A UAV-mounted elastic wave detection system and method for concrete structures

Abstract

This invention provides a UAV-mounted system and method for detecting elastic waves in concrete structures, belonging to the field of non-destructive testing in civil engineering. The system includes a UAV, a dual-arm robotic device, a firing device, a receiving device, a vision acquisition device, and a control device. The dual-arm robotic device is positioned below the UAV, with the firing and receiving devices located at their ends. The vision acquisition device collects images and depth information to generate three-dimensional data. The control device plans the location of points, controls the robotic arms to their positions, and determines the coupling state; upon successful detection, synchronous triggering occurs. The firing device excites elastic waves, the receiving device collects the vibration signals, and the control device extracts features and combines them with the three-dimensional data to obtain defect information. This invention combines non-contact access to high-altitude concrete structures with contact elastic wave detection, solving the problems of high risk and low efficiency in traditional high-altitude testing. The dual robotic arms reduce interference, and combined with vision modeling, achieve precise positioning, improving testing safety, automation, and positioning accuracy.

Description

Technical Field

[0001] This invention mainly relates to the field of non-destructive testing technology in civil engineering, specifically to an unmanned aerial vehicle (UAV) mounted elastic wave testing system and method for concrete structures. Background Technology

[0002] Concrete structures are widely used in infrastructure projects such as buildings, bridges, tunnels, and water conservancy. During long-term service, they are susceptible to damage from loads, environmental erosion, and material aging, resulting in internal damage and external defects. These hidden dangers are difficult to identify with the naked eye in their early stages. If they are not detected and addressed in a timely manner, they will continue to develop and reduce the structural safety and durability. Currently, routine inspections of concrete structures are still mainly conducted manually on-site. For high-altitude, large-span, edge-prone, and inaccessible areas, scaffolding, aerial work platforms, or suspended scaffolds are usually required. This not only makes the work process cumbersome and inefficient but also poses a high safety risk.

[0003] Current conventional inspection methods largely rely on manual, handheld devices to collect data point by point. The distribution of inspection points depends on human experience, making it difficult to achieve uniform and standardized full coverage, resulting in poor objectivity and consistency of inspection results. Furthermore, traditional inspection methods primarily rely on single signal acquisition, lacking effective integration of structural spatial location information and inspection data, and failing to intuitively present the spatial distribution and development trends of defects. In addition, some automated inspection equipment suffers from complex operation, poor adaptability, and lagging data processing, making it difficult to meet the needs of rapid, accurate, and intelligent on-site inspection, and unable to provide efficient and reliable technical support for structural safety assessment and maintenance decisions. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide an unmanned aerial vehicle (UAV) mounted elastic wave detection system and method for concrete structures, which addresses the shortcomings of the prior art.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: A UAV-mounted concrete structure elastic wave detection system includes a UAV, a dual-arm device, a firing device, a MEMS receiving device, a visual acquisition device, and a control device; the dual-arm device is installed below the UAV, the firing device is installed at the end of the first arm of the dual-arm device, the MEMS receiving device is installed at the end of the second arm of the dual-arm device, the visual acquisition device is installed on the top of the UAV, and the control device is installed inside the UAV fuselage; The visual acquisition device is used to acquire surface images and depth information of the concrete structure to be tested when the UAV hovers over the area to be tested, and to generate raw three-dimensional environmental data based on the surface images and depth information of the concrete structure to be tested. The control device is used to plan detection points based on the original three-dimensional environment data and generate robotic arm motion control commands. The dual robotic arm device is used to move the firing device and the MEMS receiving device to the detection point based on the robotic arm motion control command. The control device is also used to determine the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested. If the coupling state is qualified, a synchronous trigger command is output. The firing device is used to apply an impact to the surface of the concrete structure to be tested based on the synchronous triggering command, so as to excite the generation of elastic waves inside the concrete structure. The MEMS receiving device is used to acquire the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test based on the synchronous triggering command, and transmit the vibration response signal to the control device. The control device is also used to extract features from the vibration response signal to obtain signal feature data related to the defects of the concrete structure under test, and to obtain defect information of the concrete structure under test based on the extracted signal feature data and the original three-dimensional environment data.

[0006] Based on the above technical solution, the present invention can be further improved as follows.

[0007] Furthermore, surface images and depth information of the concrete structure under test are acquired, and raw three-dimensional environmental data are generated based on the surface images and depth information of the concrete structure under test, including: Collect surface images and depth information of the concrete structure to be tested, and obtain two-dimensional texture images and depth point cloud data from the surface images and depth information of the concrete structure to be tested; Noise removal processing is performed on the two-dimensional texture image and the depth point cloud data respectively; The denoised 2D texture image and the denoised depth point cloud data are spatially registered to ensure that the image and the point cloud correspond in the same coordinate system. Spatial coordinate calibration is performed on the registered two-dimensional texture image and depth point cloud data, and the mapping relationship between the image pixels, depth data and real physical space of the two-dimensional texture image is established based on the spatial coordinate calibration information. Based on the mapping relationship, the calibrated two-dimensional texture image and depth point cloud data are used to perform three-dimensional reconstruction, generating original three-dimensional environment data containing spatial coordinate information.

[0008] Furthermore, based on the original 3D environment data, detection points are planned, and robotic arm motion control commands are generated, including: The original three-dimensional environmental data is processed by structural region segmentation to obtain the key detection area data and the conventional detection area data of the concrete structure to be tested. Regional features are extracted from the key detection area data and the regular detection area data respectively to obtain key detection area feature data and regular detection area feature data; According to the preset detection density rules, the detection points are deployed for the feature data of the key detection area and the feature data of the regular detection area to obtain the initial detection point coordinate data. The initial detection point coordinate data is spatially validated to obtain valid detection point coordinate data. Based on the effective detection point coordinate data and the kinematic parameters of the dual robotic arm device, the robotic arm motion trajectory is planned to obtain the robotic arm motion trajectory data. Based on the robotic arm motion trajectory data, corresponding robotic arm motion control commands are generated to drive the dual robotic arm device to move the firing device and MEMS receiving device sequentially to each effective detection point.

[0009] Furthermore, the coupling status between the firing device and the MEMS receiving device and the surface of the concrete structure under test is determined. If the coupling status is qualified, a synchronous trigger command is output, including: The system collects first contact force data between the firing device and the surface of the concrete structure under test, as well as second contact force data between the MEMS receiving device and the surface of the concrete structure under test. The first contact force data is compared with a preset first contact force threshold, and the second contact force data is compared with a preset second contact force threshold to obtain the contact force comparison result; Based on the contact force comparison results, the contact stability between the firing device and the MEMS receiving device is determined, and the contact stability determination result is obtained. Based on the contact stability determination result, the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested is determined, and the coupling state determination result is obtained. When the coupling state determination result is qualified, a synchronization trigger command for synchronously triggering the firing device and the MEMS receiving device is generated and output.

[0010] Furthermore, based on the synchronous triggering command, the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired, and the vibration response signal is transmitted to the control device, including: Based on the synchronous trigger command, the MEMS sensing unit is activated to collect the original vibration simulation signal generated by the propagation of elastic waves on the surface of the concrete structure under test. The original vibration simulation signal is amplified using low-noise amplification to obtain the amplified vibration simulation signal; The amplified vibration simulation signal is subjected to anti-aliasing filtering to obtain a filtered vibration simulation signal; The filtered vibration simulation signal is subjected to analog-to-digital conversion to obtain a digital vibration response signal. Electromagnetic interference suppression processing is performed on the digital vibration response signal to obtain a pure vibration response signal; The pure vibration response signal is sent to the control device through the data transmission interface to complete the transmission of the vibration response signal.

[0011] Furthermore, feature extraction is performed on the vibration response signal to obtain signal feature data related to the defects in the concrete structure under test. Based on the extracted signal feature data and the original three-dimensional environmental data, defect information of the concrete structure under test is obtained, including: The vibration response signal is optimized using a CNN-LSTM hybrid network model, which includes a one-dimensional convolutional neural network module, a long short-term memory network module, and a softmax classifier. The one-dimensional convolutional neural network module is used to receive the preprocessed elastic wave time-domain waveform sequence as input, perform sliding convolution operation in the time domain, extract the morphological features of the waveform within a local time window, and form a feature sequence arranged in time. The long short-term memory network module is used to perform time series modeling based on the time evolution law of waveform features in the feature sequence, so as to obtain a global time series representation reflecting the defect echo sequence. The Softmax classifier is used to perform classification processing based on the global temporal representation and output the probability distribution of the defects type of the concrete structure to be tested; the defect type includes at least one of surface cracks, shallow spalling, internal shallow peeling or horizontal cracks, and internal deep hollowing.

[0012] Furthermore, when the maximum predicted probability of the disease type is lower than the preset confidence threshold, the control device generates an adjustment command to drive the dual robotic arm device to fine-tune the contact posture or change the firing parameters, and perform a retest on the same measuring point.

[0013] Furthermore, the control device is also used to perform report generation and disease classification and early warning processing after outputting the probability distribution of the types of defects in the concrete structure to be tested, including: The vibration response signals at each effective detection point are subjected to feature quantization processing to extract the signal feature parameters of the measured wave velocity values; Based on the measured wave velocity value and the preset reference wave velocity value of intact concrete, the comprehensive damage index of each detection point is calculated. Clustering algorithms are used to aggregate spatially adjacent and similarly damaged anomaly detection points to obtain diseased area aggregate data, which includes the area of ​​the diseased area, the centroid coordinates of the diseased area, and the average degree of damage in the diseased area. A concrete structure inspection report is generated based on the comprehensive damage index and aggregated data of the diseased areas. The comprehensive damage index is compared with a preset grading threshold to obtain the disease grading result. Based on the disease grading result, the corresponding early warning level and maintenance suggestion information are matched. The early warning level includes Level 1 minor early warning, Level 2 moderate early warning and Level 3 severe early warning.

[0014] Furthermore, the dual robotic arm device is a multi-degree-of-freedom lightweight robotic arm, and a vibration isolation and damping structure is provided between the dual robotic arm device and the UAV; The sensing end face of the MEMS receiver is provided with an adaptive flexible dry coupling layer, the firing device has a built-in recoil damping module, and the MEMS receiver is provided with an electromagnetic shielding encapsulation shell.

[0015] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: A method for detecting elastic waves in concrete structures mounted on a UAV, employing the UAV-mounted elastic wave detection system for concrete structures as described above, includes the following steps: When the drone hovers over the area to be tested, it collects surface images and depth information of the concrete structure to be tested through a visual acquisition device, and generates raw three-dimensional environmental data based on the surface images and depth information. The control device plans the detection points based on the original three-dimensional environment data and generates control commands for the robotic arm's movements. Based on the robotic arm motion control commands, the dual robotic arm device drives the firing device and the MEMS receiving device to move to the detection point. The control device determines the coupling status between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested. If the coupling status is qualified, a synchronous trigger command is output. The firing device applies an impact to the surface of the concrete structure under test based on the synchronous triggering command, thereby exciting the generation of elastic waves inside the concrete structure. The vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired by the MEMS receiving device based on the synchronous triggering command, and the vibration response signal is transmitted to the control device. The vibration response signal is feature extracted by the control device to obtain signal feature data related to the defects of the concrete structure under test, and the defect information of the concrete structure under test is obtained based on the signal feature data and the original three-dimensional environment data.

[0016] The beneficial effects of this invention are as follows: By constructing an integrated detection system composed of a drone, a dual-arm robotic device, a firing device, a MEMS receiving device, a visual acquisition device, and a control device, a non-contact approach to high-altitude concrete structures and contact elastic wave detection are organically combined, fundamentally solving the problems of traditional high-altitude detection requiring scaffolding, high operational risks, and low efficiency. The system adopts a dual-arm separate structure, allowing the firing and receiving devices to be independently deployed, effectively reducing the interference of vibration recoil and mechanical crosstalk on the detection signal. Combined with visual 3D modeling and autonomous point planning, it achieves full coverage and precise positioning of the detection area. This embodiment constructs a complete hardware and control architecture for non-destructive testing of high-altitude concrete, improving the safety, automation, and spatial positioning accuracy of the detection operation, and providing a stable and reliable hardware foundation and data support for subsequent accurate identification of defects. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the overall structure of the UAV-mounted concrete structure elastic wave detection system provided in an embodiment of the present invention. Figure 2 A schematic diagram of the firing device structure of the UAV-mounted concrete structure elastic wave detection system provided in an embodiment of the present invention; Figure 3 A schematic diagram of the MEMS receiving device of the UAV-mounted concrete structure elastic wave detection system provided in an embodiment of the present invention. Figure 4 Synchronization triggering and timestamp latching timing diagram of the UAV-mounted concrete structure elastic wave detection system provided in this embodiment of the invention; Figure 5 The flowchart illustrates the combined detection process of variable source distance detection (SASW) and impact echo IE in the UAV-mounted concrete structure elastic wave detection system provided in this embodiment of the invention.

[0018] In the attached diagram, the component names represented by each label are as follows: 1. Unmanned Aerial Vehicle (UAV); 2. Dual-Arm Device; 21. First Arm; 22. Second Arm; 3. Firing Device; 31. Tubular Shell; 32. Electromagnetic Coil; 33. Ferromagnetic Strike Pin; 34. Return Spring; 35. Recoil Damping Module; 4. MEMS Receiver; 41. MEMS Accelerometer Chip; 42. Electromagnetic Shielding Encapsulation Shell; 43. In-situ Signal Conditioning Circuit; 44. Digital Interface Module; 45. Universal Adjustment Mechanism; 46. Vibration Isolation Damping Layer; 47. Adaptive Flexible Dry Coupling Layer; 5. Visual Acquisition Device; 6. Control Device; 61. Hardware Synchronization Controller; 7. Wireless Communication Module; 8. Display and Recording Unit. Detailed Implementation

[0019] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0020] The purpose of this invention is to provide an unmanned aerial vehicle (UAV)-based elastic wave detection system and method for concrete structures. This system addresses the difficulties in detecting high-level or hard-to-reach concrete components, the limitations of traditional manual detection methods, and poor coupling stability in existing technologies. It achieves high-precision, multi-dimensional, and visualized real-time detection of internal defects in concrete structures, providing accurate three-dimensional spatial data support for infrastructure health assessment and maintenance decisions, and ensuring structural safety. The following detailed description uses several specific embodiments.

[0021] like Figure 1 As shown, this embodiment of the invention provides a UAV-mounted concrete structure elastic wave detection system, including a UAV 1, a dual-arm device 2, a firing device 3, a MEMS receiver 4, a vision acquisition device 5, and a control device 6; the dual-arm device 2 is installed below the UAV 1, the firing device 3 is installed at the end of the first arm 21 of the dual-arm device 2, the MEMS receiver 4 is installed at the end of the second arm 22 of the dual-arm device 2, the vision acquisition device 5 is installed on the top of the UAV 1, and the control device 6 is installed inside the fuselage of the UAV 1; The visual acquisition device 5 is used to acquire surface images and depth information of the concrete structure to be tested when the UAV 1 hovers over the area to be tested, and to generate raw three-dimensional environmental data based on the surface images and depth information of the concrete structure to be tested. The control device 6 is used to plan detection points based on the original three-dimensional environment data and generate robotic arm motion control commands. The dual robotic arm device 2 is used to drive the firing device 3 and the MEMS receiving device 4 to the detection point based on the robotic arm motion control command. The control device 6 is also used to determine the coupling state between the firing device 3 and the MEMS receiving device 4 and the surface of the concrete structure to be tested. If the coupling state is qualified, a synchronous trigger command is output. The firing device 3 is used to apply an impact to the surface of the concrete structure to be tested based on the synchronous triggering command, so as to excite the generation of elastic waves inside the concrete structure. The MEMS receiving device 4 is used to acquire the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test based on the synchronous triggering command, and transmit the vibration response signal to the control device 6. The control device 6 is also used to extract features from the vibration response signal to obtain signal feature data related to the defects of the concrete structure under test, and to obtain defect information of the concrete structure under test based on the extracted signal feature data and the original three-dimensional environment data.

[0022] In addition, the system also includes a hardware synchronization controller 61, a wireless communication module 7, and a display and recording unit 8. The hardware synchronization controller 61 is integrated inside the control device 6 and is used to realize the synchronous triggering of the firing device 3, the MEMS receiving device 4, and the visual acquisition device 5. The wireless communication module 7 is set on the fuselage of the UAV 1 and is used to establish a wireless data transmission channel between the control device 6 and the display and recording unit 8. The display and recording unit 8 is a ground measurement and control terminal, used to receive and display and store detection data and disease information in real time, and to complete the visualization and recording of detection results.

[0023] In the above embodiments, an integrated detection system consisting of a drone, a dual-robotic arm device, a firing device, a MEMS receiving device, a visual acquisition device, and a control device is constructed. This system organically combines non-contact access to high-altitude concrete structures with contact elastic wave detection, fundamentally solving the problems of traditional high-altitude detection requiring scaffolding, high operational risks, and low efficiency. The system adopts a dual-robotic arm separate structure, allowing the firing and receiving devices to be deployed independently, effectively reducing the interference of vibration recoil and mechanical crosstalk on the detection signal. Combined with visual 3D modeling and autonomous point planning, it achieves full coverage and precise positioning of the detection area. This embodiment constructs a complete hardware and control architecture for non-destructive testing of high-altitude concrete, improving the safety, automation, and spatial positioning accuracy of the detection operation, and providing a stable and reliable hardware foundation and data support for subsequent accurate identification of defects.

[0024] Preferably, the surface image and depth information of the concrete structure to be tested are acquired, and raw three-dimensional environmental data are generated based on the surface image and depth information of the concrete structure to be tested, including: Collect surface images and depth information of the concrete structure to be tested, and obtain two-dimensional texture images and depth point cloud data from the surface images and depth information of the concrete structure to be tested; Noise removal processing is performed on the two-dimensional texture image and the depth point cloud data respectively; The denoised 2D texture image and the denoised depth point cloud data are spatially registered to ensure that the image and the point cloud correspond in the same coordinate system. Spatial coordinate calibration is performed on the registered two-dimensional texture image and depth point cloud data, and the mapping relationship between the image pixels, depth data and real physical space of the two-dimensional texture image is established based on the spatial coordinate calibration information. Based on the mapping relationship, the calibrated two-dimensional texture image and depth point cloud data are used to perform three-dimensional reconstruction, generating original three-dimensional environment data containing spatial coordinate information.

[0025] In the above embodiments, to ensure accurate correspondence between the detection data and the 3D model position, the system establishes a rigid body coordinate transformation chain consisting of the world coordinate system, the visual map coordinate system, the body coordinate system, and the end-sensor coordinate system, and achieves strict spatiotemporal matching through a microsecond-level synchronization mechanism. The absolute coordinates Pdata of the MEMS receiving device in 3D space are calculated in real time using kinematic chain equations. , in, This is a fixed offset of the MEMS sensor center relative to the end flange. for t The real-time transformation matrix of the robotic arm's end effector relative to the main body is obtained by the forward kinematics calculation of the robotic arm's joint encoder. The system calculates the body pose matrix for visual SLAM or RTK-GPS. It uses matrix operations to solve spatial mapping relationships in real time, enabling the detection data to be mapped to a 3D digital twin model with centimeter-level accuracy.

[0026] In the above embodiments, by acquiring, denoising, spatially registering, calibrating coordinates, and reconstructing the surface images and depth information of the concrete structure, original 3D environmental data containing precise physical coordinates is generated, achieving high-precision 3D digital modeling of the area to be tested. This process effectively eliminates environmental noise and abnormal interference in the image and point cloud data, completes a unified spatial mapping of texture information and depth data, ensuring a strict correspondence between the detection points and the spatial location of the structure, and eliminating positioning deviations caused by manual marking. 3D reconstruction provides a precise data foundation for detection path planning, spatial location of defects, and result visualization, significantly improving the rationality of detection point layout and the accuracy of defect location marking, enhancing the traceability and engineering practicality of the detection results.

[0027] Preferably, based on the original three-dimensional environment data, detection points are planned, and robotic arm motion control commands are generated, including: The original three-dimensional environmental data is processed by structural region segmentation to obtain the key detection area data and the conventional detection area data of the concrete structure to be tested. Regional features are extracted from the key detection area data and the regular detection area data respectively to obtain key detection area feature data and regular detection area feature data; According to the preset detection density rules, the detection points are deployed for the feature data of the key detection area and the feature data of the regular detection area to obtain the initial detection point coordinate data. The initial detection point coordinate data is spatially validated to remove unreachable and duplicate points, thus obtaining valid detection point coordinate data. Based on the effective detection point coordinate data and the kinematic parameters of the dual robotic arm device, the robotic arm motion trajectory is planned to obtain the robotic arm motion trajectory data. Based on the robotic arm motion trajectory data, corresponding robotic arm motion control commands are generated to drive the dual robotic arm device to move the firing device and MEMS receiving device sequentially to each effective detection point.

[0028] Specifically, the original three-dimensional environmental data is processed by structural region segmentation, which involves dividing the surface of the concrete structure to be tested into key inspection areas and routine inspection areas based on the normal vector, curvature, and depth abrupt change characteristics of the three-dimensional point cloud. The key inspection areas include structural corners, joints, variable cross-sections, stress concentration areas, and areas with visible surface defects, while the routine inspection areas are the main areas with flat structures and no obvious abnormalities, thereby obtaining the data of key inspection areas and routine inspection areas.

[0029] Regional features were extracted from the data of key detection areas and regular detection areas respectively. The geometric dimensions, surface flatness, curvature distribution, spatial orientation and boundary contour features of the areas were extracted to obtain the feature data of key detection areas and regular detection areas.

[0030] The detection points are laid out according to the preset detection density rules: for key detection areas, the points are densely distributed with a small spacing (e.g., 10cm×10cm); for regular detection areas, the points are distributed with a regular density with a larger spacing (e.g., 20cm×20cm) to obtain the initial detection point coordinate data.

[0031] Spatial rationality verification of the initial detection point coordinate data: Combining the movement range, posture constraints and three-dimensional environmental obstacle information of the dual robotic arms, points that exceed the working space of the robotic arms, interfere with obstacles, cannot fit the posture, or have overlapping spatial positions are eliminated, and points that can be stably contacted, have compliant postures, and have no risk of collision are retained to obtain valid detection point coordinate data.

[0032] Specifically, based on the effective detection point coordinate data and the kinematic parameters of the dual-arm robotic device, the robotic arm motion trajectory is planned to obtain the robotic arm motion trajectory data. The specific processing procedure is as follows: The control device uses a greedy algorithm or shortest path planning rule to traverse and sort all measurement points based on the three-dimensional spatial coordinates of the effective detection points, so that the movement path of the two robotic arms from the current position to the next measurement point is the shortest and the spatial turning angle is the smallest, thereby reducing redundant movement and flight disturbance.

[0033] The control device uses a greedy algorithm or shortest path planning rule to traverse and sort all measurement points based on the three-dimensional spatial coordinates of the effective detection points. The specific implementation steps are as follows: Using the current position of the two robotic arms as the starting point for traversal, read the three-dimensional coordinates of all valid detection points.

[0034] The system calculates the 3D Euclidean distance between the current point and all untraversed measurement points in real time, and selects the nearest measurement point as the next target point.

[0035] Simultaneously calculate the spatial turning angle from the current robotic arm posture to the next measurement point, and prioritize the measurement point with the smallest turning angle and the smoothest movement direction to avoid body disturbance caused by large-angle turning.

[0036] Mark the traversed measurement points as complete, and use the newly arrived measurement point as the current point. Repeat the distance calculation and angle filtering until all valid measurement points have been traversed.

[0037] Generate a sequential sequence of measurement points, ensuring that the dual robotic arms move from their current positions to the next measurement point with the shortest path, the smallest turning angle, and the smoothest trajectory. This significantly reduces redundant motion and flight disturbances, improving the stability and efficiency of high-altitude operations.

[0038] The control device performs inverse kinematics calculations for each target measurement point based on the kinematic parameters such as joint length, rotation range, and limit angle of the dual robotic arm device. It converts the spatial rectangular coordinates of the measurement point into the rotation angle, angular velocity, and acceleration information of each joint of the dual robotic arm, thereby obtaining the joint control quantities that satisfy the motion constraints.

[0039] The control device performs fifth-order polynomial interpolation or trapezoidal velocity planning on the motion path between adjacent measuring points, so that the start-up, operation and stopping process of the robotic arm is smooth and stable, avoiding sudden speed changes and impact vibrations, and ensuring that the hovering attitude of the UAV is not disturbed by the movement of the robotic arm.

[0040] During trajectory generation, the control device combines the original 3D environment data to perform real-time collision detection, determining whether the robotic arm's movement path interferes with the drone's body, surrounding structures, or obstacles; it automatically eliminates unreachable points or fine-tunes their attitude to ensure the trajectory is safe and feasible.

[0041] The control device constrains the coordinated trajectory of the first and second robotic arms, ensuring that the firing device and the MEMS receiving device move synchronously, arrive at the measurement point at the same time, and make contact at the same time, thus ensuring that the coupling state of the two is consistent and improving the detection synchronization and signal stability.

[0042] After the above processing, the final data of the robotic arm motion trajectory, which includes the coordinates of each measuring point, joint angle sequence, motion speed, acceleration and arrival time, is generated and used to drive the dual robotic arm device to complete continuous automated inspection operations.

[0043] Preferably, the system utilizes the independent movement capability of dual robotic arms to execute a variable source-detector distance scanning strategy, namely Surface Wave Spectrum Analysis (SASW). By dynamically adjusting the source-detector distance d between the firing point and the receiving point, broadband elastic wave data is acquired. Small source-detector distance high-frequency scanning is used for shallow layer detection. The source-detector distance is set to a small value, and the firing device excites high-frequency short waves, concentrating the elastic wave energy in the shallow surface layer. The control device analyzes the high-frequency peak value based on the impact echo principle to identify surface cracks and shallow layer spalling. Large source-detector distance low-frequency scanning is used for deep layer detection. Keeping the MEMS receiving device stationary, the firing device is moved to gradually increase the source-detector distance, acquiring low-frequency Rayleigh wave signals to achieve deep structure detection.

[0044] Preferably, the system employs SASW dispersion inversion technology to process multi-spacing echo data. The control device utilizes the dispersion characteristics of Rayleigh waves for analysis and uses an inversion algorithm to calculate the profile information of shear wave velocity Vs as a function of depth. The system first calculates the phase velocity of the elastic wave signal, constructs a phase velocity-wavelength curve, and then uses the damped least squares method to transform the dispersion curve into a shear wave velocity-depth profile. This allows for the analysis of the hardness gradient variation along the depth direction within the concrete, extending the assessment from single defect location to depth-oriented material performance evaluation.

[0045] In the above embodiments, structural region segmentation, feature extraction, point placement, spatial verification, and trajectory planning are performed based on 3D environmental data, forming a full-process intelligent decision-making mechanism from region division to robotic arm control. This enables differentiated deployment strategies for intensive detection in key areas and uniform detection in regular areas. By verifying the accessibility and rationality of initial points, invalid and duplicate points are eliminated, improving detection efficiency and equipment movement safety. A smooth and stable motion trajectory is generated by combining the kinematic parameters of the dual robotic arms, ensuring that the firing and receiving devices accurately reach the target measurement point, reducing the impact of robotic arm movement on the drone's hovering attitude, and improving the stability of high-altitude contact operations and the execution accuracy of detection points.

[0046] Preferably, the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested is determined. If the coupling state is qualified, a synchronous trigger command is output, including: The system collects first contact force data between the firing device and the surface of the concrete structure under test, as well as second contact force data between the MEMS receiving device and the surface of the concrete structure under test. The first contact force data is compared with a preset first contact force threshold, and the second contact force data is compared with a preset second contact force threshold to obtain the contact force comparison result; Based on the contact force comparison results, the contact stability between the firing device and the MEMS receiving device is determined, and the contact stability determination result is obtained. Based on the contact stability determination result, the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested is determined, and the coupling state determination result is obtained. When the coupling state determination result is qualified, a synchronization trigger command for synchronously triggering the firing device and the MEMS receiving device is generated and output.

[0047] Specifically, pressure sensors respectively installed at the front end of the firing device and the MEMS receiving device are used to collect the first contact force data between the firing device and the surface of the concrete structure to be tested, as well as the second contact force data between the MEMS receiving device and the surface of the concrete structure to be tested. The first contact force data is compared with a preset first contact force threshold (e.g., 3N-8N), and the second contact force data is compared with a preset second contact force threshold (e.g., 3N-8N) to obtain the contact force comparison result; Based on the contact force comparison results, it is determined whether the contact force remains stable within the threshold range for a preset time (e.g., 0.3s-1s) and the fluctuation amplitude is less than the preset limit. The contact stability between the firing device and the MEMS receiving device is then determined, and the contact stability determination result is obtained. Based on the contact stability determination result, the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure under test is determined to be qualified only when both contact forces are stable and qualified, and the coupling state determination result is obtained. When the coupling state determination result is qualified, a synchronization trigger command for synchronously triggering the firing device and the MEMS receiving device is generated and output.

[0048] This embodiment provides a specific method for non-destructive testing of defects in high-rise concrete structures using the aforementioned system, detailing the entire process from physical signal acquisition to final result presentation. This process is completed collaboratively by airborne autonomous logic and ground-based manual monitoring, ensuring the validity and traceability of the test data.

[0049] Physical acquisition is synchronized with spatiotemporal data acquisition. Specifically, at the start of the detection task, the drone is controlled to fly to and hover in front of the concrete surface to be tested. The visual acquisition device scans the area to be tested and establishes a local coordinate system. Based on a preset measurement point grid, such as 10cm×10cm, the control device controls the dual robotic arms to extend under the guidance of a force control algorithm, so that the firing device at the end of the first robotic arm and the MEMS receiving device at the end of the second robotic arm simultaneously contact the concrete surface. At this time, the system executes closed-loop force control logic: only when the feedback values ​​of the pressure sensors at the ends of the two robotic arms are stable within a preset range, such as 5N, and remain so for a certain period of time, the robotic arms enter a compliant control mode. This not only compensates for the slight shaking of the drone but also automatically conforms to the wall angle using a universal joint mechanism. Then, the flight attitude and the joint stiffness of the robotic arms are locked, confirming the "coupling ready" state.

[0050] At this point, the hardware synchronization controller, such as an FPGA synchronization controller, sends a pulse to the firing device and simultaneously sends a start signal to the MEMS acquisition circuit. Within the same microsecond of sending the trigger signal, the hardware synchronization controller latches the current system timestamp and triggers the vision acquisition device to acquire the current frame image and the UAV's RTK coordinates. For example... Figure 4 As shown, this mechanism ensures a rigid binding between waveform data, visual images, and spatial coordinates, eliminating the spatiotemporal misalignment caused by wireless transmission delay.

[0051] Simultaneously, a variable source ranging scanning strategy is employed, utilizing the independent movement capabilities of dual robotic arms to perform surface spectral analysis (SASW). For example... Figure 5 As shown, broadband elastic wave data containing rich dispersion information is obtained by dynamically adjusting the straight-line distance between the firing point and the receiving point, i.e., the source-detection distance d. The specific steps are as follows: Small source-detection distance high-frequency scanning (shallow detection): Control the dual robotic arms to move closer, set the source-detection distance d to a small value, for example, d = 10cm, the firing device triggers a high-frequency shortwave, and the MEMS receiving device acquires the signal. At this time, the elastic wave energy is mainly concentrated in the shallow surface layer of concrete. The control device uses the impact echo (IE) principle to focus on analyzing the high-frequency reflection peaks in the signal spectrum. If abnormal concentration of spectral energy or phase reversal is found, it is determined that there are surface cracks or shallow spalling defects in the shallow concrete layer, for example, within a depth of 0-10cm. Large source-detection distance low-frequency scanning (deep detection): Keep the MEMS receiving device stationary at the current measurement point position, control the first robotic arm to move the firing device along the measurement line direction, and sequentially adjust the source-detection distance d to a larger value of multiple gradients, for example, setting d = 20cm, 40cm, and 60cm sequentially, firing and acquiring data at each distance point. As the source-detection distance increases, high-frequency components attenuate. MEMS receivers primarily pick up low-frequency Rayleigh waves (R waves) with longer wavelengths. Since longer waves have a deeper penetration depth, this step is used to detect structural features deep within concrete, such as within a depth range of 10-50 cm.

[0052] like Figure 4 The diagram shown is a timing diagram of synchronous triggering and timestamp latching in this embodiment. The hardware synchronization controller 61 outputs three trigger signals at the same synchronous triggering time t0: triggering the firing device 3 to perform an impact action, triggering the MEMS receiving device 4 to collect vibration response data, and triggering the visual acquisition device 5 to collect structural images and pose information. All elastic wave data, vibration response data, visual / 3D environmental data, and pose information are latched at the same timestamp t0, realizing microsecond-level spatiotemporal binding of detection data and spatial position, completely eliminating the positioning offset caused by wireless transmission delay, and ensuring accurate and error-free marking of the defect location.

[0053] When inspecting the side of the viaduct pier, a synchronous trigger is used to ensure that a certain set of elastic wave signals accurately corresponds to the measuring point with coordinates (X=8.62m, Y=5.17m, Z=16.33m) in the three-dimensional point cloud, without misalignment or drift.

[0054] In the above embodiments, a closed-loop coupling qualification judgment mechanism is constructed through contact force acquisition, threshold comparison, stability determination, and coupling state confirmation to ensure that the firing device and the MEMS receiving device can maintain a stable and reliable contact state with the concrete surface even in high-altitude disturbance environments. By independently detecting and jointly judging contact forces on both sides, signal distortion caused by poor contact on one side is avoided. Detection is only triggered after the coupling state meets the standard, effectively improving the consistency between elastic wave excitation and signal acquisition. This embodiment ensures signal acquisition quality from the source, reduces invalid detection and repetitive operations, and improves the success rate and data reliability of high-altitude detection operations.

[0055] Preferably, the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired based on the synchronous trigger command, and the vibration response signal is transmitted to the control device, including: Based on the synchronous trigger command, the MEMS sensing unit is activated to collect the original vibration simulation signal generated by the propagation of elastic waves on the surface of the concrete structure under test. The original vibration simulation signal is amplified using low-noise amplification to obtain the amplified vibration simulation signal; The amplified vibration simulation signal is subjected to anti-aliasing filtering to obtain a filtered vibration simulation signal; The filtered vibration simulation signal is subjected to analog-to-digital conversion to obtain a digital vibration response signal. Electromagnetic interference suppression processing is performed on the digital vibration response signal to obtain a pure vibration response signal; The pure vibration response signal is sent to the control device through the data transmission interface to complete the transmission of the vibration response signal.

[0056] Specifically, to suppress random noise, the system defaults to a multi-fire superposition strategy, that is, while keeping the contact state unchanged, the firing device fires N times continuously at a frequency of 5-10Hz, and the MEMS receiving device simultaneously collects N sets of waveform data and temporarily stores them in the airborne buffer.

[0057] Data processing and feature calculations are performed after analog-to-digital conversion, mainly by the airborne edge computing unit, which converts the original waveform into engineering physical quantities.

[0058] Since the rotor rotation and airflow disturbance of the UAV generate background noise, the system first uses the time-domain superposition averaging method to process the signal. It fires and acquires data N times (e.g., 5-10 times) at the same measurement point to obtain N sets of original time-domain waveforms Si(t), and then performs linear superposition averaging according to the following formula: , This operation cancels out random environmental noise, enhances coherent structural reflection signals, and significantly improves the signal-to-noise ratio.

[0059] Then the averaged signal A digital bandpass filter is used. The high-pass cutoff frequency is set to 1.5kHz to filter out low-frequency mechanical sway signals and aerodynamic noise generated by the rotor during drone hovering; the low-pass cutoff frequency is set to 30kHz to 50kHz to filter out high-frequency electronic noise and meet anti-aliasing requirements. For extreme conditions with low signal-to-noise ratio, wavelet transform is further used for threshold denoising to suppress residual noise while preserving the transient characteristics of the impact echo.

[0060] Perform a Fast Fourier Transform on the preprocessed time-domain signal to obtain the amplitude spectrum A(f), and search for the frequency corresponding to the maximum energy peak in the amplitude spectrum, denoted as the main frequency. Based on the principle of shock echo, and utilizing the transient elastic wave resonance theory in plate structures, the depth of the reflecting interface is inverted using the following formula. D : , The system performs decision-level fusion of the inversion results with the surface texture features extracted by the visual acquisition device: if the analysis shows... If the signal shifts significantly to higher frequencies (calculated depth D < 10 cm) and the visual image shows that the surface of the area is intact and without visible cracks, it is determined to be an internal shallow peel or horizontal crack. If the signal shows obvious low-frequency bending vibration patterns and there are no visual abnormalities, it is determined to be a large-area deep hollow inside. If the signal decays rapidly in the time domain and there is no obvious main peak in the frequency domain, and the visual image identifies the characteristics of an open crack, it is directly determined to be a surface crack, and depth inversion is no longer performed.

[0061] In the above embodiments, high-fidelity acquisition and digital transmission of weak elastic wave vibration signals are achieved through synchronous triggering, analog signal amplification, anti-aliasing filtering, analog-to-digital conversion, and electromagnetic interference suppression. In-situ amplification and filtering effectively improve the signal-to-noise ratio of weak signals, while electromagnetic interference suppression reduces electromagnetic noise from the UAV motor and firing pulses, resulting in purer and more stable acquired vibration response signals. Digital output reduces signal transmission loss and interference, ensuring the control device can acquire high-quality raw waveform data, providing a precise and reliable data source for subsequent defect feature extraction and intelligent identification, and significantly improving the system's signal acquisition capabilities in complex high-altitude environments.

[0062] Preferably, feature extraction is performed on the vibration response signal to obtain signal feature data related to the defects in the concrete structure under test. Based on the extracted signal feature data and the original three-dimensional environmental data, defect information of the concrete structure under test is obtained, including: The vibration response signal is optimized using a CNN-LSTM hybrid network model, which includes a one-dimensional convolutional neural network module, a long short-term memory network module, and a softmax classifier. The one-dimensional convolutional neural network module is used to receive the preprocessed elastic wave time-domain waveform sequence as input, perform sliding convolution operation in the time domain, extract the morphological features of the waveform within a local time window, and form a feature sequence arranged in time. The long short-term memory network module is used to perform time series modeling based on the time evolution law of waveform features in the feature sequence, so as to obtain a global time series representation reflecting the defect echo sequence. The Softmax classifier is used to perform classification processing based on the global temporal representation and output the probability distribution of the defects type of the concrete structure to be tested; the defect type includes at least one of surface cracks, shallow spalling, internal shallow peeling or horizontal cracks, and internal deep hollowing.

[0063] Specifically, in the disease identification process, a one-dimensional convolutional neural network (1D-CNN) module is constructed by the control device as a feature extractor. This module takes the preprocessed single or superimposed averaged elastic wave time-domain waveform sequence as input, performs sliding convolution operations in the time domain, and automatically extracts the morphological features of the waveform within a local time window through multiple convolution kernels and nonlinear mapping, thus replacing the traditional manually designed parameter extraction process. To facilitate integration with subsequent time-series modeling, the output of the 1D-CNN module is no longer represented by a single "feature vector," but rather forms a feature sequence or feature map arranged in time, denoted as... ,in, Indicates the waveform number 1 Local feature representation corresponding to each time location.

[0064] Subsequently, the feature sequence F output by the 1D-CNN module is input sequentially into the Long Short-Term Memory (LSTM) network module for temporal modeling. Given that the propagation of elastic waves in concrete has a strict arrival order, the LSTM module receives the feature vector at each time step. By combining historical hidden states with recursive updates, the hidden state sequence is output. This allows for the modeling of the temporal evolution of waveform features. In this process, LSTM uses a gated memory mechanism to suppress disordered scattering interference caused by dense reinforcing bars, making the model more inclined to retain effective echo information with a stable propagation order, thus obtaining a global temporal representation reflecting the defect echo sequence. To form the final representation vector for classification, this embodiment preferably uses the final hidden state h. M As a sequence representation, or by pooling the hidden state sequences to obtain the representation vector h*.

[0065] Finally, the representation vector h* output by the LSTM is input into the fully connected layer, and the probability distribution of the disease type is output through the Softmax classifier. Let the output of the fully connected layer be the score vector z for each class, then the Softmax probability output is: in, Number of disease categories For the first The predicted probability of each class is used, and the final diagnostic category is the category with the highest probability. The output disease types include surface cracks, shallow peeling, internal shallow peeling or horizontal cracks, and internal deep hollowing. The Softmax output is the probability distribution of each of these categories.

[0066] Meanwhile, the control device is configured with adaptive retest logic: when the maximum classification probability is maxed out... k p k When the confidence level is below the preset confidence threshold (e.g., 85%), the current data is deemed insufficiently valid. Instead of outputting a diagnostic result, the control device generates an adjustment command to drive the dual robotic arm device to fine-tune the contact posture or change the firing parameters. Automatic retesting is then performed on the same measurement point until the inference confidence level meets the requirements.

[0067] In the above embodiments, a CNN-LSTM hybrid network model is used to perform deep feature mining and defect classification on elastic wave time-series signals. A one-dimensional convolutional neural network is used to extract local waveform morphological features, and a long short-term memory network is combined to capture the temporal evolution of the signal, enabling intelligent identification of various typical defects such as surface cracks, shallow spalling, internal peeling, and deep hollowing. This method replaces traditional manual interpretation, reduces subjective errors, and improves the accuracy and generalization ability of defect identification. It can accurately extract defect features from complex noise signals, achieve quantitative identification of hidden defects inside concrete structures, and provide an objective and efficient intelligent analysis method for structural health assessment.

[0068] Preferably, when the maximum predicted probability of the disease type is lower than the preset confidence threshold, the control device generates an adjustment command to drive the dual robotic arm device to fine-tune the contact posture or change the firing parameters, and perform a retest on the same measuring point.

[0069] In the above embodiments, through confidence threshold judgment and automatic retesting mechanism, when the confidence of the defect identification result is insufficient, the robotic arm is driven to fine-tune its posture or optimize the firing parameters, and the same measurement point is collected and analyzed a second time. This effectively eliminates misjudgments and omissions caused by poor contact, signal interference, or single acquisition errors. This closed-loop feedback mechanism improves the robustness of the system under complex working conditions, ensures that the output defect identification result has high confidence, reduces invalid data and erroneous judgments, improves the accuracy of detection results and the reliability of engineering decisions, and enables the system to adapt to complex detection scenarios such as high-altitude strong interference and uneven surfaces.

[0070] Preferably, the control device is further configured to perform report generation and disease classification and early warning processing after outputting the probability distribution of the types of defects in the concrete structure to be tested, including: The vibration response signals at each effective detection point are subjected to feature quantization processing to extract the signal feature parameters of the measured wave velocity values; Based on the measured wave velocity value and the preset reference wave velocity value of intact concrete, the comprehensive damage index of each detection point is calculated. Clustering algorithms are used to aggregate spatially adjacent and similarly damaged anomaly detection points to obtain diseased area aggregate data, which includes the area of ​​the diseased area, the centroid coordinates of the diseased area, and the average degree of damage in the diseased area. A concrete structure inspection report is generated based on the comprehensive damage index and aggregated data of the diseased areas. The comprehensive damage index is compared with a preset grading threshold to obtain the disease grading result. Based on the disease grading result, the corresponding early warning level and maintenance suggestion information are matched. The early warning level includes Level 1 minor early warning, Level 2 moderate early warning and Level 3 severe early warning.

[0071] In the above embodiments, the control device uses the Kriging interpolation algorithm to spatially smooth the discrete detection data, generates a continuous disease distribution cloud map, analyzes the spatial correlation of diseases through the variogram, and calculates the predicted value of unknown nodes using the following formula: , in, To convert discrete wave velocity and defect depth data into a continuous scalar field, unbiased weighting coefficients calculated based on spatial geometric relationships are used. Simultaneously, the system utilizes deep learning algorithms such as PointNet++ to perform semantic segmentation on the 3D point cloud, automatically identifying structural components such as beams, slabs, and columns. Defect attributes, including defect depth and comprehensive damage index, are written into the LAS point cloud format, achieving multi-dimensional fusion of geometric, semantic, and defect information. Finally, UV unwrapping technology is used to render the defect heatmap as a texture onto the surface of the 3D model, supporting transparency adjustment to achieve a transparent 3D defect visualization effect.

[0072] Specifically, the system incorporates an intelligent report generation module, enabling automated closed-loop processing and end-to-end data traceability from data analysis to project delivery. First, feature quantization is performed. The control device analyzes the elastic wave signal at each measuring point, extracting three types of feature parameters: dominant frequency offset, spectral energy attenuation rate, and wave velocity value. Combined with visually acquired information on surface crack width and length, the Comprehensive Damage Index (CDI) is calculated using the following formula: , in, The measured wave velocity at the current measuring point. This index serves as a reference wave velocity for intact concrete areas and directly reflects the degree of material stiffness degradation.

[0073] Subsequently, the system uses DBSCAN or K-Means clustering algorithms to merge spatially adjacent and similarly damaged abnormal measurement points into diseased areas, and calculates and records the total area, centroid coordinates and average damage degree of each diseased area.

[0074] Finally, the system performs tiered early warning and decision support by inputting the CDI into a pre-set tiered model and setting thresholds A=15 and B=30 according to relevant specifications. Level 1 Warning (Slight): 0≤CDI<15, corresponding to slight weathering or micro-cracks on the surface, marked in green, and it is recommended to continue observation; Level 2 Warning (Medium): 15≤CDI<30, corresponding to shallow voids or medium-sized cracks, marked in yellow, and recommended to be included in the maintenance plan; Level 3 Warning (Severe): CDI≥30, corresponding to deep through cracks, severe hollowing or steel corrosion, marked in red, and immediate reinforcement is recommended.

[0075] Meanwhile, the system has a built-in maintenance knowledge base that can automatically match repair suggestions based on the type of disease and visual texture characteristics; for level 3 warning areas, it can further recommend corresponding repair processes such as epoxy resin grouting and pressure grouting based on the crack morphology.

[0076] In the above embodiments, the entire process from data processing to project delivery is automated through wave velocity feature quantification, comprehensive damage index calculation, disease area clustering, test report generation, and hierarchical early warning. The damage index calculated based on wave velocity differences intuitively reflects the degree of concrete stiffness degradation. Spatial clustering integrates discrete measuring points into continuous disease areas, facilitating the assessment of the disease range and severity. The three-level early warning mechanism and corresponding maintenance recommendations directly provide decision-making basis for project operation and maintenance. The automatically generated test reports reduce manual processing costs, improve the standardization and engineering application level of test results, and achieve integrated output of testing, assessment, early warning, and reporting.

[0077] Preferably, the dual robotic arm device is a multi-degree-of-freedom lightweight robotic arm, and a vibration isolation and damping structure is provided between the dual robotic arm device and the UAV; The sensing end face of the MEMS receiver is provided with an adaptive flexible dry coupling layer, the firing device has a built-in recoil damping module, and the MEMS receiver is provided with an electromagnetic shielding encapsulation shell.

[0078] In the above embodiments, the overall stability and environmental adaptability of the system are improved through structural optimization design of multi-degree-of-freedom lightweight dual robotic arms, vibration isolation and damping structures, adaptive flexible dry coupling layers, low-recoil firing modules, and electromagnetic shielding shells. The vibration isolation structure effectively isolates the interference of UAV fuselage vibrations on the detection signal; the adaptive coupling layer improves the sensor's fit and coupling consistency on rough and inclined surfaces; the recoil damping module reduces the disturbance of firing reaction force on flight attitude; and the electromagnetic shielding structure suppresses electromagnetic noise. These structural improvements synergistically enhance the system's signal quality, operational stability, and detection accuracy in complex high-altitude environments, extend equipment lifespan, and enhance the system's engineering applicability and durability.

[0079] like Figure 1 , Figure 2 , Figure 3 As shown in the figure, the present invention provides a structure of a UAV-mounted concrete structure elastic wave detection system, which mainly includes: UAV body 1, dual robotic arm device 2, firing device 3, MEMS receiving device 4, visual acquisition device 5, and control device 6. The system further includes a wireless communication module 7 and a display and recording unit 8 (ground measurement and control terminal).

[0080] In this embodiment, the main body of the drone is mostly made of industrial grade, possessing six-way obstacle avoidance and wind-resistant hovering capabilities. It is equipped with four rotor arms, enabling the drone to fly to higher areas of the concrete structure under test. The lower part of the drone body consists of a drone fuselage load interface, a multi-functional mounting base, and a dual-arm device. The multi-functional mounting base is located at the center of the drone's underside. The upper connecting plate is made of high-modulus carbon fiber composite material and is connected to the pre-reserved mounting points on the drone's underside using four high-strength alloy bolts. The lower mounting slide is I-shaped, with dovetail groove rails and spring pins on both sides. During use, the two robotic arm bases are pushed along the rails into the preset slots of the mounting base and locked by the spring pins to achieve a rigid connection. Both the upper connecting plate and the lower mounting slide have corresponding slots that match the neck dimensions of the high-damping silicone shock absorber balls. The shock absorber balls, relying on the elastic deformation of the rubber material, are inserted into these slots with an interference fit, forming a physical isolation layer to reduce impact and maintain flight stability. To prevent the shock absorber balls from breaking due to aging or violent operation, causing the equipment to fall, each shock absorber ball is fitted with an anti-fall limit pin at its central axis.

[0081] The dual-arm device includes a first robotic arm 21 and a second robotic arm 22, which are designed with independent degrees of freedom. The first robotic arm 21 is equipped with a firing device 3 at its end, and the second robotic arm 22 is equipped with a MEMS receiving device 4 at its end.

[0082] The visual acquisition device 5 is installed on the top or front of the UAV 1 and uses a binocular depth camera or lidar to acquire the surface three-dimensional information of the structure under test in real time.

[0083] The control device 6 is embedded inside the fuselage of the UAV 1. As the airborne edge computing core, it is responsible for coordinating the control of the robotic arm's movements, synchronously collecting data, and performing preliminary signal processing.

[0084] The display and recording unit 8 (ground control terminal) establishes a two-way connection with the airborne control device 6 through the wireless communication module 7. It is used to receive and store the transmitted detection data and visual images in real time, and send flight and detection control commands to the UAV to realize basic human-machine interaction.

[0085] Design of a vibration damping and attitude adaptive module at the end of a dual-arm robotic device. This embodiment focuses on the design of a highly stable mechanical mount, aiming to solve the challenges of vibration coupling and contact force control during high-altitude operations of UAVs.

[0086] like Figure 3 As shown, a vibration damping and attitude adaptive module is also provided between the second robotic arm 22 and the MEMS receiving device 4, including a universal adjustment mechanism 45, a vibration damping layer 46, and an adaptive flexible dry coupling layer 47.

[0087] The omnidirectional adjustment mechanism 45 is connected between the end of the second robotic arm 22 and the sensor module. Its upper end is fixed to the end interface of the robotic arm with screws, while its lower end uses a ball joint or dual-axis flexible hinge structure and is connected to a metal base plate, giving the receiving device at least ±30° of passive rotational freedom in the pitch and lateral directions. When the end of the robotic arm contacts a non-flat concrete surface such as an arc or incline, the omnidirectional mechanism automatically deflects under the contact reaction force, ensuring that the central axis of the MEMS receiving device 4 is as perpendicular as possible to the measured surface, eliminating signal attenuation caused by contact angle deviation.

[0088] The vibration damping layer 46 is made of high loss factor butyl rubber, in the form of a ring or rectangular gasket structure, with a thickness of 3.0 mm. It is fixed to the metal base plate of the universal mechanism with industrial-grade high-strength double-sided tape, and the lower end is bonded to the back of the metal shielding shell of the MEMS receiver 4. It can effectively block the transmission of high-frequency mechanical vibration to the sensor and significantly improve the signal-to-noise ratio of the detection signal.

[0089] The adaptive flexible dry coupling layer 47 is made of highly elastic silicone, with a thickness controlled between 1 mm and 3 mm, and is attached to the front end of the MEMS receiver 4. Under the preset pressure applied by the robotic arm, the coupling layer undergoes micro-rheological changes to fill the tiny pits and rough textures on the concrete surface, ensuring good vibration reception and reducing the reflection coefficient of elastic waves at the interface between the concrete and the sensor.

[0090] The specific design of the end-firing device and the receiving device. This embodiment details the specific structure of the firing device and the MEMS receiving device that are in direct contact with concrete, aiming to achieve high signal-to-noise ratio detection.

[0091] The firing device 3 employs a low-reaction-force electromagnetic pulse firing mechanism. For example... Figure 2 As shown, the mechanism includes a tubular housing 31, an electromagnetic coil 32, a ferromagnetic firing pin 33, a return spring 34, and a recoil damping module 35. The tubular housing 31 is fixed to the end of the first robotic arm 21; the electromagnetic coil 32 is wound inside the tubular housing and generates an instantaneous magnetic field through pulsed current; the ferromagnetic firing pin 33 is coaxially arranged at the center of the electromagnetic coil, with a hemispherical hard steel head embedded at its end. When a trigger command is received, the electromagnetic coil generates an instantaneous strong magnetic field to drive the firing pin to accelerate linearly and impact the concrete surface to excite elastic waves; the return spring 34 is used to pull the ferromagnetic firing pin back to its initial position after power is cut off; the recoil damping module 35 is located at the rear of the tubular housing and is used to absorb the reverse momentum generated at the moment of firing of the firing pin to maintain the stability of the UAV's hovering attitude.

[0092] like Figure 3As shown, the MEMS receiving device 4 is an anti-interference integrated microelectromechanical system (MEMS) sensing module. The core components of this module are: a MEMS accelerometer chip 41, encapsulated within the module core, possessing wideband response characteristics, used to acquire weak vibration signals from the concrete surface; an electromagnetic shielding enclosure 42, made of high-permeability material, encasing the MEMS chip to shield against transient electromagnetic pulse interference from the triggering device; an in-situ signal conditioning circuit 43, integrated with the MEMS chip on the same PCB substrate, including a low-noise amplifier and an anti-aliasing filter, used to directly amplify and purify the analog signal at the sensor end, improving the signal-to-noise ratio; and a digital interface module 44, including an analog-to-digital converter and communication protocol logic, used to convert the acquired vibration data into digital signals and transmit them to the control device 6. The signal transmission sequence is as follows: mechanical vibration is converted into a weak analog electrical signal by the MEMS accelerometer chip, then amplified by the low-noise amplifier, filtered by the anti-aliasing filter, and finally converted into an anti-interference digital signal for transmission to the airborne control device.

[0093] Example 2: This embodiment of the invention also provides a UAV-mounted method for detecting elastic waves in concrete structures, using the UAV-mounted elastic wave detection system for concrete structures as described above, including the following steps: When the drone hovers over the area to be tested, it collects surface images and depth information of the concrete structure to be tested through a visual acquisition device, and generates raw three-dimensional environmental data based on the surface images and depth information. The control device plans the detection points based on the original three-dimensional environment data and generates control commands for the robotic arm's movements. Based on the robotic arm motion control commands, the dual robotic arm device drives the firing device and the MEMS receiving device to move to the detection point. The control device determines the coupling status between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested. If the coupling status is qualified, a synchronous trigger command is output. The firing device applies an impact to the surface of the concrete structure under test based on the synchronous triggering command, thereby exciting the generation of elastic waves inside the concrete structure. The vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired by the MEMS receiving device based on the synchronous triggering command, and the vibration response signal is transmitted to the control device. The vibration response signal is feature extracted by the control device to obtain signal feature data related to the defects of the concrete structure under test, and the defect information of the concrete structure under test is obtained based on the signal feature data and the original three-dimensional environment data.

[0094] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0095] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0096] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0097] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0098] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0099] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A UAV-mounted elastic wave detection system for concrete structures, characterized in that, The device includes a drone, a dual-arm robotic assembly, a firing device, a MEMS receiver, a vision acquisition device, and a control device. The dual-arm robotic assembly is mounted below the drone, the firing device is mounted at the end of the first arm of the dual-arm robotic assembly, the MEMS receiver is mounted at the end of the second arm of the dual-arm robotic assembly, the vision acquisition device is mounted on the top of the drone, and the control device is mounted inside the drone fuselage. The visual acquisition device is used to acquire surface images and depth information of the concrete structure to be tested when the UAV hovers over the area to be tested, and to generate raw three-dimensional environmental data based on the surface images and depth information of the concrete structure to be tested. The control device is used to plan detection points based on the original three-dimensional environment data and generate robotic arm motion control commands. The dual robotic arm device is used to move the firing device and the MEMS receiving device to the detection point based on the robotic arm motion control command. The control device is also used to determine the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested. If the coupling state is qualified, a synchronous trigger command is output. The firing device is used to apply an impact to the surface of the concrete structure to be tested based on the synchronous triggering command, so as to excite the generation of elastic waves inside the concrete structure. The MEMS receiving device is used to acquire the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test based on the synchronous triggering command, and transmit the vibration response signal to the control device. The control device is also used to extract features from the vibration response signal to obtain signal feature data related to the defects of the concrete structure under test, and to obtain defect information of the concrete structure under test based on the extracted signal feature data and the original three-dimensional environment data.

2. The UAV-mounted elastic wave detection system for concrete structures according to claim 1, characterized in that, Acquire surface images and depth information of the concrete structure to be measured, and generate raw 3D environmental data based on the surface images and depth information of the concrete structure to be measured, including: Collect surface images and depth information of the concrete structure to be tested, and obtain two-dimensional texture images and depth point cloud data from the surface images and depth information of the concrete structure to be tested; Noise removal processing is performed on the two-dimensional texture image and the depth point cloud data respectively; The denoised 2D texture image and the denoised depth point cloud data are spatially registered to ensure that the image and the point cloud correspond in the same coordinate system. Spatial coordinate calibration is performed on the registered two-dimensional texture image and depth point cloud data, and the mapping relationship between the image pixels, depth data and real physical space of the two-dimensional texture image is established based on the spatial coordinate calibration information. Based on the mapping relationship, the calibrated two-dimensional texture image and depth point cloud data are used to perform three-dimensional reconstruction, generating original three-dimensional environment data containing spatial coordinate information.

3. The UAV-mounted elastic wave detection system for concrete structures according to claim 1, characterized in that, Based on the original 3D environment data, detection points are planned, and robotic arm motion control commands are generated, including: The original three-dimensional environmental data is processed by structural region segmentation to obtain the key detection area data and the conventional detection area data of the concrete structure to be tested. Regional features are extracted from the key detection area data and the regular detection area data respectively to obtain key detection area feature data and regular detection area feature data; According to the preset detection density rules, the detection points are deployed for the feature data of the key detection area and the feature data of the regular detection area to obtain the initial detection point coordinate data. The initial detection point coordinate data is spatially validated to obtain valid detection point coordinate data. Based on the effective detection point coordinate data and the kinematic parameters of the dual robotic arm device, the robotic arm motion trajectory is planned to obtain the robotic arm motion trajectory data. Based on the robotic arm motion trajectory data, corresponding robotic arm motion control commands are generated to drive the dual robotic arm device to move the firing device and MEMS receiving device sequentially to each effective detection point.

4. The UAV-mounted elastic wave detection system for concrete structures according to claim 1, characterized in that, Determine the coupling status between the firing device and the MEMS receiver and the surface of the concrete structure under test. If the coupling status is qualified, output a synchronous trigger command, including: The system collects first contact force data between the firing device and the surface of the concrete structure under test, as well as second contact force data between the MEMS receiving device and the surface of the concrete structure under test. The first contact force data is compared with a preset first contact force threshold, and the second contact force data is compared with a preset second contact force threshold to obtain the contact force comparison result; Based on the contact force comparison results, the contact stability between the firing device and the MEMS receiving device is determined, and the contact stability determination result is obtained. Based on the contact stability determination result, the coupling state between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested is determined, and the coupling state determination result is obtained. When the coupling state determination result is qualified, a synchronization trigger command for synchronously triggering the firing device and the MEMS receiving device is generated and output.

5. The UAV-mounted elastic wave detection system for concrete structures according to claim 1, characterized in that, Based on the synchronous triggering command, the vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired, and the vibration response signal is transmitted to the control device, including: Based on the synchronous trigger command, the MEMS sensing unit is activated to collect the original vibration simulation signal generated by the propagation of elastic waves on the surface of the concrete structure under test. The original vibration simulation signal is amplified using low-noise amplification to obtain the amplified vibration simulation signal; The amplified vibration simulation signal is subjected to anti-aliasing filtering to obtain a filtered vibration simulation signal; The filtered vibration simulation signal is subjected to analog-to-digital conversion to obtain a digital vibration response signal. Electromagnetic interference suppression processing is performed on the digital vibration response signal to obtain a pure vibration response signal; The pure vibration response signal is sent to the control device through the data transmission interface to complete the transmission of the vibration response signal.

6. The UAV-mounted elastic wave detection system for concrete structures according to claim 1, characterized in that, Feature extraction is performed on the vibration response signal to obtain signal feature data related to the defects of the concrete structure under test. Based on the extracted signal feature data and the original three-dimensional environment data, defect information of the concrete structure under test is obtained, including: The vibration response signal is optimized using a CNN-LSTM hybrid network model, which includes a one-dimensional convolutional neural network module, a long short-term memory network module, and a softmax classifier. The one-dimensional convolutional neural network module is used to receive the preprocessed elastic wave time-domain waveform sequence as input, perform sliding convolution operation in the time domain, extract the morphological features of the waveform within the local time window, and form a feature sequence arranged in time. The long short-term memory network module is used to perform time series modeling based on the time evolution law of waveform features in the feature sequence, so as to obtain a global time series representation reflecting the defect echo sequence. The Softmax classifier is used to perform classification processing based on the global temporal representation and output the probability distribution of the defects type of the concrete structure to be tested; the defect type includes at least one of surface cracks, shallow spalling, internal shallow peeling or horizontal cracks, and internal deep hollowing.

7. The UAV-mounted elastic wave detection system for concrete structures according to claim 6, characterized in that, When the maximum predicted probability of the disease type is lower than the preset confidence threshold, the control device generates an adjustment command to drive the dual robotic arm device to fine-tune the contact posture or change the firing parameters, and perform a retest on the same measuring point.

8. The UAV-mounted elastic wave detection system for concrete structures according to claim 7, characterized in that, The control device is further configured to, after outputting the probability distribution of the types of defects in the concrete structure to be tested, perform report generation and defect classification and early warning processing, including: The vibration response signals at each effective detection point are subjected to feature quantization processing to extract the signal feature parameters of the measured wave velocity values; Based on the measured wave velocity value and the preset reference wave velocity value of intact concrete, the comprehensive damage index of each detection point is calculated. Clustering algorithms are used to aggregate spatially adjacent anomaly detection points with similar damage characteristics to obtain disease area aggregate data. The disease area aggregate data includes the disease area, the centroid coordinates of the disease area, and the average damage degree of the disease area. A concrete structure inspection report is generated based on the comprehensive damage index and aggregated data of the diseased areas. The comprehensive damage index is compared with a preset grading threshold to obtain the disease grading result. Based on the disease grading result, the corresponding early warning level and maintenance suggestion information are matched. The early warning level includes Level 1 minor early warning, Level 2 moderate early warning and Level 3 severe early warning.

9. The UAV-mounted elastic wave detection system for concrete structures according to any one of claims 1 to 8, characterized in that, The dual robotic arm device is a multi-degree-of-freedom lightweight robotic arm, and a vibration isolation and damping structure is provided between the dual robotic arm device and the drone. The sensing end face of the MEMS receiver is provided with an adaptive flexible dry coupling layer, the firing device has a built-in recoil damping module, and the MEMS receiver is provided with an electromagnetic shielding encapsulation shell.

10. A method for detecting elastic waves in concrete structures mounted on a drone, employing the elastic wave detection system for concrete structures as described in any one of claims 1 to 9, characterized in that, Includes the following steps: When the drone hovers over the area to be tested, it collects surface images and depth information of the concrete structure to be tested through a visual acquisition device, and generates raw three-dimensional environmental data based on the surface images and depth information. The control device plans the detection points based on the original three-dimensional environment data and generates control commands for the robotic arm's movements. Based on the robotic arm motion control commands, the dual robotic arm device drives the firing device and the MEMS receiving device to move to the detection point. The control device determines the coupling status between the firing device and the MEMS receiving device and the surface of the concrete structure to be tested. If the coupling status is qualified, a synchronous trigger command is output. The firing device applies an impact to the surface of the concrete structure under test based on the synchronous triggering command, thereby exciting the generation of elastic waves inside the concrete structure. The vibration response signal generated by the propagation of the elastic wave in the concrete structure under test is acquired by the MEMS receiving device based on the synchronous triggering command, and the vibration response signal is transmitted to the control device. The vibration response signal is feature extracted by the control device to obtain signal feature data related to the defects of the concrete structure under test, and the defect information of the concrete structure under test is obtained based on the signal feature data and the original three-dimensional environment data.

Images