Detection device suitable for tunnel lining quality and use method

Abstract

The invention discloses a detection device suitable for tunnel lining quality and a using method, the device comprises a base, a stand column, a control host, a processing host, a cross beam, a T-shaped swing arm, an ultrasonic sensor, a motor and other parts, non-contact detection is performed on a tunnel lining through the ultrasonic sensor, and damage to the structure by a traditional drilling coring method is avoided. The use method comprises the steps of device preparation, calibration before detection, detection operation and data acquisition, data analysis, result display and the like, and tunnel lining thickness calculation and intelligent defect identification are realized by utilizing multi-modal data fusion, curved surface adaptive fitting and a deep learning algorithm. The motor drives the T-shaped swing arm to drive the ultrasonic sensor to swing back and forth to form a fan-shaped scanning surface, the detection efficiency and accuracy are improved, meanwhile, real-time data feedback and automatic analysis are achieved, the device is suitable for narrow space operation, and the intelligent level of tunnel lining quality detection is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering inspection applications, and in particular to an inspection device and method for testing the quality of tunnel lining. Background Technology

[0002] As a support structure, the quality of tunnel lining directly affects tunnel safety and service life. Currently, the industry commonly uses methods such as ground-penetrating radar (GPR), percussion testing, and core drilling for inspection. GPR detects internal defects through electromagnetic wave reflection imaging, but is limited by high equipment costs and complex operation; percussion testing relies on manual experience to determine hollow areas, resulting in low efficiency and high subjectivity; while core drilling provides accurate results, it is a destructive method and cannot comprehensively cover the inspection area. In recent years, with the expansion of tunnel construction, the demand for efficient, accurate, and non-destructive inspection technologies has become increasingly urgent.

[0003] Common methods for inspecting tunnel lining quality cannot balance efficiency, accuracy, and economy. In particular, there is a lack of portable equipment suitable for confined spaces (such as subway shield tunnels), and most technologies are difficult to achieve real-time data feedback and automated analysis. Furthermore, destructive testing can easily lead to damage to the lining structure, which cannot meet the working requirements of tunnel engineering inspection applications. Therefore, a testing device and method for tunnel lining quality are proposed. Summary of the Invention

[0004] This invention provides the following technical solution: a testing device suitable for the quality of tunnel lining, comprising: The base has columns connected to its four corners on its upper surface, and top seats are installed on the top of each column. A support frame is connected to the inner side of each column, and a control host is installed on the top of the support frame. The control host integrates a signal generator and a data acquisition card. The omnidirectional wheels are installed at the four bottom corners of the base. The processing host is installed on the top rear side of the base. The top of the processing host is equipped with a display. The processing host is equipped with a data analysis module, which has a built-in thickness calculation algorithm and defect identification model. A crossbeam is welded to the top center of the top seat. A mounting platform is installed on the top of the crossbeam. A first hinge seat is installed on the front side of the top of the mounting platform. A T-shaped swing arm is inserted into the first hinge seat through a rotating shaft. An ultrasonic sensor is installed on the front side of the T-shaped swing arm, and a gear is coaxially connected to the rear of the first hinge seat at a position corresponding to the rear end of the T-shaped swing arm pivot. The motor is installed at the rear of the mounting platform. A second hinge seat is installed on the top rear side of the mounting platform. A transmission rod is coaxially connected to the front end of the motor rotor. A connecting frame is installed on the top outer side of the mounting platform. A transmission frame is installed on the top front side of the connecting frame via a bearing. A sector tooth is installed on the upper front side of the transmission frame. The transmission frame has a transmission groove inside, and a plug rod is connected to the outer front part of the transmission rod.

[0005] This invention provides a method for using a tunnel lining quality testing device. Based on the above-mentioned testing device suitable for tunnel lining quality, the method includes the following steps: S1 device preparation: Move the device to the designated detection position using the casters at the bottom of the base; S2 pre-test calibration: Start the control host and processing host, perform system self-test, and confirm that the signal generator, data acquisition card, data analysis module and display are working properly. Then start the ultrasonic sensor to perform calibration scan. S3 testing operation and data acquisition: The motor is started, and the ultrasonic sensor on the T-shaped swing arm is driven to reciprocate and scan through the transmission rod and transmission frame. The host control collects ultrasonic signals in real time and performs preliminary processing through the data acquisition card. The processing host further processes the collected data through the data analysis module, including thickness calculation and defect identification. S4 Data Analysis: The processing host integrates ultrasonic testing data with tunnel lining design drawings and construction records through a multimodal data fusion module. It also automatically fits the inner surface shape of the tunnel lining based on the ultrasonic testing data using a surface adaptive fitting algorithm. Simultaneously, it uses a deep learning defect identification algorithm to intelligently analyze the fused data and automatically identify and classify the defect types of the tunnel lining. S5 Results Display and Report Generation: The host computer displays the analysis results as a 3D model on the monitor and generates a detailed inspection report based on the defect identification results.

[0006] Preferably, each of the columns is welded with a reinforcing column, and the number of the reinforcing columns is 6-12 sets. The top of the top seat is equipped with handles on both the front and rear sides, and the top of the T-shaped swing arm is arc-shaped.

[0007] Preferably, a data cable is inserted into the interface of the ultrasonic sensor, and a wire spool is connected to the middle of the front side of the crossbeam. The data cable passes through the inside of the wire spool, and the number of ultrasonic sensors is 3-6 sets.

[0008] Preferably, the length of the data cable is 2-4 times the distance between the ultrasonic sensor and the control host, and the bottom interface of the data cable is connected to the data receiving port of the control host.

[0009] Preferably, the data interfaces of the control host and the processing host are connected by wires, the gear meshes with the sector teeth, the inside of the mounting platform is provided with a through slot corresponding to the position of the transmission frame, and the inside of the processing host is equipped with a multimodal data fusion module.

[0010] Preferably, in step S2, after the control host and processing host complete the system self-test, a multi-level verification mechanism is used to perform benchmark calibration on a standard test block of known thickness to establish the correspondence between ultrasonic propagation time and thickness. Then, 3-5 locations are selected on the surface of the tunnel lining to be tested for local calibration. Finally, the ultrasonic sensor channels are verified one by one through the self-diagnostic program built into the control host.

[0011] Preferably, the multimodal data fusion in step S4 adopts spatiotemporal alignment technology, which matches the real-time acquired ultrasonic data with the spatial coordinate system of the design drawings and eliminates measurement errors through feature point registration algorithm.

[0012] Preferably, the surface adaptive fitting algorithm in step S4 includes curvature continuity optimization processing, that is, by analyzing the time delay characteristics of ultrasonic reflection signals, the microscopic geometry of the inner surface of the lining is reconstructed, and for areas with incomplete detection data, an interpolation algorithm based on physical constraints is used for inference.

[0013] Preferably, the deep learning defect identification algorithm in step S4 adopts a multi-scale feature extraction architecture, that is, it processes the original waveform, spectral features and time-frequency analysis results separately through parallel convolutional neural networks, and then fuses the feature representations at different levels through an attention mechanism.

[0014] In summary, compared with the prior art, the present invention provides a detection device and method for tunnel lining quality, which has the following beneficial effects: 1. This invention enables ultrasonic testing of tunnel lining structures using an added ultrasonic sensor. The signal generator and data acquisition card inside the control unit process the data collected by the ultrasonic sensor, and the thickness calculation algorithm and defect identification model inside the processing unit further process the data to obtain the final test data. This avoids destructive damage to the lining structure, solves the structural damage problem caused by traditional core drilling, ensures the integrity of the tunnel support system, and, through multi-probe collaborative scanning and surface adaptive fitting algorithm, can accurately measure the lining thickness and identify internal defects. This overcomes the shortcomings of manual tapping method, which is highly subjective and has large errors, and the test results are more objective and reliable. The added control unit and processing unit can process ultrasonic signals in real time and automatically generate thickness calculation and defect classification results through the data analysis module, avoiding the lag of traditional methods that rely on manual interpretation, and realizing intelligent and real-time testing. 2. This invention uses a motor to drive the rotation of the transmission rod and the insertion rod, which in turn drives the transmission frame and the sector teeth to swing through the transmission groove. This, in turn, drives the T-shaped swing arm to swing left and right through the gears. This swinging motion allows the ultrasonic sensor to form a sector-shaped scanning surface during ultrasonic sensor detection. A single swing can cover the transverse cross-section of the tunnel lining, reducing the number of device movements. Furthermore, the swinging motion allows the ultrasonic sensor to scan the lining surface at different angles, enabling the acquisition of more comprehensive reflected signal data. Combined with multi-probe collaborative work, it can effectively reduce detection blind spots and improve the accuracy of identifying defects such as voids and cracks. At the same time, the swinging ultrasonic sensor can better conform to the curved shape of the tunnel lining, ensuring that the ultrasonic signal is incident perpendicularly, reducing signal attenuation or misjudgment, and improving the reliability of thickness measurement. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the structure of the present invention.

[0016] Figure 2 This is a schematic diagram of the host structure of the present invention.

[0017] Figure 3 This is a schematic diagram of the beam structure of the present invention.

[0018] Figure 4 This is a schematic diagram of the top structure of the mounting platform of the present invention.

[0019] Figure 5 This is a schematic diagram of the connecting frame structure of the present invention.

[0020] Figure 6 This is a schematic diagram of the transmission frame structure of the present invention.

[0021] Figure 7 This is a flowchart of the method of the present invention.

[0022] Explanation of reference numerals in the attached figures: 1. Base; 2. Column; 3. Top mount; 4. Support frame; 5. Control host; 6. Casters; 7. Processing host; 8. Display; 9. Reinforcing column; 10. Crossbeam; 11. Cable spool; 12. Mounting platform; 13. First hinge seat; 14. T-shaped swing arm; 15. Gear; 16. Ultrasonic sensor; 17. Data cable; 18. Through slot; 19. Connecting frame; 20. Motor; 21. Second hinge seat; 22. Sector gear; 23. Transmission rod; 24. Transmission frame; 25. Transmission groove; 26. Insert rod; 27. Handle. Detailed Implementation

[0023] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] This invention provides a technical solution: a detection device suitable for tunnel lining quality, comprising a base 1, a column 2, a top seat 3, a support frame 4, a control host 5, casters 6, a processing host 7, a display 8, a reinforcing column 9, a crossbeam 10, a cable reel 11, a mounting platform 12, a first hinge seat 13, a T-shaped swing arm 14, a gear 15, an ultrasonic sensor 16, a data cable 17, a through groove 18, a connecting frame 19, a motor 20, a second hinge seat 21, a sector tooth 22, a transmission rod 23, a transmission frame 24, a transmission groove 25, an insertion rod 26, and a handle 27. Please see Figure 1 The base 1 has four columns 2 connected to the upper surface of each of the four corners. Each column 2 has a top seat 3 installed at the top. The inner side of the column 2 is connected to a support frame 4. The top of the support frame 4 is equipped with a control host 5. The control host 5 integrates a signal generator and a data acquisition card. Each column 2 has a reinforcing column 9 welded between it. The number of reinforcing columns 9 is 6-12. The top of the top seat 3 has handles 27 installed on both the front and rear sides. The top of the T-shaped swing arm 14 is arc-shaped. The casters 6 are mounted at the four bottom corners of the base 1. Please refer to [link / reference]. Figure 2 The processing host 7 is installed on the top rear side of the base 1. The top of the processing host 7 is equipped with a display 8. The processing host 7 is equipped with a data analysis module, which has a built-in thickness calculation algorithm and defect identification model. Please see Figure 3 The crossbeam 10 is welded to the top center of the top seat 3. A mounting platform 12 is installed on the top of the crossbeam 10. (See attached image.) Figure 4 and Figure 5A first hinge seat 13 is installed on the top front side of the mounting platform 12, and a T-shaped swing arm 14 is inserted into the first hinge seat 13 through a rotating shaft. An ultrasonic sensor 16 is installed on the front side of the T-shaped swing arm 14. A gear 15 is coaxially connected to the rear of the first hinge seat 13 at a position corresponding to the rear end of the T-shaped swing arm 14 shaft. A data cable 17 is inserted into the interface of the ultrasonic sensor 16. A wire drum 11 is connected to the middle of the front side of the crossbeam 10. The data cable 17 passes through the inside of the wire drum 11. There are 3-6 sets of ultrasonic sensors 16. The length of the data cable 17 is 2-4 times the distance between the ultrasonic sensor 16 and the control host 5. The bottom interface of the data cable 17 is connected to the data receiving port of the control host 5. Please see Figure 6 The motor 20 is installed at the rear of the mounting platform 12. A second hinge seat 21 is installed on the rear top of the mounting platform 12. A transmission rod 23 is coaxially connected to the front end of the rotor of the motor 20. A connecting frame 19 is installed on the outer top of the mounting platform 12. A transmission frame 24 is installed on the front top of the connecting frame 19 through a bearing. A sector tooth 22 is installed on the upper front side of the transmission frame 24. The transmission frame 24 has a transmission groove 25 inside. A plug rod 26 is connected to the outer front side of the transmission rod 23. The data interface of the control host 5 and the processing host 7 is connected by a wire. The gear 15 meshes with the sector tooth 22. A through groove 18 is opened inside the mounting platform 12 at a position corresponding to the transmission frame 24. A multimodal data fusion module is installed inside the processing host 7.

[0025] Please see Figure 7 This invention provides a method for using a tunnel lining quality testing device. Based on the above-mentioned testing device suitable for tunnel lining quality, the method includes the following steps: S1 device preparation: The device is moved to the designated detection position using the casters 6 at the bottom of the base 1; S2 pre-test calibration: The control host 5 and processing host 7 are started to perform system self-test, confirming that the signal generator, data acquisition card, data analysis module and display 8 are working properly. Then, the ultrasonic sensor 16 is started to perform calibration scan. After the control host 5 and processing host 7 complete the system self-test, a multi-level verification mechanism is used to perform benchmark calibration on a standard test block of known thickness to establish the correspondence between ultrasonic propagation time and thickness. Then, 3-5 positions are selected on the surface of the tunnel lining to be tested for local calibration. Finally, the self-diagnostic program built into the control host 5 is used to verify each channel of the ultrasonic sensor 16. S3 testing operation and data acquisition: The motor 20 is started, which drives the ultrasonic sensor 16 on the T-shaped swing arm 14 to reciprocate and scan through the transmission rod 23 and the transmission frame 24. The host 5 collects ultrasonic signals in real time and performs preliminary processing through the data acquisition card. The processing host 7 further processes the collected data through the data analysis module, including thickness calculation and defect identification. S4 Data Analysis: The processing host 7 integrates ultrasonic detection data with tunnel lining design drawings and construction record data through a multimodal data fusion module. It also automatically fits the inner surface shape of the tunnel lining based on the ultrasonic detection data using a surface adaptive fitting algorithm. At the same time, it uses a deep learning defect recognition algorithm to intelligently analyze the fused data and automatically identify and classify the defect types of the tunnel lining. Multimodal data fusion employs spatiotemporal alignment technology. This technology matches real-time acquired ultrasonic data with the spatial coordinate system of the design drawings and eliminates measurement errors through a feature point registration algorithm. In the spatial benchmark establishment stage, the system first loads the original design drawing data of the tunnel lining. These drawings contain the planar coordinates and elevation information of the tunnel axis, as well as detailed geometric parameters of the lining cross-section. The processing host 7 uses a built-in coordinate transformation module to convert the two-dimensional CAD coordinate system of the design drawings into a three-dimensional engineering coordinate system suitable for on-site inspection. Simultaneously, the system establishes a correspondence between the ultrasonic sensor array 16 and the tunnel mileage markers based on the initial positioning information of the detection device. During this process, the system intelligently identifies key features of the tunnel, such as construction joints and expansion joints, as reference points for subsequent data matching. The time series synchronization stage mainly addresses the time dimension alignment of the detection data. The system uses a high-precision clock synchronization mechanism to ensure that the timestamps of the ultrasonic acquisition data strictly correspond to the device's motion trajectory records. When the motor 20 drives the T-shaped swing arm 14 to perform reciprocating scanning, the system records the precise moment of each ultrasonic pulse emission in real time and establishes a correlation with the angular displacement sensor data of the transmission mechanism. This spatiotemporal correlation mechanism ensures that each set of detection data accurately corresponds to the specific spatial location of the tunnel lining, laying the foundation for subsequent multi-source data fusion. Feature point matching and calibration is the core of this technology. The system first extracts significant feature points from the ultrasonic detection data, such as obvious reflective interfaces on the lining surface and points of abrupt changes in material properties. These feature points are intelligently matched with key structural nodes in the design drawings. During the matching process, the system employs an adaptive search algorithm to find the optimal spatial correspondence while considering allowable errors in tunnel construction. For important structural parts, the system performs multiple verifications to ensure the reliability of the matching results. This feature point registration effectively eliminates spatial position deviations caused by construction deviations, equipment installation errors, and other factors. The error compensation optimization process further improves the accuracy of data fusion. The system establishes a dynamic error compensation model, calculating the error distribution pattern of the entire detection area based on the known deviations of the matching points. For minor deviations in local areas, the system uses a smooth interpolation method for correction; for systematic deviations, it performs overall adjustment through coordinate transformation. Simultaneously, the system combines actual measurement data from construction records to make necessary corrections to the theoretical coordinates of the design drawings, achieving the best fusion effect for the multi-source data. The entire fusion process achieved millimeter-level spatial alignment accuracy, providing a reliable data foundation for subsequent defect identification; The adaptive surface fitting algorithm includes curvature continuity optimization processing. This involves reconstructing the microscopic geometry of the lining's inner surface by analyzing the time delay characteristics of the ultrasonic reflection signal. For areas with incomplete detection data, a physical constraint-based interpolation algorithm is used for estimation. In the raw data preprocessing stage, the system first extracts the time delay features of the acquired ultrasonic reflection signals. After digital filtering, the reflected signals received by each ultrasonic sensor 16 accurately identify the reflection interface positions of each medium layer. By calculating the time difference between ultrasonic wave transmission and reception, and combining this with the material's sound velocity characteristics, the system constructs a preliminary ranging data point cloud. These data points record the spatial location information of each detection point on the lining's inner surface. During preprocessing, the system removes outliers, eliminating erroneous ranging data caused by signal interference or equipment noise, ensuring the data quality for subsequent reconstruction. Curvature continuity optimization reconstruction is the core of this algorithm. The system employs a progressive surface generation strategy, gradually constructing a complete lining inner surface model starting from the initial detection points. During reconstruction, the system pays particular attention to maintaining the geometric continuity of the surface, using intelligent algorithms to ensure smooth transitions between adjacent areas. For areas with significant curvature variations, such as the tunnel arch, the system automatically increases sampling density to improve reconstruction accuracy. Simultaneously, the algorithm references the theoretical surface shape in the design drawings, optimizing the reconstructed surface while ensuring the accuracy of the measured data. This dual constraint based on measured data and engineering experience ensures that the final reconstructed surface conforms to both the actual detection results and the structural rationality requirements. For areas with incomplete detection data, the system employs an intelligent compensation mechanism based on physical constraints. When data loss due to occlusion or signal attenuation is detected in certain areas, the system first analyzes the distribution characteristics of surrounding valid data and establishes a local geometric trend model. Then, based on the mechanical properties and material properties of the tunnel structure, reasonable physical constraints are applied, such as maximum allowable curvature and minimum structural thickness. Under these constraints, the system uses an adaptive interpolation algorithm to infer the surface morphology of the missing areas. For critical structural components, the system provides a data reliability assessment, alerting engineers to potential uncertainties. The deep learning defect identification algorithm adopts a multi-scale feature extraction architecture, that is, it processes the original waveform, spectral features and time-frequency analysis results through parallel convolutional neural networks, and then fuses the feature representations at different levels through an attention mechanism. The specific process of the above method is as follows: Original waveform processing branch: This branch of the network receives the original ultrasonic signal as input and extracts local and global features from the signal through a series of convolutional layers, pooling layers and activation functions. The convolutional layers are responsible for capturing local patterns in the signal, the pooling layers reduce the dimensionality of the features, and the activation functions introduce nonlinearity to enhance the network's expressive power. Spectral Feature Processing Branch: This branch network receives the spectral features of the ultrasonic signal as input. The spectral features are extracted from the original signal using methods such as Fourier transform, reflecting the signal's characteristics in the frequency domain. Similarly, through convolutional layers, pooling layers, and activation functions, useful information in the spectral features is extracted. This branch network can capture specific patterns of the signal in the frequency domain, playing an important role in identifying certain types of defects. Time-frequency analysis result processing branch: This branch network receives the time-frequency analysis results of the ultrasonic signal as input. The time-frequency analysis results are obtained through methods such as wavelet transform, and simultaneously reflect the characteristics of the signal in the time and frequency domains. Through convolutional layers, pooling layers, and activation functions, multi-scale features in the time-frequency analysis results are extracted. This branch network can capture the joint features of the signal in the time and frequency domains, which is of great significance for identifying complex and variable defect patterns. After the parallel convolutional neural network processes the features of the three branches, the algorithm fuses these feature representations from different levels through an attention mechanism. The attention mechanism dynamically weights inputs from different features or different time steps to highlight important information and suppress noise and redundant information. The algorithm first calculates the attention weight for each feature branch. These weights reflect the importance of different feature branches in defect identification. Then, the features of the three branches are weighted and fused according to the attention weights to obtain a comprehensive feature representation. This step can effectively integrate feature information at different levels and scales, improving the accuracy and robustness of defect identification. Finally, the algorithm inputs the fused feature representation into a fully connected layer and a classifier to achieve automatic defect identification and classification. The fully connected layer further integrates and transforms the fused features to obtain a high-dimensional feature vector. The classifier then outputs the defect type and confidence level based on this feature vector. The classifier can use methods such as the soft maximum function to implement multi-classification tasks. S5 Results Display and Report Generation: The host computer 7 displays the analysis results as a three-dimensional model on the monitor 8, and generates a detailed inspection report based on the defect identification results.

[0026] This solution utilizes an added ultrasonic sensor 16 to perform ultrasonic testing on the tunnel lining structure. The signal generator and data acquisition card inside the control host 5 process the data collected by the ultrasonic sensor 16, and the thickness calculation algorithm and defect identification model inside the processing host 7 further process the data to obtain the final test data. This avoids destructive damage to the lining structure, solves the structural damage problem caused by traditional core drilling methods, and ensures the integrity of the tunnel support system. Furthermore, through multi-probe collaborative scanning and surface adaptive fitting algorithms, the lining thickness can be accurately measured and internal defects can be identified, overcoming the subjectivity and large error of manual tapping methods. The test results are more objective and reliable. The added control host 5 and processing host 7 can process ultrasonic signals in real time and automatically generate thickness calculation and defect classification results through the data analysis module, avoiding the lag of traditional methods that rely on manual interpretation, and realizing intelligent and real-time testing.

[0027] This solution uses a motor 20 to drive the rotation of the transmission rod 23 and the insertion rod 26, which in turn drives the transmission frame 24 and the sector tooth 22 to swing through the transmission groove 25. This, in turn, drives the T-shaped swing arm 14 to swing back and forth through the gear 15. This swinging motion allows the ultrasonic sensor 16 to form a sector scanning surface during detection. A single swing can cover the transverse cross-section of the tunnel lining, reducing the number of device movements. Furthermore, the swinging motion allows the ultrasonic sensor 16 to scan the lining surface at different angles, enabling the acquisition of more comprehensive reflected signal data. Combined with the collaborative work of multiple probes, this effectively reduces blind spots and improves the accuracy of identifying defects such as voids and cracks. At the same time, the swinging ultrasonic sensor 16 can better conform to the curved shape of the tunnel lining, ensuring that the ultrasonic signal is incident perpendicularly, reducing signal attenuation or misjudgment, and improving the reliability of thickness measurement.

[0028] 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.

[0029] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A device for detecting the quality of a tunnel lining, characterized in that, Include: The upper surface of the base (1) is connected with the column (2) at four corners, the top of the column (2) is installed with the top seat (3), the inner side of the column (2) is connected with the support frame (4), the top of the support frame (4) is installed with the control host (5), the inside of the control host (5) is integrated with signal generator and data acquisition card; The universal wheel (6) is installed at the bottom of the base (1) at four corners, the top of the base (1) is installed with the processing host (7) at the rear side, the top of the processing host (7) is provided with the display (8), the inside of the processing host (7) is installed with data analysis module, the data analysis module is built-in thickness calculation algorithm and defect identification model; The crossbeam (10) is welded at the top of the top seat (3) at the middle position, the top of the crossbeam (10) is installed with the mounting table (12), the top of the mounting table (12) is installed with the first hinged seat (13) at the front side, the inside of the first hinged seat (13) is inserted with T-shaped swing arm (14) through the rotating shaft; The ultrasonic sensor (16) is installed at the front side of the T-shaped swing arm (14), the rear part of the first hinged seat (13) is coaxially connected with the gear (15) corresponding to the rear end of the rotating shaft of the T-shaped swing arm (14); The motor (20) is installed at the rear part of the mounting table (12), the top of the mounting table (12) is installed with the second hinged seat (21) at the rear side, the front end of the rotor of the motor (20) is coaxially connected with the transmission rod (23), the top of the mounting table (12) is installed with the connecting frame (19) at the outer side, the top of the connecting frame (19) is installed with the transmission frame (24) through the bearing at the front side; The sector tooth (22) is installed at the front part of the top side of the transmission frame (24), the inside of the transmission frame (24) is provided with the transmission groove (25), the front part of the outer side of the transmission rod (23) is connected with the plug rod (26).

2. The device for detecting the quality of tunnel lining according to claim 1, characterized in that: The reinforcing column (9) is welded between the columns (2), the number of the reinforcing column (9) is 6-12 groups, the top of the top seat (3) is installed with the handle (27) at the front and rear sides, the top of the T-shaped swing arm (14) is arc-shaped.

3. The device for detecting the quality of tunnel lining according to claim 1, characterized in that: The interface of the ultrasonic sensor (16) is inserted with the data line (17), the front side of the crossbeam (10) is connected with the wire cylinder (11), the data line (17) penetrates the inside of the wire cylinder (11), the number of the ultrasonic sensor (16) is 3-6 groups.

4. The device for detecting the quality of tunnel lining according to claim 3, characterized in that: The length of the data line (17) is 2-4 times of the distance between the ultrasonic sensor (16) and the control host (5), the bottom interface of the data line (17) is connected with the data receiving socket of the control host (5).

5. The device for detecting the quality of tunnel lining according to claim 1, characterized in that: The data interfaces of the control host (5) and the processing host (7) are connected through wires, the gear (15) is engaged with the sector tooth (22), the inside of the mounting table (12) is provided with the through slot (18) corresponding to the position of the transmission frame (24), the inside of the processing host (7) is installed with the multi-modal data fusion module.

6. A method of using a device for detecting the quality of a tunnel lining, based on the device for detecting the quality of a tunnel lining according to any one of claims 1-5, characterized in that, Include the following steps: S1 device preparation: The device is moved to the designated detection position by the universal wheels (6) at the bottom of the base (1); S2 pre-detection calibration: Start the control host (5) and the processing host (7), perform system self-checking, confirm that the signal generator, the data acquisition card, the data analysis module and the display (8) are working properly, and then start the ultrasonic sensor (16) to perform calibration scanning; S3 detection operation and data acquisition: Start the motor (20), drive the ultrasonic sensor (16) on the T-shaped swing arm (14) to swing back and forth through the transmission rod (23) and the transmission frame (24), control the host (5) to collect ultrasonic signals in real time, and perform preliminary processing through the data acquisition card, and the processing host (7) further processes the collected data through the data analysis module, including thickness calculation and defect identification; S4 data analysis: The processing host (7) fuses the ultrasonic detection data with the design drawing data and construction record data of the tunnel lining through the multi-modal data fusion module, and automatically fits the inner surface shape of the tunnel lining according to the ultrasonic detection data through the curved surface adaptive fitting algorithm, and intelligently analyzes the fused data through the deep learning defect identification algorithm, automatically identifies and classifies the defect types of the tunnel lining; S5 result display and report generation: The analysis results are displayed on the display (8) in the form of a three-dimensional model through the processing host (7), and a detailed detection report is generated according to the defect identification results.

7. A method of using a tunnel lining quality detection apparatus according to claim 6, wherein: In step S2, after the control host (5) and the processing host (7) perform system self-checking, a multi-level verification mechanism is adopted to calibrate the reference on the standard test block with known thickness, establish the corresponding relationship between the ultrasonic propagation time and the thickness, then select 3-5 positions on the surface of the tunnel lining to be detected for local calibration, and finally check the ultrasonic sensor (16) channel one by one through the self-diagnosis program built in the control host (5).

8. The method of using a tunnel lining quality detection device of claim 6, wherein: The multi-modal data fusion in step S4 adopts a space-time alignment technology, which matches the real-time collected ultrasonic data with the spatial coordinate system of the design drawing, and eliminates measurement errors through a feature point registration algorithm.

9. The method of using a tunnel lining quality detection device of claim 6, wherein: The curved surface adaptive fitting algorithm in step S4 includes curvature continuity optimization processing, that is, by analyzing the time delay characteristics of the ultrasonic reflection signal, the micro-geometric shape of the lining inner surface is reconstructed, and for the area where the detection data is incomplete, an interpolation algorithm based on physical constraints is used for speculation.

10. The method of using a tunnel lining quality detection device of claim 6, wherein: The deep learning defect identification algorithm in step S4 adopts a multi-scale feature extraction architecture, that is, the original waveform, spectral features and time-frequency analysis results are processed by a parallel convolutional neural network respectively, and then different levels of feature representations are fused through an attention mechanism.

Images