Loess tunnel surrounding rock deformation monitoring method

Abstract

The invention discloses a loess tunnel surrounding rock deformation monitoring method, and relates to the technical field of civil engineering and geological engineering, and the method comprises the steps: integrating a spiral winding type optical fiber sensor, a double-cavity humidity compensation vibrating wire sensor and a microseismic array, and capturing surrounding rock strain, vibration and geological activities; the vibrating wire sensor suppresses humidity interference through a silicone oil damping medium and self-adaptive excitation frequency, and the optical fiber sensor is fixed through a pre-embedded silica gel sleeve to adapt to surrounding rock deformation; edge computing nodes are deployed on the inner wall of the tunnel, an FPGA chip and a lightweight GRU model are integrated, optical fiber strain, a micro-seismic energy spectrum and laser point cloud displacement field data are fused in real time, the strain gradient is analyzed, and a crack propagation probability cloud picture is generated; redundant optical fiber link switching is combined with a six-degree-of-freedom mechanical arm to realize breakpoint self-repairing, and a self-cleaning air curtain is integrated to inhibit dust adhesion; and constructing a geological parameter library based on a BIM-GIS fusion platform, driving a finite element-discrete element coupling model to dynamically update boundary conditions, and predicting surrounding rock deformation and collapse risks.

Description

Technical Field

[0001] The present invention relates to the technical fields of civil engineering and geological engineering, and specifically relates to a method for monitoring the deformation of surrounding rock in loess tunnels. Background Technique

[0002] Loess is widely distributed in China. With the advancement of infrastructure construction such as transportation and water conservancy, the quantity and scale of loess tunnels, including railway tunnels, highway tunnels, and hydraulic tunnels, are continuously increasing. The special physical and mechanical properties of loess (low strength, high porosity, easy collapsibility) lead to a significantly higher risk of surrounding rock deformation in tunnels than that in ordinary rock and soil tunnels. There is an urgent need for targeted monitoring methods to ensure project safety.

[0003] During the construction process of loess tunnels, deformation problems such as crown settlement, side wall convergence, and floor heave are likely to occur. In severe cases, it may trigger collapse accidents. During the operation period, affected by factors such as groundwater seepage and loess collapsibility, the long-term stability of the surrounding rock is threatened. Accurately monitoring the deformation trend is the core means to prevent disasters and optimize the support plan. Traditional manual monitoring has low efficiency and poor real-time performance, and it is difficult to capture sudden deformations. With the popularization of technologies such as the Internet of Things, big data, and artificial intelligence, automated, high-precision, and real-time monitoring technologies have become the industry trend, promoting the upgrade of loess tunnel monitoring towards the intelligent direction.

[0004] However, traditional monitoring methods have great limitations. Contact measurement has low efficiency. Total stations and convergence meters require manual operation, and data updating lags behind the rapid deformation stage. Point sensors have insufficient coverage and can only capture local deformations, making it difficult to reflect the overall dynamics of the surrounding rock. The adaptability of sensor technology is insufficient, being sensitive to environmental interference. Changes in loess humidity easily cause drift of vibrating wire sensors, and the layout of fiber optic sensors is easily damaged by construction interference. The long-term stability is poor. The monitoring period is as long as several months to several years, and existing devices such as resistance strain gauges are prone to aging and failure. Summary of the Invention

[0005] To solve the above technical problems, a method for monitoring the deformation of surrounding rock in loess tunnels is provided. This technical solution solves the limitations of the above traditional monitoring methods and the insufficient adaptability of sensor technology.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is as follows: A method for monitoring the deformation of surrounding rock in loess tunnels, comprising: S1. Arrange annular monitoring sections at intervals along the longitudinal direction of the tunnel. Deploy a distributed fiber optic sensor array, a micro-vibration sensor, and a humidity-compensated vibrating wire sensor in each section; S2. The optical fiber sensor adopts a spirally wound packaging structure, is covered with a shear-resistant flexible sheath, and is fixed to the surface of the surrounding rock through a pre-buried silicone sleeve; the humidity compensation vibrating string sensor has a built-in double-cavity sealing structure, the outer cavity is filled with silicone oil damping medium, and the excitation frequency range is set to suppress the interference of loess humidity; S3. Deploy edge computing nodes on the inner wall of the tunnel, integrate FPGA chips and lightweight GRU neural network models, and integrate optical fiber strain data, microseismic event energy spectrum, and 3D laser scanning point cloud displacement field in real time; use wavelet packet transform to denoise the optical fiber data, extract characteristic components of specific frequency bands, and calculate the surrounding rock strain gradient distribution; combine the spatiotemporal evolution of microseismic events to generate a crack extension probability cloud map; S4, through the redundant optical fiber link switching mechanism and the robot arm auxiliary deployment device, the sensor breakpoints caused by construction vibration can be repaired in real time; the sensor shell is made of titanium alloy plating and nano-hydrophobic coating, and the surface is integrated with a self-cleaning air curtain device to resist dust adhesion; S5. Based on the BIM-GIS fusion model, a tunnel geological parameter database is constructed, including initial ground stress, collapsibility coefficient and permeability curve; the boundary conditions of the finite element-discrete element coupling model are regularly updated to predict the deformation of the surrounding rock and the risk level of collapse.

[0007] Preferably, the S1 specifically includes: The layout of the circular monitoring section is based on the longitudinal spacing control, and the spacing is dynamically adjusted according to the collapsibility grade of the loess; the layout is densely packed within the given range of the tunnel entrance; Sensors deployed in each section must cover the vault, waist, side walls and invert areas; control the radial deviation of the section and align the three-dimensional data space.

[0008] Preferably, the S1 specifically includes: The microseismic sensor array layout includes multiple three-component microseismic sensors deployed in each section in a regular tetrahedron topology distribution. The sensors are embedded in the ends of the anchor bolts to ensure the anchoring depth. Adaptive trigger thresholds are used, and the background noise amplitude and 6dB are used as trigger thresholds. The humidity compensated vibrating string sensor uses a humidity interference suppression mechanism, dynamically adjusts the excitation frequency range, locks humidity-insensitive frequencies through frequency sweeping, has a built-in micro humidity sensor, and provides real-time feedback of compensation parameters. It uses an improved Kalman filter to fuse vibration frequency, temperature and humidity, and pore water pressure data to output effective stress values.

[0009] Preferably, the S2 specifically includes: The fiber optic sensor is packaged in a spiral winding manner, and the predetermined winding angle is locked after verification by finite element simulation. The spiral radius matches the curvature of the tunnel section; The anti-shear flexible sheath uses a polyurethane matrix doped with a certain proportion of carbon nanotubes; a fluorosilane modified layer is coated on the surface; when installing the embedded silicone sleeve, a spiral spring buffer structure is embedded.

[0010] Preferably, the S2 specifically includes: The humidity compensation type vibrating wire sensor has a double-chamber seal. The inner chamber is vacuum-sealed, and the vibrating wire is made of beryllium bronze; the outer chamber is filled with silicone oil damping medium, which is synchronized with the thermodynamic response of the pore water in the loess. Humidity interference suppression, multi-frequency excitation scanning, dynamically adjusting the excitation frequency, identifying the humidity-insensitive frequency points through the frequency response curve; a built-in micro humidity sensor for real-time feedback, and realizing the correction of the vibration frequency-stress conversion through the in-situ compensation algorithm. Signal processing enhancement, the Kalman filter fuses the vibration frequency, temperature, and pore water pressure data to output the effective stress value, and the sampling interval is dynamically adjusted.

[0011] Preferably, the S3 specifically includes: The core processor of the edge computing node is an FPGA chip of the Xilinx Zynq UltraScale+ series, integrating a quad-core ARM Cortex-A53 and a programmable logic unit; supporting the synchronous operation of fiber optic data processing and microseismic analysis; a fiber optic demodulation device, integrating a four-channel phase-sensitive optical time domain reflectometer, and the microseismic signal acquisition is a 24-bit high-precision ADC; the protection is set in an IP68 waterproof and explosion-proof box, with an active heat dissipation air duct built-in and an anti-electromagnetic interference shielding layer. The lightweight GRU neural network model architecture includes an input layer, six-channel fiber optic strain, microseismic three-axis energy spectrum, and 3D components of the laser displacement field; an output layer, crack propagation probability and stability coefficient.

[0012] Preferably, the S3 specifically includes: Perform wavelet packet transform on the fiber optic data, select the db4 wavelet basis function, and improve the signal-to-noise ratio based on threshold denoising; calculate the strain gradient, calculate the spatial derivative of the strain based on the difference method, and generate a cross-sectional strain gradient distribution cloud map. Microseismic-laser data fusion, localize and map the energy of microseismic events, and the wave velocity model is anisotropic velocity inversion; calculate the energy spectrum, generate a three-dimensional matrix of energy-frequency-time through short-time Fourier transform; for the three-dimensional laser point cloud, improve the iterative closest point algorithm and introduce normal vector constraints; calculate the displacement vector based on the difference between adjacent frame point clouds to generate a displacement field. Construct a microseismic energy-strain gradient-displacement correlation matrix, and extract the precursor characteristics of surrounding rock failure through principal component analysis. Divide the space-time grid, the probability model includes inputs, strain gradient, microseismic cumulative energy, and displacement rate; output, probability calculation based on the Bayesian network, and generate a crack propagation probability cloud map.

[0013] Preferably, S4 specifically includes: Deploy three independent optical fiber links at each monitoring section, and the switching priority is dynamically adjusted according to the signal attenuation value; based on the optical time domain reflectometry principle, distinguish the shear fracture of construction machinery and the fiber breakage due to surrounding rock creep through the break point type recognition algorithm. The robotic arm-assisted layout device is a six-degree-of-freedom robotic arm, with an optical fiber fusion splicer and a coating module integrated at the end; adopt a dual-mode navigation system, combining laser SLAM and UWB positioning; adopt an intelligent repair strategy for break point repair, and the repair process is OTDR positioning, break point cleaning, fusion splicing, and coating reinforcement; set an autonomous obstacle avoidance algorithm, receive the pose data of construction machinery in real time, and plan a collision-free path.

[0014] Preferably, S4 specifically includes: The sensor housing adopts a titanium alloy coating structure, and the coating process is vacuum plasma spraying of a titanium alloy layer; and a nano-hydrophobic coating is set, including a fluorosilicon polymer matrix doped with SiO2 nanoparticles, with a self-healing function, and the coating microcracks trigger the nanoparticle migration and filling mechanism. Self-cleaning air curtain device, the air curtain forms an axial airflow barrier covering the full angle through an annular porous nozzle and a high-pressure air pump; implement an energy-saving control strategy, start and stop adaptively according to the dust concentration, start when the PM10 is greater than the established concentration, and standby when it is less than the established concentration.

[0015] Preferably, S5 specifically includes: The BIM model level is modeled with LOD400 accuracy, integrating the tunnel structure, the pose of construction machinery, and the topology of monitoring equipment; dynamically bind the driving mileage and geological parameters, and associate the construction progress; the GIS geological database includes a spatial database, drilling data, and geophysical profiles; update the in-situ stress field through microseismic events and fiber strain inversion. The tunnel geological parameter library includes the initial in-situ stress, which is measured by the hydraulic fracturing method and generated as a non-uniform field by a three-dimensional inversion algorithm; the collapsibility coefficient, which is calibrated based on indoor compression tests and on-site soaking load tests; the permeability curve, which is modeled by combining transient permeability tests and CT scans of pore structures. Update the finite element-discrete element coupling model, based on the model coupling algorithm, divide the continuous-discontinuous domain, the finite element region is the undisturbed surrounding rock; the discrete element region is the fracture zone and the potential fracture network; the data exchange interface is synchronized regularly through stress-displacement boundary conditions, and the fracture propagation path is fed back to the GIS database. Dynamically update the boundary conditions, the construction disturbance factors include the excavation unloading effect, and adjust the in-situ stress release coefficient according to the driving speed; the support effect, the pre-tightening force of the bolt and the lining stiffness are input into the FEM model in real time; the environmental interaction parameters include seepage-stress coupling, and the permeability curve drives the update of the pore water pressure field; humidity-creep correlation, and the collapsibility coefficient corrects the rheological constitutive equation of the surrounding rock.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention proposes to arrange annular monitoring sections at intervals along the longitudinal direction of the tunnel, integrating a spiral-wound optical fiber sensor, a double-chamber humidity-compensated vibrating wire sensor, and a microseismic array to form a multi-dimensional signal capture network; the optical fiber sensor adopts a shear-resistant flexible sheath and a pre-embedded silica gel sleeve fixing structure, and improves the strain transfer efficiency through spiral winding encapsulation to adapt to the dynamic deformation of the surrounding rock; the vibrating wire sensor effectively suppresses the interference of the high-humidity loess environment on the monitoring signal through a silicone oil damping medium and an adaptive excitation frequency adjustment mechanism; the sensor housing adopts a titanium alloy coating and a self-repairing nano-hydrophobic coating, combined with a self-cleaning air curtain device, significantly reducing the risk of dust adhesion and corrosion, and ensuring the long-term stability of the equipment in harsh environments.

[0017] Regarding the problem of sensor breakpoints caused by construction vibrations, the system adopts a redundant optical fiber link dynamic switching mechanism and a six-degree-of-freedom robotic arm for collaborative repair; the robotic arm locates the breakpoints through laser SLAM and UWB dual-mode navigation, and completes the integrated repair of fusion and coating, supporting collision-free path planning in complex construction environments. At the same time, a geological parameter database is constructed based on the BIM-GIS fusion platform, integrating the initial in-situ stress field, collapsibility coefficient, and permeability curve, driving the dynamic update of the boundary conditions of the finite element-discrete element coupling model, predicting the deformation trend of the surrounding rock and the risk level of collapse, and providing a quantitative basis for construction control and emergency plans. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 is a flowchart of a method for monitoring the deformation of the surrounding rock of a loess tunnel; Figure 2 is a flowchart of S1; Figure 3 is a flowchart of S2; Figure 4 is a flowchart of S3. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0019] The following description is used to disclose the present invention so that those skilled in the art can implement the present invention. The preferred embodiments described below are only examples, and those skilled in the art can think of other obvious variations.

[0020] Referring to Figure 1 as shown, a method for monitoring the deformation of the surrounding rock of a loess tunnel includes: S1. Arrange annular monitoring sections at intervals along the longitudinal direction of the tunnel, and deploy a distributed optical fiber sensor array, a micro-vibration sensor, and a humidity-compensated vibrating wire sensor at each section; S2. The optical fiber sensor adopts a spirally wound packaging structure, is covered with a shear-resistant flexible sheath, and is fixed to the surface of the surrounding rock through a pre-buried silicone sleeve; the humidity compensation vibrating string sensor has a built-in double-cavity sealing structure, the outer cavity is filled with silicone oil damping medium, and the excitation frequency range is set to suppress the interference of loess humidity; S3. Deploy edge computing nodes on the inner wall of the tunnel, integrate FPGA chips and lightweight GRU neural network models, and integrate optical fiber strain data, microseismic event energy spectrum, and 3D laser scanning point cloud displacement field in real time; use wavelet packet transform to denoise the optical fiber data, extract characteristic components of specific frequency bands, and calculate the surrounding rock strain gradient distribution; combine the spatiotemporal evolution of microseismic events to generate a crack extension probability cloud map; S4, through the redundant optical fiber link switching mechanism and the robot arm auxiliary deployment device, the sensor breakpoints caused by construction vibration can be repaired in real time; the sensor shell is made of titanium alloy plating and nano-hydrophobic coating, and the surface is integrated with a self-cleaning air curtain device to resist dust adhesion; S5. Based on the BIM-GIS fusion model, a tunnel geological parameter database is constructed, including initial ground stress, collapsibility coefficient and permeability curve; the boundary conditions of the finite element-discrete element coupling model are regularly updated to predict the deformation of the surrounding rock and the risk level of collapse.

[0021] It should be noted that the construction of a multimodal sensing network, full-area coverage and complementary verification, distributed optical fiber capturing continuous deformation, microseismic sensors locating local damage, and vibrating string sensors verifying load transfer, the three form a multi-dimensional data closed loop; the humidity compensation mechanism and self-cleaning device work together to resist the wetting and high dust environment interference of loess to ensure data reliability.

[0022] Dynamic adaptive adjustment: the monitoring section spacing and finite element model parameters are updated in conjunction with the loess subsidence level to achieve dynamic adaptation of “monitoring-model-early warning”.

[0023] Edge computing responds in real time, and FPGA chips complete filtering and feature extraction of high-frequency data, reducing transmission bandwidth requirements; lightweight GRU models enable initial risk assessment on-site; Deep optimization in the cloud: after receiving the pre-processed data from the edge nodes, the BIM-GIS model updates the boundary conditions of the coupling model and iteratively optimizes the long-term prediction accuracy; Full life cycle closed-loop management: Construction period: Robotic arm-assisted deployment and redundant fiber switching ensure network integrity; Operational period: Self-cleaning air curtains and nano-coatings extend equipment life, and digital twin models support long-term health monitoring; Disaster response: The three-level early warning mechanism triggers the emergency plan, and the crack probability cloud map is combined to guide the rescue route.

[0024] Encapsulation and materials for sensor anti-interference, spiral-wound optical fiber reduces shear stress concentration, and the tensile strength of the flexible sheath reaches 85 MPa; the corrosion resistance of the titanium alloy coating is increased by 300%, the contact angle of the nano-hydrophobic coating is > 150°, and the dust adhesion rate is < 0.3%.

[0025] Dynamic compensation algorithm, the vibration frequency-humidity decoupling algorithm reduces the stress measurement error from ±2%FS to ±0.5%FS.

[0026] Data fusion depth, wavelet packet transform extracts strain gradient features, microseismic energy spectrum correlates with crack propagation rate, laser point cloud quantifies the non-uniformity of the displacement field, and multi-source heterogeneous data is spatio-temporally correlated through the GRU model; Model coupling accuracy, the FEM-DEM model introduces a collapsibility coefficient to dynamically correct the rheological constitutive equation, and the collapse prediction accuracy is > 90%.

[0027] Refer to Figure 2 As shown, the specific S1 includes: Layout of the circular monitoring section, controlled according to the longitudinal spacing, and dynamically adjusted according to the collapsibility grade of loess; densified layout within the established range of the tunnel portal; Each section needs to deploy sensors to cover the crown, arch waist, side wall and invert areas; control the radial deviation of the section and align the three-dimensional data space; Layout of the microseismic sensor array, including deploying multiple three-component microseismic sensors at each section, distributed in a regular tetrahedron topology, with the sensors embedded at the end of the anchor bolt to ensure the anchoring depth; using an adaptive trigger threshold, with the background noise amplitude and 6 dB as the trigger threshold; Humidity compensation type vibrating wire sensor, using a humidity interference suppression mechanism, the dynamic adjustment range of the excitation frequency, locking the humidity-insensitive frequency point through frequency sweeping, with a built-in micro humidity sensor to provide real-time feedback of compensation parameters; using an improved Kalman filter to fuse vibration frequency, temperature, humidity and pore water pressure data to output the effective stress value.

[0028] It should be noted that the dynamic spacing control mechanism, the collapsibility grade-spacing mapping rule:

[0029] Principle of densified layout at the portal, within 50 m of the portal, due to the severe stress redistribution, a 2 m spacing is adopted and verified by superimposing the laser displacement field; Sensor deployment topology: Crown: 2 groups of optical fibers + 1 vibrating wire sensor (monitoring tensile deformation); Arch waist: 3 groups of optical fibers + 2 microseismic sensors (capturing shear slip); Side wall / invert: 3 groups of optical fibers + 1 vibrating wire sensor (monitoring extrusion and floor heave).

[0030] Spatial calibration technology, based on a three-dimensional laser scanner to establish a cross-section coordinate system, and the radial deviation is fine-tuned by a robotic arm to ensure the spatial consistency of multi-source data.

[0031] Microseismic sensor array: Improved tetrahedral topological positioning accuracy. A regular tetrahedron is formed by 4 three-component sensors. The fracture source positioning error is compressed from 1.5 m to 0.5 m through a wave velocity inversion algorithm; anti-interference design, embedded installation at the end of the anchor rod (anchoring depth ≥ 1.5 m) isolates the vibration of construction machinery, and the signal-to-noise ratio is increased by 20 dB; Adaptive trigger threshold algorithm, dynamic noise baseline, updating the background noise amplitude (in the 0.1 - 10 Hz frequency band) every 10 seconds, trigger threshold = baseline value + 6 dB (suppressing 80% of false alarm events); Event classification mechanism: Type A events (energy > 1 kJ): Immediate warning; Type B events (100 J - 1 kJ): Associated strain gradient analysis; Type C events (< 100 J): Only recorded and not stored in the database.

[0032] Humidity-compensated vibrating wire sensor: Dual-engine for humidity interference suppression, active suppression: Dynamically scan the excitation frequency (40 - 60 Hz), and lock 52 Hz as the humidity-insensitive frequency point through the slope analysis of the frequency response curve (drift rate < 0.02%RH / 10% humidity change); Passive compensation: A micro humidity sensor (accuracy of ±1%RH) corrects the vibration frequency - stress conversion formula in real time:

[0033] In the formula, is the stress, which is the output quantity in the vibration frequency - stress conversion formula and reflects the stress state of the material or structure under the change of vibration frequency; is the proportionality coefficient, which is used to convert the frequency change into a stress change and reflects the inherent characteristics of the material or structure; is the current vibration frequency, representing the currently measured vibration frequency; is the reference vibration frequency, usually the initial vibration frequency measured under standard humidity conditions; is the humidity sensitivity coefficient, indicating the degree of influence of humidity change on the vibration frequency - stress conversion; is the current relative humidity, representing the relative humidity value in the current environment; is the reference relative humidity, usually the initial relative humidity value measured under standard humidity conditions; Kalman filter data fusion, input parameters: vibration frequency (main data source), temperature (PT1000 sensor, ±0.1°C), pore water pressure (piezoelectric ceramic sensor, 0 - 2MPa range); output optimization, the comprehensive error of the effective stress value after fusion ≤ ±0.5%FS, and the sampling rate is dynamically switched according to the deformation rate (steady state 10Hz / mutation 100Hz).

[0034] Multi-source data cross-validation, strain - microseismic correlation. When the strain gradient of a certain section > 1.2 με / mm and is accompanied by type B microseismic events, it is determined as a high-risk area for crack initiation; Stress-displacement coupling. When the sudden change in stress output by the vibrating wire sensor > 15%, trigger the 3D laser scanner to perform encrypted sampling on this area (point cloud density → 5000 points / m²).

[0035] Refer to Figure 3 As shown, the S2 specifically includes: The optical fiber sensor is encapsulated in a spiral winding. After being verified by finite element simulation, the established winding angle is locked, The spiral radius matches the curvature of the tunnel section; The anti-shear flexible sheath uses a polyurethane matrix doped with a certain proportion of carbon nanotubes; a fluorosilane modified layer is coated on the surface; when installing the embedded silicone rubber sleeve, a spiral spring buffer structure is embedded; The humidity compensation type vibrating wire sensor has a double-chamber seal. The inner chamber is vacuum-sealed, and the vibrating wire is made of beryllium bronze; the outer chamber is filled with silicone oil damping medium, which is synchronized with the thermodynamics response of the pore water in the loess; Humidity interference suppression, multi-frequency excitation scanning, dynamically adjust the excitation frequency, identify the humidity-insensitive frequency points through the frequency response curve; an internal micro humidity sensor provides real-time feedback, and through the in-situ compensation algorithm, realize the correction of the vibration frequency - stress conversion; Signal processing enhancement, the Kalman filter fuses the vibration frequency, temperature, and pore water pressure data, outputs the effective stress value, and the sampling interval is dynamically adjusted.

[0036] It should be noted that the optical fiber sensor encapsulation technology: Verified by finite element simulation. Through ANSYS Workbench to simulate the surrounding rock deformation conditions, the optimal spiral winding angle is determined to be 15° ± 2°, ensuring that the strain transfer efficiency > 95%; Curvature adaptive matching, the spiral radius is dynamically adjusted according to the tunnel section (the radius of the vault is 1.2m / the side wall is 0.8m), suppressing the micro-bending loss of the optical fiber (< 0.02dB / km); Buffer structure design, a spiral spring (stiffness coefficient 5kN / m) is embedded in the silicone rubber sleeve, allowing a displacement buffer of ±5mm, reducing the risk of optical fiber breakage caused by construction impact; Anti-shear sheath, the polyurethane matrix is doped with 15% carbon nanotubes, and the tensile strength is increased to 85MPa; Surface modification layer, a fluorosilane coating (thickness 50μm) achieves superhydrophobic characteristics (contact angle > 150°), and the dust adhesion rate is reduced by 90%.

[0037] Humidity compensation type vibrating wire sensor: The inner cavity is vacuum encapsulated, and the beryllium bronze vibrating wire (elastic modulus 128GPa) has a vacuum degree ≤ 10 -3 Pa, and the frequency drift rate < 0.001Hz / ℃; the outer cavity damping medium, the silicone oil viscosity is 500cSt, and the thermal expansion coefficient (CTE = 2.1×10 -4 / ℃) is synchronized with the pore water of loess to suppress the cross-interference of temperature and humidity; Dynamic anti-interference mechanism, multi-frequency excitation scanning: sweep frequency locking of 40 - 60Hz to lock the humidity-insensitive frequency point, 52Hz ± 0.1Hz, and the analysis error of the slope of the frequency response curve < 0.5%; Miniature humidity sensor integration, high-precision capacitive sensor, sampling rate 10Hz, and real-time feedback of data to the Kalman filter.

[0038] Multi-source data fusion algorithm, Kalman filter model, input parameters: vibration frequency (main), temperature (PT1000 ± 0.1℃), pore water pressure (piezoelectric ceramic ± 0.2%FS); weight distribution, the weight of vibration frequency is 80% under steady state, and the weight of pore water pressure is increased to 30% during the mutation stage; output accuracy, the comprehensive error of effective stress ≤ ±0.5%FS.

[0039] Sampling rate adaptive strategy: Steady state mode: 10Hz sampling (energy consumption 5W), used for long-term deformation monitoring; Mutation mode: 100Hz sampling (energy consumption 25W), and the triggering condition is strain rate > 0.1με / s.

[0040] Refer to Figure 4 As shown, the specific content of S3 includes: The core processor of the edge computing node is an FPGA chip of the Xilinx Zynq UltraScale+ series, integrating a quad-core ARM Cortex-A53 and a programmable logic unit; supporting the synchronous operation of fiber optic data processing and microseismic analysis; Fiber optic demodulation equipment, integrating a four-channel phase-sensitive optical time domain reflectometer, and the microseismic signal acquisition is a 24-bit high-precision ADC; the protection setting is an IP68 waterproof and explosion-proof box, with an active heat dissipation air duct built-in and an anti-electromagnetic interference shielding layer; Lightweight GRU neural network model architecture, including an input layer, six-channel fiber optic strain, three-axis energy spectrum of microvibrations, and 3D components of the laser displacement field; an output layer, crack propagation probability, stability coefficient; Perform wavelet packet transform on the fiber optic data, select the db4 wavelet basis function, and perform threshold denoising to improve the signal-to-noise ratio; calculate the strain gradient, calculate the spatial derivative of the strain based on the difference method, and generate a contour map of the cross-sectional strain gradient distribution; Microseismic-laser data fusion, locate microseismic events and map the energy, and the wave velocity model is for anisotropic velocity inversion; calculate the energy spectrum, and generate a three-dimensional matrix of energy-frequency-time through short-time Fourier transform; for the three-dimensional laser point cloud, improve the iterative closest point algorithm and introduce normal vector constraints; calculate the displacement vector based on the difference between adjacent frame point clouds to generate a displacement field; Construct a microseismic energy-strain gradient-displacement correlation matrix, and extract the precursor characteristics of surrounding rock failure through principal component analysis; Divide the space-time grid, and the probability model includes inputs such as strain gradient, cumulative microseismic energy, and displacement rate; the output is probability calculation based on a Bayesian network to generate a probability cloud map of crack propagation.

[0041] It should be noted that for the edge computing node: Xilinx Zynq UltraScale+ MPSoC: Four-core ARM Cortex-A53 (main frequency 1.5 GHz) runs a lightweight GRU model and system control; the programmable logic unit realizes preprocessing of fiber optic data (5 times acceleration of FFT operation) and real-time filtering of microseismic signals (parallelization of FIR filters); for the data throughput capacity, the fiber optic channel is 1.2 Gbps, the microseismic signal sampling rate is 5 kHz / channel, and the delay < 5 ms; IP68 protection and heat dissipation, and the active heat dissipation air duct (dual fan redundancy) maintains the chip temperature < 65 °C; the electromagnetic shielding layer (galvanized steel plate + conductive foam) suppresses high-frequency interference from construction equipment (> 1 GHz attenuation 30 dB).

[0042] Analysis of multi-source signals: Optimization of wavelet packet transform, db4 wavelet basis, matching the short-time mutation characteristics of the strain signal of loess surrounding rock; adaptive threshold denoising, dynamically adjust the threshold based on the SURE Shrink algorithm, and the signal-to-noise ratio is increased from 12 dB to 28 dB; Calculation of strain gradient, combining the difference method with cubic spline interpolation, the spatial resolution is 0.1 m, and the gradient error < 0.5 με / mm; the contour map rendering uses HSV color mapping, and the gradient > 1.5 με / mm is marked as the red warning area.

[0043] Microseismic-laser data fusion: Microseismic anisotropic wave velocity model, based on the P-wave velocity matrix (V_p = 3500 m / s horizontally and 3200 m / s vertically), the positioning accuracy reaches 0.3 m; the short-time Fourier transform window length is 50 ms to generate an energy-frequency-time matrix; Laser point cloud displacement field construction, improvement of the ICP algorithm, introduction of normal vector constraint (weight 0.7), registration error < 0.2 mm; differential calculation of the displacement vector field (adjacent frame time interval 1 s), rate > 2 mm / min triggers an early warning.

[0044] Lightweight GRU model: Input layer: six-channel fiber optic strain (mean / variance / gradient), micro-vibration three-axis energy spectrum (E_x, E_y, E_z), 3D components of laser displacement (Δx, Δy, Δz); Compression of the GRU model, with 1.2M parameters, quantized to INT8 precision, inference time < 20 ms; the activation function uses LeakyReLU (α = 0.1) to prevent gradient disappearance; Identification of precursors of surrounding rock failure, principal component analysis, extraction of the first 3 principal components (cumulative contribution rate > 85%) from the correlation matrix (micro-seismic energy × strain gradient × displacement rate); mapping of feature vectors to a low-dimensional space, and clustering of abnormal points (Mahalanobis distance > 3σ) is determined as a high-risk area; Dynamic early warning of Bayesian network, spatio-temporal grid division: spatial grid 0.5m × 0.5m × 0.5m, time window 10 minutes; probability calculation includes prior probability, training with historical collapse data (P(crack) = 0.15); likelihood function, strain gradient > 1.2 με / mm (likelihood ratio 3.2), micro-seismic energy > 50 J (likelihood ratio 2.8), displacement rate > 1 mm / min (likelihood ratio 2.1); Posterior probability output, grading of the probability cloud map of crack propagation (green < 10%, yellow 10% - 30%, orange 30% - 60%, red > 60%).

[0045] The specific content of S4 includes: Deploy three independent fiber optic links at each monitoring section, and the switching priority is dynamically adjusted according to the signal attenuation value; based on the principle of optical time domain reflectometry, through the breakpoint type recognition algorithm, distinguish between construction machinery shear fracture and surrounding rock creep fiber breakage; The robotic arm-assisted laying device is a six-degree-of-freedom robotic arm, with an optical fiber fusion splicer and a coating module integrated at the end; a dual-mode navigation system is adopted, combining laser SLAM and UWB positioning; an intelligent repair strategy is used for breakpoint repair, and the repair process is OTDR positioning, breakpoint cleaning, fusion splicing, and coating reinforcement; an autonomous obstacle avoidance algorithm is set, and the pose data of construction machinery is received in real time to plan a collision-free path; The sensor housing adopts a titanium alloy coating structure, and the coating process is vacuum plasma spraying of a titanium alloy layer; and a nano-hydrophobic coating is set, including a fluorosiloxane polymer matrix doped with SiO2 nanoparticles, with a self-healing function, and the coating microcracks trigger the nanoparticle migration and filling mechanism; Self-cleaning air curtain device. The air curtain forms an axial airflow barrier covering the full angle through an annular porous nozzle and a high-pressure air pump. Implement an energy-saving control strategy, start and stop adaptively according to the dust concentration, start when the PM10 is greater than the established concentration, and standby when it is less than the established concentration.

[0046] It should be noted that the three-link dynamic switching mechanism: Priority strategy, dynamically allocate link weights according to the real-time signal attenuation value (unit: dB / km): Main link (attenuation < 0.2 dB / km): Carry 90% of the data traffic; Standby link 1 (attenuation 0.2 - 0.5 dB / km): Load balancing 5%; Standby link 2 (attenuation > 0.5 dB / km): Only for heartbeat detection, no data transmission.

[0047] Breakpoint type identification: Mechanical shear fracture: The signal drops suddenly (slope > 10 dB / μs), and the reflection peak presents the characteristics of "steep - flat"; Surrounding rock creep fiber breakage: The signal drops slowly (slope < 2 dB / μs), and the reflection peak is accompanied by multiple micro fluctuations.

[0048] Intelligent layout and repair of robotic arms, six-degree-of-freedom collaborative operation: Laser SLAM positioning accuracy: ±3 mm (static), ±8 mm (dynamic construction environment); UWB compensation positioning: Activated when the dust concentration > 50 mg / m³, positioning error ≤ ±15 cm.

[0049] Closed-loop intelligent repair process: 1. OTDR positioning: Breakpoint coordinate error < ±0.1 m; 2. Plasma cleaning: The cleanliness of the fiber end face reaches the Telcordia GR-326 standard; 3. Fusion splicing control: Fusion splicing loss ≤ 0.03 dB; 4. Ultraviolet coating reinforcement: Coating thickness 80 μm, tensile strength > 50 N.

[0050] Sensor anti-interference: Titanium alloy-nano coating composite protection, vacuum plasma spraying process, titanium alloy layer thickness 150 μm, porosity < 0.5%, bonding strength > 70 MPa; Corrosion resistance, no pitting after 3000 hours of salt spray test (ASTM B117).

[0051] Self-repairing nano-hydrophobic coating, with a fluorosilicon polymer matrix, its contact angle > 155°, rolling angle < 5°; SiO2 nanoparticle migration mechanism, microcracks (> 5 μm) trigger particle directional filling, repair efficiency > 95%.

[0052] Air curtain barrier energy-saving control: Airflow coverage optimization, annular nozzle design: 24 micro-holes (hole diameter 0.3mm), airflow velocity 30m / s; coverage angle, axial 360°, radial ±45°, eliminating dead corners of dust adhesion; Adaptive start-stop strategy, PM10 threshold: start threshold > 75μg / m³ (typical dust concentration in tunnel construction); energy-saving mode: standby power consumption < 5W, full-power operation power consumption 120W (high-pressure air pump + control module).

[0053] The specific content of S5 includes: The BIM model is modeled with LOD400 accuracy, integrating the tunnel structure, the poses of construction machinery, and the topology of monitoring equipment; dynamically bind the tunneling mileage and geological parameters, and associate with the construction progress; the GIS geological database includes a spatial database, borehole data, and geophysical exploration profiles; update the in-situ stress field through microseismic events and fiber optic strain inversion; The tunnel geological parameter library includes the initial in-situ stress, measured data by the hydraulic fracturing method and a non-uniform field generated by a three-dimensional inversion algorithm; the collapsibility coefficient, calibrated based on indoor compression tests and on-site immersion load tests; the permeability curve, modeled by combining transient permeability tests and CT scans of pore structures; Update the finite element-discrete element coupling model, based on the model coupling algorithm, divide the continuous-discontinuous domain, the finite element region is the undisturbed surrounding rock; the discrete element region is the fracture zone and the potential fracture network; the data exchange interface, synchronize regularly through stress-displacement boundary conditions, and feedback the fracture propagation path to the GIS database; Dynamically update the boundary conditions, the construction disturbance factors include the excavation unloading effect, adjust the in-situ stress release coefficient according to the tunneling speed; the support effect, input the bolt pre-tightening force and the lining stiffness into the FEM model in real time; the environmental interaction parameters, including seepage-stress coupling, and update the pore water pressure field driven by the permeability curve; humidity-creep correlation, correct the rheological constitutive equation of the surrounding rock with the collapsibility coefficient.

[0054] It should be noted that the BIM level and data linkage: LOD400 refined modeling: the structure model includes the arrangement of tunnel lining steel bars and the details of construction joints; dynamically bind the pose of the tunneling machine (GNSS + IMU positioning) and the topological relationship of monitoring equipment; associate with the construction progress in real time, synchronize and update the support parameters of the geological database, such as updating the surrounding rock grade every 1m of advancement; The GIS geological spatial database integrates borehole data (hole spacing ≤ 20m), geophysical exploration profiles (seismic wave CT resolution 0.5m × 0.5m), and microseismic event heat maps, and generates a three-dimensional in-situ stress field through the Kriging interpolation algorithm.

[0055] Multi-source data-driven update mechanism, fiber optic strain inversion, based on distributed strain data, calculate the expansion rate of the plastic zone of the surrounding rock; Microseismic event correlation. Within 24 hours after blasting, intensive microseismic events (>50 times) trigger a re-evaluation of the in-situ stress field, and the lateral pressure coefficient λ (range 0.6 - 1.3) is corrected.

[0056] Construction of geological parameter database: Initial in-situ stress field modeling, with a dual-engine of field measurement - inversion. Five benchmark points of the principal stress directions are measured by the hydraulic fracturing method; a three-dimensional inversion algorithm (based on FLAC3D) generates a non-uniform field; non-uniformity characterization, the maximum horizontal principal stress gradient ≤ 0.5 MPa / 10 m, and the vertical stress is distributed according to the gradient of γ = 20 kN / m³. Coefficient of collapsibility: Laboratory test: Consolidometer compression test (pressure 50 - 400 kPa) to obtain δ_s = 0.015 - 0.083; Field verification: 10 m × 10 m soaking load test, correction coefficient β = 1.2 (loess stratum).

[0057] Permeability curve, measured by a transient permeameter, k = 1×10 -6 -5×10 -5 cm / s; CT scan to construct a pore network model; Finite element - discrete element cross-scale coupling: Model zoning and data interaction, continuous - discontinuous domain division: FEM region: Undisturbed surrounding rock (element size 0.2 - 0.5 m), adopting the Drucker - Prager elastoplastic structure; DEM region: Fracture zone (particle size 10 - 50 mm), simulating frictional sliding through the Hertz - Mindlin contact model; Boundary synchronization interface, exchanging stress - displacement data through the MPCCI platform every 1 hour, and mapping the crack propagation path to the GIS database (update frequency 5 minutes / time).

[0058] Fracture network prediction algorithm, expansion criterion. When the fracture rate of particle contact force chains in the DEM region > 30%, it is determined as a potential fracture initiation area; path optimization, the A* algorithm calculates the optimal crack propagation direction and cross - validates with the microseismic event location results.

[0059] Dynamic boundary conditions and multi - physical field coupling: Quantification of construction disturbance factors, excavation unloading effect. According to the stratum loss rate V_L = 0.5% - 1.2%, dynamically adjust the in - situ stress release coefficient α (range 0.3 - 0.8); Real - time feedback of the support effect, the pre - tightening force of bolts (80 - 150 kN) is collected by fiber Bragg grating sensors; the lining stiffness (C30 → C50 concrete) is input into the model according to the age function; Update of environmental interaction parameters: Seepage-stress coupling, the permeability curve drives the update of the pore water pressure field, the Darcy flow model calculates the seepage force, and an early warning is triggered when the gradient ≥ 5 kPa / m; Humidity-creep correlation, the collapsibility coefficient corrects the parameters of the Burgers rheological model to predict the long-term deformation of the surrounding rock.

[0060] Closed-loop cross-model data: Forward drive: BIM construction progress → update the initial conditions of FEM-DEM → predict crack propagation → feedback to the GIS early warning system; Reverse verification: microseismic event location → correct the DEM crack parameters → optimize the in-situ stress field → guide the adjustment of roadheader parameters.

[0061] In summary, the advantages of the present invention are as follows: Innovation in anti-interference ability, humidity suppression. The double-chamber vibrating wire sensor realizes signal fidelity in the high-humidity loess environment through silicone oil damping medium and adaptive excitation frequency adjustment, solving the problem of measurement distortion caused by humidity fluctuations of traditional sensors; dust protection, the combination of titanium alloy coating and self-repairing nano-hydrophobic coating with self-cleaning air curtain forms multiple protection barriers, significantly reducing the risk of dust adhesion and equipment corrosion, and adapting to the extreme environment of tunnel construction.

[0062] Accurate signal capture, the spiral-wound optical fiber encapsulation improves the strain transfer efficiency. Combined with wavelet packet denoising and strain gradient analysis, it accurately identifies the signals of the initiation and propagation of microcracks in the surrounding rock, breaking through the sensitivity limit of traditional straight-line laid optical fibers.

[0063] The dynamic switching of redundant optical fiber links and the cooperation of a six-degree-of-freedom robotic arm achieve millisecond-level switching of breakpoints and autonomous repair within a short time, avoiding construction interruptions caused by manual intervention and ensuring the continuity of monitoring; the dual-mode navigation of the robotic arm supports collision-free path planning under complex working conditions, improving the repair efficiency and safety.

[0064] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the principle of the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for monitoring the deformation of surrounding rock in a loess tunnel, characterized in that, include: S1. Annular monitoring sections are arranged at intervals along the longitudinal direction of the tunnel, and distributed optical fiber sensor arrays, micro-vibration sensors and humidity-compensated vibrating-wire sensors are deployed in each section; S2. The optical fiber sensor adopts a spirally wound packaging structure, is covered with a shear-resistant flexible sheath, and is fixed to the surface of the surrounding rock through a pre-buried silicone sleeve; the humidity compensation vibrating string sensor has a built-in double-cavity sealing structure, the outer cavity is filled with silicone oil damping medium, and the excitation frequency range is set to suppress the interference of loess humidity; S3. Deploy edge computing nodes on the inner wall of the tunnel, integrate FPGA chips and lightweight GRU neural network models, and integrate optical fiber strain data, microseismic event energy spectrum, and 3D laser scanning point cloud displacement field in real time; use wavelet packet transform to denoise the optical fiber data, extract characteristic components of specific frequency bands, and calculate the surrounding rock strain gradient distribution; combine the spatiotemporal evolution of microseismic events to generate a crack extension probability cloud map; S4, through the redundant optical fiber link switching mechanism and the robot arm auxiliary deployment device, the sensor breakpoints caused by construction vibration can be repaired in real time; the sensor shell is made of titanium alloy plating and nano-hydrophobic coating, and the surface is integrated with a self-cleaning air curtain device to resist dust adhesion; S5. Based on the BIM-GIS fusion model, a tunnel geological parameter database is constructed, including initial ground stress, collapsibility coefficient and permeability coefficient curves; the boundary conditions of the finite element-discrete element coupling model are regularly updated to predict the surrounding rock deformation and landslide risk level.

2. The method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 1, characterized in that, The S1 specifically includes: The layout of the circular monitoring section is based on the longitudinal spacing control, and the spacing is dynamically adjusted according to the collapsibility grade of the loess; the layout is densely packed within the given range of the tunnel entrance; Sensors deployed in each section must cover the vault, waist, side walls and invert areas; control the radial deviation of the section and align the three-dimensional data space.

3. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 2, characterized in that, The S1 specifically includes: The microseismic sensor array layout includes multiple three-component microseismic sensors deployed in each section in a regular tetrahedron topology distribution. The sensors are embedded in the ends of the anchor bolts to ensure the anchoring depth. Adaptive trigger thresholds are used, and the background noise amplitude and 6dB are used as trigger thresholds. The humidity compensated vibrating string sensor uses a humidity interference suppression mechanism, dynamically adjusts the excitation frequency range, locks humidity-insensitive frequencies through frequency sweeping, has a built-in micro humidity sensor, and provides real-time feedback of compensation parameters. It uses an improved Kalman filter to fuse vibration frequency, temperature and humidity, and pore water pressure data to output effective stress values.

4. The method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 3, characterized in that, The S2 specifically includes: The fiber optic sensor is packaged in a spiral winding manner, and the predetermined winding angle is locked after verification by finite element simulation. The spiral radius matches the curvature of the tunnel section; The shear-resistant flexible sheath adopts a polyurethane matrix doped with a certain proportion of carbon nanotubes; the surface is coated with a fluorosilane modified layer; when the embedded silicone casing is installed, a spiral spring buffer structure is embedded.

5. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 4, characterized in that The S2 specifically includes: The humidity compensation vibrating wire sensor has double chamber sealing, the inner chamber is vacuum sealed, and the vibrating wire is made of beryllium bronze; the outer chamber is filled with silicone oil damping medium, which is synchronized with the thermodynamic response of loess pore water; Humidity interference suppression, multi-frequency excitation scanning, dynamically adjust the excitation frequency, identify humidity-insensitive frequency points through the frequency response curve; built-in miniature humidity sensor, real-time feedback, through in-situ compensation algorithm, realize the correction of vibration frequency-stress conversion; Signal processing enhancement, the Kalman filter fuses vibration frequency, temperature, and pore water pressure data, outputs the effective stress value, and dynamically adjusts the sampling interval.

6. The method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 5, characterized in that, The specific content of S3 includes: The core processor of the edge computing node is an FPGA chip of the Xilinx Zynq UltraScale+ series, integrating a quad-core ARM Cortex-A53 and a programmable logic unit; supports synchronous operation of fiber optic data processing and microseismic analysis; fiber optic demodulation device, integrating a four-channel phase-sensitive optical time domain reflectometer, and the microseismic signal acquisition is a 24-bit high-precision ADC; the protection setting is an IP68 waterproof and explosion-proof box, with an active heat dissipation air duct built-in and an electromagnetic interference shielding layer. The lightweight GRU neural network model architecture includes an input layer, six-channel fiber optic strain, micro-vibration three-axis energy spectrum, and 3D components of the laser displacement field; an output layer, crack propagation probability, and stability coefficient.

7. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 6, characterized in that, The specific content of S3 includes: Perform wavelet packet transform on the fiber optic data, select the db4 wavelet basis function, and based on threshold denoising, improve the signal-to-noise ratio; calculate the strain gradient, calculate the spatial derivative of the strain based on the difference method, and generate a cross-sectional strain gradient distribution cloud map. Microseismic-laser data fusion, localize and map the energy of microseismic events, and the wave velocity model is anisotropic velocity inversion; energy spectrum calculation, generate a three-dimensional matrix of energy-frequency-time by short-time Fourier transform; three-dimensional laser point cloud, improve the iterative closest point algorithm, and introduce normal vector constraints; calculate the displacement vector based on the difference between adjacent frame point clouds to generate a displacement field. Construct a microseismic energy-strain gradient-displacement correlation matrix, and extract the precursor characteristics of surrounding rock failure through principal component analysis. Divide the space-time grid, the probability model includes inputs, strain gradient, microseismic cumulative energy, and displacement rate; output, probability calculation based on the Bayesian network, and generate a crack propagation probability cloud map.

8. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 7, characterized in that, The specific content of S4 includes: Deploy three independent fiber optic links for each monitoring section, and dynamically adjust the switching priority according to the signal attenuation value; based on the optical time domain reflectance principle, through the breakpoint type recognition algorithm, distinguish between construction machinery shear fracture and surrounding rock creep fiber breakage. The robotic arm-assisted laying device is a six-degree-of-freedom robotic arm, with an optical fiber fusion splicer and a coating module integrated at the end; adopts a dual-mode navigation system, combining laser SLAM and UWB positioning; adopts an intelligent repair strategy for breakpoint repair, and the repair process is OTDR positioning, breakpoint cleaning, fusion splicing, and coating reinforcement; set an autonomous obstacle avoidance algorithm, receive the pose data of construction machinery in real time, and plan a collision-free path.

9. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 8, characterized in that, The specific content of S4 includes: The sensor housing adopts a titanium alloy plating structure, and the plating process is vacuum plasma spraying of a titanium alloy layer; and a nano-hydrophobic coating is set, including a fluorosilicon polymer matrix doped with SiO2 nanoparticles, with a self-healing function, and the coating microcracks trigger the nanoparticle migration and filling mechanism. Self-cleaning air curtain device. The air curtain forms an axial airflow barrier covering the full angle through an annular porous nozzle and a high-pressure air pump. An energy-saving control strategy is implemented, with the dust concentration adaptively starting and stopping. It starts when the PM10 is greater than the established concentration and stands by when it is less than the established concentration.

10. A method for monitoring the deformation of surrounding rock in a loess tunnel according to claim 9, characterized in that, The specific steps of S5 include: The BIM model is modeled with LOD400 accuracy, integrating the tunnel structure, the pose of construction machinery, and the topology of monitoring equipment. Dynamically bind the tunneling mileage and geological parameters, and correlate with the construction progress. The GIS geological database includes a spatial database, borehole data, and geophysical profiles. Update the in-situ stress field through microseismic events and fiber optic strain inversion. The tunnel geological parameter library includes the initial in-situ stress, which is generated by measured data using the hydraulic fracturing method and a three-dimensional inversion algorithm to form a non-uniform field; the collapsibility coefficient, which is calibrated based on indoor compression tests and in-situ immersion load tests; the permeability coefficient curve, which is modeled by combining transient permeability tests and CT scans of pore structures. Update the finite element-discrete element coupling model. Based on the model coupling algorithm, divide the continuous-discontinuous domain. The finite element region is the undisturbed surrounding rock; the discrete element region is the fracture zone and the potential fracture network. The data exchange interface synchronizes regularly through stress-displacement boundary conditions, and the fracture propagation path is fed back to the GIS database. Dynamically update the boundary conditions. The construction disturbance factors include the excavation unloading effect, and adjust the in-situ stress release coefficient according to the tunneling speed; the support effect, with the pre-tightening force of the anchor bolt and the lining stiffness input into the FEM model in real time; the environmental interaction parameters, including seepage-stress coupling, and the permeability coefficient curve drives the update of the pore water pressure field; humidity-creep correlation, and the collapsibility coefficient corrects the rheological constitutive equation of the surrounding rock.

Images