Tunnel lining internal defect detection method and system based on radar antenna

Abstract

This invention discloses a method and system for detecting internal defects in tunnel lining based on a radar antenna, belonging to the field of data measurement technology. The method includes: S1, setting a guide rail on the lining surface to couple a radar antenna to move at a constant speed; S2, triggering multi-frequency electromagnetic wave transmission at a fixed spatial interval and receiving the echo; S3, setting a reference medium block to collect calibration signals and calculate the gain coefficient; S4, applying gain compensation and separating the dataset by frequency; S5, determining the depth of the structural reflection interface based on the first dataset; S6, using this depth as a benchmark to extract candidate anomaly clusters within a preset depth range in the second dataset; S7, calculating the spatial energy concentration E and waveform phase consistency P index of each cluster, and jointly determining the defect. This invention has the advantages of stable constraints, multi-frequency defect identification, and anti-interference.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data measurement, in particular to a tunnel lining internal defect detection method and system based on a radar antenna. BACKGROUND

[0002] In the application of geological radar technology in tunnel engineering lining, especially in the non-destructive testing of defects, the physical coupling state (including distance, contact pressure, pitch angle, etc.) of the existing handheld antenna scanning antenna and the secondary lining inner surface changes uncontrollably due to differences in manual operation and the unevenness of the lining surface, resulting in random fluctuations in electromagnetic wave transmission efficiency and receiving sensitivity, unpredictable attenuation or distortion of the echo signal amplitude, not only reducing the defect recognition accuracy, but also making the data of different sections and different secondary detection lose comparability due to coupling differences.

[0003] Secondly, the strong reflection signals generated by the multi-layer complex medium interface of the lining internal reinforcement mesh, surrounding rock contact surface and construction joint will form strong event interference in the radar profile. The energy of these structural reflection signals far exceeds that of the weak echo signals of small defects such as cavities and loose bands, not only covering the defect characteristics, but also possibly causing false anomalies due to the Fresnel zone effect when detecting at high frequencies; while low-frequency detection can identify deep structural interfaces, it sacrifices the resolution ability for shallow small defects.

[0004] Although the existing defect recognition method attempts to improve the signal-to-noise ratio through time-frequency analysis, clutter suppression and other post-processing algorithms, it fails to solve the coupling stability problem at the signal collection source and does not effectively integrate the complementary advantages of multi-frequency data: when the coupling state is unstable, the extraction of amplitude-sensitive features (such as defect energy concentration) is severely distorted, and the robustness of phase consistency and other indicators also decreases significantly; at the same time, single-frequency data analysis cannot balance the depth positioning accuracy and shallow detail resolution, resulting in small defects (such as voids) behind deep structural interfaces being easily overwhelmed by high-energy interface reflections or random noise, causing missed detection or false alarms.

[0005] Therefore, the above-mentioned coupling uncontrollability and multi-scale signal interlacing complex problems restrict the reliable detection ability of geological radar for defects with a size less than 5cm in the range of 20-50cm lining thickness. SUMMARY

[0006] In view of the technical problems existing in the above background art, the present application provides a tunnel lining internal defect detection method and system based on a radar antenna.

[0007] A method for detecting internal defects in tunnel lining based on a radar antenna includes: S1, setting up a guide rail on the inner surface of the tunnel lining, coupling a radar antenna to the guide rail via a sliding bracket, and controlling the radar antenna to move at a constant speed along the guide rail while maintaining a preset posture; S2, triggering the radar antenna at fixed spatial intervals during its movement, controlling it to emit at least two detection electromagnetic waves with different center frequencies at different times or simultaneously, and synchronously receiving the corresponding echo signal sequences; S3, setting a reference medium block with known electromagnetic properties on the detection path of the guide rail, collecting calibration echo signals from the reference medium block when the radar antenna passes through it, and calculating compensation for the current coupling state based on a comparison between the amplitude of the calibration echo signal and a preset standard amplitude. S4. Apply the gain coefficient to the echo signal sequence for amplitude compensation and separate it by frequency to obtain at least a first echo dataset corresponding to a first center frequency and a second echo dataset corresponding to a second center frequency; S5. Based on the first echo dataset, determine the depth information of a major structural reflection interface inside the tunnel lining; S6. Using the depth information of the major structural reflection interface as a reference, extract candidate anomalous signal clusters located within a preset depth range behind the reference in the second echo dataset; S7. For each candidate anomalous signal cluster, calculate its spatial energy concentration index and waveform phase consistency index, and based on the joint judgment result of the two indices, determine whether there is a defect inside the tunnel lining.

[0008] Optionally, S1 further includes: the sliding bracket forms a three-dimensional constraint fit with the guide rail through at least three rollers, wherein the pressure of at least one roller is adjustable, and the radar antenna is mounted on the sliding bracket through an elastic damping element.

[0009] Optionally, S2 further includes: at least two different center frequencies, including a low-frequency band with a center frequency less than or equal to 500 MHz and a mid-frequency band with a center frequency between 800 MHz and 1.2 GHz, wherein the first echo dataset corresponds to the low-frequency band and the second echo dataset corresponds to the mid-frequency band.

[0010] Optionally, S3 further includes: the surface of the reference dielectric block is flush with the inner surface of the lining, and the gain coefficient is simultaneously applied to compensate for the first echo dataset and the second echo dataset.

[0011] Optionally, S5 further includes: inverting the depth distribution of the main structure reflection interface by performing hyperbolic fitting on the reflection phase axis with the strongest energy in the first echo dataset.

[0012] Optionally, the hyperbola fitting model is: Where t is the two-way travel time of the echo, x is the horizontal position coordinate of the antenna with the origin directly above the reflection point, d is the depth of the reflection interface, and v is the propagation speed of the electromagnetic wave in the lining.

[0013] Optionally, S6 further includes: the preset depth range is related to the design thickness of the tunnel lining, and the extraction process includes initial screening based on amplitude threshold and clustering based on spatial proximity.

[0014] Optionally, S7 further includes: the calculation method for the spatial energy concentration index E is as follows: Where s_k is the echo complex signal of the k-th sampling point within the candidate abnormal signal cluster, Σ represents the summation over all sampling points within the cluster, and L is the spatial length covered by the cluster; the waveform phase consistency index P is calculated as follows: , where s_k and s_{k+1} are the echo complex signals of two adjacent sampling points within the cluster, Re() represents taking the real part, |·| represents taking the modulus, and max represents taking the maximum value.

[0015] Optionally, the condition for determining a defect is that the spatial energy concentration index E is greater than or equal to the first threshold and the waveform phase consistency index P is greater than or equal to the second threshold at the same time.

[0016] A radar antenna-based tunnel lining internal defect detection system is also provided to implement the aforementioned radar antenna-based tunnel lining internal defect detection method. The system includes: a mechanically stable scanning module, which includes a guide rail, a sliding support, and a drive unit, for supporting and controlling the stable movement of the radar antenna; a multi-frequency radar control and acquisition module, for controlling the radar antenna to emit electromagnetic waves of at least two frequencies at fixed spatial intervals and receive echo signals; a signal processing module configured for steps S3, S4, S5, S6, and S7; and a result output module, for outputting defect information.

[0017] The beneficial effects of this invention are reflected in: In the entire method for detecting internal defects in tunnel lining based on radar antennas, firstly, the mechanical design of guide rails and three-dimensional constraint rollers, combined with elastic shock-absorbing installation, physically constrains the antenna attitude and movement trajectory, compressing the fluctuation amplitude of the antenna-lining coupling distance to the millimeter level and the angle deviation to less than 1°. This fundamentally suppresses random changes in coupling caused by human operation and surface unevenness, ensuring that the amplitude fluctuation of the echo signal originates from the internal differences of the medium rather than external interference, and significantly improving the stability and repeatability of the signal source. Furthermore, multi-frequency collaborative acquisition (e.g., 400MHz low frequency and 900MHz mid frequency) combined with a frequency band division processing strategy utilizes the penetrability of low frequencies to obtain accurate depth benchmarks for deep structural interfaces (e.g., inversion error of less than 3cm at the lining-surrounding rock interface). This then guides the mid-frequency data to focus and search in a pre-defined high-risk area (e.g., a depth window of 10-30cm behind the interface). This avoids the computational burden of blind searching across the entire profile and captures minute anomalies through high-frequency resolution. Simultaneously, online gain compensation (referencing PTFE block calibration) eliminates the impact of residual coupling fluctuations on amplitude, ensuring that both low-frequency benchmark positioning and mid-frequency defect signal extraction are based on reliable amplitude. Furthermore, the joint criterion of spatial energy concentration (E) and phase consistency (P) is calculated based on the compensated mid-frequency signal. The E index filters out low-energy diffuse interference (e.g., localized damp areas), and the P index filters out high-energy random noise (e.g., boulder reflections), retaining only candidate clusters with both energy focusing and waveform continuity (e.g., E≥1500V). 2 The void size of / m and P≥0.85 significantly reduces false alarms caused by clutter and noise, and improves the specificity of identifying small defects (such as 5cm voids). Attached Figure Description

[0018] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the accompanying drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0019] Figure 1 This is a schematic diagram illustrating the steps of the method for detecting internal defects in tunnel lining based on a radar antenna according to the present invention. Figure 2 This is a schematic diagram of a portion of the process of the method for detecting internal defects in tunnel lining based on a radar antenna according to the present invention. Figure 3 This is a schematic diagram of another part of the process of the method for detecting internal defects in tunnel lining based on radar antenna of the present invention. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, 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. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0021] Therefore, the following detailed description of the embodiments of the invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the invention without inventive effort are within the scope of protection of the invention.

[0022] It should be noted that similar labels and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, the terms "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0023] like Figure 1 , Figure 2 and Figure 3 As shown, a method for detecting internal defects in tunnel lining based on a radar antenna is provided. In one embodiment, the method includes: S1. A guide rail is laid on the inner surface of the tunnel lining, and the radar antenna is coupled to the guide rail through a sliding bracket. The radar antenna is controlled to move along the guide rail at a constant speed while maintaining a preset posture. S2. During the movement of the radar antenna, it is triggered at fixed spatial intervals to control it to transmit at least two detection electromagnetic waves with different center frequencies in a time-division or simultaneously, and to receive the corresponding echo signal sequence synchronously. S3. A reference medium block with known electromagnetic properties is set on the detection path of the guide rail. When the radar antenna passes by, the calibration echo signal from the reference medium block is collected. Based on the comparison between the amplitude of the calibration echo signal and the preset standard amplitude, the gain coefficient used to compensate for the current coupling state is calculated. S4. Apply the gain coefficient to the echo signal sequence to perform amplitude compensation and separate it by frequency to obtain at least a first echo dataset corresponding to a first center frequency and a second echo dataset corresponding to a second center frequency. S5. Based on the first echo dataset, determine the depth information of a major structural reflection interface inside the tunnel lining; S6. Based on the depth information of the main structural reflection interface, extract candidate abnormal signal clusters located within a preset depth range behind the reference in the second echo dataset. S7. For each candidate abnormal signal cluster, calculate its spatial energy concentration index and waveform phase consistency index, and based on the joint judgment result of the two indices, determine whether there are defects inside the tunnel lining.

[0024] In this embodiment, it should be noted that in S1, a guide rail is laid on the inner surface of the tunnel lining. The radar antenna is coupled to the rail via a sliding bracket and its uniform movement is controlled to maintain a preset attitude. This is a key mechanical measure to solve the problem of antenna coupling instability. The guide rail (e.g., an I-shaped cross-section) is attached to the inner wall of the lining by a reliable fixing method (e.g., bolts), providing a precise straight reference path for antenna movement.

[0025] Furthermore, the sliding support employs at least three roller mechanisms to form a three-dimensional spatial constraint with the track. The upper rollers clamp the sides of the track flanges, limiting lateral displacement and pitch, while the lower rollers support the bottom of the web, constraining its downward movement. This multi-point contact ensures that the support does not overturn or swing significantly during movement. In particular, the pressure of at least one roller can be adjusted through design (such as an eccentric shaft) to accommodate possible slight deformations or installation errors in the track, ensuring stable and controllable rolling friction and preventing the support from jamming due to excessive pressure or exacerbating vibration due to insufficient pressure. The radar antenna body is not directly and rigidly fixed to the support, but is installed through elastic damping elements (such as rubber pads of specific hardness and thickness) with appropriate stiffness and damping characteristics. This flexible connection can effectively absorb microscopic unevenness of the lining surface or high-frequency micro-impact vibrations transmitted from the drive, preventing such disturbances from being directly transmitted to the antenna body and causing attitude changes.

[0026] During operation, the drive unit controls the sliding bracket to move the antenna along the track at a pre-set constant speed. Simultaneously, the movement is monitored to ensure the antenna maintains its designed relative position and angle with the track (i.e., preset attitude) throughout the entire journey; for example, the antenna's bottom surface is parallel to the track plane and at a constant distance. This step physically eliminates human-induced jitter, uneven pressure, and distance variations caused by handheld scanning, significantly suppressing fluctuations in electromagnetic wave transmission efficiency and reception sensitivity caused by random changes in coupling state. This establishes a stable and reliable geometric and contact condition foundation for subsequent signal acquisition, ensuring the initial quality of the echo signal.

[0027] In S2, during the stable movement of the radar antenna, the radar is triggered at fixed spatial intervals to control the antenna to transmit at least two detection electromagnetic waves with different center frequencies at different times or simultaneously, and to synchronously receive the corresponding echo signal sequences. This is the core control logic for achieving complementary acquisition of multi-frequency data. The triggering mechanism is strictly synchronized with the movement displacement; for example, a trigger pulse is generated for every unit length of movement (specific value omitted) to ensure that the echo samples are uniformly distributed and accurately positioned in space.

[0028] Furthermore, the selection of transmission frequency combinations needs to balance depth detection capability and shallow resolution. This typically includes a low-frequency band (with a lower center frequency, such as 500MHz or less) and a mid-frequency band (with a higher center frequency, such as between 800MHz and 1.2GHz). Low-frequency electromagnetic waves have strong penetration capabilities but limited spatial resolution, making them suitable for detecting deep interfaces; mid-frequency electromagnetic waves have better detail resolution but are more sensitive to small anomalies in shallow layers. Upon triggering, the antenna can be controlled to radiate detection pulses at the selected center frequency sequentially or synchronously. Time-division multiplexing avoids crosstalk between frequency bands, while simultaneous transmission improves efficiency but requires isolation circuitry. Regardless of the transmission mode, each trigger synchronously receives and records the echo signal at the corresponding transmission frequency, forming a serialized dataset indexed by spatial location.

[0029] Furthermore, the key to S2 lies in the close integration of physical scanning and electronic triggering: uniform motion ensures spatial consistency of sampling intervals; fixed-interval triggering overcomes the spatial distortion of data caused by uneven manual scanning speeds; and multi-frequency coordinated emission covers the spectral range sensitive to defect features at different scales. The resulting multidimensional echo sequences are spatially precisely aligned, with each frequency component independent, containing rich information about the reflections of multiple layers of media within the tunnel lining from shallow to deep. This provides a foundation for subsequent data analysis in different frequency bands. Low-frequency data may capture strong reflections at deep structural interfaces, while mid-frequency data more easily reveals detailed features of shallow or minute dielectric anomalies.

[0030] In S3, a reference medium block with known electromagnetic properties is fixedly installed on the detection path of the guide rail. When the antenna moves past the block, its reflected echo is collected as a calibration signal, and the gain coefficient is calculated accordingly to compensate for the small changes in the actual coupling state of the antenna. This is a direct correction method to solve the problem of residual coupling fluctuations affecting the signal amplitude.

[0031] Furthermore, the reference dielectric block must possess a stable dielectric constant and low loss, and its material (such as a specific polymer) must maintain constant and reliable electromagnetic properties under temperature and humidity variations. Its size must be sufficient to cover the antenna's main lobe illumination range, and its surface must be installed flush with the inner surface of the lining, ensuring that its reflection effect on the radar profile matches the actual lining background and avoiding the introduction of highly discontinuous interference. When the radar antenna passes directly above the reference block, the reflected echo at a specific frequency (often an intermediate frequency) is recorded. Due to its fixed position and uniform material, the returned signal primarily reflects the current antenna radiation-coupling efficiency. The measured echo amplitude of the reference block (e.g., the average amplitude in the central region) is compared with the standard amplitude of the same material under ideal coupling conditions calibrated in the laboratory, and the gain compensation coefficient (K) is calculated from the ratio.

[0032] Furthermore, the calculations considered the antenna radiation field distribution characteristics and the block geometry to select the solution region. The K value essentially reflects the attenuation ratio relative to the reference coupling efficiency under the current operating conditions, theoretically caused by residual coupling deviations such as small changes in the antenna-lining gap and slight differences in contact pressure. The core value of this step lies in providing an online position-dependent absolute amplitude reference: the physical existence of the reference block ensures the authenticity of the calibration; its known position allows for precise spatial positioning of the calibration point; and the K calculation establishes a dynamically corrected reference.

[0033] Furthermore, this gain coefficient will be applied in real time to all echo signals captured along the entire detection path to offset the amplitude attenuation or fluctuation of the received signal caused by factors that are difficult to completely eliminate, such as minor track unevenness, residual vibration of the damping system, and ambient temperature drift. This ensures that the echo amplitude information used in subsequent analysis can reflect the true dielectric difference inside the medium rather than coupling error, thereby improving the reliability of amplitude-related characteristic parameter extraction.

[0034] In S4, the calculated gain coefficient (K) is uniformly applied to the original echo signal sequence acquired throughout the entire moving path for amplitude compensation, and independent low-frequency and medium-frequency echo datasets are obtained according to the transmission frequency. This is a standardization and organization step in data preprocessing.

[0035] Furthermore, all original echo signals (including all frequency components and all location point records) are multiplied by the gain coefficient K to perform global amplitude compensation processing, which is equivalent to amplifying or attenuating the waveform value at the signal level to offset the gain loss or enhancement caused by the coupling state revealed by S3, so that the overall amplitude level of the compensated signal is restored to the expected value close to the ideal coupling state, and the amplitude fluctuations caused by the slight differences in coupling at different locations are smoothly corrected.

[0036] Furthermore, the full dataset after amplitude correction needs to be separated and reassembled based on the transmission frequency attribute: echoes from all spatial triggering locations at the same transmission frequency are extracted and independently constituted as a complete dataset corresponding to that center frequency. For example, all echoes in the low-frequency band form the first echo dataset (D_low), and all echoes in the mid-frequency band form the second echo dataset (D_mid). The separation process must strictly maintain the original spatial location index, ensuring that the location of each sampling point in the dataset strictly corresponds to the spatial coordinates defined by the S2 fixed-interval triggering.

[0037] Furthermore, S4 closely relies on the results of S2 and S3: the original signal spatial / frequency information provided by S2 forms the basis for separation; the gain coefficient provided by S3 serves as the basis for compensation. Compensation restores the authenticity of the echo's physical amplitude, while separation organization optimizes the data architecture according to frequency characteristics. The resulting frequency-independent datasets each reflect the structural response of the tunnel lining at different depths: low-frequency (e.g., D_low) data, due to its deep penetration, better reflects the reflection characteristics of the deep interface; mid-frequency (e.g., D_mid) data, with its higher resolution, retains more shallow detail information. The complete spatial alignment of D_low and D_mid establishes a mutually correlated and amplitude-reliable data source foundation for subsequent frequency-band-specific analyses (S5 uses D_low, S6 / S7 uses D_mid).

[0038] In S5, based on low-frequency echo datasets (such as D_low), the depth distribution information of the most dominant structural reflection interfaces inside the lining is identified and quantified. This is a key step in locating deep stable boundaries to provide a geometric benchmark for subsequent analysis.

[0039] Furthermore, D_low data, due to its use of lower-frequency electromagnetic waves with stronger penetrating power, can capture the deepest and most significant primary reflection interface features (usually the interface between the lining and the surrounding rock behind, or the boundary of a large structure) with a relatively strong signal-to-noise ratio. This interface's phase axis (reflection wave continuity) appears as a distinct hyperbola or a near-horizontal linear high-energy band on the radar profile. The processing algorithm automatically searches for the strongest reflection phase axis in the global or local region of D_low and performs continuous spatial tracking along this axis.

[0040] Furthermore, a hyperbolic mathematical model is used to fit the searched strong reflection trajectory. This model uses the interface depth as the core parameter for inversion and solution. The inversion process uses iterative optimization to minimize the error between the model's predicted travel time and the actual travel time. After successful fitting, the depth estimate of the main reflection interface at each antenna sampling position can be output, forming a continuous spatial depth distribution curve.

[0041] Furthermore, S5 fully leverages the advantages of low-frequency data: lower frequencies are less sensitive to shallow, fine structures, thus avoiding interference and highlighting the most significant deep interfaces; the amplitude of D_low, after compensation by S4, is more reliable, making the search for the strongest reflection more robust. The obtained depth information (d) serves as an important input reference for subsequent S6, representing the location of the deep boundary of the lining matrix. The error must be controlled within a certain range to ensure practicality. Depth curve continuity assessment (such as depth gradient variation constraints) can further screen stable segments as reliable references. This establishes a spatially varying, robust geometric reference surface, providing a clear depth-based coordinate basis for subsequent exploration of potential anomalies located near or behind this reference surface in mid-frequency data.

[0042] In S6, the depth information of the main structural reflection interface obtained in S5 is used as the reference plane. In the separated intermediate frequency echo dataset (such as D_mid), the signal region located within a limited depth range behind the reference plane is searched, and candidate anomalous signal clusters that may characterize defects are extracted. This is an intelligent preliminary screening process that narrows the detection range and focuses the analysis of deep target areas. First, based on the design thickness of the tunnel lining and the characteristics of typical defect locations, a reasonable depth range around the reference depth (d) is preset (such as a certain offset range behind d). This range is then transformed into the time domain of the intermediate frequency radar data to form a focusing time window (time and depth are related through wave velocity).

[0043] Furthermore, within the D_mid data subset defined by this time window, an initial amplitude-based screening is first performed: local extrema points whose echo amplitudes significantly exceed the statistical level of background noise (e.g., multiples of the background root mean square); these points indicate possible dielectric discontinuities or scattering sources. Then, a spatial proximity clustering algorithm is applied: extrema points that are spatially adjacent (e.g., horizontal intervals less than a set value) and whose amplitudes meet the criteria are grouped into clusters, forming candidate anomalous signal clusters with spatial continuity. Each cluster corresponds to a set of k spatially correlated anomalous sampling points on the radar profile.

[0044] Furthermore, S6 is closely related to S4 and S5: D_mid provides a high-resolution signal; the S5 depth (d) serves as a precise spatial reference to lock the area of ​​interest. A preset depth range focuses on the risk area behind the interface to avoid blind searching across the entire profile; an amplitude threshold eliminates random noise; spatial clustering aggregates scattered points to improve the signal-to-noise ratio. This outputs a series of candidate clusters, each representing a potential anomalous area on the radar profile. However, these clusters are not necessarily real defects; they may contain signals from residual clutter, isolated noise, or non-defective reflectors (such as pebbles), requiring further refined identification. This provides a pre-screened list of target areas for feature extraction and discrimination in S7, improving analysis efficiency and the targeting of subsequent processing.

[0045] In S7, for each candidate anomalous signal cluster identified in S6, the spatial energy concentration index (E) and the phase consistency index (P) of waveform similarity are calculated to evaluate its signal aggregation characteristics. The joint decision result of these two indices is then used to determine whether the corresponding cluster indicates a real defect. This is the decision-making step for achieving quantitative identification of defect features and suppression of misjudgments.

[0046] Furthermore, the E index characterizes the degree of concentration of reflected energy from all signal sampling points within a cluster along its coverage space length (the calculation formula is the same as in the original scheme), reflecting the spatial focusing of energy in abnormal areas. Real defects tend to form high E-value distributions due to their large dielectric contrast and concentrated size. The P index, on the other hand, evaluates the phase similarity of echo waveforms between adjacent sampling points within a cluster (the calculation formula is the same as in the original scheme). It calculates the correlation between complex signals at adjacent points and takes the maximum value as the P value, reflecting the spatial consistency and smoothness of the waveform. Real defect reflective surfaces typically have good phase continuity under good signal-to-noise ratios, leading to higher P values.

[0047] Furthermore, before calculation, a Hilbert transform is typically performed on the original signal to obtain a complex signal form for calculating phase information. The E and P indices are calculated using information from all points within the cluster (E) and the relationships between adjacent points (P), respectively. A joint judgment threshold condition is set based on E and P (e.g., requiring both to be no less than a certain value), and a defect response is only confirmed when the cluster simultaneously meets both the E and P thresholds.

[0048] Furthermore, S7 relies on the output of S6 and the processing of S4: candidate clusters define the analysis range; D_mid, after compensation by S4, has better complex signal quality, which is beneficial for phase analysis. The E index combats noise and isolated points but requires significant defect energy; the P index identifies phase abrupt changes and is sensitive to waveform consistency. The combination of the two is complementary: low-energy clutter groups are excluded due to their low E value; high-energy random noise clusters are filtered out due to their low P value caused by phase discontinuity; only high-E and high-P candidate clusters that simultaneously exhibit high energy aggregation and phase continuity in a local spatial region are judged as defects, thereby reducing false targets and improving detection reliability. The final output includes a list of defect locations and ranges, as well as information on the judgment criteria, providing input to the results module.

[0049] In summary, firstly, by combining the mechanical design of the guide rail and the three-dimensional constraint roller with the elastic shock-absorbing installation, the antenna attitude and movement trajectory are constrained from a physical perspective. This reduces the fluctuation range of the antenna-lining coupling distance to the millimeter level and the angular deviation to less than 1°. This fundamentally suppresses the random changes in coupling caused by human operation and surface unevenness, ensuring that the amplitude fluctuation of the echo signal originates from the internal differences of the medium rather than external interference, thus significantly improving the stability and repeatability of the signal source. Furthermore, multi-frequency collaborative acquisition (e.g., 400MHz low frequency and 900MHz mid frequency) combined with a frequency band division processing strategy utilizes the penetrability of low frequencies to obtain accurate depth benchmarks for deep structural interfaces (e.g., inversion error of less than 3cm at the lining-surrounding rock interface). This then guides the mid-frequency data to focus and search in a pre-defined high-risk area (e.g., a depth window of 10-30cm behind the interface). This avoids the computational burden of blind searching across the entire profile and captures minute anomalies through high-frequency resolution. Simultaneously, online gain compensation (referencing PTFE block calibration) eliminates the impact of residual coupling fluctuations on amplitude, ensuring that both low-frequency benchmark positioning and mid-frequency defect signal extraction are based on reliable amplitude. Furthermore, the joint criterion of spatial energy concentration (E) and phase consistency (P) is calculated based on the compensated mid-frequency signal. The E index filters out low-energy diffuse interference (e.g., localized damp areas), and the P index filters out high-energy random noise (e.g., boulder reflections), retaining only candidate clusters with both energy focusing and waveform continuity (e.g., E≥1500V). 2 The void size of / m and P≥0.85 significantly reduces false alarms caused by clutter and noise, and improves the specificity of identifying small defects (such as 5cm voids).

[0050] In one implementation, S1 further includes: The sliding bracket forms a three-dimensional constraint fit with the guide rail through at least three rollers, wherein the pressure of at least one roller is adjustable, and the radar antenna is mounted on the sliding bracket through an elastic damping element.

[0051] In this embodiment, it should be noted that in S1, the sliding support uses at least three rollers (for example, in the case of an I-beam rail, three rollers are used: two upper rollers and one lower roller) to form a three-dimensional constraint fit with the guide rail (such as an I-shaped cross-section), which is the core of physical stability. The two upper rollers (the mounting shaft can be an eccentric shaft design) respectively clamp the two sides of the upper flange of the rail, restricting the lateral movement (X-direction displacement) and rotation around the direction of travel (i.e., pitch angle change) of the support; the lower roller supports and contacts the bottom of the rail web, preventing the support from sinking due to gravity or vibration (Z-direction displacement).

[0052] Furthermore, the clamping force of the upper roller can be finely adjusted by adjusting the rotation angle of the eccentric shaft (for example, rotating 10° can change the clamping force by about 20%) to accommodate slight local deformation of the track or installation errors, ensuring stable rolling friction. This avoids both excessive pressure causing jamming (such as resistance exceeding the drive motor load) and insufficient pressure causing bracket wobbling. The radar antenna body is not rigidly fixed to the main structure of the bracket, but is mounted on the top bracket of the bracket using shock-absorbing elements with specific elasticity and damping characteristics (such as rubber pads with a Shore hardness of 60A and a thickness of 5mm). This elastic mounting method can effectively absorb micro-undulations on the lining surface (such as waviness at the level of 0.5mm) and high-frequency micro-vibrations transmitted from the drive (such as slight vibrations above 50Hz caused by motor cogging effect), forming a mechanical filtering effect.

[0053] Furthermore, when the support encounters minor obstacles or uneven tracks, the damping elements deform (e.g., the rubber pad compresses by 1mm), allowing the antenna to float slightly up and down, thus buffering the impact and preventing rigid impacts from being directly transmitted to the antenna, causing it to deviate from the preset posture (e.g., the parallelism between the antenna's bottom surface and the track plane). The three-dimensional constraints of the rollers, combined with elastic damping, ensure that during the antenna's uniform movement at 0.2m / s, the distance change between its bottom plane and the track reference plane is controlled within ±1mm, and the spatial posture (horizontal and pitch) fluctuation is less than 1°. From a physical perspective, this maximizes the stability of the coupling distance and angle between the antenna's radiating surface and the lining surface, creating the basic conditions for consistent electromagnetic wave transmission and reception.

[0054] In one implementation, S2 further includes: The at least two different center frequencies include a low-frequency band with a center frequency of less than or equal to 500 MHz and a mid-frequency band with a center frequency between 800 MHz and 1.2 GHz, the first echo dataset corresponding to the low-frequency band and the second echo dataset corresponding to the mid-frequency band.

[0055] In this embodiment, it should be noted that the definition of "at least two different center frequencies including a low-frequency band with a center frequency less than or equal to 500MHz (such as 400MHz in the example scheme) and a mid-frequency band with a center frequency between 800MHz and 1.2GHz (such as 900MHz in the example scheme)" in S2 is based on the actual needs and physical characteristics of tunnel lining detection.

[0056] Among them, 400MHz low-frequency electromagnetic waves have a relatively high penetration depth in common concrete media (theoretically, it can reach more than 1m), and their wavelength is relatively long (about 150mm in concrete with a relative permittivity of 9). The energy can easily penetrate the lining to reach deep depths, and it has a strong reflection capability (strong signal energy) at deep macroscopic structural interfaces (such as the interface between the lining and the surrounding rock, which is usually 200-500mm or even deeper). Therefore, it can form a significant in-phase axis in the D_low dataset.

[0057] However, its resolution is limited by wavelength (the target feature size needs to be larger than about 1 / 4 wavelength, or about 40 mm, to be effectively distinguishable), and its ability to detect shallow, small defects (such as pores and voids smaller than 50 mm) is weak. 900 MHz mid-frequency electromagnetic waves, with a wavelength of about 70 mm under the same dielectric parameters in concrete (theoretical resolution of about 15-25 mm), can provide finer detail resolution and are more sensitive to small dielectric anomalies (such as honeycomb texture, small voids, and localized looseness) in shallow layers (such as within 50-200 mm below the lining surface). However, its penetration ability is relatively low (the effective detection depth in concrete may drop to 300-400 mm), and it is susceptible to multiple reflections in areas with high-density steel reinforcement.

[0058] Furthermore, defining the low-frequency band (400MHz) as the first center frequency for generating the first echo dataset (D_low) clearly aims to stably acquire deep reference interfaces; defining the mid-frequency band (900MHz) as the second center frequency for generating the second echo dataset (D_mid) aims to focus on shallow high-resolution details. This combination and division of frequency bands (D_low for deep reference localization, D_mid for fine detection of shallow anomalies) is a key choice for effectively fusing deep coverage and resolving details.

[0059] In one implementation, S3 further includes: The surface of the reference medium block is flush with the inner surface of the lining, and the gain coefficient is simultaneously applied to compensate for the first echo dataset and the second echo dataset.

[0060] In this embodiment, it should be noted that in S3, it is crucial to mount the surface of the reference dielectric block (such as a PTFE block with dimensions of 300mm x 300mm x 50mm) exactly flush with the inner surface of the tunnel lining. This ensures that the calibrated reflected signal originates from a known and flat interface, truly simulating the reflection environment of the actual lining surface. The dielectric constant (about 2.1) and loss tangent of PTFE are stable and significantly different from those of concrete (about 6 - 9), making the amplitude of its reflected signal much higher than that of the ordinary lining surface, facilitating identification and extraction. Flush mounting of the surface avoids additional interface reflections or the formation of false hyperbolic diffraction interference introduced by the protrusion or depression of the block (for example, a 1-cm protrusion may generate obvious unexpected echoes), ensuring the clear positioning and identifiability of this reference signal on the radar section (manifested as a strong-amplitude flat in-phase axis that appears locally).

[0061] Furthermore, once the gain coefficient (K = A_std / A_ref, for example, if the actually measured A_ref may be 180mV and the laboratory-calibrated A_std is 200mV, then K ≈ 1.111) is calculated, it is synchronously applied to amplitude compensation for the echo signal sequences of all frequencies collected throughout the process. This means that whether it is the low-frequency (400MHz) dataset D_low or the mid-frequency (900MHz) dataset D_mid, the data used in subsequent processing are those amplified or attenuated by K times (for example, multiplying the original 900mV signal by K to become 1000mV, and the 120mV of 400MHz becomes 133mV).

[0062] Furthermore, this global synchronous compensation processing is based on a reasonable assumption: that is, the change in the coupling state during the movement of the antenna (such as slight warping of the antenna, slight aging of the rubber pad, and minute changes in the average gap caused by extremely small bending of the track) has a similar impact on its radiation and reception efficiency at all operating frequencies. Although the near-field distribution characteristics of electromagnetic waves of different frequencies are slightly different, within the allowable range of engineering accuracy (the goal is to compensate for the main gain drift rather than pursue strict physical matching), this single gain coefficient for synchronous compensation of dual-band echoes is an effective and practical solution, ensuring that the signal amplitudes in the two datasets D_low and D_mid are corrected as a whole and consistently affected by coupling fluctuations, providing a comparable and reliable amplitude basis for subsequent use of D_low for depth reference identification and D_mid for defect energy assessment.

[0063] In one embodiment, S5 further includes: By performing hyperbola fitting on the reflection in-phase axis with the strongest energy in the first echo dataset, the depth distribution of the main structural reflection interface is inversely obtained.

[0064] Furthermore, the model for hyperbola fitting is: Where t is the two-way travel time of the echo, x is the horizontal position coordinate of the antenna with the origin directly above the reflection point, d is the depth of the reflection interface, and v is the propagation speed of the electromagnetic wave in the lining.

[0065] In this embodiment, it should be noted that the process in S5, "fitting a hyperbola to the strongest reflection phase axis in the first echo dataset (D_low, e.g., 400MHz data)," is a key mathematical tool for inverting the depth of the reflection interface of the main structure of the lining. The model uses the standard radar equations to describe the hyperbolic travel time of the point reflector: .

[0066] Where t(x) represents the two-way propagation time (in nanoseconds, ns) of the detected signal from transmission, to the reflection point at depth d, and back to the antenna when the antenna is at a certain horizontal position x. The x-coordinate takes the position directly above the reflection point as the origin (assuming the interface is horizontal or approximately horizontal, x=0 is the vertex). d is the actual depth (in meters, m) of the main reflecting interface that we need to solve for. v represents the propagation speed of electromagnetic waves in the tunnel lining medium (usually concrete) (in meters per nanosecond, m / ns), which is a known parameter or obtained through calibration (an empirical value of 0.12 m / ns is used in this example).

[0067] Furthermore, in practice, the strongest, continuous reflection phase axis (usually representing the lining-surrounding rock interface) is identified from the D_low data. This phase axis exhibits an approximately hyperbolic, convex shape on the radar profile. The fitting process involves recording the arrival time t_i of the strong reflection signal at each spatial location x_i along the scanning trajectory. Then, using the aforementioned mathematical formula, the model parameters d (and v, if fitted as variables) are adjusted through nonlinear least squares optimization or a similar algorithm to minimize the mean square error between the calculated theoretical travel time t_model(x_i) and the actually measured travel time t_i for all x_i (e.g., minimizing the mean square error). Minimum).

[0068] Furthermore, once the fit converges, the model parameter d becomes the depth estimate of the reflective interface below the location x_i (corresponding to a specific point on the track). This inversion is performed on all points along the entire detection path, ultimately yielding a spatially distributed depth curve d(x) of the main structural reflective interface. This curve has physical meaning, representing the morphology of the tunnel's true structural boundary (the bottom of the lining or the contact surface with the surrounding rock). Its accuracy is affected by the v-value error and the data signal-to-noise ratio (in this example, a continuity constraint |Δd / Δx| < 0.05 is used to screen reliable sections), providing a depth benchmark for S6 to pinpoint potential defect locations.

[0069] In one implementation, S6 further includes: The preset depth range is related to the design thickness of the tunnel lining, and the extraction process includes initial screening based on amplitude threshold and clustering based on spatial proximity.

[0070] In this embodiment, it should be noted that in S6, "the preset depth range is related to the design thickness of the tunnel lining" indicates that the search area for the candidate abnormal signal cluster is not fixed, but depends on the depth d(x) of the main structural reflection interface obtained in S5, and takes into account the actual design parameters of the project (e.g., a tunnel design lining thickness of 350 mm). Common dangerous defects (such as voids behind the lining) are located near the contact surface between the lining and the surrounding rock, or in the surrounding rock cavity not far after the contact surface.

[0071] Therefore, the preset depth range is usually set within a certain distance behind the reference surface d(x). In the example scheme, it is specifically set as a time window [t_min, t_max] = [2(d(x) + 0.1) / v, 2(d(x) + 0.3) / v] (v is taken as 0.12m / ns). The basis for converting the depth offset into a time window is the relationship between two-way travel time and depth (time = 2 * depth / v). 0.1m and 0.3m represent the depth range of 10cm to 30cm behind the reference interface (i.e., assuming the lining thickness d(x), searching for the location from d(x) + 0.1m to d(x) + 0.3m deep). This range covers the area where typical defects behind the lining may occur (within the designed lining thickness and the shallow surrounding rock immediately behind it).

[0072] Furthermore, the extraction process involves two steps: First, an initial screening based on an amplitude threshold is performed. Within a D_mid (e.g., 900MHz) data subset within a preset time window, the average level of background noise (root mean square RMS value) is calculated. Then, all sampling points whose amplitude values ​​are greater than a specific multiple of the background noise RMS value (3 times in this example; if the background noise RMS is 10mV, the threshold is set to 30mV) are identified as significant outliers (the purpose of amplitude threshold screening is to eliminate most random noise).

[0073] Furthermore, spatial proximity-based clustering is performed: all significant outliers are traversed, and points that are less than a set proximity distance (5cm in this example) in the horizontal direction are considered spatially adjacent and grouped into the same cluster (using a simple distance clustering method similar to DBSCAN). A cluster typically consists of multiple spatially closely adjacent outliers (k points, each spaced 2cm apart). This process eliminates isolated strong noise points and combines spatially clustered suspected outliers into meaningful candidate target regions (which may represent a small hole, a loose region, etc.).

[0074] Furthermore, the d(x) provided by S5 precisely focuses this search on high-risk areas, amplitude filtering removes most clutter noise, and spatial clustering improves the spatial coherence of suspected areas, providing a pre-processed, high-quality candidate target list for the precise identification of S7.

[0075] In one implementation, S7 further includes: The spatial energy concentration index E is calculated as follows: Where s_k is the echo complex signal of the kth sampling point in the candidate abnormal signal cluster, Σ represents the summation of all sampling points in the cluster, and L is the spatial length covered by the cluster; The waveform phase consistency index P is calculated as follows: , where s_k and s_{k+1} are the echo complex signals of two adjacent sampling points within the cluster, Re() represents taking the real part, |·| represents taking the modulus, and max represents taking the maximum value.

[0076] In this embodiment, it should be noted that, in S7, the calculation formula for the spatial energy concentration index E is defined as follows: Its physical meaning is to assess the concentration of reflected energy relative to spatial length within the spatial range occupied by a candidate anomaly signal cluster. In the formula, s_k represents the complex radar echo signal at the location of the k-th spatial sampling point within the candidate anomaly cluster after Hilbert transform.

[0077] Furthermore, complex signal representation ( (j is the imaginary unit used to represent complex signals) contains all the amplitude and phase information of the signal at that point. This represents the square of the instantaneous energy of the echo signal at that point (the unit is usually volts squared V). 2 ), reflecting the instantaneous power value of the reflection intensity at that point. Σ represents the instantaneous energy |s_k| for all spatial sampling points (k=1,2,...,N) within the cluster. 2 Perform summation (Σ|s_k|) 2 The total energy value of the cluster is obtained. L represents the actual spatial length (in meters, m) covered by the candidate anomaly cluster in the radar scanning direction (i.e., the longitudinal direction of the tunnel). For example, for a cluster containing 20 adjacent points, if the spatial sampling interval is 2 cm, then L = 0.38 m. Finally, the E value is the total instantaneous energy of the cluster divided by the spatial length it covers, that is, the concentrated energy level per unit spatial length (in V). 2 / m). For example, if the total energy of a cluster is 600V. 2 If the coverage length L is 0.4m, then E = 1500V 2 / m. Real defects (such as a 60mm diameter cavity) have a large difference in dielectric constant compared to the surrounding concrete, and their volume is relatively concentrated, resulting in a relatively concentrated spatial energy reflection (manifested as a high Σ|s_k| even at small L values). 2 Typically, a higher E value is calculated for low-energy clutter regions or large-area non-defect reflections with diffused energy (such as large, slightly damp areas). 2 It may not be low, but because its spatial coverage length L is relatively large (e.g., 3m), its E value (Σ|s_k| 2 / L) will be very low. The amplitude compensation of S4 ensures the reliability of the s_k amplitude, so that the calculated E value can truly reflect the target characteristics rather than coupled fluctuations. The E index is used to measure whether a candidate target has enough concentrated* energy to indicate a potentially significant reflector (such as a cavity).

[0078] Furthermore, the formula for calculating the waveform phase consistency index P is defined as follows: This metric focuses on evaluating the phase continuity and similarity between the signal waveforms of two adjacent spatial sampling points within a candidate anomalous signal cluster, along the scanning trajectory. In the formula, s_k and s_{k+1} represent the echo complex signals obtained after the Hilbert transform, specifically the k-th and the next (k+1) spatial sampling point positions within the cluster. The operator "・" represents the complex dot product (inner product). For two complex numbers A=a+jb and B=c+jd, the dot product is A・B=ac+bd (real part)+j(...), but Re(...) in the formula only takes the real part of the result.

[0079] Furthermore, the denominator |s_k|・|s_{k+1}| is the product of the magnitudes (amplitudes) of the complex signals at two adjacent points. The numerator Re(s_k・s_{k+1}) geometrically represents the real part of the dot product of two complex vectors, which is essentially equivalent to the normalized cosine correlation function of the two signals (the correlation value with a lag of 0 at point k). Therefore, the fractional value... The mathematical essence of this is to calculate the instantaneous complex correlation coefficient (or the real part of the normalized cross-correlation at zero hysteresis) between two adjacent sampling points. The value of this coefficient ranges from [-1, 1]. The closer the value is to 1, the more consistent the phase and the more similar the waveforms of the echo signals at the two adjacent points are; if it is close to 0 or -1, it indicates that the phases are inconsistent or the waveforms are very different.

[0080] Furthermore, the `max{}` operator iterates through all adjacent point pairs within a cluster (all point pairs from k to k+1), taking the maximum value among these instantaneous correlation coefficients as the phase consistency index P for that cluster. In this example, the P threshold is set to 0.85. Echoes from a clear, spatially continuous reflective interface (such as the top or bottom of a cavity) typically exhibit very high waveform similarity and phase continuity between adjacent points within the cluster, resulting in a high P value (close to 1). In contrast, "pseudo-clusters" formed by aggregated random noise points, or scattering points formed by sharp foreign objects (such as pebbles), often exhibit strong randomness between adjacent points, with poor phase correlation (weak coherence), leading to lower calculated P values. The Hilbert transform ensures the acquisition of signal phase information, and the compensation of S4 guarantees signal quality. The P index effectively evaluates the spatial self-similarity (coherence) of signals within a candidate cluster, indicating whether it originates from a regular, defect-like reflector with a smooth or continuous interface, helping to eliminate high-energy but chaotic interference signals. E and P are combined (e.g., E>=1500V in the example). 2 The condition that / m and P>=0.85 provides a more reliable defect criterion.

[0081] In one implementation, the condition for determining a defect is that the spatial energy concentration index E is greater than or equal to a first threshold and the waveform phase consistency index P is greater than or equal to a second threshold at the same time.

[0082] In this embodiment, it should be noted that the dual threshold condition for judging defects requires that the spatial energy concentration index (E) and the waveform phase consistency index (P) reach preset threshold values ​​(E≥E_th and P≥P_th) simultaneously, so as to exclude misjudgment based on a single feature by using physical quantification standards.

[0083] For example: When a candidate cluster (e.g., 20 spatially adjacent sampling points covering a length L = 0.4m) is transformed by Hilbert to obtain a complex signal sequence: the spatial energy concentration E is calculated as: the sum of the instantaneous energy of all points within the cluster and Σ|s_k| 2 =600V 2 Then E=600V 2 / 0.4m=1500V 2 / m (satisfies E_th=1500V) 2 / m threshold requirement).

[0084] Waveform phase consistency P calculation: Traversing adjacent points within the cluster (19 pairs), the maximum real part of the dot product of the complex signal appears at points k and k+1: s_k=3.2+4.1j→|s_k|=5.2V; s_{k+1}=2.9+3.8j→|s_{k+1}|=4.8V; Re(s_k·s_{k+1})=(3.2×2.9)+(4.1×3.8)=9.28+15.58=24.86; Normalized value: 24.86 / (5.2×4.8)≈0.996→ Taking the maximum value of this value at all adjacent points within the cluster, P=0.996 (far exceeding P_th=0.85).

[0085] This cluster simultaneously satisfies the E and P thresholds (1500 ≥ 1500 and 0.996 ≥ 0.85), and is therefore classified as a genuine defect (e.g., a void with a diameter of 5 cm). Conversely, if a cluster has high energy but low phase coherence: E = 1800 V... 2 / m (exceeding E_th), but the phase jump between adjacent points is drastic, with a maximum P=0.4 (below 0.85) → excluded (e.g., reflection from metal debris); high phase but low energy concentration: P=0.92 (exceeding P_th), but Σ|s_k| 2 =200V 2 Coverage L=0.5m, E=400V 2 / m (below 1500) → excluded (e.g., weak noise).

[0086] This joint criterion, through the physical coupling verification of energy spatial clustering (E) and waveform continuity (P), avoids misjudgment caused by single image or phase features (interface interference from high-energy noise or phase smoothing), ensuring that defect judgment simultaneously meets the requirements of reflection intensity concentration and waveform coherence, and improving the specificity of identifying minute anomalies.

[0087] A radar antenna-based tunnel lining internal defect detection system is also provided to implement the aforementioned radar antenna-based tunnel lining internal defect detection method, comprising: The mechanically stabilized scanning module includes a guide rail, a sliding bracket, and a drive unit, which is used to support and control the stable movement of the radar antenna. A multi-frequency radar control and acquisition module is used to control the radar antenna to transmit electromagnetic waves of at least two frequencies at fixed spatial intervals and to receive echo signals. Signal processing module, configured for S3, S4, S5, S6 and S7; The results output module is used to output defect information.

[0088] In this embodiment, it should be noted that an "I"-shaped aluminum alloy guide rail is installed on the secondary lining surface of the tunnel and fixed with expansion bolts. The sliding bracket is coupled to the rail via three rollers, with two upper rollers clamping the two sides of the upper flange of the rail and one lower roller supporting the bottom of the web. The mounting shaft of the upper roller is designed as an eccentric shaft, and the clamping force can be adjusted by rotation. The radar antenna is mounted on the top of the bracket via a 5mm thick, Shore A hardness 60A damping rubber pad. The antenna integrates 400MHz (low frequency) and 900MHz (intermediate frequency) dual transmit / receive units. At a distance of 1.5m from the starting point, a 300mm×300mm×50mm polytetrafluoroethylene (PTFE) block is pasted as a reference medium block, with its surface flush with the lining. The system also includes a drive motor, a control unit, a data acquisition card, and an embedded industrial control computer (signal processing module).

[0089] Further, the detection process is as follows: S1 and S2: The system is started, and the sliding support moves at a constant speed of 0.2 m / s. The system triggers a data acquisition every 2 cm of movement. Each time it is triggered, the antenna sequentially transmits 400 MHz and 900 MHz pulses (1 ms apart) and receives the echoes.

[0090] S3: When the antenna passes the PTFE block, its 900MHz echo is collected. The average amplitude of the 10 center sampling points is taken as A_ref. This is compared with the pre-stored standard amplitude A_std to calculate the gain coefficient. .

[0091] S4: Multiply all echo data by K compensation and separate them into a low-frequency dataset D_low and a mid-frequency dataset D_mid.

[0092] S5: Perform hyperbolic fitting on the strongest in-phase axis in D_low. (v is taken as 0.12 m / ns), the inversion depth d(x) is used, and the continuous segment with |Δd / Δx|<0.05 is determined as the main structural reflection interface.

[0093] S6: In D_mid, define a time window [2(d+0.1) / v, 2(d+0.3) / v] based on d. Search for extreme points within this window whose amplitude exceeds 3 times the root mean square of the background noise, and cluster points with a spatial distance of less than 5cm to form candidate clusters.

[0094] S7: For each candidate cluster, first perform a Hilbert transform on its echo signal to obtain a complex signal.

[0095] Calculate the spatial energy concentration .

[0096] Calculate waveform phase consistency .

[0097] If E≥1500V2 If the value is / m and P≥0.85, it is considered a defect.

[0098] In this embodiment, it should be noted that the specific method of performing the operation of the above-mentioned radar antenna-based tunnel lining internal defect detection system has been described in detail in the embodiments of the radar antenna-based tunnel lining internal defect detection method, and will not be elaborated here.

[0099] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0100] It should also be noted that the various specific technical features described in the above embodiments can be combined in any suitable manner without contradiction. To avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0101] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

[0102] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be covered within the scope of the claims and specification of the present invention.

Claims

1. A method for detecting internal defects in tunnel lining based on a radar antenna, characterized in that, include: S1. A guide rail is laid on the inner surface of the tunnel lining, and the radar antenna is coupled to the guide rail through a sliding bracket. The radar antenna is controlled to move along the guide rail at a constant speed while maintaining a preset posture. S2. During the movement of the radar antenna, it is triggered at fixed spatial intervals to control it to transmit at least two detection electromagnetic waves with different center frequencies in a time-division or simultaneously, and to receive the corresponding echo signal sequence synchronously. S3. A reference medium block with known electromagnetic properties is set on the detection path of the guide rail. When the radar antenna passes by, the calibration echo signal from the reference medium block is collected. Based on the comparison between the amplitude of the calibration echo signal and the preset standard amplitude, the gain coefficient used to compensate for the current coupling state is calculated. S4. Apply the gain coefficient to the echo signal sequence to perform amplitude compensation and separate it by frequency to obtain at least a first echo dataset corresponding to a first center frequency and a second echo dataset corresponding to a second center frequency. S5. Based on the first echo dataset, determine the depth information of a major structural reflection interface inside the tunnel lining; S6. Based on the depth information of the main structural reflection interface, extract candidate abnormal signal clusters located within a preset depth range behind the reference in the second echo dataset. S7. For each candidate abnormal signal cluster, calculate its spatial energy concentration index and waveform phase consistency index, and based on the joint judgment result of the two indices, determine whether there are defects inside the tunnel lining.

2. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S1 also includes: The sliding bracket forms a three-dimensional constraint fit with the guide rail through at least three rollers, wherein the pressure of at least one roller is adjustable, and the radar antenna is mounted on the sliding bracket through an elastic damping element.

3. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S2 also includes: The at least two different center frequencies include a low-frequency band with a center frequency of less than or equal to 500 MHz and a mid-frequency band with a center frequency between 800 MHz and 1.2 GHz, the first echo dataset corresponding to the low-frequency band and the second echo dataset corresponding to the mid-frequency band.

4. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S3 also includes: The surface of the reference medium block is flush with the inner surface of the lining, and the gain coefficient is simultaneously applied to compensate for the first echo dataset and the second echo dataset.

5. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S5 also includes: The depth distribution of the main structural reflection interface is obtained by inverting the reflection axis with the strongest energy in the first echo dataset through hyperbolic fitting.

6. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 5, characterized in that, The hyperbola fitting model is as follows: Where t is the two-way travel time of the echo, x is the horizontal position coordinate of the antenna with the origin directly above the reflection point, d is the depth of the reflection interface, and v is the propagation speed of the electromagnetic wave in the lining.

7. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S6 also includes: The preset depth range is related to the design thickness of the tunnel lining, and the extraction process includes initial screening based on amplitude threshold and clustering based on spatial proximity.

8. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 1, characterized in that, S7 also includes: The spatial energy concentration index E is calculated as follows: Where s_k is the echo complex signal of the kth sampling point in the candidate abnormal signal cluster, Σ represents the summation of all sampling points in the cluster, and L is the spatial length covered by the cluster; The waveform phase consistency index P is calculated as follows: , where s_k and s_{k+1} are the echo complex signals of two adjacent sampling points within the cluster, Re() represents taking the real part, |·| represents taking the modulus, and max represents taking the maximum value.

9. The method for detecting internal defects in tunnel lining based on a radar antenna according to claim 8, characterized in that, The conditions for determining a defect are that the spatial energy concentration index E is greater than or equal to the first threshold and the waveform phase consistency index P is greater than or equal to the second threshold at the same time.

10. A radar antenna-based tunnel lining internal defect detection system, used to implement the radar antenna-based tunnel lining internal defect detection method according to any one of claims 1 to 9, characterized in that, include: The mechanically stabilized scanning module includes a guide rail, a sliding bracket, and a drive unit, which is used to support and control the stable movement of the radar antenna. A multi-frequency radar control and acquisition module is used to control the radar antenna to transmit electromagnetic waves of at least two frequencies at fixed spatial intervals and to receive echo signals. Signal processing module, configured for S3, S4, S5, S6 and S7; The results output module is used to output defect information.

Images