Tunnel rock mass dynamic disaster monitoring system based on digital twinning

Abstract

The invention provides a tunnel rock mass dynamic disaster monitoring system based on digital twinning, and relates to the technical field of tunnel rock mass dynamic disaster monitoring based on digital twinning. By integrating a multi-mode sensing network, the system comprises a self-adaptive infrared structured light scanner, a distributed micro-seismic sensing array and a fiber grating strain sensor; comprehensive and real-time monitoring of the rock mass state is realized; meanwhile, an intelligent analysis algorithm based on deep learning is introduced, abnormal conditions are automatically recognized, a rockburst risk comprehensive index is generated, in addition, a physical entity and a virtual model are combined through a dynamic digital twinborn modeling technology, and visual analysis of rock mass state changes is achieved; and finally, an edge-cloud cooperative communication technology is adopted to ensure low delay and reliability of data transmission and ensure that the monitoring system can quickly respond and timely take emergency measures.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel rock mass dynamic disaster monitoring based on digital twins, and in particular to a tunnel rock mass dynamic disaster monitoring system based on digital twins. Background Art

[0002] In modern tunneling and mining operations, rockbursts pose a serious threat to production safety. Rockbursts can not only cause casualties but can also damage equipment and halt production, resulting in significant economic losses. Traditional rockburst monitoring methods rely heavily on manual observation and discrete sensor deployments, making it difficult to achieve comprehensive, real-time monitoring of rock mass conditions. Especially in complex geological conditions, monitoring accuracy and timeliness are poor, making it impossible to effectively warn of potential rockburst risks. Therefore, a new monitoring system is urgently needed that integrates multiple sensing technologies and intelligently analyzes them to enhance dynamic monitoring of rock mass conditions and promptly detect and assess rockburst risks.

[0003] Existing technologies primarily rely on single sensors or devices, typically employing traditional geological radar, strain gauges, and acoustic emission monitoring. While these methods can provide some feedback on rock mass conditions, they often suffer from information silos. Even with data collection from multiple sensors, manual analysis and judgment are still required, making real-time monitoring and feedback difficult to achieve. Furthermore, existing technologies typically focus on monitoring a specific parameter and lack the ability to comprehensively assess the multi-dimensional and multi-factorial aspects of the rock mass, resulting in low prediction accuracy. This approach is prone to omissions and false alarms when faced with complex geological conditions, impacting project safety and decision-making effectiveness.

[0004] The deficiency of existing technologies is that they cannot achieve comprehensive and real-time monitoring of rock mass conditions, especially the lack of comprehensive analysis capabilities under dynamic changes and complex situations.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] The purpose of the present invention is to provide a tunnel rock dynamic disaster monitoring system based on digital twin to solve the problems raised in the above background technology.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A tunnel rock mass dynamic disaster monitoring system based on digital twins, specifically comprising the following steps:

[0009] A multimodal sensing network layer, including an adaptive infrared structured light scanning subsystem that uses self-calibration and adaptive phase-shift interferometry to acquire tunnel cross-sectional deformation data; a distributed microseismic sensing array for monitoring the amplitude, frequency, and location of microseismic events; fiber Bragg grating strain sensors for acquiring rock mass strain energy data; and a multispectral gas analysis module for detecting the concentration of hazardous gases, including methane and hydrogen sulfide.

[0010] The data preprocessing module uses deep learning-based anomaly detection and dynamic compensation algorithms to perform noise suppression and self-calibration on the collected data.

[0011] Dynamic digital twin modeling layer, used to build multi-level digital twins including geometric models, physical field models and data-driven models;

[0012] The intelligent inversion analysis layer trains a global rockburst risk prediction model through a federated learning framework;

[0013] Adaptive early warning and decision-making layer, based on the comprehensive rockburst risk index predicted by the global rockburst risk prediction model, triggers graded early warnings and generates emergency response plans;

[0014] The edge-cloud collaborative communication layer is used to deploy lightweight digital twin models at edge computing nodes and achieve low-latency data transmission through 5G slicing technology;

[0015] The calculation formula of the rockburst risk comprehensive index CRRI is:

[0016]

[0017] Among them, σ eq is the equivalent stress output by the digital twin model, σ c is the critical compressive strength of rock mass calibrated in the laboratory, E r is the cumulative release strain of microseismic events within 24 hours, E c is the critical strain energy threshold for inversion of historical rockburst events, A i ,f i ,D i are the amplitude, main frequency and distance from the tunnel face of the i-th microseismic event within 24 hours, N is the total number of microseismic events monitored within 24 hours, Δu max is the maximum deformation measured by the infrared structured light scanning subsystem within 24 hours, u c is the critical deformation threshold that is dynamically adjusted according to the surrounding rock type within 24 hours, VB-Causal(q ij ) is the disaster chain intensity coefficient output by variational Bayesian causal inference within 24 hours, q ijrepresents the causal influence coefficient between the i-th microseismic event and the j-th type of disaster event within 24 hours. β, γ, and δ are all weight coefficients dynamically calibrated through multi-objective evolutionary optimization.

[0018] Furthermore, the adaptive infrared structured light scanning subsystem in the multimodal sensing network layer includes a self-calibrating infrared camera and a variable-focus grating, and uses adaptive phase-shift interferometry technology to improve anti-interference capabilities. The infrared light wavelength is 940nm, and the grating period of the variable-focus grating is dynamically adjustable from 0.5 to 5μm. The maximum deformation of the tunnel section is calculated using the following formula:

[0019]

[0020] Where f is the equivalent focal length of the infrared camera, d is the period of the variable focus grating, and D max is the maximum pixel displacement of the deformation on the camera imaging plane;

[0021] The distributed microseismic sensing array uses a cellular topology network consisting of MEMS triaxial accelerometers with a frequency response of 0.1Hz-10kHz and piezoelectric ceramic sensors with a sensitivity of 5mV / g. The node spacing is less than or equal to 3 meters. The location of microseismic events is analyzed using a time difference positioning algorithm, with a positioning error of less than or equal to 0.5m.

[0022] The fiber Bragg grating strain sensor achieves distributed strain monitoring of 200 measuring points per kilometer based on femtosecond laser writing technology, with a spatial resolution of 5 mm and is combined with a temperature compensation algorithm to eliminate thermal drift errors.

[0023] Furthermore, the multimodal sensing network layer also includes a multispectral gas analysis module, which integrates 400-700nm visible light, 8-14μm thermal infrared, and 1.5-2.5μm short-wave infrared cameras to identify gas and hydrogen sulfide concentrations through spectral absorption peaks; the data preprocessing module uses a deep residual shrinkage network to suppress noise from multi-source data;

[0024] Furthermore, the equivalent stress σ eq The calculation formula is:

[0025]

[0026] Among them, σ1, σ2, and σ3 are the principal stress components output by the digital twin model, which are obtained by solving the thermal-mechanical-fluid coupling equations using the finite element method.

[0027] Furthermore, the microseismic event cumulatively releases strain energy E r The calculation formula is:

[0028]

[0029] in, and are the stress rate and strain rate of the i-th microseismic event at time t, which are collected and integrated in real time by the microseismic sensor array. i is the duration of the i-th microseismic event.

[0030] Furthermore, the deformation critical threshold u c The dynamic adjustment method is:

[0031]

[0032] Among them, u base is the basic threshold corresponding to the surrounding rock category, and k is the stress coupling coefficient.

[0033] Furthermore, the variational Bayesian causal inference coefficient VB-Causal(q ij ) is calculated as follows:

[0034]

[0035] Among them, T is the total number of time windows of historical data, the sum of the length of each time window is 24 hours, v is the time window index, Microseism i is the characteristic vector of the i-th microseismic event, i.e., the amplitude A i , main frequency f i and distance D i , Disaster j is the type of disaster triggered, P(Disaster j |Microseism i ) is the conditional probability output by the Bayesian network.

[0036] Furthermore, the weight coefficients β, γ, and δ are dynamically calibrated through multi-objective evolutionary optimization, and the objective function is:

[0037] minimizeFPR+FNR+0.5·EnergyCost

[0038] Among them, FPR is the false alarm rate, FNR is the false negative rate, and EnergyCost is the normalized value of system energy consumption.

[0039] Furthermore, in the edge-cloud collaborative communication layer:

[0040] Microseismic event data is transmitted via 5G URLLC slices with a latency of less than or equal to 50ms. Deformation and stress data is transmitted via eMBB slices with a bandwidth utilization rate of less than or equal to 20%. The federated learning model gradient is encrypted and aggregated via blockchain smart contracts, with a block generation interval of less than or equal to 5 seconds.

[0041] Furthermore, the hierarchical warning rules of the adaptive warning and decision-making layer are:

[0042] The first-level warning is yellow, CRRI ≥ 0.6, triggering an audible and visual alarm; the second-level warning is orange, CRRI ≥ 0.8, automatically limiting the excavation speed to 50%; the third-level warning is red, CRRI ≥ 1.0, activating the emergency support device and generating a disaster avoidance path.

[0043] Compared with the prior art, the present invention has the following beneficial effects:

[0044] The present invention adopts a multimodal sensing network, combined with multiple sensing technologies such as adaptive infrared structured light scanning, distributed microseismic sensing array, and fiber Bragg grating strain sensor. Through this integrated sensing network, multi-dimensional data such as rock deformation, stress, and microseismic activity can be obtained simultaneously to comprehensively reflect the current state of the rock mass. An intelligent analysis algorithm based on deep learning is introduced, and by performing real-time analysis on the collected multi-dimensional data, the algorithm can automatically identify abnormal situations and potential risks and generate a comprehensive rockburst risk index.

[0045] The present invention combines physical entities with virtual models through dynamic digital twin modeling technology to reflect the state changes of the rock mass in real time. Digital twin technology enables all monitoring data to be visualized in a virtual environment. At the same time, by simulating the behavior of the rock mass under different conditions, this feature solves the problem of lack of comprehensive assessment of the changing rock mass environment in the existing technology and improves the ability to predict potential risks. The system's edge-cloud collaborative communication technology achieves low-latency data transmission through 5G slicing, ensuring the real-time and reliability of monitoring data.

[0046] Through a series of technical features such as the integration of multimodal sensor networks, intelligent analysis of multidimensional data, dynamic digital twin modeling, and edge-cloud collaborative communication, the present invention successfully solves the shortcomings of existing technologies in the background technology, significantly improves the accuracy and real-time performance of dynamic rock mass monitoring, thereby effectively reducing the risk of rock bursts and ensuring the safety and economy of underground projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 This is a flow chart of the overall system module of the present invention;

[0048] Figure 2 This is a fitting curve diagram of the total number of microseismic events-aggregation term of the present invention;

[0049] Figure 3 This is a fitting curve diagram of the equivalent stress-stress-energy coupling term of the present invention;

[0050] Figure 4This is a fitting curve diagram of the critical strain energy threshold-stress-energy coupling term of the present invention;

[0051] Figure 5 This is a fitting curve diagram of the actual displacement length-maximum deformation of the tunnel section of the present invention;

[0052] Figure 6 This is the equivalent focal length-maximum deformation curve of the present invention;

[0053] Figure 7 This is a fitting curve diagram of the critical deformation ratio-maximum deformation of the tunnel section of the present invention;

[0054] Figure 8 This is the fitting curve of the comprehensive index of rockburst risk of the disaster chain causal enhancement item of the present invention. DETAILED DESCRIPTION

[0055] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below with reference to specific embodiments.

[0056] It should be noted that, unless otherwise defined, the technical or scientific terms used in the present invention should have the usual meanings understood by people with ordinary skills in the field to which the present invention belongs. The "first", "second" and similar words used in the present invention do not indicate any order, quantity or importance, but are only used to distinguish different components. "Include" or "comprise" and similar words mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects. "Connect" or "connected" and similar words are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. "Up", "down", "left", "right" and the like are only used to indicate relative position relationships. When the absolute position of the object being described changes, the relative position relationship may also change accordingly.

[0057] Example:

[0058] See also Figures 1-8 , the present invention provides a technical solution:

[0059] A tunnel rock mass dynamic disaster monitoring system based on digital twins, specifically comprising the following steps:

[0060] The multimodal sensing network layer includes an adaptive infrared structured light scanning subsystem that uses self-calibration and adaptive phase-shift interferometry to acquire tunnel cross-section deformation data, a distributed microseismic sensing array to monitor the amplitude, frequency, and location of microseismic events, and fiber Bragg grating strain sensors to acquire rock mass strain energy data.

[0061] The adaptive infrared structured light scanning subsystem in the multimodal sensing network layer includes:

[0062] A self-calibrating infrared camera and a variable-focus grating are used, and adaptive phase-shift interferometry technology is used to improve anti-interference capabilities. The infrared light wavelength is 940nm, and 940nm infrared structured light is projected onto the tunnel cross-section. The grating period of the variable-focus grating is dynamically adjustable from 0.5 to 5μm, forming alternating light and dark interference fringes. A self-calibrating infrared camera with a resolution of 0.1mm / pixel is used to capture the deformed fringe images, and the ambient temperature and humidity parameters are simultaneously recorded.

[0063] The maximum deformation of the tunnel section is calculated using the following formula:

[0064]

[0065] Where f is the equivalent focal length of the infrared camera, d is the period of the variable focus grating, and D max is the maximum pixel displacement of the deformation variable on the camera imaging plane; the maximum deformation variable is obtained every 10 seconds through infrared structured light scanning;

[0066] Δu max The maximum deformation of the tunnel section measured during the monitoring process reflects the degree of deformation of the tunnel under external influences. The magnitude of the maximum deformation is directly related to the stability of the surrounding rock. The larger the deformation, the more serious the deformation of the tunnel, which may cause safety hazards such as rock bursts.

[0067] In infrared imaging systems, the focal length is a key parameter that affects imaging clarity and accuracy.

[0068] The grating period d is a parameter that affects the accuracy of the interference fringes. A smaller period can provide higher resolution. The larger the grating period, the greater the deformation that can be captured. Therefore, Δu max will increase accordingly;

[0069] The maximum pixel displacement D corresponding to the deformation max Reflects the actual displacement detected by the sensor in the image, D max Increasing means that the deformation is magnified in the image, making Δu max Increase.

[0070] The distributed microseismic sensing array uses a cellular topology network consisting of MEMS triaxial accelerometers with a frequency response of 0.1Hz-10kHz and piezoelectric ceramic sensors with a sensitivity of 5mV / g. The node spacing is less than or equal to 3 meters. The location of microseismic events is analyzed using a time difference positioning algorithm, with a positioning error of less than or equal to 0.5m.

[0071] The MEMS triaxial accelerometer can measure acceleration in different directions and provide dynamic responses to microseismic events. The piezoelectric ceramic sensor has high sensitivity and can effectively capture tiny vibration signals.

[0072] The fiber Bragg grating strain sensor achieves distributed strain monitoring of 200 measuring points per kilometer based on femtosecond laser writing technology, with a spatial resolution of 5 mm and is combined with a temperature compensation algorithm to eliminate thermal drift errors.

[0073] Visible light cameras capture images of the working face and identify the distribution of rock cracks. Thermal infrared cameras monitor areas of abnormal temperature. The cameras identify gases through absorption spectra. The gas absorption peaks are at 1.65μm and 2.3μm, and the hydrogen sulfide absorption peaks are at 1.57μm and 1.98μm. Gas concentrations are calculated based on the Beer-Lambert law.

[0074] A multi-spectral gas analysis module for detecting the concentration of harmful gases. The multimodal sensing network layer also includes a multi-spectral gas analysis module that integrates 400-700nm visible light, 8-14μm thermal infrared, and 1.5-2.5μm short-wave infrared cameras to identify gas and hydrogen sulfide concentrations through spectral absorption peaks.

[0075] Dynamic digital twin modeling layer, used to build multi-level digital twins including geometric models, physical field models and data-driven models;

[0076] The geometric model is constructed using a RIEGL VZ-400i 3D laser scanner with a point density of 500 points / square meter or more and a scanning frequency of 30 Hz. Isosurfaces are extracted from the point cloud to generate a triangular mesh model. The steps include: dividing the space into cubic voxels, calculating isosurfaces based on the interpolation of the voxel vertex values, generating triangular facets to represent the tunnel cross-section, and dynamically refining the mesh based on the stress gradient.

[0077] The geometric model generates a triangular mesh model of the tunnel cross section through high-precision 3D laser scanning, providing a high-resolution spatial structural benchmark and serving as the physical space anchor for subsequent physical field simulation and data-driven analysis. Its stress gradient-based mesh encryption technology adaptively increases mesh density in high-risk areas, such as fault zones, avoiding computational redundancy and distortion associated with traditional uniform meshes in complex geology.

[0078] The physical field model is constructed by using the finite element method to solve the thermal-mechanical-fluid coupling equation and the discrete element method to simulate the rock mass crack expansion and contact force model;

[0079] The physical field model establishes a direct relationship between engineering phenomena and mechanical parameters through physical equations, such as the stress balance equation and the crack propagation criterion, avoiding the "black box" defects of data-driven models. The thermal-mechanical-fluid coupling model can analyze the complex causes of surrounding rock instability in high ground temperature and high water pressure environments, such as the superposition of ground stress and seepage corrosion effects.

[0080] The data-driven model is constructed using PIT (Physics-Informed Transformer): the input data is multimodal sensor data, i.e. deformation, stress, microseismic data, etc., the loss function is pre-trained, and the digital twin state is synchronized every 1 second;

[0081] The PIT model uses multimodal sensor data, such as deformation, stress, and microseismic data, as input, constrains physical laws through a pre-trained loss function, and synchronously updates the state of the digital twin every second. Traditional physical models rely on linear assumptions, such as elastic constitutive models, while PIT uses deep neural networks to capture high-dimensional nonlinear characteristics, such as the spatiotemporal correlation of microseismic event sequences. A dynamic update mechanism enables the digital twin to respond to instantaneous changes in sensor data, such as stress redistribution triggered by sudden changes in gas concentration.

[0082] The geometric model provides the "skeleton" for the physical field model. The triangular mesh model defines the spatial distribution of the finite element mesh and discrete element particles, and is the discretization carrier for solving the physical field equations. For example, the crack propagation simulation requires initializing the discrete element contact relationship based on the initial defect position in the geometric model.

[0083] The physical field model provides "physiological rules" for the data-driven model. Physical field equations, such as stress conservation and fracture energy release rate constraints, are embedded in the pre-trained loss function of the PIT model to ensure that the data-driven model does not deviate from physical interpretability. For example, the spatial clustering of microseismic events must conform to the physical laws of stress concentration areas.

[0084] The data-driven model injects "dynamic perception" into geometric and physical models, allowing real-time correction of geometric models. Deformation data monitored by sensors drives the mesh refinement module to dynamically adjust local mesh density, such as refining the mesh when the deformation rate exceeds a threshold. Parameter calibration of the physical field model: Using data such as microseismic dominant frequency and gas concentration, material parameters in the physical field model, such as rock friction coefficient and permeability, are optimized through PIT inversion.

[0085] The geometric model provides spatial coordinate mapping for microseismic events and gas accumulation, supporting three-dimensional visualization analysis of disaster sources. The physical field model calculates the surrounding rock stress accumulation rate through a thermal-mechanical coupling model and determines the rockburst risk level based on the strain energy density threshold. The data-driven model associates discrete sensor data, such as microseismic amplitude and gas concentration, with physical field simulation results, namely the stress field distribution, to generate a global risk state vector. Through a federated learning framework, it uses data from multiple construction sites to train a global rockburst model, improving the prediction generalization capability of a single tunnel scenario. The synchronized twin state updates every second drive the adaptive early warning system. For example, when the PIT predicts that the stress concentration factor in a certain area exceeds a critical value, it triggers an anchor reinforcement instruction.

[0086] These three types of models form a "space-mechanism-data" ternary coupling system in the dynamic digital twin modeling layer: the geometric model is the digital carrier of the physical space, the physical field model provides theoretical constraints on engineering behavior, and the data-driven model complements the nonlinear characteristics of complex systems through real-time data and machine learning. The data-driven model is the core of the system's intelligence, and its significance lies in upgrading traditional physical simulations into "perceptible, adaptive, and evolvable" digital organisms.

[0087] The implementation of the global rockburst risk prediction model relies on the collaboration of three elements: a geometric model locates risk areas, a physical field model quantifies stress accumulation patterns, and a data-driven model integrates real-time data and cross-site experience to ultimately output a comprehensive index and trigger an early warning.

[0088] The intelligent inversion analysis layer trains a global rockburst risk prediction model through a federated learning framework;

[0089] The encrypted gradient aggregation of the federated learning framework is based on Paillier homomorphic encryption, and the NSGA-II algorithm is used for multi-objective evolutionary optimization. Based on the PPO algorithm, the action space is {number of LSTM layers, number of PIT heads, convolution kernel size} reinforcement learning strategy.

[0090] Adaptive early warning and decision-making layer, based on the comprehensive rockburst risk index predicted by the global rockburst risk prediction model, triggers graded early warnings and generates emergency response plans;

[0091] The edge-cloud collaborative communication layer deploys a lightweight digital twin model at edge computing nodes and achieves low-latency data transmission through 5G slicing technology;

[0092] The calculation formula of the rockburst risk comprehensive index CRRI is:

[0093]

[0094] Among them, σ eq is the equivalent stress output by the digital twin model, reflecting the stress state of the rock mass, σc E is the critical compressive strength of rock mass calibrated in the laboratory, which indicates the maximum stress that the rock mass can withstand without damage. r is the cumulative strain released during microseismic events, reflecting the energy release of the rock mass. c The critical strain energy threshold value derived from the inversion of historical rock burst events is used to evaluate the energy storage capacity of the rock mass;

[0095] A i ,f i ,D i are the amplitude, main frequency and distance from the tunnel face of the i-th microseismic event; A i ,f i ,D i Each microseismic event is recorded in real time with a time accuracy of less than or equal to 10ms. The event data is saved for 24 hours and automatically archived after exceeding the period.

[0096] A i A larger amplitude may cause the rock mass to be subjected to greater energy impact, thereby increasing CRRI, f i A higher main frequency may indicate more frequent rock stress changes and increase CRRI; the distance D between the microseismic event location and the tunneling face i , affecting energy propagation and impact range. The greater the distance, the smaller the impact may be. In the denominator, add D i It will reduce CRRI;

[0097] N is the total number of microseismic events monitored, Δu max is the maximum deformation measured by the infrared structured light scanning subsystem, reflecting the deformation of the rock mass, u c is the critical deformation threshold that is dynamically adjusted according to the surrounding rock type, VB-Causal(q ij ) is the disaster chain intensity coefficient output by variational Bayesian causal inference, which quantifies the causal relationship between different disasters, q ij represents the causal influence coefficient between the i-th microseismic event and the j-th type of disaster event within 24 h. β, γ, and δ are all weight coefficients dynamically calibrated through multi-objective evolutionary optimization to adjust the influence of different factors on CRRI.

[0098] The Comprehensive Rockburst Risk Index (CRRI) reflects the overall risk level of rockburst in a tunnel or rock mass. This index integrates multiple factors, such as stress state, microseismic activity, and deformation degree, to provide a quantitative risk assessment result. A larger CRRI value indicates a higher rockburst risk and requires timely implementation of appropriate safety measures.

[0099] Stress-energy coupling term Traditional rockburst criteria, such as the Russenes criterion, only consider the stress ratio σ eq / σ c or energy ratio E r / E c The exponential term exp(·) describes the nonlinear amplification effect of energy release on risk, which is more consistent with the sudden change characteristics of rock burst.

[0100] Microseismic spatiotemporal clustering term Introducing the amplitude-frequency-distance joint weight of microseismic events Quantify the spatiotemporal clustering of microseismic activity, indicating the accumulation of rock mass fractures, and overcome the limitations of traditional frequency-energy statistical methods;

[0101] Critical deformation ratio The cross-section deformation is associated with the empirical threshold to dynamically reflect the degree of degradation of the surrounding rock stability; the causal enhancement term of the disaster chain 1+δ·VB-Causal(q ij ), quantify the coupling probability of multiple disasters through VB-Causal (Variational Bayesian Causal Inference) and dynamically improve the threshold sensitivity;

[0102] β controls the accumulated strain energy E r Contribution to CRRI reflects the cumulative effect of rock mass energy release and controls the spatiotemporal aggregation of microseismic events. Contribution to CRRI, quantifying the temporal and spatial density of microseismic activity; δ controls the hazard chain coupling term 1+δ·VB-Causal(q ij ) on CRRI, reflecting the chain risk of secondary disasters;

[0103] The relative size of the weight coefficient depends on the sensitivity priority of the actual monitoring scenario: if the rock mass is sensitive to energy accumulation, such as deep tunnels: β>γ>δ; if microseismic activity is frequent but scattered, such as fault zones: γ>β>δ; if the disaster chain effect is significant, such as gas outburst areas: δ>β≈γ;

[0104] The weight coefficients β, γ, and δ are updated in real time through multi-objective evolutionary optimization, and β, γ, and δ>0, ensuring that the model achieves the Pareto optimal balance between false alarm rate, false negative rate, and energy consumption;

[0105] β increases, energy release term Exponential growth, significant increase in CRRI, increase in γ, and spatiotemporal aggregation of microseismic events Enhanced, CRRI is more sensitive to dense microseisms, δ increases, and the disaster chain coupling term 1+δ·VB-Causal(q ij )Amplify the risk of secondary disasters, and CRRI is more likely to trigger high-level warnings.

[0106] The specific data of some rock mass parameters and microseismic spatiotemporal aggregation items are shown in Table 1.

[0107] Table 1

[0108]

[0109] The total number of each microseismic event is random. The microseismic spatiotemporal aggregation term reflects the spatiotemporal aggregation of microseismic activity. The larger the amplitude and main frequency, the larger the microseismic spatiotemporal aggregation term. If the distance from the tunnel face increases, the microseismic spatiotemporal aggregation term will decrease accordingly.

[0110] The data preprocessing module uses deep learning-based anomaly detection and dynamic compensation algorithms to perform noise suppression and self-calibration on collected data.

[0111] The data preprocessing module uses a deep residual shrinkage network to suppress noise from multi-source data;

[0112] The network structure is input → wavelet packet decomposition → soft threshold shrinkage → residual connection → output; calculate the loss function and compensate based on the timing of the sliding window;

[0113] The equivalent stress σ eq The calculation formula is:

[0114]

[0115] Among them, σ1, σ2, and σ3 are the principal stress components output by the digital twin model, obtained by solving the thermal-mechanical-fluid coupling equation using the finite element method. The latest equivalent stress is obtained from the digital twin model every second, and the finite element simulation calculates the stress field every second to ensure real-time performance.

[0116] σ eq The calculated equivalent stress of the rock mass is an important indicator for measuring the state of the rock mass subjected to external forces. The equivalent stress reflects the comprehensive stress state of the rock mass under the action of various principal stresses σ1, σ2, and σ3. A larger value indicates that the pressure on the rock mass is closer to its critical compressive strength, increasing the risk of rock burst.

[0117] The various principal stresses σ1, σ2, and σ3 represent the stresses applied in different directions, which are usually calculated by the finite element method. The changes in the principal stress components will directly affect σ eq , if any principal stress increases, it will usually lead to σ eq Increase, thereby increasing the risk of rock mass instability;

[0118] Thermal-mechanical-fluid coupled analysis involves three main physical phenomena: heat is related to temperature changes, heat conduction, and heat convection; force is related to the forces, stress states, and deformations of materials; and flow involves the movement and flow of fluids, such as groundwater flow and gas migration.

[0119] The coupling relationship between these phenomena: the effect of heat on stress, temperature changes will cause thermal expansion or contraction of materials, thus causing stress changes; the effect of stress on flow, the stress state in the soil will affect the pore water pressure and fluid flow; the effect of flow on temperature and stress, the flow of fluid will carry away heat, change the temperature distribution of the material, and may also lead to stress redistribution;

[0120] In the calculation of equivalent stress of rock mass, the problem of thermal-mechanical-fluid coupling is particularly important because the mechanical properties of geomaterials, such as strength and stiffness, often change with temperature and moisture. When moisture flows in the soil, it causes changes in pore pressure, which in turn affects the stress state of the rock mass. Heat sources, such as chemical reactions inside the rock mass or external environmental changes, can lead to uneven temperature distribution, thereby causing thermal stress changes. By solving the thermal-mechanical-fluid coupling equation by the finite element method, these factors can be comprehensively considered to accurately calculate the equivalent stress σ under different principal stresses σ1, σ2, and σ3. eq .

[0121] The microseismic event cumulatively releases strain energy E r The calculation formula is:

[0122]

[0123] in, and are the stress rate and strain rate of the i-th microseismic event, respectively, which are collected and integrated in real time by the microseismic sensor array. N is the number of microseismic events, t i is the duration of the i-th microseismic event;

[0124] E r It represents the cumulative strain energy released by all microseismic events during the monitoring period, reflecting the energy release experienced by the rock mass due to microseismic events. r The size of E is directly related to the fatigue degree and damage state of the rock mass. A larger cumulative strain energy value usually means that the rock mass has experienced more microseismic events, which may lead to structural degradation of the rock mass. r The larger the value, the more intense the microseismic activity experienced by the rock mass, which increases the risk of rock burst and other structural instability.

[0125] N represents the total number of microseismic events recorded during the monitoring process. The greater the number of microseismic events, the greater the number of E r Increases, indicating that the rock mass has experienced more strain energy accumulation; stress rate The higher the stress, the faster the rock mass is subjected to stress changes, and the more strain energy will be released in the microseismic event. The increase of E r will also increase; the strain rate The higher the value, the faster the material deforms, resulting in more strain energy release. The increase of E r The value of will also increase.

[0126] The specific data of some rock mass parameters and stress-energy coupling terms are shown in Table 2.

[0127] Table 2

[0128]

[0129]

[0130] Reflects the stress state and energy release of the rock mass in different microseismic events.

[0131] The deformation critical threshold u c The dynamic adjustment method is:

[0132]

[0133] Among them, u base It is the basic threshold corresponding to the surrounding rock type, which is set according to different levels of surrounding rock, such as level III surrounding rock u base =15mm, k=0.2 is the stress coupling coefficient. The existence of this coefficient makes u c It is dynamically adjusted as the stress state changes. The larger the k value, the greater the influence of stress on the deformation threshold.

[0134] u c It represents the critical threshold value of surrounding rock deformation after dynamic adjustment, reflecting the safe deformation limit of surrounding rock under different stress states, u c The size of u indicates the maximum deformation that the surrounding rock can withstand under specific stress conditions. It is an important indicator for evaluating the stability of the surrounding rock. c A larger value of usually indicates that the deformation limit of the surrounding rock under the current stress conditions is higher, but it may also reflect that the bearing capacity of the surrounding rock is reduced, and attention should be paid to the stability of the rock mass;

[0135] Equivalent stress σ eq The larger the value, the higher the pressure on the rock mass, which leads to u c increases, which means that under greater stress conditions, the deformation limit of the surrounding rock will also be higher; the critical compressive strength σ c The larger the value, the stronger the bearing capacity of the rock mass, which usually leads to u c relative increase.

[0136] The specific data for calculating the critical deformation ratio are shown in Table 3.

[0137] Table 3

[0138]

[0139]

[0140] The equivalent focal length in the table is inversely proportional to the maximum deformation of the tunnel section. The smaller the focal length, the higher the imaging accuracy. Therefore, in this data, the equivalent focal length fluctuates within a certain range. The deformation critical ratio data represents the ratio of the maximum deformation to the critical deformation threshold, reflecting the stability of the surrounding rock. A larger value indicates more severe deformation.

[0141] The variational Bayesian causal inference coefficient VB-Causal(q ij ) is calculated as follows:

[0142]

[0143] Where T is the total number of time windows of historical data, the sum of the length of each time window is 24 hours, v is the time window index, Microseism i is the characteristic vector of the i-th microseismic event, i.e., the amplitude A i , main frequency f i and distance D i , Disaster j The type of disaster that is triggered, such as flooding, landslide, gas outburst, P (Disaster j |Microseism i ) is the conditional probability output by the Bayesian network, which represents the probability of disaster j occurring when microseismic event i occurs;

[0144] VB-Causal(q ij The causal network is updated every 5 minutes based on the latest monitoring data, and the Bayesian network training cycle is 5 minutes to ensure the timeliness of causal reasoning;

[0145] q ij It represents the causal relationship strength of the i-th microseismic event to the j-th type of disaster, that is, the relative probability of the microseismic event causing a specific disaster. When q ij When it is close to 1, it indicates that microseismic event i is almost certain to cause disaster j. ij When it is close to 0, it indicates that there is no significant correlation between microseismic event i and disaster j; ij The larger it is, the stronger the causal relationship between microseismic event i and disaster j is, and the early warning system should give priority to such events;

[0146] Amplitude A i Refers to the energy intensity of microseismic events, reflecting the kinetic energy released by rock mass rupture. If the disaster j is a landslide or rock burst, A i The larger the P(Disaster j |Microseismi ) is larger, and q ij The larger the distance, the greater the distance D. i Comprehensive judgment;

[0147] Main frequency f i The main frequency of the microseismic signal reflects the scale of the rupture. High frequency corresponds to small-scale rupture, and low frequency corresponds to large-scale rupture. High-frequency microseismic signals, i.e., f i The larger the value, the more likely it is that the local rock mass will fall off, the higher the probability of rock burst will be, and the lower the frequency of microseismic earthquake will be. i The smaller it is, the more likely it is that cracks are expanding on a large scale and the probability of collapse is increasing;

[0148] Distance D i Refers to the spatial distance between the microseismic event and the tunnel face, which determines its direct impact range on the construction area. i The smaller the earthquake, the more likely it is to directly cause a disaster. ij The larger the distance D i When the larger the value, the correlation strength decays, q ij The smaller;

[0149] The conditional probability P(Disaster j |Microseism i ) Based on historical data and feature learning, quantify the causal relationship between microseismic events and disasters, P (Disaster j |Microseism i ) is larger, q ij The larger the value, the normalized denominator ensures that the total probability is 1, avoiding misjudgment caused by fluctuations in absolute probability values;

[0150] T is the time window containing complete construction cycle data (excavation, support, monitoring), ensuring that causal statistics cover the reasonable time range of disaster triggering; the numerator in the formula counts the number of times microseismic event i triggers disaster j in all time windows in history, and the denominator counts the total number of occurrences of microseismic event i in all time windows;

[0151] The formula breaks through the traditional static weight allocation method by dynamically correlating microseismic event characteristics with disaster types, thereby improving early warning accuracy.

[0152] The weight coefficients β, γ, and δ are dynamically calibrated through multi-objective evolutionary optimization, and the objective function is:

[0153] minmize FPR+FNR+0.5·Energy Cost

[0154] Among them, FPR is the false alarm rate, FNR is the false alarm rate, and Energy Cost is the normalized value of system energy consumption;

[0155] By minimizing the false positive rate (FPR), false negative rate (FNR), and normalized energy consumption, the overall system performance is optimized, reflecting the efficiency and reliability of the system in terms of early warning, decision-making, and energy use. In this case, the smaller the value of the objective function, the higher the accuracy and energy efficiency of the system, which is the goal of optimization.

[0156] The false positive rate (FPR) is the probability of mistakenly labeling a normal situation as a disaster. The higher the FPR, the more serious the false alarm situation is, the lower the reliability of the system is, and the value of the objective function will increase. The false negative rate (FNR) is the probability of failing to successfully detect an actual disaster. The higher the FNR, the more serious the missed alarm situation is, the higher the potential disaster risk is, and the value of the objective function will increase.

[0157] Energy Cost is the energy consumption of the system during operation. The higher the energy consumption, the lower the efficiency of the system and the larger the value of the objective function. Therefore, the optimization goal is to reduce energy consumption while ensuring the accuracy of the system.

[0158] In the edge-cloud collaborative communication layer:

[0159] Microseismic event data is transmitted via 5G URLLC slices with a latency of less than or equal to 50ms. Deformation and stress data is transmitted via eMBB slices with a bandwidth utilization rate of less than or equal to 20%. Federated learning model gradients are encrypted and aggregated via blockchain smart contracts, with a block generation interval of less than or equal to 5 seconds.

[0160] Ensuring latency below 50ms, low-latency transmission ensures the system quickly responds to microseismic events, enabling timely early warning and decision-making. Through eMBB slicing transmission, the system remains stable even under high data traffic, avoiding data transmission delays caused by excessive bandwidth usage. Blockchain smart contract implementation ensures data privacy while ensuring the transparency and security of model updates through blockchain technology, driving system intelligence.

[0161] The hierarchical warning rules of the adaptive warning and decision-making layer are:

[0162] The first-level warning is yellow, with a CRRI ≥ 0.6, triggering an audible and visual alarm, indicating a low risk level but requiring vigilance. The second-level warning is orange, with a CRRI ≥ 0.8, automatically limiting the excavation speed to 50%, indicating a high risk level and requiring measures to reduce the risk. The third-level warning is red, with a CRRI ≥ 1.0, activating emergency support devices and generating evacuation routes, indicating an extremely high risk. Emergency measures must be taken immediately to ensure personnel safety.

[0163] The CRRI comprehensively assesses the current disaster risk situation and provides a basis for early warning decisions. Higher values ​​indicate greater disaster risk, and the warning level should be increased accordingly. The increase in warning level is positively correlated with the increase in CRRI. As the CRRI value increases, the warning response measures become more urgent and stringent. This dynamic adjustment mechanism ensures that the system can respond promptly to different risk levels to ensure the safety of personnel and equipment.

[0164] The specific data for calculating the critical deformation ratio are shown in Table 4.

[0165] Table 4

[0166]

[0167]

[0168] Through data analysis, the disaster chain causal enhancement item reflects the causal relationship between microseismic events and disasters. The larger its value, the more significant the increase in the comprehensive rockburst risk index. In microseismic event 3, the disaster chain enhancement item is only 1.42, and the CRRI is 1.79. The enhancement item is higher than the average value (1.3), indicating that the correlation between microseismic events and disasters is strong, which directly pushes the CRRI into the third-level warning. Microseismic event 5 shows that the medium enhancement item is superimposed on the high stress-energy coupling item (0.8), causing the CRRI to approach the second-level warning threshold. The disaster chain causal enhancement item is adjusted dynamically by the threshold. The value sensitivity amplifies the probability of microseismic events triggering rock bursts (for example, when the enhancement term is equal to 1.55 in sample 6, the CRRI reaches 2.51); the stress-energy-microseismic coupling term is the core driving factor of CRRI, and its value is exponentially positively correlated with the rock burst risk. For example, in microseismic event 12, the coupling term is equal to 1.5, and the CRRI is 3.04: the coupling term exceeds the threshold (1.2), causing the CRRI to far exceed the third-level warning threshold; the product relationship between the stress ratio and the energy ratio dominates the risk amplification effect (for example, in microseismic event 31, the coupling term is 1.6, and the CRRI is 3.7;

[0169] The critical deformation ratio acts as a multiplication factor. When its value exceeds 0.9, it has a significant amplification effect on the CRRI. For example, microseismic event 6 represents a deformation ratio of 1.2 and a CRRI of 2.51. The deformation ratio exceeds the critical value (1.0) and is superimposed on the high coupling term (1.35), triggering extreme risk.

[0170] Level 3 warning (red): The coupling term must be greater than or equal to 1.1, the enhancement term must be greater than or equal to 1.4, and the deformation ratio must be greater than or equal to 1.0 (such as microseismic events 10 and 12). Level 2 warning (orange): Usually driven by a coupling term greater than or equal to 0.8 or a deformation ratio greater than or equal to 0.9 (such as microseismic event 5). Level 1 warning (yellow): None of the parameters have exceeded the threshold, but vigilance against the density of microseismic events is required, such as samples 2 and 9.

[0171] The above formulas are all dimensionless and numerical calculations. The formulas are obtained by collecting a large amount of data and performing software simulation to obtain the most recent real situation. The preset parameters in the formulas are set by technicians in this field according to actual conditions.

[0172] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed by hardware or software depends on the specific application and design constraints of the technical solution.

[0173] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, and may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment as needed.

[0174] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A tunnel rock mass dynamic disaster monitoring system based on digital twin, characterized by: The specific steps include: A multimodal sensing network layer, including an adaptive infrared structured light scanning subsystem that uses self-calibration and adaptive phase-shift interferometry to acquire tunnel cross-sectional deformation data; a distributed microseismic sensing array for monitoring the amplitude, frequency, and location of microseismic events; fiber Bragg grating strain sensors for acquiring rock mass strain energy data; and a multispectral gas analysis module for detecting the concentration of hazardous gases, including methane and hydrogen sulfide. The data preprocessing module uses deep learning-based anomaly detection and dynamic compensation algorithms to perform noise suppression and self-calibration on the collected data. Dynamic digital twin modeling layer, used to build multi-level digital twins including geometric models, physical field models and data-driven models; The intelligent inversion analysis layer trains a global rockburst risk prediction model through a federated learning framework; Adaptive early warning and decision-making layer, based on the comprehensive rockburst risk index predicted by the global rockburst risk prediction model, triggers graded early warnings and generates emergency response plans; Edge-cloud collaborative communication layer, used to deploy lightweight digital twin models on edge computing nodes and achieve low-latency data transmission through 5G slicing technology; The calculation formula of the rockburst risk comprehensive index CRRI is: Among them, σ eq is the equivalent stress output by the digital twin model, σ c is the critical compressive strength of rock mass calibrated in the laboratory, E r is the cumulative release strain of microseismic events within 24 hours, E c is the critical strain energy threshold for inversion of historical rockburst events, A i ,f i ,D i are the amplitude, main frequency and distance from the tunnel face of the i-th microseismic event within 24 hours, N is the total number of microseismic events monitored within 24 hours, Δu max is the maximum deformation measured by the infrared structured light scanning subsystem within 24 hours, u c is the critical deformation threshold that is dynamically adjusted according to the surrounding rock type within 24 hours, VB-Causal(q ij ) is the disaster chain intensity coefficient output by variational Bayesian causal inference within 24 hours, q ij represents the causal influence coefficient between the i-th microseismic event and the j-th type of disaster event within 24 hours. β, γ, and δ are all weight coefficients dynamically calibrated through multi-objective evolutionary optimization.

2. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1 is characterized in that: The adaptive infrared structured light scanning subsystem in the multimodal sensing network layer includes a self-calibrating infrared camera and a variable-focus grating, and uses adaptive phase-shift interferometry technology to improve anti-interference capabilities. The infrared light wavelength is 940nm, and the grating period of the variable-focus grating is dynamically adjustable from 0.5 to 5μm. The maximum deformation of the tunnel section is calculated using the following formula: Where f is the equivalent focal length of the infrared camera, d is the period of the variable focus grating, and D max is the maximum pixel displacement of the deformation on the camera imaging plane; The distributed microseismic sensing array uses a cellular topology network consisting of MEMS triaxial accelerometers with a frequency response of 0.1Hz-10kHz and piezoelectric ceramic sensors with a sensitivity of 5mV / g. The node spacing is less than or equal to 3 meters. The location of microseismic events is analyzed using a time difference positioning algorithm, with a positioning error of less than or equal to 0.5m. The fiber Bragg grating strain sensor achieves distributed strain monitoring of 200 measuring points per kilometer based on femtosecond laser writing technology, with a spatial resolution of 5 mm and is combined with a temperature compensation algorithm to eliminate thermal drift errors.

3. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1 is characterized in that: The multimodal sensing network layer also includes a multispectral gas analysis module that integrates 400-700nm visible light, 8-14μm thermal infrared, and 1.5-2.5μm shortwave infrared cameras to identify gas and hydrogen sulfide concentrations through spectral absorption peaks; the data preprocessing module uses a deep residual shrinkage network to suppress noise in multi-source data.

4. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1 is characterized in that: The equivalent stress σ eq The calculation formula is: Among them, σ1, σ2, and σ3 are the principal stress components output by the digital twin model, which are obtained by solving the thermal-mechanical-fluid coupling equations using the finite element method.

5. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1 is characterized in that: The microseismic event cumulatively releases strain energy E r The calculation formula is: in, and are the stress rate and strain rate of the i-th microseismic event at time t, which are collected and integrated in real time by the microseismic sensor array. i is the duration of the i-th microseismic event.

6. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1, characterized in that: The deformation critical threshold u c The dynamic adjustment method is: Among them, u base is the basic threshold corresponding to the surrounding rock category, and k is the stress coupling coefficient.

7. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1 is characterized in that: The variational Bayesian causal inference coefficient VB-Causal(q ij ) is calculated as follows: Where T is the total number of time windows of historical data, the sum of the length of each time window is 24 hours, v is the time window index, Microseism i is the characteristic vector of the i-th microseismic event, i.e., the amplitude A i , main frequency f i and distance D i , Disaster j is the type of disaster triggered, P(Disaster j |Microseism i ) is the conditional probability output by the Bayesian network.

8. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1, characterized in that: The weight coefficients β, γ, and δ are dynamically calibrated through multi-objective evolutionary optimization, and the objective function is: minimize FPR+FNR+0.5·Energy Cost Among them, FNR is the false alarm rate, FNR is the missed alarm rate, and Energy Cost is the normalized value of system energy consumption.

9. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1, characterized in that: In the edge-cloud collaborative communication layer: Microseismic event data is transmitted via 5G URLLC slices with a latency of less than or equal to 50ms. Deformation and stress data is transmitted via eMBB slices with a bandwidth utilization rate of less than or equal to 20%. The federated learning model gradient is encrypted and aggregated via blockchain smart contracts, with a block generation interval of less than or equal to 5 seconds.

10. The tunnel rock mass dynamic disaster monitoring system based on digital twin according to claim 1, characterized in that: The hierarchical warning rules of the adaptive warning and decision-making layer are: The first-level warning is yellow, CRRI ≥ 0.6, triggering an audible and visual alarm; the second-level warning is orange, CRRI ≥ 0.8, automatically limiting the excavation speed to 50%; The third-level warning is red, CRRI ≥ 1.0, the emergency support device is activated and a disaster avoidance route is generated.

Images