A method and system for predicting roof fall risk based on fiber optic acoustic sensing.

Abstract

This invention discloses a method and system for predicting roof fall risk based on fiber optic acoustic sensing, relating to the field of roadway risk prediction technology. The method includes: deploying distributed single-mode sensing fibers along the roadway roof direction; sequentially connecting the distributed single-mode sensing fibers to exposed sections of anchor bolts, exposed sections of anchor cables, and roof monitoring sections to generate monitoring zones; acquiring initial dynamic response signals based on the monitoring zones and determining the baseline of the first-order bending natural frequency of the roof; acquiring dynamic response signals according to the monitoring zones and extracting the main peak frequencies of the exposed anchor bolt sections, the main peak frequencies of the exposed anchor cables, and the first-order bending natural frequency of the roof based on the dynamic response signals; and dynamically correcting the frequency-stress correlation data and the baseline of the first-order bending natural frequency of the roof by combining on-site verification records and monitoring records, thereby improving the correspondence between the roadway roof fall risk identification results and the actual support status, as well as the stability of early warning under continuous monitoring conditions.

Description

Technical Field

[0001] This invention relates to the field of roadway risk prediction technology, specifically to a method and system for predicting roof fall risk based on fiber optic acoustic wave sensing measurement. Background Technology

[0002] As coal mine roadways develop towards deeper mining, high-intensity tunneling, and complex geological conditions, the coupling relationship between the roof support status and the surrounding rock response continues to strengthen. Roof fall risk monitoring has gradually shifted from single component inspection to continuous perception of the coordinated status of anchor bolts, anchor cables, and the roof.

[0003] Currently, in the process of monitoring the risk of roof fall in the roadway roof support area, the monitoring data often comes from the vibration response of the support components and the vibration response of the roof structure, respectively. The two differ in spatial range, response form and change rhythm, which easily leads to the problem of a lack of unified correspondence between various response signals after the monitoring zone is divided. In particular, when the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, the first-order bending natural frequency of the roof and the baseline of the first-order bending natural frequency of the roof are all involved in the risk analysis, if there is a lack of continuous constraints on the frequency-force correlation data and the component force characterization value, it is easy to cause a deviation between the collaborative diagnosis results and the actual support state, which in turn affects the reflection of the risk probability and risk classification results on the true state of the monitoring zone.

[0004] Secondly, in the process of early warning and management of roof fall risk in roadways, the on-site verification records, monitoring records and subsequent response results after the early warning are triggered are often separated from the previous risk assessment process. This makes it difficult to correct the frequency stress correlation data and the baseline of the first-order bending natural frequency of the roof in a timely manner with changes in the support status during long-term operation. If the original judgment criteria are still used after the monitoring zone experiences support loosening, component fracture, treatment recovery or changes in surrounding rock conditions, the subsequent early warning results and treatment priorities may be mismatched with the current monitoring status, and further weaken the stability of the risk identification process in continuous monitoring scenarios. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a method and system for predicting roof collapse risk based on fiber optic acoustic wave sensing measurement.

[0006] A method for predicting roof collapse risk based on fiber optic acoustic wave sensing, the method comprising:

[0007] Distributed single-mode sensing optical fibers are laid along the direction of the roadway roof. The distributed single-mode sensing optical fibers are connected in sequence to the exposed sections of anchor bolts, exposed sections of anchor cables, and roof monitoring sections to generate monitoring zones. Initial dynamic response signals are collected based on the monitoring zones, and the baseline of the first-order bending natural frequency of the roof is determined.

[0008] Dynamic response signals are collected based on the monitoring zones, and the main peak frequencies of the exposed sections of anchor bolts, anchor cables, and the first-order bending natural frequency of the top plate are extracted based on the dynamic response signals. Frequency-force correlation data are generated based on the main peak frequencies of the exposed sections of anchor bolts and anchor cables.

[0009] Based on the frequency-force correlation data, the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, the first-order bending natural frequency of the top plate, and the baseline of the first-order bending natural frequency of the top plate, the component stress characterization value and collaborative diagnosis results are generated, and the risk probability and risk classification results are generated based on the collaborative diagnosis results.

[0010] Based on the collaborative diagnosis results, risk probability, and risk classification results, early warning results and governance priorities are generated. The on-site verification records and monitoring records corresponding to the early warning results are retrieved, and the frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate are corrected.

[0011] Furthermore, the steps for generating monitoring zones are as follows:

[0012] Distributed single-mode sensing optical fibers are laid along the direction of the tunnel roof, and continuous deployment paths are established at corresponding locations in the exposed sections of anchor bolts, exposed sections of anchor cables, and roof monitoring sections.

[0013] Distributed single-mode sensing optical fibers are fixed to the monitoring section of the top plate by a surface composite coating, and the distributed single-mode sensing optical fibers are connected to the exposed sections of the anchor bolts and anchor cables by a clamping mechanism to form a series continuous coupling structure.

[0014] The exposed sections of anchor bolts, exposed sections of anchor cables, and monitoring sections of the top plate that are in the same continuous deployment path are combined to generate monitoring zones.

[0015] Furthermore, the steps for determining the baseline of the first-order bending natural frequency of the top plate are as follows:

[0016] Start the distributed fiber optic acoustic wave sensor host to collect the initial dynamic response signal of each monitoring zone, and extract the roof response signal corresponding to the roof monitoring section from the initial dynamic response signal.

[0017] The top plate response signal is filtered to obtain the first-order bending natural frequency of the top plate.

[0018] The first-order bending natural frequency of the roof corresponding to each monitoring zone is associated with the monitoring zone to generate the baseline of the first-order bending natural frequency of the roof.

[0019] Furthermore, the steps for extracting the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable are as follows:

[0020] Extract the response signals of exposed anchor bolt sections and exposed anchor cable sections from the dynamic response signals according to the monitoring zones;

[0021] For the response signals of the exposed sections of the anchor bolts and anchor cables, component frequency band screening and time-frequency analysis are performed to obtain candidate peak frequencies;

[0022] A peak shape consistency check is performed on the candidate peak frequencies, and the main peak frequencies of the exposed sections of the anchor bolt and the anchor cable are output. The peak shape consistency check is implemented through a quality gating check.

[0023] Furthermore, the steps for extracting the first-order bending natural frequency of the top plate are as follows:

[0024] According to the monitoring zones, extract the roof response signal corresponding to the roof monitoring section from the dynamic response signal;

[0025] The roof response signal is subjected to roof frequency band screening and multi-point coherence analysis to obtain candidate roof frequency components;

[0026] Modal identification is performed on the candidate frequency components of the top plate, and the first-order bending natural frequency of the top plate is output.

[0027] Furthermore, the steps for generating frequency-force correlation data are as follows:

[0028] Select field calibration points in the exposed sections of anchor bolts and anchor cables, and apply tension perturbations to the field calibration points step by step;

[0029] The main peak frequencies of the exposed sections of the anchor bolt and the anchor cable corresponding to the tension perturbation were collected simultaneously, and the applied tension corresponding to the tension perturbation was monotonically fitted with the main peak frequencies of the exposed sections of the anchor bolt and the anchor cable.

[0030] The monotonic fitting results are correlated with the corresponding monitoring zones to generate frequency-force correlation data.

[0031] Furthermore, the steps for generating component stress characterization values ​​and collaborative diagnostic results are as follows:

[0032] Retrieve frequency-force correlation data, and based on the frequency-force correlation data, the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable, generate anchor bolt force characterization values ​​and anchor cable force characterization values, and integrate them to obtain component force characterization values;

[0033] The stress characterization values ​​of the component, the first-order bending natural frequency of the top plate, and the baseline of the first-order bending natural frequency of the top plate are spatiotemporally registered to generate collaborative diagnostic results.

[0034] Furthermore, the steps for generating risk probability and risk classification results are as follows:

[0035] During the process of generating collaborative diagnostic results, the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point are retrieved, and time-series input data are constructed based on the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point at continuous sampling time.

[0036] Input the time-series input data into a predetermined time-series deep learning model to obtain the risk probability;

[0037] Risk classification results are generated by matching risk probabilities with collaborative diagnostic results.

[0038] Furthermore, the steps for generating early warning results and treatment priorities, and correcting the frequency-stress correlation data and the baseline of the first-order bending natural frequency of the top plate are as follows:

[0039] Based on the collaborative diagnosis results, risk probability, and risk classification results, early warning results and governance priorities are generated.

[0040] Retrieve the on-site verification records and monitoring records corresponding to the early warning results, and extract the on-site verification results and subsequent response results;

[0041] The frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate were corrected based on the on-site verification results and subsequent response results.

[0042] A roof fall risk prediction system based on fiber optic acoustic wave sensing measurement, used in any one of the roof fall risk prediction methods based on fiber optic acoustic wave sensing measurement, the system comprising:

[0043] Zoning baseline module: Distributed single-mode sensing optical fibers are laid along the direction of the roadway roof. The distributed single-mode sensing optical fibers are connected in sequence to the exposed sections of anchor bolts, exposed sections of anchor cables and roof monitoring sections to generate monitoring zones. Based on the monitoring zones, the initial dynamic response signals are collected and the first-order bending natural frequency baseline of the roof is determined.

[0044] Frequency-force correlation module: Collects dynamic response signals based on monitoring zones, and extracts the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the top plate based on the dynamic response signals. Generates frequency-force correlation data based on the main peak frequencies of the exposed sections of the anchor bolt and the anchor cable.

[0045] Diagnostic grading module: Based on frequency-force correlation data, the main peak frequency of exposed anchor bolt section, the main peak frequency of exposed anchor cable section, the first-order bending natural frequency of top plate, and the baseline of the first-order bending natural frequency of top plate, the component stress characterization value and collaborative diagnosis results are generated, and the risk probability and risk grading results are generated based on the collaborative diagnosis results.

[0046] Monitoring baseline module: Generates early warning results and treatment priorities based on collaborative diagnosis results, risk probability and risk classification results, and retrieves the on-site verification records and monitoring records corresponding to the early warning results to correct the frequency-force correlation data and the first-order bending natural frequency baseline of the top plate.

[0047] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0048] This invention, in a roof fall risk monitoring scenario where the roof support status and roof structure response change synchronously, unifies the correspondence between monitoring zones, the main peak frequency of exposed anchor bolt sections, the main peak frequency of exposed anchor cable sections, the first-order bending natural frequency of the roof, and the baseline of the first-order bending natural frequency of the roof. This allows the frequency-force correlation data and the component force characterization values ​​to simultaneously reflect the support component status and the overall roof response, reducing the deviation caused by judging based on a single vibration characteristic. Consequently, the collaborative diagnosis results are more consistent with the actual force status of the monitoring zones, thus establishing a continuous and consistent data foundation for the generation of risk probability and risk classification results.

[0049] Furthermore, this invention also establishes a correlation between collaborative diagnostic results, risk probability, and risk classification results and early warning results, treatment priorities, on-site verification records, and monitoring records. This allows the frequency-force correlation data and the baseline of the first-order bending natural frequency of the roof to be corrected based on the on-site verification results and subsequent monitoring status. This improves the alignment between early warning results and treatment priorities with the current monitoring zone status and maintains consistency between risk judgment criteria and subsequent handling feedback. As a result, it enhances the continuous identification capability and long-term monitoring stability of roadway roof collapse risks.

[0050] In summary, this invention establishes a collaborative determination relationship between the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the roof. It also dynamically corrects the frequency-stress correlation data and the baseline of the first-order bending natural frequency of the roof by combining on-site verification records and monitoring records. This improves the correspondence between the roadway roof fall risk identification results and the actual support status, as well as the early warning stability under continuous monitoring conditions. Attached Figure Description

[0051] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0052] Figure 1 This is a flowchart of a roof fall risk prediction method based on fiber optic acoustic wave sensing measurement provided in Embodiment 1 of the present invention;

[0053] Figure 2 This is a block diagram of a roof fall risk prediction system based on fiber optic acoustic wave sensing measurement provided in Embodiment 2 of the present invention;

[0054] Figure 3 The schematic diagram of a roadway for a method for predicting roof fall risk based on fiber optic acoustic wave sensing measurement provided in Embodiment 1 of the present invention. Detailed Implementation

[0055] 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, 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.

[0056] Example 1

[0057] Please see Figure 1 As shown in the figure, this embodiment discloses a method for predicting roof collapse risk based on fiber optic acoustic wave sensing measurement. The method includes:

[0058] S11: Distributed single-mode sensing optical fibers are laid along the direction of the roadway roof. The distributed single-mode sensing optical fibers are connected to the exposed sections of anchor bolts, exposed sections of anchor cables and roof monitoring sections in sequence to generate monitoring zones. The initial dynamic response signals are collected based on the monitoring zones and the baseline of the first-order bending natural frequency of the roof is determined.

[0059] Specifically, the steps for generating monitoring partitions are as follows:

[0060] S111: Lay distributed single-mode sensing optical fibers along the direction of the tunnel roof, and establish continuous deployment paths at corresponding locations of exposed anchor bolt sections, exposed anchor cable sections, and roof monitoring sections;

[0061] In one specific embodiment, roadway site preparation is performed, referring to... Figure 3 As shown:

[0062] Figure 3 In the diagram, 101 is the anchor bolt, 102 is the anchor bolt tray, 103 is the roof plate, 104 is the distributed optical fiber, 105 is the anchor cable, 106 is the anchor cable tray, 107 is the special clamping mechanism, 108 is the base plate, 109 is the distributed optical fiber acoustic wave sensor host, 110 is the artificial intelligence module, 111 is the calibration and incremental learning module, 112 is the ground, 113 is the tunnel, and 114 is the shaft.

[0063] First, clean the roof slab 103 base surface along the direction of tunnel 113 to ensure there are no loose rocks or debris;

[0064] Subsequently, distributed single-mode sensing optical fibers 104 are laid along the roof 103 of the roadway. To ensure the continuity of monitoring, the distributed single-mode sensing optical fibers 104 must cover all exposed sections of anchor bolts 101, exposed sections of anchor cables 105, and roof monitoring sections between adjacent support components during the layout planning. This establishes a continuous layout path that spans multiple support components and rock masses. It should be noted that armored optical fibers are preferred for the distributed single-mode sensing optical fibers 104 to cope with the complex environment of coal mines.

[0065] S112: The distributed single-mode sensing fiber is fixed to the monitoring section of the top plate through a surface composite coating, and the distributed single-mode sensing fiber is connected to the exposed sections of the anchor bolt and anchor cable through a clamping mechanism to form a series continuous coupling structure.

[0066] In one specific embodiment, in the top plate monitoring section, the distributed single-mode sensing fiber 104 is applied to the surface of the top plate 103 by high-pressure airless spraying process or fixed by a slot.

[0067] The surface composite coating used for fixing adopts an epoxy / fiber reinforced or equivalent high modulus system, with a thickness controlled at 0.8±0.2mm, and has acid and alkali resistance level 5, flame retardancy, antistatic and corrosion resistance.

[0068] It should be noted that the surface composite coating is used to improve the vibration transmission stiffness, ensuring that the small vibrations of the top plate can be efficiently coupled to the optical fiber;

[0069] To absorb the stress caused by the sinking of the top plate, a relaxation arc of 0.5-1m is reserved every 10m for the optical fiber.

[0070] In the exposed sections of anchor bolt 101 and anchor cable 105, a special clamping mechanism 107 is used to achieve axial high stiffness and low additional damping coupling between the optical fiber and the component.

[0071] The clamping mechanism 107 includes a rigid clamping ring, an axial guide groove, a limiting / anti-loosening structure, a silicone buffer layer, the silicone buffer layer being used to prevent damage to the optical fiber, and a fixing assembly consisting of an inner steel ring and an outer steel ring.

[0072] It should be noted that this threaded clamping-elastic compensation composite clamp ensures the axial elongation of the anchor bolt / anchor cable. ;

[0073] In the above manner, the distributed single-mode sensing fiber 104 sequentially couples the exposed section of the anchor bolt 101, the exposed section of the anchor cable 105, and the top plate 103, ultimately forming a series continuous monitoring chain of "anchor bolt - anchor cable - top plate" as a series continuous coupling structure.

[0074] S113: Combine exposed sections of anchor bolts, exposed sections of anchor cables, and monitoring sections of the top plate that are in the same continuous deployment path to generate monitoring zones;

[0075] Monitoring zones are set up according to the direction of the roadway. Each zone consists of a multi-point coherent array composed of several anchor bolts 101, anchor cables 105, and a roof monitoring section spanning a certain length. The division of zones follows the principle of "same quality in the same zone, different strategies in different zones", and is based on the function of the working face, the geological conditions, and the monitoring requirements.

[0076] It should be noted that: the functions of the working face include but are not limited to the coal face, return airway, and intersection; the geological conditions include but are not limited to lithology, burial depth, and geological structure; and the monitoring requirements include but are not limited to high-risk areas and conventional areas.

[0077] Specifically, the steps for determining the baseline of the first-order bending natural frequency of the top plate are as follows:

[0078] S114: Start the distributed fiber optic acoustic sensor host, collect the initial dynamic response signal of each monitoring zone, and extract the roof response signal corresponding to the roof monitoring section from the initial dynamic response signal.

[0079] Activate the phase-sensitive optical time-domain reflectometry (OTDR) system on the ground or underground. -OTDR) Distributed fiber optic acoustic wave sensor host 109, host 109 is connected to distributed single-mode sensing fiber optic cable 104, its spatial resolution Determined by the following formula:

[0080] ;

[0081] In the formula, The spatial resolution of the distributed fiber optic acoustic wave sensor host 109, The speed of light in a vacuum. The refractive index of the optical fiber. The duration of the laser pulse;

[0082] It should be noted that the host is set with a pulse repetition frequency of 2–10kHz and a spatial resolution of 0.5–2m. In the initial stage of system operation, the distributed fiber optic acoustic wave sensor host 109 collects the initial dynamic response signals of each section along the line. Based on the physical location and fiber optic channel mapping relationship recorded by the fiber optic deployment, the roof response signal corresponding to the roof monitoring section is extracted.

[0083] S115: Filter the top plate response signal to obtain the first-order bending natural frequency of the top plate;

[0084] In one specific embodiment, the extracted roof response signal is bandpass filtered at 5–50 Hz, and the energy distribution of the extracted signal is extracted using short-time Fourier transform or wavelet analysis.

[0085] To eliminate local noise interference, a multi-point coherence analysis criterion is used to calculate the frequency. Multichannel coherence coefficient at the location ;

[0086] The calculation formula is expressed as follows:

[0087] ;

[0088] For frequency The multi-channel coherence coefficient at the location, In frequency Place, No. The first channel and the first Cross-power spectral density between channels;

[0089] , The first The first channel, the first The self-power spectral density of each channel, The total number of measurement points within the partition is used to calculate the multi-channel coherence coefficient. With a predetermined multi-channel coherence coefficient threshold When making a comparison, > Furthermore, when the spatial distribution is continuous, this frequency is considered to be the true first-order bending natural frequency of the top plate, which is preferred. 0.7 can be used.

[0090] It should be noted that the multi-channel coherence coefficient threshold... The determination is based on the historical monitoring records, on-site calibration results, and background noise levels corresponding to the monitoring zones, and is used to distinguish between the overall structural modal response and the local random disturbance response.

[0091] Continuous spatial distribution refers to a frequency characteristic, such as the vibration frequency of the top plate, which not only appears at a single measuring point, but also along the deployment path of the distributed single-mode sensing fiber 104, can be synchronously detected at multiple adjacent continuous measuring points (channels) with highly consistent signal responses. This characteristic is used to distinguish between overall structural modes and local random noise: the true natural frequency will cause coordinated vibration of the entire monitoring section, forming a continuous energy strip in space; while isolated local interference or single-point equipment noise is discrete and discontinuous in spatial distribution.

[0092] S116: Establish a correlation between the first-order bending natural frequency of the roof corresponding to each monitoring zone and the monitoring zone to generate a baseline of the first-order bending natural frequency of the roof.

[0093] The first-order bending natural frequencies of the top plate in the initial state of each partition, obtained through coherent verification, are recorded as follows: Using this as the baseline for the first-order bending natural frequency of the roof, all the first-order bending natural frequency baselines of the roof are associated with their corresponding monitoring zone unique codes and stored in the calibration and incremental learning module 111.

[0094] S12: Collect dynamic response signals according to the monitoring zones, and extract the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the top plate based on the dynamic response signals. Generate frequency-force correlation data based on the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable.

[0095] Specifically, the steps for extracting the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable are as follows:

[0096] S121: Extract the response signals of exposed anchor bolt sections and exposed anchor cable sections from the dynamic response signals according to the monitoring zones;

[0097] During the operation, a monitoring mode of "passive seismic source as the main source and active excitation as the auxiliary source" was adopted.

[0098] The distributed fiber optic acoustic sensor host 109 continuously collects dynamic response signals from each monitoring zone, which include passive source responses generated by mining equipment, environmental vibrations, etc.

[0099] Based on the monitoring zones generated in step S11, the corresponding spatial measurement point locations are used. The signals of the exposed sections of anchor bolt 101 and anchor cable 105 are extracted in a directional manner, namely the response signals of the exposed sections of anchor bolt and anchor cable.

[0100] S122: Perform component frequency band screening and time-frequency analysis on the response signals of the exposed sections of the anchor bolt and the anchor cable to obtain candidate peak frequencies;

[0101] Prioritize component frequency band screening: Based on component characteristics, filter within a predetermined component frequency band, wherein the screening frequency band for anchor bolt 101 is 10–100Hz, and the screening frequency band for anchor cable 105 is 5–80Hz.

[0102] Then, perform time-frequency analysis: Perform synchronous compressed wavelet transform or continuous wavelet transform on the response signals of the exposed sections of the anchor bolt and anchor cable, and calculate the modulus of the wavelet transform coefficients. ;

[0103] in, The spatial measurement point locations are corresponding to the distributed single-mode sensing fiber 104. For frequency, Using time sampling points, maximum points are identified along the energy ridge to obtain candidate peak frequencies.

[0104] S123: Perform peak shape consistency verification on the candidate peak frequencies and output the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable. The peak shape consistency verification is achieved through quality gating check.

[0105] To prevent local noise peaks, transient disturbance peaks, or single-point spurious peaks from being misjudged as the true main vibration peaks of components, a peak shape consistency check is performed. This is specifically achieved by performing a quality gating check on the candidate peak frequencies.

[0106] The quality gating checks include signal-to-noise ratio verification, continuity verification of adjacent time sampling points, consistency verification of responses of adjacent spatial measuring points corresponding to the same exposed section of anchor bolt or exposed section of anchor cable, and peak stability verification.

[0107] It should be noted that the signal-to-noise ratio verification is used to determine whether the peak energy corresponding to the candidate peak frequency is significantly higher than the background noise level, so as to exclude low-confidence peaks caused by environmental noise or stray vibrations of equipment.

[0108] The continuity verification of adjacent time sampling points is used to determine whether the candidate peak frequency can be stably maintained in the same frequency neighborhood during continuous time sampling, so as to exclude transient interference peaks that only occasionally appear at a single moment.

[0109] The consistency verification of the response of adjacent spatial measuring points corresponding to the same exposed section of anchor bolt or exposed section of anchor cable is used to determine whether the candidate peak frequency forms a continuous response distribution along multiple adjacent spatial measuring points corresponding to the exposed section of anchor bolt or exposed section of anchor cable, so as to exclude local abnormal peaks that only appear in a single isolated spatial measuring point.

[0110] The peak shape stability verification is used to determine whether the peak shape, peak width and energy ridge changes corresponding to the candidate peak frequency remain stable, so as to eliminate false peaks with peak shape divergence, abrupt peak width changes or discrete energy distribution.

[0111] For the candidate peak frequencies that pass the above quality gating check, the main peak frequency is determined according to the position corresponding to the maximum value of the wavelet transform coefficient modulus on the frequency axis, expressed as:

[0112] ;

[0113] In the formula, Indicates spatial location ,time The main peak frequency at that location, Indicates frequency Find the independent variable corresponding to the maximum value. The wavelet transform coefficient modulus is used to determine the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable, which are determined by the above formula after passing the peak shape consistency check.

[0114] Specifically, the steps for extracting the first-order bending natural frequency of the top plate are as follows:

[0115] S124: Extract the roof response signal corresponding to the roof monitoring section from the dynamic response signal according to the monitoring zone;

[0116] Based on the monitoring partition generated in step S11, the signal of the stable coupling section between the distributed single-mode sensing fiber 104 and the composite coating on the surface of the top plate 103 is extracted from the dynamic response signal, namely the top plate response signal.

[0117] S125: Perform roof frequency band screening and multi-point coherent analysis on the roof response signal to obtain the roof candidate frequency components;

[0118] For the bending mode of the top plate 103, a frequency band of 5–50 Hz is selected, and the multi-channel coherence coefficient is calculated using a multi-point coherence array within the partition. The formula for calculating the multi-channel coherence coefficient is the same as that in step S115.

[0119] The final output multi-channel coherence coefficient is denoted as m, multi-channel coherence coefficient m and multi-channel coherence coefficient threshold Perform a comparison, if m> When the spatial distribution is continuous, it indicates that the frequency is not a random disturbance at a certain point, but rather the synchronous vibration of the entire roof. This set of frequencies that spans multiple continuous spatial locations within the same time period is identified as the candidate frequency components of the roof.

[0120] It should be noted that: S115 is used to generate the baseline of the first-order bending natural frequency of the top plate in the initial state, and S125 is used to determine the validity of the candidate frequency components of the top plate at the current sampling time.

[0121] S126: Perform modal identification on the candidate frequency components of the top plate and output the first-order bending natural frequency of the top plate.

[0122] In one specific embodiment, the logic for performing modality recognition is as follows:

[0123] Based on the candidate frequency components of the roof obtained in step S125, the geological exploration data, rock sample test records, or surrounding rock mechanical parameter records corresponding to the monitoring zone are first retrieved to obtain the elastic modulus of the roof. Elastic modulus of top plate Mechanical parameters used to characterize the stiffness of materials in the roof rock mass;

[0124] Based on the spatial range, roof thickness record, and roof geometric dimension record of the roof monitoring section corresponding to the monitoring zone, the equivalent thickness and equivalent width of the roof monitoring section are determined, and the moment of inertia of the section is calculated based on the equivalent thickness and equivalent width. The moment of inertia of the cross section is a geometric parameter used to characterize the resistance of the cross section of the roof monitoring section to bending deformation, and is used to reflect the cross section stiffness distribution characteristics of the roof monitoring section under bending vibration.

[0125] It should be noted that: the equivalent thickness is the effective thickness parameter of the roof monitoring section that actually participates in bending vibration within the current monitoring range, used to characterize the bending range of the roof monitoring section along the thickness direction; the equivalent width is the effective width parameter of the roof monitoring section that actually participates in bending vibration in the transverse direction, used to characterize the bending range of the roof monitoring section along the transverse direction.

[0126] Retrieve the physical property records or rock sample density test records of the corresponding rock strata in the monitoring zone to obtain the density of the roof rock mass. Density of the top rock mass Physical property parameters used to characterize the mass of the roof rock mass per unit volume are used to characterize the mass distribution characteristics of the roof monitoring section in the calculation of the first-order bending natural frequency of the roof.

[0127] The cross-sectional area is determined based on the equivalent thickness and equivalent width of the roof monitoring section. Cross-sectional area Geometric parameters used to characterize the transverse cross-sectional dimensions of the roof monitoring section;

[0128] Based on the layout relationship between adjacent support components within the same monitoring zone, the support spacing records, and the coverage boundary of the roof monitoring section, the support spacing is determined. Support spacing Structural parameters used to characterize the effective support constraint range on both sides of the roof monitoring section, and to reflect the effective vibration span of the roof monitoring section under support conditions.

[0129] Based on the elastic modulus, moment of inertia, density, cross-sectional area, and support spacing of the roof, the roof monitoring section is considered as an equivalent beam simply supported at both ends. Then, according to the candidate frequency components of the roof selected and retained in step S125, the corresponding first-order bending natural frequency of the roof is generated. ;

[0130] Represented as:

[0131] ;

[0132] In the formula, This is the elastic modulus of the top plate, in Pa. The moment of inertia of the cross section is expressed in units of 1000 m / s. , For density, For cross-sectional area, This refers to the support spacing.

[0133] Specifically, the steps for generating frequency-force correlation data are as follows:

[0134] S127: Select field calibration points in the exposed sections of anchor bolts and anchor cables, and apply tension perturbation to the field calibration points step by step;

[0135] The calibration and incremental learning module 111 selects representative field calibration points on the anchor bolts 101 and anchor cables 105 in the target monitoring zone;

[0136] The tension applied to the above-mentioned components using hydraulic or mechanical devices in progressively increasing or decreasing steps is denoted as tension perturbation. .

[0137] S128: Synchronously collect the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable corresponding to the tension perturbation, and perform monotonic fitting between the applied tension corresponding to the tension perturbation and the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable.

[0138] While applying tension perturbation, the distributed fiber optic acoustic sensor host 109 synchronously records the corresponding main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable.

[0139] Based on string vibration theory, an initial mapping relationship between the applied tension and the main peak frequency is established, and a monotonic correction fitting is performed using the tension perturbation results from the field calibration points, expressed as:

[0140] ;

[0141] In the formula, The tension applied by the anchor bolt or anchor cable. The effective vibration length is determined by the length of the exposed section. This refers to the main peak frequency collected in real time. The equivalent linear density (kg / m) depends on the density and cross-sectional area of ​​the component material.

[0142] The monotonic fitting formula is output as the monotonic fitting result.

[0143] S129: Establish a correlation between the monotonic fitting results and the corresponding monitoring zones to generate frequency-force correlation data;

[0144] Calculate the 95% confidence interval corresponding to the monotonic fitting result to ensure that the mapping relationship is strictly monotonic in a physical sense, so as to eliminate the risk of false alarm caused by multi-valued mapping;

[0145] The monotonic fitting results generated in step S128 and their corresponding confidence intervals are dynamically correlated with the monitoring partitions to which they belong, and the monotonic fitting results are stored and maintained in a partitioned manner according to the different dimensions of the monitoring partitions.

[0146] It should be noted that the different dimensions of the monitoring zones include, but are not limited to, working face function and geological conditions. Working face function includes coal mining face, return airway, etc., and geological conditions include hard rock, soft rock or fractured zone, etc.

[0147] By locking the monotonic fitting results of each zone with its geographical and geological information, frequency stress correlation data covering the entire tunnel is finally generated.

[0148] S13: Generate component stress characterization values ​​and collaborative diagnosis results based on frequency-force correlation data, main peak frequency of exposed anchor bolt section, main peak frequency of exposed anchor cable section, first-order bending natural frequency of top plate, and baseline of first-order bending natural frequency of top plate, and generate risk probability and risk classification results based on collaborative diagnosis results.

[0149] Specifically, the steps for generating component stress characterization values ​​and collaborative diagnostic results are as follows:

[0150] S131: Retrieve frequency-force correlation data, and generate anchor bolt force characterization values ​​and anchor cable force characterization values ​​based on the frequency-force correlation data, the main peak frequency of the exposed section of the anchor bolt, and the main peak frequency of the exposed section of the anchor cable, and integrate them to obtain the component force characterization values.

[0151] The initial tension is calculated according to the monotonic fitting formula in step S128, and a temperature-gradient compensation term is introduced. To eliminate spurious frequency drift caused by thermal expansion and contraction, a force characterization value is generated; the calculation formula is as follows:

[0152] ;

[0153] In the formula, These are the stress characterization values ​​for the anchor bolt or the anchor cable. This is based on the uncompensated force value, i.e., the applied tension calculated in step S128. The temperature compensation function is pre-established based on the field calibration results. The temperature gradient along the 104 axis of the distributed single-mode sensing fiber. For the ambient reference temperature, the temperature compensation function Used to set ambient reference temperature and temperature gradient Converted into stress compensation value;

[0154] The calculated stress characterization values ​​of anchor bolts and anchor cables are encapsulated and combined to generate component stress characterization values ​​that reflect the current status of support components in the monitoring zone.

[0155] It should be noted that: ambient reference temperature To determine the ambient temperature baseline value of the monitoring zone at the current sampling time, the ambient temperature monitoring record of the corresponding temperature sensor for that monitoring zone is retrieved.

[0156] The ambient temperature monitoring records corresponding to the monitoring zone are collected by temperature sensors deployed in the roadway monitoring area. The temperature sensors are set along the distributed single-mode sensing fiber deployment path and are synchronized with the distributed fiber acoustic wave sensing host 109 in time.

[0157] Temperature gradient The temperature change rate is calculated based on the temperature difference between adjacent temperature measuring points within the same monitoring zone and the distance between adjacent temperature measuring points, and is used to characterize the temperature change rate along the direction of the distributed single-mode sensing fiber.

[0158] Temperature compensation function During the on-site calibration phase, the main peak frequency shift corresponding to the same applied tension was collected under different ambient reference temperatures and temperature gradients. The ambient reference temperature, temperature gradient, and the force compensation value corresponding to the frequency shift were then subjected to multivariate fitting to generate the value.

[0159] S132: Spatiotemporally register the component stress characterization value, the first-order bending natural frequency of the top plate, and the baseline of the first-order bending natural frequency of the top plate to generate collaborative diagnostic results.

[0160] In one specific embodiment, the artificial intelligence module 110 spatiotemporally aligns the component stress characterization values ​​and the first-order bending natural frequency of the top plate with the same monitoring zone and the same sampling time.

[0161] By calculating the rate of change of the current frequency relative to the baseline, collaborative diagnostic results are generated. This is used to determine whether there is a risk of structural degradation, and its criterion formula is expressed as: If and only if and ;

[0162] In the formula, This is an event identification flag in the collaborative diagnostic results. A value of 1 indicates a risk event, and a value of 0 indicates no risk. This refers to the peak frequency of the exposed section of the anchor bolt or the peak frequency of the exposed section of the anchor cable during the current time period. This refers to the reference maximum frequency corresponding to the main peak frequency of the exposed section of the anchor bolt or the main peak frequency of the exposed section of the anchor cable. The threshold for the maximum rate of change of frequency. The first-order bending natural frequency of the top plate. The baseline for the first-order bending natural frequency of the top plate. The threshold for the rate of change of the inherent frequency.

[0163] It should be noted that the reference maximum frequency is the main peak frequency of the exposed section of the anchor bolt or the main peak frequency of the exposed section of the anchor cable. The generation logic is as follows: during the initial calibration stage or the stable working condition stage, the main peak frequency of the exposed section of the anchor bolt or the main peak frequency of the exposed section of the anchor cable corresponding to the monitoring zone is recorded, and the recorded value is used as the reference maximum frequency of the corresponding component.

[0164] Threshold of main peak frequency change The frequency range of the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable under normal working conditions was preset based on the field calibration sample.

[0165] natural frequency change threshold The range of the first-order bending natural frequency fluctuation of the roof slab under the initial calibration stage and historical stable working conditions of the monitoring zone is preset, and is preferably taken as 5%.

[0166] Specifically, the steps for generating risk probability and risk classification results are as follows:

[0167] S133: Retrieve the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point during the process of generating collaborative diagnostic results, and construct time-series input data based on the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point at continuous sampling time.

[0168] In a specific embodiment, the component stress characterization value, the rate of change of the current main peak frequency relative to the main peak frequency baseline, and the rate of change of the first-order bending natural frequency of the top plate relative to the first-order bending natural frequency baseline of the top plate are retrieved at each measuring point in the same monitoring zone at continuous sampling time.

[0169] The component stress characteristics, main peak frequency change rate, and first-order bending natural frequency change rate of the top plate at each of the above measuring points at continuous sampling times are arranged in chronological order to construct a three-dimensional feature tensor. Its dimensions are defined as time step, number of measurement points, and number of features, and the three-dimensional feature tensor is... As time-series input data, the number of features corresponds to the component stress characterization value, the rate of change of the main peak frequency and the rate of change of the first-order bending natural frequency of the top plate.

[0170] S134: Input the time series input data into a predetermined time series deep learning model to obtain the risk probability;

[0171] Input timing data Input a pre-trained temporal deep learning model, which extracts spatial features through convolutional layers, captures temporal failure modes through a long short-term memory network, performs probability calibration at the end of the model through a Softmax activation function, and outputs a risk probability distribution vector.

[0172] ;

[0173] In the formula, For risk probability, , , These represent the posterior probabilities corresponding to low, medium, and high risk levels, respectively.

[0174] It should be noted that the actual input layer of the model is time-series input data. The number of features in the text corresponds to the component's stress characterization value, the rate of change of the main peak frequency, and the rate of change of the first-order bending natural frequency of the top plate.

[0175] The generation steps of the time-series deep learning model are as follows:

[0176] Historical monitoring data is acquired and divided into a model training set and a model test set. The historical monitoring data includes the component stress characterization value, main peak frequency change rate, first-order bending natural frequency change rate of the top plate, and corresponding risk classification results at each measuring point at continuous sampling time.

[0177] It should be noted that the risk classification results are obtained by manual or rule-based annotation based on the on-site verification records, handling records, and post-handling continuous monitoring records corresponding to historical early warning events.

[0178] Configure an initial classification network, which includes convolutional layers, a long short-term memory network, and a classification output layer; construct a three-dimensional feature tensor from the component stress characterization values, main peak frequency change rate, and first-order bending natural frequency change rate of the top plate in the model training set according to the time order and measurement point order. The three-dimensional feature tensor is used as the input data of the initial classification network, and the risk classification results corresponding to the model training set are used as the output data of the initial classification network. The initial classification network is trained to obtain the initial temporal deep learning model.

[0179] The initial temporal deep learning model is validated using a model test set. An initial temporal deep learning model with a classification accuracy greater than or equal to the preset accuracy is output and used as the temporal deep learning model.

[0180] S135: Based on the risk probability and collaborative diagnosis results, perform corresponding matching to generate risk classification results;

[0181] Specifically, the risk classification results include high risk, medium risk, and low risk. The logic for generating the risk classification results is to retrieve the posterior probability of high risk from the risk probabilities. ;

[0182] when ≥80%, and collaborative diagnostic results When =1, it is judged as high risk in the risk classification result;

[0183] When 60%≤ <80% and collaborative diagnostic results =1, or <60% but collaborative diagnostic results When =1, it is determined to be medium risk in the risk classification results;

[0184] when ≤60% and collaborative diagnostic results When the value is 0, it is considered low risk;

[0185] The final output includes a risk classification result containing risk level and risk probability.

[0186] S14: Generate early warning results and governance priorities based on collaborative diagnosis results, risk probability and risk classification results, and retrieve the on-site verification records and monitoring records corresponding to the early warning results to correct the frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate.

[0187] Specifically, the steps for generating early warning results and treatment priorities, and correcting frequency-stress correlation data and the baseline of the first-order bending natural frequency of the top plate are as follows:

[0188] S141: Generate early warning results and governance priorities based on collaborative diagnosis results, risk probabilities, and risk classification results;

[0189] In a specific embodiment, the artificial intelligence module 110 matches the generated risk classification results with the preset early warning response rules to generate an early warning result; the early warning result includes risk location information and risk level color indicators, wherein the risk location information is generated based on the monitoring partition code and the corresponding spatial measurement point location, and the risk level color indicators correspond to green for low risk, yellow for medium risk, and red for high risk.

[0190] After generating the early warning results, governance priorities are generated based on the severity of the risk classification results and the magnitude of the risk probability.

[0191] For multiple monitoring zones with the same risk classification results, they are sorted from high to low risk probability;

[0192] For multiple monitoring zones with the same risk probability, the monitoring zone with the collaborative diagnosis result of a risk event should be given higher governance priority.

[0193] S142: Retrieve the on-site verification records and monitoring records corresponding to the early warning results, and extract the on-site verification results and subsequent response results;

[0194] In a specific embodiment, the calibration and incremental learning module 111 retrieves the monitoring records of the monitoring partition before and after the warning is triggered, as well as the on-site verification records provided by the operation and maintenance personnel, based on the monitoring partition corresponding to the warning result and the warning trigger time.

[0195] Extract the on-site verification results from the on-site verification records. The on-site verification results include, but are not limited to, whether the components are broken, whether the support is loose, and whether there is delamination.

[0196] Subsequent response results are extracted from the continuous monitoring records after the treatment. These subsequent response results include the change in the stress characterization value of the component, the rise in the main peak frequency of the exposed section of the anchor bolt or the main peak frequency of the exposed section of the anchor cable, and the change in the first-order bending natural frequency of the top plate relative to the baseline of the first-order bending natural frequency of the top plate.

[0197] S143: Correct the frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate based on the on-site verification results and subsequent response results;

[0198] In a specific embodiment, the calibration and incremental learning module 111 writes back the on-site verification results and subsequent response results to the corresponding monitoring partition;

[0199] When subsequent response results show that the variation amplitude of the first-order bending natural frequency of the roof after treatment is less than the preset frequency fluctuation threshold at multiple consecutive sampling times, and the deviation of the component stress characterization value of the corresponding monitoring zone from the historical stable working condition average at multiple consecutive sampling times is less than the preset stress fluctuation threshold, the first-order bending natural frequency of the roof corresponding to that period is recalibrated as the baseline of the first-order bending natural frequency of the roof of that monitoring zone. ;

[0200] When the on-site verification results show that the component is loose or broken, and the component stress characterization value at the corresponding continuous sampling time in the monitoring record drops more than the preset stress change threshold, while the first-order bending natural frequency of the top plate does not drop more than the preset frequency change threshold at the current sampling time or within the preset lag sampling window, or when the on-site verification results show that the support status has recovered but the component stress characterization value and the first-order bending natural frequency of the top plate do not show a synchronous recovery trend, it is determined that the frequency stress correlation data of the current monitoring zone is inconsistent with the actual monitoring status, and the recalibration process is triggered.

[0201] In the recalibration process, the updated peak frequencies of the exposed anchor bolt sections and the exposed anchor cable sections, as well as the corresponding on-site verification results, are retrieved for the monitoring zone. The frequency-force correlation data of the monitoring zone is then re-established to complete the dynamic correction of the frequency-force correlation data of the monitoring zone.

[0202] Example 2

[0203] Please see Figure 2 As shown, based on a unified inventive concept, this embodiment discloses a roof fall risk prediction system based on fiber optic acoustic wave sensing measurement, the system comprising:

[0204] S21: Distributed single-mode sensing fiber is laid along the direction of the roadway roof. The distributed single-mode sensing fiber is connected to the exposed section of the anchor bolt, the exposed section of the anchor cable and the roof monitoring section in sequence to generate a monitoring zone. The initial dynamic response signal is collected based on the monitoring zone and the first-order bending natural frequency baseline of the roof is determined.

[0205] Frequency-force correlation module S22: Collects dynamic response signals according to the monitoring zone, and extracts the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the top plate according to the dynamic response signals. Generates frequency-force correlation data based on the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable.

[0206] Diagnostic grading module S23: Generates component stress characterization values ​​and collaborative diagnostic results based on frequency-force correlation data, main peak frequency of exposed anchor bolt section, main peak frequency of exposed anchor cable section, first-order bending natural frequency of top plate and baseline of first-order bending natural frequency of top plate, and generates risk probability and risk grading results based on collaborative diagnostic results;

[0207] Monitoring baseline module S24: Generates early warning results and governance priorities based on collaborative diagnosis results, risk probability and risk classification results, and retrieves the on-site verification records and monitoring records corresponding to the early warning results, and corrects the frequency-force correlation data and the first-order bending natural frequency baseline of the top plate.

[0208] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for predicting roof collapse risk based on fiber optic acoustic wave sensing, characterized in that, The method includes: Distributed single-mode sensing fibers are laid along the roof of the tunnel. These fibers are then sequentially connected to exposed sections of anchor bolts, exposed sections of anchor cables, and roof monitoring sections to generate monitoring zones. Initial dynamic response signals are acquired based on these monitoring zones, and the baseline of the first-order bending natural frequency of the roof is determined, including: Start the distributed fiber optic acoustic wave sensor host to collect the initial dynamic response signal of each monitoring zone, and extract the roof response signal corresponding to the roof monitoring section from the initial dynamic response signal. The top plate response signal is filtered to obtain the first-order bending natural frequency of the top plate. The first-order bending natural frequency of the roof corresponding to each monitoring zone is associated with the monitoring zone to generate the first-order bending natural frequency baseline of the roof. Dynamic response signals are collected based on the monitoring zones, and the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the top plate are extracted based on the dynamic response signals. The steps for extracting the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable are as follows: Extract the response signals of exposed anchor bolt sections and exposed anchor cable sections from the dynamic response signals according to the monitoring zones; For the response signals of the exposed sections of the anchor bolts and anchor cables, component frequency band screening and time-frequency analysis are performed to obtain candidate peak frequencies; A peak shape consistency check is performed on the candidate peak frequencies, and the main peak frequencies of the exposed sections of the anchor bolt and anchor cable are output. The peak shape consistency check is implemented through a quality gating inspection. The steps for extracting the first-order bending natural frequency of the top plate are as follows: According to the monitoring zones, extract the roof response signal corresponding to the roof monitoring section from the dynamic response signal; The roof response signal is subjected to roof frequency band screening and multi-point coherence analysis to obtain candidate roof frequency components; Modal identification is performed on the candidate frequency components of the top plate, and the first-order bending natural frequency of the top plate is output. Based on the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable, frequency-force correlation data is generated. Based on the frequency-force correlation data, the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, the first-order bending natural frequency of the top plate, and the baseline of the first-order bending natural frequency of the top plate, the component stress characterization value and collaborative diagnosis results are generated, and the risk probability and risk classification results are generated based on the collaborative diagnosis results. Based on the collaborative diagnosis results, risk probability, and risk classification results, early warning results and governance priorities are generated. The on-site verification records and monitoring records corresponding to the early warning results are retrieved, and the frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate are corrected.

2. The method for predicting roof collapse risk based on fiber optic acoustic wave sensing according to claim 1, characterized in that, The steps to generate a monitoring partition are as follows: Distributed single-mode sensing optical fibers are laid along the direction of the tunnel roof, and continuous deployment paths are established at corresponding locations in the exposed sections of anchor bolts, exposed sections of anchor cables, and roof monitoring sections. Distributed single-mode sensing optical fibers are fixed to the monitoring section of the top plate by a surface composite coating, and the distributed single-mode sensing optical fibers are connected to the exposed sections of the anchor bolts and anchor cables by a clamping mechanism to form a series continuous coupling structure. The exposed sections of anchor bolts, exposed sections of anchor cables, and monitoring sections of the top plate that are in the same continuous deployment path are combined to generate monitoring zones.

3. The method for predicting roof collapse risk based on fiber optic acoustic wave sensing according to claim 2, characterized in that, The steps for generating frequency-force correlation data are as follows: Select field calibration points in the exposed sections of anchor bolts and anchor cables, and apply tension perturbations to the field calibration points step by step; The main peak frequencies of the exposed sections of the anchor bolt and the anchor cable corresponding to the tension perturbation were collected simultaneously, and the applied tension corresponding to the tension perturbation was monotonically fitted with the main peak frequencies of the exposed sections of the anchor bolt and the anchor cable. The monotonic fitting results are correlated with the corresponding monitoring zones to generate frequency-force correlation data.

4. The method for predicting roof collapse risk based on fiber optic acoustic wave sensing according to claim 3, characterized in that, The steps for generating component stress characterization values ​​and collaborative diagnostic results are as follows: Retrieve frequency-force correlation data, and based on the frequency-force correlation data, the main peak frequency of the exposed section of the anchor bolt and the main peak frequency of the exposed section of the anchor cable, generate anchor bolt force characterization values ​​and anchor cable force characterization values, and integrate them to obtain component force characterization values; The stress characterization values ​​of the component, the first-order bending natural frequency of the top plate, and the baseline of the first-order bending natural frequency of the top plate are spatiotemporally registered to generate collaborative diagnostic results.

5. The method for predicting roof collapse risk based on fiber optic acoustic wave sensing according to claim 4, characterized in that, The steps to generate risk probability and risk classification results are as follows: During the process of generating collaborative diagnostic results, the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point are retrieved, and time-series input data are constructed based on the component stress characterization value, main peak frequency change rate and top plate first-order bending natural frequency change rate of each measuring point at continuous sampling time. Input the time-series input data into a predetermined time-series deep learning model to obtain the risk probability; Risk classification results are generated by matching risk probabilities with collaborative diagnostic results.

6. The method for predicting roof collapse risk based on fiber optic acoustic wave sensing according to claim 5, characterized in that, The steps for generating early warning results and treatment priorities, and correcting frequency-stress correlation data and the baseline of the first-order bending natural frequency of the top plate are as follows: Based on the collaborative diagnosis results, risk probability, and risk classification results, early warning results and governance priorities are generated. Retrieve the on-site verification records and monitoring records corresponding to the early warning results, and extract the on-site verification results and subsequent response results; The frequency-force correlation data and the baseline of the first-order bending natural frequency of the top plate were corrected based on the on-site verification results and subsequent response results.

7. A roof fall risk prediction system based on fiber optic acoustic wave sensing, used to execute the roof fall risk prediction method based on fiber optic acoustic wave sensing as described in any one of claims 1-6, characterized in that, The system includes: Zoning baseline module: Distributed single-mode sensing optical fibers are laid along the direction of the roadway roof. The distributed single-mode sensing optical fibers are connected in sequence to the exposed sections of anchor bolts, exposed sections of anchor cables and roof monitoring sections to generate monitoring zones. Based on the monitoring zones, the initial dynamic response signals are collected and the first-order bending natural frequency baseline of the roof is determined. Frequency-force correlation module: Collects dynamic response signals based on monitoring zones, and extracts the main peak frequency of the exposed section of the anchor bolt, the main peak frequency of the exposed section of the anchor cable, and the first-order bending natural frequency of the top plate based on the dynamic response signals. Generates frequency-force correlation data based on the main peak frequencies of the exposed sections of the anchor bolt and the anchor cable. Diagnostic grading module: Based on frequency-force correlation data, the main peak frequency of exposed anchor bolt section, the main peak frequency of exposed anchor cable section, the first-order bending natural frequency of top plate, and the baseline of the first-order bending natural frequency of top plate, the component stress characterization value and collaborative diagnosis results are generated, and the risk probability and risk grading results are generated based on the collaborative diagnosis results. Monitoring baseline module: Generates early warning results and treatment priorities based on collaborative diagnosis results, risk probability and risk classification results, and retrieves the on-site verification records and monitoring records corresponding to the early warning results to correct the frequency-force correlation data and the first-order bending natural frequency baseline of the top plate.

Images