silicone foam material with a thickness of 10mm and is laid on the entire inner side of the protective shell 31, tightly fitting the high-frequency shielded ground-penetrating radar 32 antenna. The antenna detection opening 34 is located at the bottom of the protective shell 31, with dimensions of 150mm×60mm, matching the transmitting surface of the high-frequency shielded ground-penetrating radar 32 antenna to ensure unobstructed transmission and reception of radar electromagnetic waves.

[0058] The control and communication mechanism includes an onboard computing unit, a drive unit, a power supply, and a wireless communication unit. The control and communication mechanism is electrically connected to the ground-penetrating radar 32, auxiliary sensors, drive motor 121, and ducted fan 21, respectively. The onboard computing unit, drive unit, and power supply work together to achieve motion control and actuator drive for the wall-climbing robot. A communication link is established with the ground workstation via the wireless communication unit to transmit detection results, operating status, and control commands between the robot and the ground workstation.

[0059] The airborne computing unit internally deploys a signal preprocessing module, a feature extraction module, and a corrosion classification network module. The signal preprocessing module performs DC offset removal, time zero-point correction, gain adjustment, bandpass filtering, and background removal on the raw signals acquired by the ground-penetrating radar 32, generating B-Scan grayscale or pseudo-color images to eliminate signal interference introduced during robot movement. The feature extraction module extracts physical and image features from the preprocessed signals and B-Scan images. Physical features include the vertex position of the hyperbola reflecting the rebar, amplitude intensity, and waveform characteristics. Image features are automatically extracted using a convolutional neural network. The corrosion classification network module is preferably a deep residual network (ResNet), an efficient convolutional neural network (EfficientNet), a lightweight convolutional neural network (MobileNet), or an equivalent convolutional neural network classifier. It automatically analyzes and processes the extracted feature data to classify and identify the rebar as having no corrosion, light corrosion, moderate corrosion, or severe corrosion, and outputs the corresponding confidence scores.

[0060] The onboard computing unit is also connected to a depth camera and an inertial measurement unit (IMU). The depth camera is used to acquire images and depth information of the structural surface in front of or to the side of the robot, while the IMU is used to acquire the robot's acceleration and angular velocity. The onboard computing unit runs a visual-inertial SLAM algorithm to fuse the depth camera and IMU data, estimate the six-degree-of-freedom pose of the wall-climbing robot in the structural coordinate system in real time, and gradually build a three-dimensional dense or semi-dense map of the structural surface. The spatial coordinates of each corrosion detection point are recorded in the coordinate system of this map, providing a unified spatial reference for the structural mapping of the detection results and subsequent maintenance positioning.

[0061] The airborne computing unit connects to the ground workstation via wireless communication, associates the spatial location of each detection point with the corresponding corrosion identification results, generates a steel reinforcement corrosion distribution map based on the mapping coordinate system, and displays it on the ground workstation interface with different colors or symbols, clearly showing the spatial distribution and severity of the corrosion area.

[0062] A negative pressure adsorption, radar scanning, AI recognition and 3D modeling Figure 4 The integrated internal steel reinforcement corrosion detection method includes the following steps.

[0063] Step 1: Wall-climbing robot attachment, SLAM mapping and detection path planning

[0064] The ducted fan 21 in the adsorption mechanism 2 is activated, allowing the wall-climbing robot to attach to the surface of the reinforced concrete structure. The depth camera and IMU are turned on, and the onboard computing unit runs the visual-inertial SLAM algorithm. During the initial movement of the robot along the structure surface, the map is initialized, and the 3D map of the structure surface and the robot pose are updated in real time. The robot's crawling path and detection grid are planned on the map or building structure diagram, and the motion module is controlled to drive the robot to crawl at a constant speed along the preset path or stop at a fixed point.

[0065] Step 2: Follow-up scanning and multimodal data acquisition

[0066] As the wall-climbing robot moves at a constant speed along the preset path, the high-frequency shielded ground-penetrating radar 32 antenna continuously performs B-Scan profile scanning along with the robot. Auxiliary sensors synchronously collect environmental temperature and humidity data. The onboard computing unit records radar signals, encoder position information, and timestamps according to the set sampling frequency, measurement point spacing, and scanning speed. The sampling frequency is no less than 6 times the center frequency of the radar antenna to ensure high-density data acquisition.

[0067] The robot's crawling speed is matched with the radar triggering frequency to ensure the uniformity and integrity of the detection grid; for each preset detection area, the robot is controlled to perform at least two reciprocating scans to improve the reliability of data acquisition.

[0068] Step 3: Signal Preprocessing and Feature Extraction

[0069] The airborne computing unit performs preprocessing operations such as DC offset removal, time zero-point correction, gain adjustment, bandpass filtering, and background removal on the acquired raw radar signals to generate B-Scan grayscale or pseudo-color images. Then, physical features and image features are extracted from the preprocessed signals and B-Scan images. The physical features include the vertex position of the hyperbola reflecting the steel bars, amplitude intensity, waveform features, etc. The image features are automatically extracted through a convolutional neural network.

[0070] After completing routine preprocessing, the airborne computing unit performs polarization feature extraction and correction on the dual-polarization radar data: Reinforcement azimuth angle calculation: Based on the phase relationship between the echoes from the common-polarization channel and the cross-polarization channel, an angle compensation criterion is introduced to solve for the angle α between the reinforcement axis and the radar polarization direction. Alford rotation correction: Using the solved reinforcement azimuth angle α, an Alford rotation linear transformation is performed on the scattering matrix to eliminate the influence of the reinforcement's relative radar azimuth on the polarization features, obtaining the corrected fully polarized waveform of the reinforcement at a 0° azimuth angle. Polarization attribute extraction: Polarization decomposition is performed on the corrected fully polarized waveform to extract the following multi-dimensional polarization features: polarization coherence matrix T3, polarization covariance matrix C3, Pauli decomposition coefficients, and polarization similarity attributes. Polarization feature fusion: The above polarization attributes are fused with the original physical features and image features at the feature level to form a feature vector with richer dimensions and clearer physical meaning, which is then input into the subsequent classification network.

[0071] Step 4: Corrosion Classification and Result Determination

[0072] A convolutional neural network is used to construct a corrosion classification network. The airborne computing unit inputs the extracted physical features and image features into the corrosion classification network to obtain the classification results of the steel reinforcement corrosion status (such as "no corrosion", "light corrosion", "moderate corrosion", "severe corrosion" or "suspected corrosion") and their confidence scores. When the confidence score is lower than the preset threshold, the detection point is marked as "suspected corrosion", prompting manual review or controlling the robot to repeat the scan in the area.

[0073] A domain adaptive module is introduced to address the issue of feature distribution differences among different concrete structures (different mix proportions, different rebar diameters, and different cover thicknesses):

[0074] Domain discriminator: An adversarial training mechanism is introduced during network training to force the feature extractor to learn domain-invariant features, so that the feature distributions of the source domain (scenes with existing labeled data) and the target domain (new detection scene) are as close as possible.

[0075] Rapid adaptive mechanism: Upon arrival at a new testing site, operators can select 1-2 known locations (such as healthy areas) for rapid scanning. The model then fine-tunes a small number of parameters to achieve rapid adaptation with a small number of samples.

[0076] Adaptive amplitude thresholding: This statistically optimized amplitude thresholding method dynamically adjusts the classification threshold based on the statistical characteristics (mean, variance) of the radar signal in the current detection scene, rather than using a fixed threshold. Research shows that this method can accurately pinpoint potential corrosion areas.

[0077] In this embodiment, the ground-penetrating radar 32 uses a 1.8GHz pulse source to drive a shielded butterfly antenna via a T / R switch, emitting nanosecond-level electromagnetic waves close to the concrete surface. Upon encountering reinforcing steel or a rust layer, the reflected electromagnetic wave echo is instantly captured by the same antenna. The original radar echo signal is amplified by low noise and then input to a 14-bit, 1GS / s sampling rate analog-to-digital converter for digital sampling. The digital signal after analog-to-digital conversion is transmitted to a field-programmable gate array (FPGA), where zero-time correction, exponential gain compensation, 0.5-4GHz bandpass filtering, and background removal are sequentially performed. The processed multi-channel echo signal is reconstructed and mapped to grayscale values ​​to generate a B-Scan profile grayscale image. Subsequently, Jetson Orin... Nano performs non-destructive testing on reinforcing steel in the test area by fusing physical features (hyperbola vertices, peak amplitude, half-peak area) with image features using a detection algorithm. It outputs the corrosion level and confidence score in real time. If the confidence score is below 0.6, a retest is automatically triggered; if the confidence score is ≥0.6, the results are stored in the database and a new 3D distribution map is generated. The entire "transmit-switch-receive-process-display" chain is integrated into a 1.2 kg, 120 mm × 80 mm × 45 mm aluminum shell. The robot, moving at a constant speed of 0.15 m / s, refreshes one corrosion cloud map per second, achieving efficient, non-destructive, and quantitative identification of internal corrosion in concrete reinforcing steel. After scanning, the robot returns to the starting point, the negative pressure is turned off, a test report is generated, and the raw data is automatically uploaded to the system.

[0078] Step 5: Recording Results and Mapping

[0079] The airborne computing unit combines the robot's pose, movement trajectory, and ground-penetrating radar 32 antenna parameters in the SLAM map to determine the spatial coordinates of each detection point. It then associates these coordinates with the corresponding corrosion classification results to form a steel reinforcement corrosion detection dataset. The detection results are mapped onto a 3D structural map and its corresponding building structure diagram. The ground workstation draws each detection point on the 3D map of the structural surface and its corresponding building structure diagram. Different corrosion levels are marked with different highlight colors or symbols, and suspected corrosion areas are represented by special colors or dashed boundaries. This enables the visualization and statistical analysis of the steel reinforcement corrosion detection results.

[0080] Step 6: Damage Verification and Model Calibration

[0081] Representative points are selected in each preset area, and core samples are obtained by drilling. The thickness of the protective layer and the actual degree of corrosion of the steel bars are measured. The real data obtained from the damage verification are compared with the radar detection results of the corresponding points one by one. The measured error of each indicator is calculated, the detection model parameters are corrected, and the detection accuracy is further improved.

[0082] This application describes a wall-climbing robot with good adsorption stability, capable of adapting to uneven reinforced concrete surfaces, maintaining the continuity of the inspection process, providing a wide inspection coverage and high efficiency. It can complete automatic scanning inspection, multimodal data acquisition, deep learning processing and intelligent recognition on the same wall-climbing platform, and achieve precise correlation between inspection results and structural spatial position through visual inertial SLAM mapping, which is beneficial for subsequent maintenance positioning, thereby improving inspection efficiency and reliability and reducing the risks of high-altitude operations.

[0083] This application also includes a negative pressure adsorption, radar scanning, AI recognition, and 3D modeling. Figure 4 An integrated internal steel reinforcement corrosion detection system, which employs the aforementioned negative pressure adsorption, radar scanning, AI recognition, and 3D modeling... Figure 4 The integrated internal steel reinforcement corrosion detection method includes a real-time monitoring module, a control and interaction module, a waveform presentation module, and an identification and feedback module.

[0084] The preferred embodiments of the present invention have been described in detail above. However, the present invention is not limited to the specific details in the above embodiments. Within the scope of the technical concept of the present invention, various equivalent transformations can be made to the technical solutions of the present invention, and these equivalent transformations all fall within the protection scope of the present invention.

Claims

1. A wall-climbing robot for detecting internal steel reinforcement corrosion, integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping, characterized in that... This includes the robot body, the adsorption mechanism, and the corrosion detection mechanism; The robot body includes a robot chassis and motion mechanisms located on both sides of the robot chassis. The motion mechanisms include drive motors and drive wheels connected to the drive motors. The adsorption mechanism is located at the bottom center of the robot chassis and is used to adsorb onto the surface of vertical or inclined reinforced concrete structures. The corrosion detection mechanism is located at the rear of the robot chassis, opposite the surface of the reinforced concrete structure; the corrosion detection mechanism includes a protective shell, a ground-penetrating radar and auxiliary sensors housed within the protective shell; Ground-penetrating radar scans and detects reinforced concrete structures as the wall-climbing robot moves. The protective shell has an internal electromagnetic shielding layer, and an antenna detection opening is provided at the bottom of the protective shell, which is matched with the antenna transmitting surface of the ground penetrating radar. The ground-penetrating radar antenna emitter maintains a constant distance from the surface of the reinforced concrete structure.

2. The wall-climbing robot for detecting internal steel reinforcement corrosion, integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping as described in claim 1, is characterized in that... Auxiliary sensors include temperature and humidity sensors, which are used to collect ambient temperature and humidity data to provide a basis for signal correction.

3. The wall-climbing robot for detecting internal steel reinforcement corrosion, integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping as described in claim 1, is characterized in that... The inner wall of the protective shell is equipped with a shock-absorbing buffer layer to protect the ground-penetrating radar antenna and auxiliary sensors from impacts and vibrations.

4. The wall-climbing robot for detecting internal steel reinforcement corrosion, integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping as described in claim 1, is characterized in that... It also includes a control and communication mechanism, which is electrically connected to the ground-penetrating radar, auxiliary sensors, and drive motor, respectively. The control and communication mechanism includes an onboard computing unit, a drive unit, a power supply, and a wireless communication unit; The airborne computing unit includes a signal preprocessing module, a feature extraction module, and a corrosion classification network module.

5. A method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping, using a wall-climbing robot for detecting internal steel reinforcement corrosion as described in any one of claims 1-4, characterized in that... Includes the following steps, S1. Adsorption: Adsorb the wall-climbing robot onto the surface of the reinforced concrete structure, plan the climbing path and detection area, and control the wall-climbing robot to move at a constant speed along the preset path; S2. Radar Scan: Perform B-Scan profile scanning on the reinforced concrete structure according to the set sampling frequency, measurement point spacing and scanning speed; preprocess the acquired raw radar signals to generate B-Scan grayscale or pseudo-color images. S3. Feature Extraction: Extract the physical features of the reinforcing bars from the preprocessed signal; Extracting image features of reinforcing bars from images using a convolutional neural network; S4. Corrosion Classification and Result Judgment: A convolutional neural network is used to construct a corrosion classification network. The physical features and image features of each detection point are input into the corrosion classification network, and the classification result of the steel corrosion state of the detection point and its confidence level are output. When the confidence level is lower than the preset threshold, the detection point is marked as "suspected corrosion", prompting manual review or controlling the wall-climbing robot to repeat the scan. S5. Result Recording and Map Mapping: Based on the pose, movement trajectory, and ground-penetrating radar antenna parameters of the wall-climbing robot, determine the spatial coordinates of each detection point, associate them with the classification results of the steel reinforcement corrosion status, and map them to the three-dimensional structural map and its corresponding building structure diagram to realize the marking, statistics, and location of the corrosion area.

6. The method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and three-dimensional mapping as described in claim 5, is characterized in that... It also includes S6 and damage verification. Representative points were selected in each preset area, and core samples were obtained by drilling. The thickness of the protective layer and the actual degree of corrosion of the steel bars were measured and compared with the radar detection results to correct the detection model parameters.

7. The method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and 3D mapping as described in claim 5, is characterized in that... The sampling frequency is no less than 6 times the center frequency of the bottom-penetrating radar antenna, and the scanning speed is matched with the radar trigger frequency; The robot's crawling speed is matched with the radar triggering frequency; for each preset detection area, the wall-climbing robot is controlled to perform at least two reciprocating scans.

8. The method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and three-dimensional mapping as described in claim 5, is characterized in that... The physical characteristics of reinforcing bars include the position of the apex of the hyperbola reflecting the bars, amplitude intensity, and waveform characteristics; The corrosion classification network can be a deep residual network, an efficient convolutional neural network, or a lightweight convolutional neural network.

9. The method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and three-dimensional mapping as described in claim 5, is characterized in that... S3 also includes polarization feature extraction and correction of dual-polarization radar signal data, specifically... The azimuth angle of the reinforcing bar is solved, and the corrected fully polarized waveform of the reinforcing bar at the 0° azimuth angle is obtained by Alford rotational linear transformation. Multidimensional polarization features are extracted from it, including polarization coherence matrix, polarization covariance matrix, Pauli decomposition coefficients and polarization similarity properties. Multidimensional polarization features are fused with physical features and image features at the feature level and then input into the corrosion classification network.

10. The method for detecting internal steel reinforcement corrosion integrating negative pressure adsorption, radar scanning, AI recognition, and three-dimensional mapping as described in claim 5, is characterized in that... S4 also includes the introduction of a domain adaptation module to address the differences in feature distribution between different concrete structures, including a domain discriminator, a fast adaptation mechanism, or an amplitude threshold adaptation mechanism.

Images