Tunnel concrete lining multi-modal detection method and system

Abstract

The invention relates to the technical field of tunnel lining internal defect detection, and provides a tunnel concrete lining multi-modal detection method and system. The system comprises a multi-mode sensor group and a data processing unit, the multi-mode sensor group comprises a geological radar antenna, an infrared thermal imager and an ultrasonic detection module, and the geological radar antenna, the infrared thermal imager and the ultrasonic detection module are used for obtaining internal structure information, surface temperature distribution, lining thickness and crack conditions of a tunnel lining respectively. The multi-mode sensor group is installed on a telescopic electric push rod, and the angle of the multi-mode sensor group is adjusted through a rotary joint driven by a motor so as to adapt to different detection requirements. And the data processing unit is responsible for collecting and fusing data of different sensors, analyzing the data by adopting a deep learning algorithm to identify and position lining defects, supporting wireless data transmission and uploading a detection result to a remote monitoring center in real time. Through multi-modal data fusion, the limitation of a single sensor is effectively overcome, the detection comprehensiveness, precision and intelligent level are improved, and the method is suitable for efficient and automatic detection of tunnel concrete lining.

Description

Technical Field

[0001] The present invention relates to the field of structural defect detection, and particularly to a multi-modal sensor fusion detection system integrating ground penetrating radar, infrared and ultrasonic detection, which is applicable to non-destructive detection of internal defects of concrete structures, linings and other complex engineering components. Background Art

[0002] With the continuous development of tunnel engineering construction, as an important structural part of the tunnel, the quality of the tunnel lining directly affects the safety and durability of the tunnel. At present, traditional defect detection technologies mainly rely on a single sensor (such as ground penetrating radar) or manual tapping to judge the internal condition of the structure. The limitations of single technology detection are as follows: Although ground penetrating radar has advantages in detecting lining thickness, internal cavities, etc., due to the interference of metal materials such as steel bars, misjudgment is likely to occur; Infrared thermal imagers can detect temperature anomaly areas, but due to factors such as ambient temperature and humidity changes and surface coatings, the detection results may be uncertain; Manual tapping detection relies on experience judgment, with low efficiency and subjectivity; Ultrasonic detection requires good coupling between the probe and the concrete surface, but if the surface is rough or the contact is poor, energy loss will occur, affecting the detection effect.

[0003] Therefore, there is an urgent need for a multi-modal data fusion technology to reduce the limitations of a single sensor through information complementarity between sensors and improve the accuracy and detection efficiency of defect analysis. Summary of the Invention

[0004] The purpose of the present invention is to provide a multi-modal detection method and system for tunnel concrete lining to solve the problems in the above background art and improve the comprehensiveness and accuracy of detection.

[0005] To achieve the above purpose, the technical solution of the present invention is:

[0006] A multi-modal detection method and system for tunnel concrete lining, including a multi-modal sensor group and a data processing unit, characterized in that: the multi-modal sensor group includes a ground penetrating radar antenna (1), an infrared thermal imager (2) and an ultrasonic detection module (3), which are used to respectively obtain the internal structure information of the tunnel lining, the surface temperature distribution, the lining thickness and crack conditions; the sensors of the multi-modal sensor group are respectively installed on retractable electric push rods (4-1, 4-2, 4-3), and the support devices (6-1, 6-2, 6-3) adjust the angles of the sensor group through motor-driven rotary joints (5-1, 5-2, 5-3) to adapt to different detection requirements;

[0007] The data processing unit (7) is electrically connected to the multi-modal sensor group, collects and fuses the detection data from different sensors, and analyzes the data using deep learning algorithms to identify and locate defects in the tunnel lining.

[0008] 1. As a preferred embodiment, the ground penetrating radar antenna (3) is used to detect the internal structure of the lining, and data comparison is carried out in combination with the ultrasonic detection module (2) to improve the detection accuracy of the lining thickness and anomalies.

[0009] 2. As a preferred embodiment, the infrared thermal imager (1) is used to detect temperature anomalies on the surface of the lining, and temperature field analysis is carried out through the data processing unit (7) to quickly judge whether there are potential hazards such as cavities and spalling.

[0010] 3. As a preferred embodiment, the data processing unit (7) is internally composed of an FPGA development board and an NVIDIA jetson series development board, with a built-in multi-modal data fusion algorithm, comprehensively analyzing the data from the ground penetrating radar (3), infrared thermal imager (1), and ultrasonic detection module (2), improving the accuracy of defect identification, and at the same time supporting the wireless data transmission function, and can upload the detection data to the remote monitoring center in real time for experts to conduct remote diagnosis and data analysis.

[0011] 4. As a preferred embodiment, the ground penetrating radar (3), infrared thermal imager (1), and ultrasonic detection module (2) each transmit the original signal to the FPGA development board through a high-speed interface, and signal preprocessing, filtering, and preliminary feature extraction are carried out on this board. The processed data is then transmitted to the NVIDIA Jetson development board in real time through a high-speed data bus, and the built-in GPU is used to accelerate the operation of the multi-modal data fusion algorithm.

[0012] 5. As a preferred embodiment, the multi-modal data fusion algorithm includes data preprocessing and spatio-temporal alignment, feature-level fusion, and defect detection algorithms;

[0013] In the data preprocessing and spatio-temporal alignment, a linear interpolation method is used to synchronize the time of data of different modalities, and homogeneous coordinate transformation is used to ensure that the data of all sensors are expressed in the same three-dimensional space coordinates.

[0014] The feature-level fusion uses a Transformer network to extract and fuse the features of ground penetrating radar, infrared thermal imaging, and ultrasonic data; specifically: the ground penetrating radar, infrared thermal imaging, and ultrasonic data respectively extract features through a CNN to form their respective feature vectors X IR 、X US 、X GPR, The Transformer calculates the mutual attention between data of different modalities, extracts cross-modal key features, and outputs a new fused feature X fused And fuse the confidence levels of different sensors through Dempster-Shafer evidence theory to improve the stability of defect detection.

[0015] 6. As a preferred implementation, the defect detection algorithm uses an improved DBSCAN clustering algorithm, namely VDBSCAN algorithm, and processes data based on distributed computing to improve the speed and accuracy of defect detection; specifically including: dividing the dataset D into multiple subsets D1, D2, …, D m , Execute the DBSCAN algorithm in parallel. After calculating the clustering results of each subset, merge the results of each subset to obtain the global clustering result, that is:

[0016]

[0017] The VDBSCAN algorithm can adapt to data with different density distributions to meet the detection requirements of internal defects in tunnel concrete linings. VDBSCAN ensures that low-density areas can also be correctly identified by adaptively adjusting the neighborhood radius ε and the core point density threshold MinPts, thereby improving the detection accuracy and robustness.

[0018] The formula for selecting the adaptive neighborhood radius ε is:

[0019]

[0020] Where ε i is the adaptive neighborhood radius of the data point x i . d(x i , x j ) represents the Euclidean distance between data points. k is the number of nearest neighbors, usually set to MinPts.

[0021] The MinPts is calculated through local density, and the formula is:

[0022]

[0023] Where N is the total number of samples in the dataset. δ is a small constant to prevent division by zero error. A larger MinPts is used in high-density areas (intact concrete), and a smaller MinPts is used in low-density areas (cracks, etc.) so that defects will not be ignored as noise.

[0024] 7. As a preferred implementation, the support device (6) and the rotary joint (5) respectively have the functions of adjustable height and angle to adapt to different types of tunnel structures and ensure that the multi-modal sensor group is always in the best detection position.

[0025] 8. As a preferred embodiment, the electric push rod (4) is equipped with a laser ranging module for real-time monitoring of the distance between the sensor and the tunnel lining wall surface. The laser ranging module obtains the gap data between the sensor and the detection surface and feeds back the measurement result to the control system in real time. The control system dynamically adjusts the telescopic amount of the electric push rod according to the ranging data to ensure that the sensor always maintains the best contact state with the wall surface, thereby improving the stability of the detection signal and the accuracy of data acquisition. In addition, this module can also monitor the local concave and convex changes of the wall surface, providing necessary compensation information for subsequent data analysis and further enhancing the reliability of defect identification.

[0026] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0027] 1. By adopting a multi-modal sensor group (ground penetrating radar, infrared thermal imager, ultrasonic detection module), the limitations of a single sensor are overcome, and comprehensive detection of internal defects of the tunnel lining is realized, improving the detection accuracy and reliability;

[0028] 2. By adopting an adjustable support device, the sensor group can adapt to different tunnel lining structures, ensuring a wide detection range and enhancing the system applicability;

[0029] 3. By adopting an improved VDBSCAN algorithm for intelligent clustering analysis of the detection data, adaptively adjusting the neighborhood radius ε and the core point density threshold MinPts, ensuring that both the high-density area (intact concrete) and the low-density area (defects such as cracks and cavities) can be accurately classified, thereby improving the accuracy and robustness of defect identification.

[0030] 4. With wireless data transmission and remote monitoring functions, the detection data can be uploaded and analyzed in real time, facilitating operation and maintenance management and improving the convenience and operability of the detection work. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a front three-dimensional structure schematic diagram of a multi-modal detection method and system for tunnel concrete lining provided by the present invention;

[0032] Figure 2 It is an axonometric view of a multi-modal detection method and system for tunnel concrete lining provided by the present invention;

[0033] Figure 3 It is an algorithm flowchart of a multi-modal detection system for tunnel concrete lining provided by the present invention.

[0034] Legend Explanation:

[0035] 1. Infrared thermal imager; 2. Ultrasonic detection module; 3. Ground penetrating radar antenna; 4. Retractable electric push rod; 5. Moving joint; 6. Support device; 7. Data processing unit. Detailed implementation mode

[0036] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0037] Embodiment 1, as Figures 1 to 2 shown, this embodiment provides a multi-modal detection system for tunnel concrete lining. The system is composed of a multi-modal sensor group and a data processing unit. Among them, the multi-modal sensor group includes a ground penetrating radar antenna (3), an infrared thermal imager (1), and an ultrasonic detection module (2), which are respectively used to detect the internal structure, surface temperature distribution and thickness information of the tunnel lining, and accurately analyze the defect positions.

[0038] In this embodiment, the detection system is installed on a trolley for moving detection along a predetermined trajectory in the tunnel. There is a fixed bracket on the trolley, and the equipment is installed on the bracket. The electric push rods (4-1, 4-2, 4-3) are used to adjust the height and angle of each sensor to always maintain the best detection state. The rotating joints (5-1, 5-2, 5-3) can further adjust the direction of the sensor to adapt to the detection requirements of different tunnel lining surfaces.

[0039] During the detection process: The ground penetrating radar antenna (3) emits high-frequency electromagnetic waves, receives the reflected signals after penetrating the lining, and performs background removal and signal enhancement through the data processing unit (7) to improve the detection accuracy. The infrared thermal imager (1) real-time collects the temperature distribution on the surface of the tunnel lining, and uses the temperature gradient analysis method to identify possible voids, spalls or delamination areas inside the lining. The ultrasonic detection module (2) analyzes the thickness and crack conditions of the lining through ultrasonic pulse echo, and compares and analyzes with the ground penetrating radar data to improve the detection accuracy.

[0040] The data processing unit (7) uses a multi-modal data fusion algorithm to classify the detection results and mark the defect positions. During the actual detection process, the staff can upload the real-time data to the remote monitoring center through the wireless data transmission module for further analysis and archiving.

[0041] Embodiment 2, as Figures 1 to 2As shown, in this embodiment, the multi-modal detection system for tunnel concrete lining can be used in fixed detection scenarios, such as tunnel construction sites. In this case, the system can be installed on a fixed bracket without relying on a trolley for mobile detection. In the fixed detection mode: the ground penetrating radar antenna (3), infrared thermal imager (1), and ultrasonic detection module (2) of the multi-modal sensor group are fixed on the bracket, and the height and angle of the bracket can be adjusted by electric push rods (4-1, 4-2, 4-3) so that the sensors can monitor a specific area for a long time. The data processing unit (7) continuously collects sensor data to evaluate the health status of the tunnel lining. For example, by using the infrared thermal imager to long-term monitor the temperature change trend of the lining surface, it can be judged whether there is a risk of water leakage or spalling. In this embodiment, the system can be linked with the tunnel monitoring platform to form a long-term health monitoring mechanism, providing decision-making support for tunnel operation and maintenance.

[0042] The above description is only a preferred embodiment of the present invention and does not limit the present invention in other forms. Any person skilled in the art can make appropriate modifications or substitutions based on the above technical content, and these modifications or substitutions are within the protection scope of the present invention.

Claims

1. A multi-modal detection method and system for tunnel concrete lining, including a multi-modal sensor group and a data processing unit, characterized in that: The multi-modal sensor group includes a ground penetrating radar antenna (3), an infrared thermal imager (1) and an ultrasonic detection module (2), which are used to obtain the internal structure information, surface temperature distribution, lining thickness and crack conditions of the tunnel lining respectively; each sensor of the multi-modal sensor group is installed on a telescopic electric push rod (4-1, 4-2, 4-3), and the support devices (6-1, 6-2, 6-3) adjust the angle of the sensor group through motor-driven rotary joints (5-1, 5-2, 5-3) to adapt to different detection requirements; the data processing unit (7) is electrically connected to the multi-modal sensor group, collects and fuses the detection data from different sensors, and analyzes the data using deep learning algorithms.

2. The multi-modal detection system for tunnel concrete lining according to claim 1, wherein: The ground penetrating radar antenna (3) is used to detect the internal structure of the lining and compare data with the ultrasonic detection module (2).

3. The multi-modal detection system for tunnel concrete lining according to claim 1, characterized in that: The infrared thermal imager (1) is used to detect abnormal surface temperature of the lining and perform temperature field analysis through the data processing unit (7).

4. The multi-modal detection system for tunnel concrete lining according to claim 1, wherein: The data processing unit (7) adopts a multi-modal data fusion algorithm to comprehensively analyze the data from the ground penetrating radar antenna (3), the infrared thermal imager (1) and the ultrasonic detection module (2); the data processing unit (7) supports the wireless data transmission function.

5. According to the tunnel concrete lining multi-modal detection system described in claim 1, the ground penetrating radar (3), the infrared thermal imager (1) and the ultrasonic detection module (2) respectively transmit the original signals to the FPGA development board through high-speed interfaces, and signal preprocessing, filtering and preliminary feature extraction are carried out on this board. The processed data is then transmitted to the NVIDIA Jetson development board in real time through a high-speed data bus.

6. The multimodal data fusion algorithm according to claim 4, wherein: Including data preprocessing and spatio-temporal alignment, feature-level fusion and defect detection algorithms; In the data preprocessing and spatio-temporal alignment, a linear interpolation method is used to synchronize the time of data in different modalities, and homogeneous coordinate transformation is used to ensure that the data of all sensors are expressed in the same three-dimensional space coordinates. The feature-level fusion uses a Transformer network to extract and fuse features from ground penetrating radar, infrared thermal imaging, and ultrasonic data. Specifically, the ground penetrating radar, infrared thermal imaging, and ultrasonic data respectively extract features through a CNN to form their respective feature vectors X IR , X US , X GPR . The Transformer calculates the mutual attention between different modality data, extracts the cross-modal key features, and outputs a new fused feature X fused And fuse the confidence of different sensors through the Dempster-Shafer evidence theory.

7. The defect detection algorithm according to claim 6, wherein: An improved DBSCAN clustering algorithm, namely the VDBSCAN algorithm, is adopted, and data is processed based on distributed computing to improve the speed and accuracy of defect detection; specifically, it includes: dividing the data set D into multiple subsets D1, D2, …, D m , executing the DBSCAN algorithm in parallel, and after calculating the clustering results of each subset, merging the results of each subset to obtain the global clustering result, that is: The VDBSCAN algorithm can adapt to data with different density distributions. VDBSCAN ensures that low-density regions can also be correctly identified by adaptively adjusting the neighborhood radius ε and the core point density threshold MinPts. The formula for selecting the adaptive neighborhood radius ε is: where ε i is the adaptive neighborhood radius of the data point x i . d(x i , x j ) represents the Euclidean distance between data points. k is the number of nearest neighbors, usually set to MinPts. The MinPts is calculated through local density, and the formula is: Where N is the total number of samples in the data set. δ is a small constant to prevent division by zero errors. A larger MinPts is used in high-density regions (intact concrete), and a smaller MinPts is used in low-density regions (such as cracks).

8. The multi-modal detection system for tunnel concrete lining according to claim 1, wherein: The support devices (6-1, 6-2, 6-3) and the rotary joints (5-1, 5-2, 5-3) can adjust the height and angle respectively.

9. The multi-modal detection system for tunnel concrete lining according to claim 1, wherein: The electric push rod (4) is equipped with a laser ranging module. The laser ranging module obtains the gap data between the sensor and the detection surface and feeds back the measurement result to the control system in real time. The control system dynamically adjusts the telescopic amount of the electric push rod according to the ranging data and monitors the local uneven changes of the wall surface at the same time.

Images