A microseismic monitoring and intelligent early warning system for tunnel dome rock mass fracture risk

Abstract

The present application relates to the technical field of tunnel engineering safety monitoring, comprising a dome curved surface sensor array module, a dynamic wave velocity field modeling module, a microseismic event space-time strong positioning module and a fracture risk intelligent early warning module. The dome curved surface sensor array module is used to construct a "dome-wall" three-dimensional monitoring network; the dynamic wave velocity field modeling module is used to dynamically update the three-dimensional anisotropic wave velocity field by using tunneling blasting and active seismic source; the microseismic event space-time strong positioning module is used to realize high-precision positioning and focal mechanism calculation in anisotropic medium; the fracture risk intelligent early warning module is used to extract multi-dimensional physical precursors such as energy release rate, fracture volume fractal dimension and focal mechanism consistency, and to output graded early warning through Bayesian network fusion. The present application solves the problems of signal collection distortion, poor adaptability of wave velocity model and single early warning mechanism, and realizes accurate monitoring and intelligent early warning of dome rock mass fracture risk.

Description

Technical Field

[0001] This invention relates to the field of tunnel engineering safety monitoring technology, and in particular to a microseismic monitoring and intelligent early warning system for the risk of rock mass fracture in tunnel dome. Background Technology

[0002] In tunnel engineering, especially in the construction and operation phases of deep-buried extra-long tunnels, the stability of the rock mass at the dome (arch) is crucial to engineering safety and the lives of personnel. Under the combined effects of excavation unloading, high ground stress, and blasting disturbance, the dome rock mass is highly susceptible to microscopic fractures, which gradually evolve into macroscopic spalling, rockfalling, or even collapse.

[0003] Currently, the monitoring methods for the stability of rock mass in tunnel domes are mainly divided into two categories: one is traditional contact-based monitoring, such as multi-point displacement gauges and anchor stress gauges. These methods are point-based monitoring, which are difficult to capture the spatiotemporal evolution of internal rock mass fractures, and are cumbersome to set up and easily affected by construction interference; the other is microseismic monitoring technology, which involves setting up sensors on the tunnel sidewalls or bottom to collect the elastic waves released when the rock mass fractures, thereby locating the fracture source.

[0004] However, existing microseismic monitoring technologies have significant limitations when applied to tunnel domes: First, sensors are typically deployed on the sidewalls or invert of the tunnel. For fracture signals in the rock mass above the dome, the signals are severely attenuated and the waveform distortion is complex due to the influence of wave propagation paths, surrounding rock anisotropy, and tunnel cavity effects, resulting in low positioning accuracy and a high rate of missed detection of microseismic events. Second, existing systems mostly use fixed, globally uniform wave velocity models, which cannot adapt to the actual situation where the mechanical properties of the surrounding rock change dynamically with the advancement of the tunnel face during tunnel excavation. Finally, early warning mechanisms are often based on a single microseismic event count or energy accumulation index, failing to delve into the spatiotemporal migration patterns of microseismic events and the physical mechanisms of rock mass fracture, resulting in a high false alarm rate and a disconnect between the early warning results and the actual situation on the engineering site. Summary of the Invention

[0005] The purpose of this invention is to provide a microseismic monitoring and intelligent early warning system for the risk of rock mass fracture in tunnel domes. This system solves the problems of signal acquisition distortion, poor adaptability of wave velocity models, and single early warning mechanism in the prior art, and realizes accurate monitoring and intelligent early warning of the risk of rock mass fracture in tunnel domes.

[0006] To achieve the above objectives, the present invention provides a microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel domes, comprising: The dome surface sensor array module is used to construct a three-dimensional monitoring network conforming to the rock mass within the dome surface of the tunnel to collect microseismic waveform data. The dynamic wave velocity field modeling module is connected to the dome surface sensor array module and is used to construct and dynamically update a three-dimensional anisotropic wave velocity field model that is synchronized with the mechanical state of the surrounding rock using standard seismic source events during tunnel excavation. The microseismic event spatiotemporal strong localization module is connected to the dome curved surface sensor array module and the dynamic wave velocity field modeling module, respectively. It is used to automatically identify, pick up first arrival, spatially locate, calculate the time of occurrence, and calculate the source mechanism and energy parameters of microseismic events based on the microseismic waveform data and the three-dimensional anisotropic wave velocity field model, and generate a microseismic event sequence. The intelligent early warning module for rupture risk is connected to the spatiotemporal strong localization module for microseismic events. It is used to extract and fuse the multidimensional physical features of the microseismic event sequence and output graded early warning information.

[0007] The dome surface sensor array module includes: The curved surface adaptive coupling unit is used to realize the acoustic coupling between the sensor and the curved rock mass of the dome; The conformal spatial configuration unit is used to construct the spatial geometric layout of the sensor array. Monitoring sections are set at intervals along the longitudinal direction of the tunnel. At least three sensors are arranged at equal angles along the dome circumferential curved surface in each monitoring section to form a dome circumferential array. Auxiliary sensors are arranged on both sides of the same section to form a sidewall auxiliary array. The dome circumferential array and the sidewall auxiliary array of multiple sections together constitute a three-dimensional monitoring network. The multi-parameter synchronous triggering unit is used to provide time synchronization for the entire system, and adopts a combination of time protocol and hardware pulse calibration to achieve full-channel synchronization; An anti-interference signal conditioning unit is used to ensure the transmission fidelity of micro-seismic signals in interference environments.

[0008] In the conformal spatial configuration unit, the sensor layout of the dome circumferential array is such that the main response direction of the sensor points to the center of the dome curvature, matching the dominant radiation direction of the dome fracture source; the three-dimensional monitoring network forms a three-dimensional enveloping observation geometry for the rock fracture source above the dome, providing 360-degree azimuth coverage in the horizontal direction and depth constraint in the vertical direction.

[0009] The dynamic wave velocity field modeling module includes: The standard source excitation unit is used to provide an elastic wave source with known spatial coordinates and excitation time. It adopts a dual-mode scheme that combines passive use of tunneling blasting source and active excitation of artificial source. In passive mode, the excitation time and spatial coordinates of blasting source are obtained synchronously with the blasting control system hardware. In active mode, a controllable impact source is excited as needed at preset measuring points. The multipath wave velocity extraction unit is used to extract equivalent wave velocity values ​​on multi-directional propagation paths from standard source events. For each standard source event, it calculates the path length based on the known source coordinates and the coordinates of each sensor, calculates the propagation time based on the time of earthquake occurrence and the arrival time of P-waves in each channel, calculates the equivalent wave velocity of each path, and performs confidence assessment and weighting on the wave velocity values ​​of each path. The dynamic interpolation unit for wave velocity field is used to extend discrete path wave velocity sampling values ​​into a continuous three-dimensional anisotropic wave velocity field model. It adopts the Gaussian process regression method, introduces surrounding rock stress monitoring data as covariates, constructs a joint interpolation model, and outputs a three-dimensional voxel grid wave velocity field covering the monitoring area. Each grid cell contains the full wave velocity tensor. The model iteration verification unit is used to determine the update timing and applicable validity of the wave velocity field model. It adopts an iterative update mechanism that combines residual analysis, positioning error threshold, and tunneling progress triple criteria. It retains the wave velocity sampling data of the most recently tunneled section in a sliding window manner to realize on-demand and periodic updates of the wave velocity field model.

[0010] In the dynamic interpolation unit of the wave velocity field, the joint interpolation model decomposes the wave velocity value into anisotropic components related to the propagation path direction and isotropic background components related to the stress state of the surrounding rock. By using a joint covariance function, the influence of spatial distance, path direction differences, and stress state differences on the wave velocity correlation is considered simultaneously, so that the generated wave velocity field model maintains physical consistency with the measured stress field distribution trend.

[0011] The microseismic event spatiotemporal strong localization module includes: The waveform preprocessing and event detection unit is used to identify microseismic event segments from continuous waveform data. It uses frequency band adaptive filtering to process the original waveform into frequency bands and adopts a multi-channel joint judgment mechanism. When a preset number of channels simultaneously meet the triggering conditions, it confirms that a microseismic event has been detected. The multi-channel first arrival precision picking unit is used to extract the first arrival time of P-waves for each channel of the detected microseismic event. It performs optimal frequency band scanning for the event waveform of each channel, selects the dominant frequency band with the highest signal-to-noise ratio, and runs three independent algorithms in parallel: long-short window ratio method, Akaike information criterion method, and wavelet transform modulus maxima method. The first arrival time is obtained by fusing the algorithms through a voting and verification mechanism, and the confidence level of each channel is output. An anisotropic medium positioning unit is used to invert the spatial coordinates and origin time of microseismic events. Based on the three-dimensional anisotropic wave velocity field model, it uses anisotropic ray tracing algorithm to calculate the theoretical travel time, establishes a positioning model with the weighted sum of squares of the difference between the theoretical arrival time and the actual picked-up arrival time of each channel as the objective function, and uses a hybrid strategy combining grid search and particle swarm optimization to solve the problem, outputting the three-dimensional spatial coordinates of the event, the origin time, and the positioning information ellipsoid. The focal mechanism and energy calculation unit is used to calculate the intensity and rupture mechanism parameters of microseismic events. It uses a grid search method to invert the focal mechanism solution based on the first motion polarity of multi-channel P-waves. After instrument response correction and geometric diffusion correction, it uses the Brune model to fit and calculate the seismic moment and moment magnitude. It calculates the radiated energy by integrating the displacement spectrum and calculates the energy-to-moment ratio.

[0012] In the anisotropic medium positioning unit, the weight of the objective function is determined based on the confidence level of each channel output by the multi-channel first arrival precision picking unit, and the channel with higher confidence level has a larger weight; the positioning information ellipsoid is used to characterize the positioning uncertainty.

[0013] The intelligent early warning module for rupture risk includes: The fracture energy release rate analysis unit is used to identify the unsteady acceleration characteristics of rock mass fracture from the energy dimension. It constructs the energy time series of microseismic event sequences, smooths them into continuous energy release rate curves using kernel density estimation, calculates the energy acceleration index by fitting an exponential function through a sliding window, identifies the acceleration release start point, and outputs the energy release rate value and energy acceleration index. The fractal dimension analysis unit for fracture volume is used to identify the cluster evolution characteristics of microseismic events from a spatial dimension. It uses a sliding window to extract a subset of microseismic events, uses the box counting method to calculate the spatial fractal dimension, tracks the trend of fractal dimension change, and determines the fracture cluster nucleation state when the fractal dimension continuously decreases from a high value to below a set threshold. It outputs the current fractal dimension value and the trend of change. The focal mechanism consistency analysis unit is used to identify the unified evolution characteristics of the focal mechanism of microseismic events from the mechanism dimension. It constructs a focal mechanism consistency index for microseismic events within a sliding window, and quantifies the degree of consistency by calculating the root mean square value of the spatial angle between the focal mechanism of each event and the average focal mechanism within the window. When the consistency index decreases to below a set threshold, it is determined to be in a state of rupture surface penetration, and outputs the consistency index value and the orientation of the dominant rupture surface. The multi-indicator fusion early warning output unit is used to fuse and infer the outputs of the aforementioned three analysis units and output graded early warning information. It adopts a Bayesian network to construct a three-layer inference structure, with energy acceleration index, fractal dimension and its changing trend, and source mechanism consistency index and its changing trend as observation nodes, energy acceleration release state, fracture cluster nucleation state, and fracture surface penetration state as hidden state nodes, and rock mass fracture risk level as target node. It calculates the posterior probability of risk level through Bayesian inference, outputs level three, level two, and level one early warning information, and pushes it to the monitoring terminal and mobile terminal in a multimodal manner.

[0014] Among them, in the multi-indicator fusion early warning output unit, the third-level early warning corresponds to the activation of only one of the three hidden states, indicating that attention should be increased; the second-level early warning corresponds to the activation of any two of the three hidden states, indicating that high-risk operations should be stopped and reinforcement should be prepared; the first-level early warning corresponds to the activation of all three hidden states, indicating that personnel should be evacuated immediately and the emergency plan should be activated.

[0015] This invention discloses a microseismic monitoring and intelligent early warning system for the risk of rock mass rupture in a tunnel dome. The system includes a dome curved surface sensor array module for constructing a "dome-sidewall" three-dimensional monitoring network on the tunnel dome surface to collect microseismic waveform data; a dynamic wave velocity field modeling module for constructing and dynamically updating a three-dimensional anisotropic wave velocity field model using tunneling blasting and active seismic sources; a microseismic event spatiotemporal strong localization module for performing event identification, first arrival picking, anisotropic medium localization, and source mechanism and energy calculation based on waveform data and the wave velocity field model, generating a microseismic event sequence; and a rupture risk intelligent early warning module for performing energy release rate analysis, rupture volume fractal dimension analysis, and source mechanism consistency analysis on the event sequence, and outputting graded early warning information by fusing multiple indicators through a Bayesian network. This invention significantly improves the positioning accuracy and early warning reliability of dome rock mass fracture sources by optimizing sensor spatial configuration, dynamically updating wave velocity field, and constructing multi-dimensional physical precursor indicators. It solves the problems of signal acquisition distortion, poor adaptability of wave velocity model, and single early warning mechanism in the prior art, and realizes accurate monitoring and intelligent early warning of tunnel dome rock mass fracture risk. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0017] Figure 1 This invention relates to the structural principle of a microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in a tunnel dome.

[0018] Figure 2 This is a schematic diagram of the structure of the dome curved surface sensor array module provided by the present invention.

[0019] Figure 3 This is a schematic diagram of the structure of the dynamic wave velocity field modeling module provided by the present invention.

[0020] Figure 4 This is a structural schematic diagram of the microseismic event spatiotemporal strong localization module provided by the present invention.

[0021] Figure 5 This is a structural schematic diagram of the intelligent early warning module for rupture risk provided by the present invention.

[0022] The diagram shows: 101-Dome curved surface sensor array module, 102-Dynamic wave velocity field modeling module, 103-Spatiotemporal strong localization module for microseismic events, 104-Intelligent early warning module for rupture risk, 1011-Surface adaptive coupling unit, 1012-Conformal space configuration unit, 1013-Multi-parameter synchronous triggering unit, 1014-Anti-interference signal conditioning unit, 1021-Standard source excitation unit, 1022-Multi-path wave velocity extraction unit, 1023-Dynamic interpolation unit for wave velocity field, 1024-Model iterative verification unit, 1031-Waveform preprocessing and event detection unit, 1032-Multi-channel accurate first arrival pickup unit, 1033-Anisotropic medium localization unit, 1034-Source mechanism and energy calculation unit, 1041-Rupture energy release rate analysis unit, 1042-Rupture volume fractal dimension analysis unit, 1043-Source mechanism consistency analysis unit, and 1044-Multi-index fusion early warning output unit. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0024] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0025] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0026] Please see Figures 1-5 This invention provides a microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel domes, comprising: The dome surface sensor array module 101 is used to construct a three-dimensional monitoring network conforming to the rock mass within the dome surface area of ​​the tunnel to collect microseismic waveform data. The dynamic wave velocity field modeling module 102 is connected to the dome surface sensor array module 101 and is used to construct and dynamically update a three-dimensional anisotropic wave velocity field model that is synchronized with the mechanical state of the surrounding rock using standard seismic source events during tunnel excavation. The microseismic event spatiotemporal strong localization module 103 is connected to the dome curved surface sensor array module 101 and the dynamic wave velocity field modeling module 102, respectively. It is used to automatically identify, pick up first arrival, spatially locate, calculate the time of occurrence and the source mechanism and energy parameters of microseismic events based on the microseismic waveform data and the three-dimensional anisotropic wave velocity field model, and generate a microseismic event sequence. The intelligent early warning module 104 for rupture risk is connected to the spatiotemporal strong localization module 103 for microseismic events. It is used to extract and fuse the multidimensional physical features of the microseismic event sequence and output graded early warning information.

[0027] In this embodiment, to address the signal acquisition distortion problem, the dome curved surface sensor array module 101 abandons the traditional practice of deploying sensors on the sidewalls or invert. In the section of the tunnel that has been excavated and initially supported, a monitoring section is set at predetermined intervals (e.g., 15 to 25 meters) along the longitudinal direction of the tunnel. On each monitoring section, at least three broadband microseismic sensors are deployed circumferentially within the curved surface of the tunnel dome. These sensors are tightly coupled to the dome rock mass through anchoring agents or mechanical expansion, forming an arc-shaped sensor array that fits the curved surface of the dome. At the same time, auxiliary sensors are also deployed on the sidewalls on both sides of the corresponding section. Thus, the dome array and sidewall sensors of multiple monitoring sections together construct a three-dimensional "dome-sidewall" monitoring network. This layout ensures that the sensor's main axis direction matches the dominant radiation direction of the dome fracture source. Furthermore, because the sensor is close to the potential fracture source (the rock mass above the dome), the wave propagation path is significantly shortened, effectively suppressing scattering interference from the tunnel cavity and laying the foundation for obtaining high signal-to-noise ratio raw microseismic waveform data. To adapt to the dynamic changes in surrounding rock properties during tunnel excavation, the dynamic wave velocity field modeling module 102 establishes a wave velocity field model that iteratively updates with the excavation progress. During tunnel excavation, using the new face position after each blasting as a reference, the wave velocity field is constructed using the following two methods: First, the tunneling blasting itself is used as a known seismic source. The system accurately records the blasting time (triggered by the blasting signal or synchronized with the blasting control system), and each sensor receives the elastic wave generated by the blasting. The equivalent wave velocity on the path from the current tunnel face to each sensor is calculated through inversion.

[0028] Second, during the shutdown period or non-blasting period, on the rock surface near the located working face, use an automatic drop hammer device or a small controllable seismic source to actively generate elastic waves at known coordinate points to conduct wave velocity tests.

[0029] By fusing multipath wave velocity data obtained through the two methods mentioned above, the system interpolates and constructs a three-dimensional wave velocity field model that varies with spatial location based on Gaussian process regression or similar geostatistical methods. Furthermore, as the tunnel face advances, the system periodically (e.g., every 10 meters of excavation) updates the model using newly acquired wave velocity data, discarding outdated data to ensure that the wave velocity field always reflects the true stress and fracture state of the surrounding rock.

[0030] The microseismic event spatiotemporal strong localization module 103 receives continuous waveform data from the dome surface sensor array and processes it in real time. First, using the long-short time window ratio method combined with the AIC Akaike information criterion, it automatically identifies microseismic events and accurately picks the first arrival times of P-waves (longitudinal waves) for each channel. Then, based on the real-time three-dimensional wave velocity field provided by the aforementioned dynamic wave velocity field modeling module 102, it substitutes the spatial coordinates and first arrival times of each sensor to establish a set of localization equations. Since the sensor array covers the dome surface, this set of localization equations has excellent constraints. The system uses a grid search method or particle swarm optimization algorithm to solve for the optimal solution of the objective function in the key area above the tunnel dome, thereby obtaining the three-dimensional spatial coordinates, occurrence time, and radiated energy of the microseismic event. This localization process fully considers the anisotropy and spatial non-uniformity of the wave velocity field, ensuring high accuracy of the localization results. At the same time, the system calculates the magnitude, energy, and focal mechanism of each event (through P-wave first motion polarity analysis).

[0031] Finally, the fracture risk intelligent early warning module 104 constructs a multi-dimensional early warning index system based on the continuous event sequence output by the microseismic event spatiotemporal strong positioning module 103, thereby realizing the physical-driven identification of the rock mass fracture evolution process.

[0032] Furthermore, the dome surface sensor array module 101 includes: The curved surface adaptive coupling unit 1011 is used to realize the acoustic coupling between the sensor and the curved rock mass of the dome. The conformal spatial configuration unit 1012 is used to construct the spatial geometric layout of the sensor array. Monitoring sections are set at intervals along the longitudinal direction of the tunnel. At least three sensors are arranged at equal angles along the dome circumferential curved surface in each monitoring section to form a dome circumferential array. Auxiliary sensors are arranged on both sides of the same section to form a sidewall auxiliary array. The dome circumferential array and the sidewall auxiliary array of multiple sections together constitute a three-dimensional monitoring network. The multi-parameter synchronization triggering unit 1013 is used to provide time synchronization for the entire system, and adopts a combination of time protocol and hardware pulse calibration to achieve full-channel synchronization. The anti-interference signal conditioning unit 1014 is used to ensure the transmission fidelity of micro-seismic signals in interference environments.

[0033] In this embodiment, the curved surface adaptive coupling unit 1011 is used to solve the problem of achieving stable, efficient, and non-destructive acoustic coupling between the sensor and the irregular curved rock mass of the tunnel dome. This unit consists of four parts: a coupling base, a universal adjustable ball joint, an elastic clamping assembly, and an acoustic coupling medium layer. The coupling base is a prefabricated metal disc with an anchor bolt interface on its back. It is fixed to the borehole in the dome rock mass using expansion anchor bolts or adhesive, forming the main load-bearing structure. The universal adjustable ball joint is installed on the front of the coupling base and has three rotational degrees of freedom. It can adaptively conform to the local normal direction of each point on the curved surface of the dome, ensuring that the axis of the sensor installed on it always points to the center of the dome's curvature, and that the sensor's main response direction is consistent with the dominant radiation direction of the dome's fracture source. The elastic clamping assembly is a set of preloaded springs and locking nuts installed between the ball joint and the sensor, providing a continuous and stable axial preload to compensate for the small gaps in the coupling surface caused by creep or temperature changes in the surrounding rock. The acoustic coupling medium layer is a special epoxy resin with high fluidity and low attenuation. After the sensor is installed, it is injected into the tiny gap between the bottom surface of the sensor and the rock mass. After curing, it forms a rigid coupling layer with good acoustic impedance matching, effectively transmitting high-frequency elastic waves. Through the above structure, this unit achieves a quadruple coupling guarantee of "anchoring-orientation-compression-filling" for the sensor on the complex dome surface, greatly improving the acquisition efficiency and consistency of microseismic signals, especially the high-frequency P-wave components.

[0034] The conformal spatial configuration unit 1012 addresses the insufficient spatial resolution of dome rupture sources using traditional linear sensor deployment methods. By optimizing the spatial geometry of the sensors, it enhances the ability to locate and constrain rupture sources above the dome. A three-dimensional spatial configuration scheme of "longitudinal interval sections and conformal circumferential curved surfaces" is adopted. Along the longitudinal direction of the tunnel (excavation direction), the system sets up a monitoring section at predetermined intervals (preferably 15 to 25 meters, specifically determined based on the tunnel cross-section dimensions and surrounding rock grade). On each monitoring section, this unit deploys at least three sensors at equal angular intervals along the circumferential curved surface of the tunnel dome. The spatial coordinates of these sensors form an arc-shaped array that perfectly conforms to the dome's curved surface, referred to as the dome circumferential array. Furthermore, to enhance the constraint capability in the depth direction of the rupture source (i.e., the vertical distance above the dome), this unit also adds an auxiliary sensor at the height corresponding to the dome circumferential array on each side wall of the same section, forming a side wall auxiliary array.

[0035] Thus, the dome-shaped circumferential array and the sidewall auxiliary array on multiple longitudinal sections together construct a "dome-sidewall" three-dimensional monitoring network. The beneficial effects of this configuration are: firstly, the dome-shaped circumferential array allows the sensors to be close to the potential fracture source area, greatly shortening the wave propagation path and reducing signal attenuation and waveform distortion; secondly, the circumferential arrangement gives the array a 360-degree azimuth coverage capability in the horizontal plane, which, combined with the depth constraint provided by the sidewall auxiliary array, forms a three-dimensional enveloping observation geometry for the rock fracture source above the dome, providing excellent spatial constraints for subsequent high-precision positioning.

[0036] The multi-parameter synchronization triggering unit 1013 is used to solve the problem of high-precision time synchronization between multiple sensors and between sensors and other systems (such as blasting control systems and stress monitoring systems), ensuring the accuracy of microseismic events and time difference measurements. Microseismic positioning relies on the small time difference between the first arrival and arrival of P-waves between different sensors (typically requiring an accuracy better than 0.1 milliseconds). A redundant synchronization scheme based on a high-stability crystal oscillator and pulse time synchronization is adopted. This unit sets up a high-precision synchronization time master in the tunnel. The master integrates a rubidium atomic clock or a high-stability isothermal crystal oscillator as the local time reference for the entire array. The master is connected to the data acquisition station of each monitoring section through optical fiber, uses the IEEE 1588 precision time protocol for network synchronization, and periodically sends hardware pulse signals to calibrate the clock of the data acquisition station, ensuring that the sampling clock deviation of all channels along the entire line is better than 50 nanoseconds. More importantly, this unit also sets up a blasting source synchronization interface for blasting tunneling scenarios. This interface connects to the blasting control system at the tunnel face. At the moment of blasting initiation, the blasting control system outputs a hardware trigger signal to the synchronization time host. This signal, after being precisely timestamped, is recorded along with the elastic wave data generated by the blast. This makes the blasting event itself a "standard seismic source" with known coordinates and a known precise time of origin, which can be used for subsequent wave velocity field modeling. Simultaneously, this unit also reserves hardware synchronization interfaces with third-party sensors such as automated total stations and stress monitoring systems, providing a unified time reference for multi-source data fusion.

[0037] The anti-interference signal conditioning unit 1014 addresses the impact of strong electromagnetic interference, mechanical vibration interference, and signal attenuation over long distances on the quality of the original micro-vibration waveform in tunnel construction environments. This unit employs a three-tiered anti-interference architecture: front-end amplification, opto-isolation, and floating ground shielding. First, a pre-amplifier with low noise is integrated within the sensor body to convert the original weak charge signal into a voltage signal on-site and amplify it, improving the signal-to-noise ratio during transmission. This amplifier uses a differential input circuit to effectively suppress common-mode interference. Second, an opto-isolation module is installed between the sensor and the lead-out cable. This module converts the electrical signal into an optical signal for transmission in the cable, and then converts it back into an electrical signal upon reaching the data acquisition station. Opto-isolation completely cuts off the path for electromagnetic interference to couple into the signal loop through the cable, while also solving the ground loop interference problem in long-distance transmission. Finally, the entire sensor assembly, cable, and its interfaces all adopt a double-layer floating ground shielding structure. The inner shielding layer wraps the signal line and is equipotentially connected to the sensor housing; the outer shielding layer wraps the cable assembly and is grounded at a single point to the grounding grid inside the tunnel. The inner shield is used to filter out electric field interference, and the outer shield is used to filter out magnetic field interference. The floating ground design also prevents grounding current from generating interference voltage on the shield.

[0038] Through the aforementioned three-level architecture, this unit enables the microseismic waveforms acquired in a noisy environment to maintain a high signal-to-noise ratio, providing a reliable raw data foundation for the accurate acquisition of P-wave arrivals and the accurate identification of microseismic events.

[0039] Furthermore, the dynamic wave velocity field modeling module 102 includes: The standard source excitation unit 1021 is used to provide an elastic wave source with known spatial coordinates and excitation time. It adopts a dual-mode scheme that combines passive use of tunneling blasting source and active excitation of artificial source. In passive mode, the excitation time and spatial coordinates of blasting source are obtained synchronously with the blasting control system hardware. In active mode, a controllable impact source is excited as needed at preset measuring points. The multipath wave velocity extraction unit 1022 is used to extract equivalent wave velocity values ​​on multi-directional propagation paths from a standard source event. For each standard source event, it calculates the path length based on the known source coordinates and the coordinates of each sensor, calculates the propagation time based on the earthquake time and the arrival time of the P-wave in each channel, calculates the equivalent wave velocity of each path, and performs confidence assessment and weighting on the wave velocity values ​​of each path. The wave velocity field dynamic interpolation unit 1023 is used to extend discrete path wave velocity sampling values ​​into a continuous three-dimensional anisotropic wave velocity field model. It adopts the Gaussian process regression method, introduces surrounding rock stress monitoring data as covariates, constructs a joint interpolation model, and outputs a three-dimensional voxel grid wave velocity field covering the monitoring area. Each grid cell contains the full wave velocity tensor. The model iteration verification unit 1024 is used to determine the update timing and applicable time of the wave velocity field model. It adopts an iterative update mechanism that combines residual analysis, positioning error threshold, and tunneling progress triple criteria. It retains the wave velocity sampling data of the most recently tunneled section in a sliding window manner to realize the on-demand and periodic update of the wave velocity field model.

[0040] In this embodiment, the standard seismic source excitation unit 1021 is used to actively or passively generate elastic wave sources with known spatial coordinates and precise excitation times in a controllable or calibrable manner during tunnel excavation, providing a standard reference for wave velocity measurement. This unit adopts a dual-mode seismic source scheme combining "passive utilization of tunneling blasting" and "active excitation of artificial seismic sources" to ensure that sufficient standard seismic sources are available at any stage of tunneling.

[0041] In passive mode, this unit establishes a hardware synchronization link with the blasting control system at the tunnel face. During each blast, the blasting control system outputs a high-precision trigger signal to the aforementioned "multi-parameter synchronous trigger unit 1013." This signal, after being precisely timestamped, serves as the moment of the blasting event. Simultaneously, by combining the tunneling design drawings and blasting hole layout diagram with the real-time three-dimensional coordinates of the tunnel face measured by an automated total station, the center position of the blasting source is precisely determined. Thus, each tunneling blast becomes a natural, standard seismic source with known spatial coordinates and a known moment of occurrence.

[0042] In active mode, for sections not requiring blasting or where denser wave velocity measurements are needed, this unit deploys a mobile automatic falling hammer seismic source vehicle within the tunnel. This vehicle is equipped with a controllable electromagnetic falling hammer device, which automatically rises and falls at preset measurement points (such as the arch foot or sidewall of a supported section behind the tunnel face) according to instructions, generating a repeatable and energy-stable impact source. The falling hammer device has a built-in accelerometer sensor that accurately records the hammer impact trigger time, and simultaneously obtains the precise three-dimensional coordinates of the impact point through an onboard positioning module or by connecting to a total station. The active seismic source can supplement wave velocity measurement data as needed when there is no blasting operation, especially during tunnel downtime or in sections with abrupt changes in geological conditions, providing a dense network of wave velocity sampling points.

[0043] The multi-path wave velocity extraction unit 1022 is used to extract equivalent wave velocity values ​​along different propagation paths from the propagation of elastic waves excited by a standard seismic source, constructing an original wave velocity sampling dataset. This unit employs a full-path wave velocity inversion method based on a multi-sensor array, fully leveraging the spatial coverage advantages of the "dome-sidewall" three-dimensional monitoring network. After the standard seismic source (whether it is a tunneling blast or an active drop hammer) is excited, this unit receives waveform data from all channels after the event is triggered. First, the aforementioned first-arrival picking algorithm is used to accurately extract the P-wave first arrival time for each sensor channel. Since the source coordinates and the time of origin are known, each sensor channel corresponds to a defined propagation path. The spatial geometry of this path is uniquely determined by the source coordinates and sensor coordinates.

[0044] For each propagation path, this unit calculates its equivalent wave velocity value using the formula: path length divided by propagation time (first arrival time minus the time of origin). The path length is calculated using the three-dimensional spatial coordinates of the source and the sensor, while the propagation time is extracted directly from the waveform data. Thus, a single standard source event can generate multiple path wave velocity values ​​equal to the number of sensors. For a "dome-sidewall" three-dimensional monitoring network composed of multiple sensors, a single source event can typically generate dozens of propagation path wave velocity sampling values ​​with different directions and lengths.

[0045] This unit assesses the confidence level of wave velocity values ​​for each path. The assessment criteria include: waveform signal-to-noise ratio along the path, reliability of first-arrival picking, and whether the path traverses known geological anomalies (such as fault fracture zones). High-confidence wave velocity values ​​are assigned higher weights for subsequent modeling, while low-confidence wave velocity values ​​are marked as suspicious data or directly discarded. Using this method, this unit constructs a raw dataset containing a large number of multi-path, multi-directional wave velocity samples, providing rich and reliable input for the subsequent construction of a three-dimensional wave velocity field.

[0046] The wave velocity field dynamic interpolation unit 1023 is used to interpolate and extend discrete path wave velocity samples distributed at different spatial locations into a continuous three-dimensional spatial wave velocity field model, and to achieve a preliminary correlation with the surrounding rock stress field. This unit adopts a geostatistical method based on Gaussian process regression and innovatively introduces surrounding rock stress field information as a covariate to construct a joint interpolation model coupled with the "wave velocity field-stress field". First, the spatial midpoint and path direction of each wave velocity sampling path are used as input features. Since wave velocity has direction dependence (anisotropy), this unit decomposes the wave velocity value into two parts: an anisotropic component related to the path direction and an isotropic background component. The anisotropic component reflects the directional difference of wave velocity caused by the stress state or fracture orientation of the rock mass, while the isotropic background component reflects the spatial distribution of the overall mechanical properties of the rock mass. During the interpolation process, this unit not only utilizes the wave velocity sampling values ​​provided by the multi-path wave velocity extraction unit 1022, but also integrates measured stress data from tunnel surrounding rock stress monitoring systems (such as vibrating wire stress gauges and fiber optic strain sensors). Specifically, this unit establishes a joint covariance function that simultaneously considers the influence of spatial distance, path direction differences, and stress state differences on wave velocity correlation. By maximizing the likelihood function and optimizing the hyperparameters in the covariance function, the interpolated wave velocity field is not only spatially smooth but also physically consistent with the distribution trend of the measured stress field—that is, high stress areas correspond to higher wave velocities, and low stress areas or relaxation areas correspond to lower wave velocities.

[0047] Finally, this unit outputs a three-dimensional wave velocity field model covering the entire monitoring area. This model is stored in voxel grid form, with each grid cell containing a full wave velocity tensor that can describe the velocity of elastic waves at that location in different propagation directions. This model provides a physical constraint for the subsequent localization of microseismic events that is far superior to that of traditional single wave velocity values.

[0048] The model iteration verification unit 1024 is used to verify the self-consistency of the updated wave velocity field model and determine the applicable time frame and update triggering conditions of the model, ensuring that the wave velocity field is always synchronized with the actual state of the surrounding rock. A model iteration update mechanism based on the triple criteria of "residual analysis - error threshold - tunneling progress" was established.

[0049] First, residual analysis. Each time the wave velocity field is updated using a new standard source event, this unit treats that source event as an independent test sample, not participating in the current modeling. Instead, it uses the wave velocity field model before the update to predict the propagation time from the source to each sensor, calculating the residual between the predicted time and the actual time. If the root mean square error of the residual exceeds a set threshold, it indicates that the existing wave velocity field model can no longer accurately describe the current surrounding rock condition, triggering an immediate update.

[0050] Second, the error threshold. This unit continuously monitors the location error of the most recent N standard source events (using the known source location as a verification standard, the current wave velocity field model is used for location, and the deviation between the location result and the actual location is calculated). When the average location error of multiple consecutive events exceeds the system's preset accuracy threshold, the reconstruction or iterative optimization of the wave velocity field model is triggered.

[0051] Third, tunneling progress. Under conditions of no abnormal residuals and stable positioning errors, this unit will still automatically trigger an incremental update of the wave velocity field model according to the preset tunneling progress (e.g., every 10 meters of tunneling or every three blasting cycles). During the update, a sliding window mechanism is used, retaining only the wave velocity sampling data from the most recent period (e.g., the most recent 30-meter tunneling interval) for modeling, discarding outdated data, ensuring that the model reflects the surrounding rock characteristics of the current section.

[0052] Furthermore, the microseismic event spatiotemporal strong localization module 103 includes: The waveform preprocessing and event detection unit 1031 is used to identify microseismic event segments from continuous waveform data. It uses frequency band adaptive filtering to process the original waveform into frequency bands and adopts a multi-channel joint judgment mechanism. When a preset number of channels simultaneously meet the triggering conditions, it confirms that a microseismic event has been detected. The multi-channel first arrival accurate picking unit 1032 is used to extract the first arrival time of P-waves for each channel of the detected microseismic event. It performs optimal frequency band scanning for the event waveform of each channel, selects the dominant frequency band with the highest signal-to-noise ratio, and runs three independent algorithms in parallel: long and short window ratio method, Akaike information criterion method and wavelet transform modulus maxima method. The first arrival time is obtained by fusing them through voting and verification mechanism, and the confidence level of each channel is output. Anisotropic medium positioning unit 1033 is used to invert the spatial coordinates and occurrence time of microseismic events. Based on the three-dimensional anisotropic wave velocity field model, it uses anisotropic ray tracing algorithm to calculate theoretical travel time, establishes a positioning model with the weighted sum of squares of the difference between the theoretical arrival time and the actual picked arrival time of each channel as the objective function, and uses a hybrid strategy combining grid search and particle swarm optimization to solve the problem, outputting the three-dimensional spatial coordinates of the event, the occurrence time, and the positioning information ellipsoid. The focal mechanism and energy calculation unit 1034 is used to calculate the intensity and rupture mechanism parameters of microseismic events. It uses a grid search method to invert the focal mechanism solution based on the first motion polarity of multi-channel P-waves. After instrument response correction and geometric diffusion correction, it uses the Brune model to fit and calculate the seismic moment and moment magnitude. It calculates the radiated energy by integrating the displacement spectrum and calculates the energy-to-moment ratio.

[0053] In this embodiment, the waveform preprocessing and event detection unit 1031 is used to perform real-time filtering, denoising, and event identification on the raw waveform data, efficiently filtering out microseismic event segments that may be caused by rock mass fracturing from continuous massive data, reducing the computational burden of subsequent processing. This unit first performs frequency band processing on the raw waveform data of each channel. Based on the dominant frequency characteristics of the tunnel dome rock mass fracturing signal (typically between tens of hertz and thousands of hertz), the signal is divided into three sub-bands: low-frequency, mid-frequency, and high-frequency. Each frequency band uses an adaptive filter, and the filter parameters are dynamically adjusted according to the real-time background noise level of the channel to ensure that effective signals are preserved to the maximum extent while suppressing noise. Subsequently, this unit independently runs an improved long-short window ratio detector on each frequency band. Unlike traditional methods, this detector adopts a "sliding window-dual threshold-time constraint" strategy. A lower trigger threshold is set to capture weak signals, and a higher confirmation threshold is set to eliminate noise interference. When the ratio of long to short time windows in a frequency band exceeds a low threshold, the system enters a waiting-to-trigger state; if the ratio further exceeds a high threshold within the specified time window, it is determined to be a valid event trigger. Simultaneously, this unit introduces a multi-channel joint determination mechanism: a micro-vibration event is confirmed only when at least a preset number of channels (e.g., no fewer than three channels) simultaneously meet the triggering conditions. This mechanism effectively eliminates sporadic noise interference from single channels.

[0054] The multi-channel first-arrival precision acquisition unit 1032 is used to accurately extract the P-wave first arrival time from the waveform data of each effective channel within the time window of the detected microseismic event. This is a decisive factor in subsequent positioning accuracy. For each channel's event waveform segment, optimal frequency band scanning analysis is performed. The signal-to-noise ratio (SNR) of the waveform in different frequency bands is calculated, and the frequency band with the highest SNR is selected as the "dominant frequency band" for that channel. Subsequent first-arrival acquisition processing is then performed within this frequency band. This step ensures that each channel performs first-arrival analysis in its optimal signal quality frequency band, avoiding signal distortion or noise amplification caused by fixed-band filtering. Three independent first-arrival acquisition algorithms are run in parallel: one is the improved long-short time window ratio method, which determines the first arrival by accurately calculating the rate of change of the characteristic function; the second is the Akaike information criterion method, which determines the first arrival by finding the minimum point of information in the waveform autoregressive model; and the third is the modulus maxima method based on wavelet transform, which determines the first arrival by detecting abrupt changes in the signal in the time-frequency domain.

[0055] For each channel, the three algorithms provide their respective initial arrival results. This unit then proceeds to the "voting and verification" stage: the mean and standard deviation of the three results are calculated. If the standard deviation of the three results is less than a set threshold, it indicates good consistency, and the mean is taken as the final initial arrival time. If the standard deviation exceeds the threshold, it indicates discrepancies, and the waveform envelope energy curve and instantaneous frequency curve are introduced as auxiliary criteria to eliminate the algorithm results with the greatest deviation, and then the mean of the remaining results is taken. For the very few channels that still cannot be reliably picked up, this unit marks them as "low-confidence channels." These channels do not participate in subsequent positioning calculations, but their waveform data is retained for reference.

[0056] The anisotropic medium positioning unit 1033 is used to invert and calculate the three-dimensional spatial coordinates and origin time of microseismic events using the three-dimensional anisotropic wave velocity field provided by the aforementioned multi-channel first arrival time and dynamic wave velocity field modeling module 102. First, a positioning objective function is established. For each effective channel involved in positioning, its theoretical arrival time is jointly determined by the source coordinates, the origin time, and the integrated travel time along the path from the source to the sensor. Unlike conventional methods, the theoretical travel time calculation in this unit does not use a fixed wave velocity multiplied by a straight-line distance. Instead, it calls the three-dimensional anisotropic wave velocity field model provided by the dynamic wave velocity field modeling module 102 and uses a ray tracing algorithm to solve for the actual propagation path and travel time of elastic waves in non-uniform anisotropic media. This ray tracing algorithm considers the spatial non-uniformity of wave velocity and its variation with the propagation direction, and can accurately describe the propagation law of waves in complex surrounding rocks.

[0057] The positioning objective function is defined as the weighted sum of squared differences between the theoretical arrival time and the actual pickup arrival time for all valid channels, where the weights are determined by the confidence scores of each channel output by the second unit—channels with higher confidence scores have larger weights. This unit employs a hybrid optimization strategy combining grid search and particle swarm optimization (PSO) to solve the problem. First, a coarse grid is established in the key area of ​​interest above the tunnel dome for grid search to obtain an initial solution. Then, using the initial solution as the center, PSO is used to perform a fine search in continuous space, iteratively optimizing the objective function until the convergence condition is met.

[0058] Because this module employs a "dome-sidewall" three-dimensional monitoring network, the sensors are well-distributed in both the horizontal and vertical directions, resulting in a single, sharp minimum point for the positioning objective function in space. This effectively avoids the multiple-solution problem common in traditional sidewall deployment methods. Finally, this unit outputs the three-dimensional spatial coordinates and occurrence time of the microseismic event, and provides a confidence ellipsoid of the positioning result to characterize the positioning uncertainty.

[0059] The focal mechanism and energy calculation unit 1034 is used to calculate the intensity parameters (magnitude, radiated energy) and focal mechanism of microseismic events based on the determined spatial location of the event, providing multi-dimensional physical characteristics for subsequent early warning modules. For the focal mechanism, this unit extracts the polarity of the P-wave initial motion (i.e., the positive and negative directions of the first peak of the waveform) on all effective channels. Since the sensor array of this system covers the dome surface, it can observe the same source from multiple azimuth angles. This polarity information constitutes an effective constraint on the focal mechanism (fault type and rupture surface orientation). This unit uses a grid search method to traverse possible focal mechanism solutions, calculates the degree of agreement between each candidate solution and the measured polarity, selects the solution with the highest agreement as the focal mechanism of the event, and outputs the spatial orientation of its corresponding P-wave radiation pattern.

[0060] For energy calculation, this unit first performs instrument response correction and geometric diffusion correction on the waveform of each channel to reconstruct the true displacement spectrum at the source. Then, the Brune model is used to fit the low-frequency plateau and high-frequency attenuation of the displacement spectrum to calculate the seismic moment corresponding to that channel. By synthesizing the seismic moments of all effective channels, their weighted average is taken to obtain the final seismic moment of the event. Furthermore, the moment magnitude is calculated based on the seismic moment. Simultaneously, this unit calculates the total elastic wave energy radiated by the event by integrating the displacement spectrum. In addition, this unit also calculates a key parameter—the energy-to-moment ratio. This ratio reflects the rupture nature of the source: a high ratio usually corresponds to stress-driven brittle fracture, a precursor to rock mass instability; a low ratio corresponds to relatively stable slip or microcrack closure. This parameter will serve as one of the important input features for subsequent early warning modules.

[0061] Through the above calculations, this unit assigns a complete set of parameters to each microseismic event, including: three-dimensional spatial coordinates, time of origin, moment magnitude, radiated energy, energy-to-moment ratio, focal mechanism solution (P-wave initial polarity distribution), and location information ellipsoid. These parameters together constitute a complete description of the microseismic event sequence, providing a rich data foundation for the rupture risk intelligent early warning module 104.

[0062] Furthermore, the intelligent early warning module 104 for rupture risk includes: The fracture energy release rate analysis unit 1041 is used to identify the unsteady acceleration characteristics of rock mass fracture from the energy dimension. It constructs the energy time series of microseismic event sequences, smooths them into continuous energy release rate curves using kernel density estimation, calculates the energy acceleration index by fitting an exponential function through a sliding window, identifies the acceleration release start point, and outputs the energy release rate value and energy acceleration index. The fracture volume fractal dimension analysis unit 1042 is used to identify the cluster evolution characteristics of microseismic events from the spatial dimension. It uses a sliding window to truncate a subset of microseismic events, uses the box counting method to calculate the spatial fractal dimension, tracks the trend of fractal dimension change, and when the fractal dimension continuously decreases from a high value to below a set threshold, it is determined to be a fracture cluster nucleation state, and outputs the current fractal dimension value and the trend of change. The focal mechanism consistency analysis unit 1043 is used to identify the unified evolution characteristics of the focal mechanism of microseismic events from the mechanism dimension. It constructs a focal mechanism consistency index for microseismic events within a sliding window, and quantifies the degree of consistency by calculating the root mean square value of the spatial angle between the focal mechanism of each event and the average focal mechanism within the window. When the consistency index decreases to below a set threshold, it is determined to be in a state of rupture surface penetration, and outputs the consistency index value and the orientation of the dominant rupture surface. The multi-indicator fusion early warning output unit 1044 is used to fuse and infer the outputs of the aforementioned three analysis units and output graded early warning information. It adopts a Bayesian network to construct a three-layer inference structure, with energy acceleration index, fractal dimension and its changing trend, and source mechanism consistency index and its changing trend as observation nodes, energy acceleration release state, fracture cluster nucleation state, and fracture surface penetration state as hidden state nodes, and rock mass fracture risk level as target node. It calculates the posterior probability of risk level through Bayesian inference, outputs level three, level two, and level one early warning information, and pushes it to the monitoring terminal and mobile terminal in a multimodal manner.

[0063] In this embodiment, the fracture energy release rate analysis unit 1041 is used to characterize the intensity evolution of rock mass fracture from an energy perspective, and to identify the key critical point where the fracture transitions from steady-state propagation to unsteady-state accelerated propagation. This is one of the most significant precursory features before macroscopic failure of the rock mass. An energy release rate monitoring method based on "time series nonlinear analysis" is adopted, which specifically includes three levels of technical processing.

[0064] First, this unit constructs the energy time series of microseismic events. Each event is represented as a discrete energy event point based on its occurrence time and radiated energy value. This unit uses kernel density estimation to smooth these discrete points into a continuous energy release rate curve, effectively eliminating the interference caused by random fluctuations in individual events and extracting the overall trend of energy release.

[0065] Secondly, this unit employs a sliding window technique on the energy release rate curve to calculate the growth exponent of the energy release rate within the window in real time. This growth exponent is not a simple linear slope, but a nonlinear acceleration coefficient obtained by fitting an exponential function. Specifically, this unit assumes that the energy release rate follows a power law or exponential growth pattern in the near-failure stage, and fits the growth coefficient using the least squares method. When this growth coefficient exceeds a preset threshold and the goodness of fit reaches a certain level, it indicates that the energy release has entered a nonlinear acceleration stage, which is identified as a precursor to "accelerated energy release".

[0066] Finally, this unit introduces an "accelerated release start point" identification algorithm. By reverse-searching the inflection point on the energy release rate curve, the time node at which acceleration begins is determined. This node is typically several hours to several days earlier than the occurrence of macroscopic damage, providing valuable lead time for early warning. This unit outputs two key parameters: the current energy release rate value, the energy acceleration index, and the accelerated release start timestamp, for use by subsequent fusion units.

[0067] The fracture volume fractal dimension analysis unit 1042 is used to characterize the spatial distribution features of microseismic events from a spatial dimension, identifying the process of rupture evolution from random, dispersed distribution to localized clustered distribution, which is an important indicator of the gradual formation and expansion of the rupture surface. This unit first extracts a subset of microseismic events within a fixed time window (e.g., the most recent 24 hours or the most recent 100 events), with each event represented by its three-dimensional spatial coordinates. Then, the spatial fractal dimension of this event point set is calculated using box counting. Specifically, the entire monitoring area is covered with cubic grids of different sizes, and the number of grids containing at least one event point is counted. The fractal dimension value is obtained by fitting the slope of a double logarithmic curve of grid size versus grid number.

[0068] The physical meaning of fractal dimension is as follows: when microseismic events are randomly and uniformly distributed in space, the fractal dimension is close to the upper limit of three-dimensional space; when events begin to cluster in local areas and gradually form macroscopic rupture surfaces, the event point set exhibits a self-similar structure, and the fractal dimension will decrease significantly; when the rupture surface is completely connected and the events are highly concentrated in a narrow region, the fractal dimension approaches the characteristic value of a one-dimensional linear distribution.

[0069] This unit calculates the fractal dimension within a sliding window in real time and tracks its changing trend. The key innovation lies in the fact that this unit not only calculates the absolute value of the current fractal dimension but also focuses on its rate of change and trend direction. When the fractal dimension continuously decreases from a high value and falls below a set threshold, it indicates that microseismic events have shifted from a random, dispersed distribution to a locally clustered distribution, and the rupture has entered the "nucleation" stage. Simultaneously, this unit calculates the second derivative of the fractal dimension. When the second derivative is positive, it indicates that the rate of decrease in the fractal dimension has slowed, and the rupture cluster region is becoming stable. This usually means that the macroscopic rupture surface has essentially formed and is about to enter the instability stage. This unit outputs the current fractal dimension value, the fractal dimension changing trend (decreasing / stable / increasing), and the spatial extent of the rupture cluster region for use by subsequent fusion units.

[0070] The focal mechanism consistency analysis unit 1043 is used to characterize the evolution of focal mechanisms in microseismic events from the perspective of rupture mechanism, identifying the process of rupture transformation from random orientation to unified orientation, which is direct evidence of macroscopic rupture surface connectivity. First, the focal mechanism parameters of each microseismic event are extracted, mainly the P-wave initial motion polarity distribution output by the aforementioned focal mechanism and energy calculation unit 1034. Since the dome surface sensor array provides observations from multiple azimuth angles, the polarity distribution of each event can effectively constrain its focal mechanism type and rupture surface orientation. A cumulative analysis method is used to construct a focal mechanism consistency index within a sliding window. Specifically, for each event within the window, the angle (i.e., spatial orientation difference) between its focal mechanism and the average focal mechanism of all events within the window is calculated. The root mean square value of the angles for all events is defined as the focal mechanism consistency index for that window. The smaller the index value, the more consistent the event's focal mechanism; the larger the index value, the more dispersed the event's focal mechanism.

[0071] Physically, during the stable deformation stage of a rock mass, the development direction of microfractures is random, and the focal mechanisms of microseismic events exhibit a highly dispersed state with a high consistency index. As the macroscopic fracture surface gradually forms, subsequent microfracture events mainly concentrate near this fracture surface, and their focal mechanisms tend to be consistent with the orientation of the main fracture surface, significantly reducing the consistency index. When the consistency index drops below a set threshold, it indicates that the fracture surface has been connected, and the rock mass is in a critical state of instability.

[0072] This unit not only calculates the consistency index for the current window but also constructs its time series to identify transition points from dispersion to concentration. Simultaneously, it calculates the spatial orientation of the average focal mechanism, which represents the dominant direction of the potential macroscopic damage surface, providing directional guidance for targeted on-site reinforcement. This unit outputs the current focal mechanism consistency index, consistency trend (decreasing / stabilizing), and dominant rupture surface orientation for use by subsequent fusion units.

[0073] The multi-indicator fusion early warning output unit 1044 is used to fuse and infer multiple precursor characteristic indicators output by the aforementioned three analysis units, comprehensively judge the rock mass fracture risk level, and output early warning information in an intuitive and operable manner to guide on-site engineering decisions. First, a three-layer Bayesian network structure is constructed: the bottom layer consists of observation nodes, including key parameters output by the aforementioned three units (energy acceleration index, fractal dimension, fractal dimension change trend, source mechanism consistency index, consistency change trend, etc.); the middle layer consists of latent state nodes, corresponding to three physical precursor states (energy accelerated release state, fracture cluster nucleation state, fracture surface penetration state); the top layer consists of target nodes, namely the rock mass fracture risk level.

[0074] The prior probabilities of the Bayesian network are obtained through training with historical microseismic event data and corresponding engineering site records (such as collapse and rockfall event records). When new microseismic event sequence data is input, this unit calculates the posterior probability of each hidden state under the current observation evidence using Bayesian inference, and then calculates the posterior probability distribution of the risk level.

[0075] The risk levels are divided into three levels, and their physical meanings correspond clearly to the engineering response measures: Level 3 Warning (Yellow): Triggered when only one of the three latent states is determined to be active. This indicates that rock mass fracturing activity has increased, but has not yet formed a significant spatial cluster or accelerated trend. The warning information prompts on-site personnel to pay closer attention and increase the frequency of monitoring.

[0076] Level 2 Warning (Orange): Triggered when any two of the three latent states are determined to be active. This indicates that the rupture has shifted from a random distribution to a localized clustering, and energy release is accelerating or the rupture surface is beginning to converge. The warning message advises that high-risk operations should be halted on-site, technical personnel should be organized to analyze and assess the situation, and reinforcement measures should be prepared.

[0077] Level 1 Warning (Red): Triggered when all three latent states are determined to be active. This indicates that three physical precursor conditions—accelerated energy release, nucleation of fracture clusters, and breakthrough of fracture surfaces—are simultaneously met, and macroscopic failure of the rock mass is imminent. The warning message prompts the immediate evacuation of personnel from the danger zone, activation of the emergency response plan, and implementation of emergency reinforcement or closure measures.

[0078] The output of this unit is presented in a multimodal manner: on the large screen of the tunnel monitoring center, the spatial distribution of microseismic events, energy release time series curves, fractal dimension change curves, and source mechanism rose diagrams are displayed in a three-dimensional visualization form, and the warning level is highlighted; at the same time, the warning information is pushed to the mobile terminals of on-site management personnel and the audible and visual alarm devices through the wireless communication network to ensure that the warning information is promptly transmitted to decision-making and execution personnel.

[0079] This unit also possesses self-learning and feedback optimization capabilities. After each warning event occurs, the system records the actual situation on-site (whether damage has occurred, the scale of damage, and the measures taken), and incorporates this feedback data into the training set of the Bayesian network. The network parameters are updated and optimized regularly, enabling the warning model to continuously adapt to the geological and construction conditions of specific tunnel projects and continuously improve the accuracy of warnings.

[0080] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0081] It should be understood that this application is not limited to the precise structure described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope.

Claims

1. A microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in a tunnel dome, characterized in that, include: The dome surface sensor array module is used to construct a three-dimensional monitoring network conforming to the rock mass within the dome surface of the tunnel to collect microseismic waveform data. The dynamic wave velocity field modeling module is connected to the dome surface sensor array module and is used to construct and dynamically update a three-dimensional anisotropic wave velocity field model that is synchronized with the mechanical state of the surrounding rock using standard seismic source events during tunnel excavation. The microseismic event spatiotemporal strong localization module is connected to the dome curved surface sensor array module and the dynamic wave velocity field modeling module, respectively. It is used to automatically identify, pick up first arrival, spatially locate, calculate the time of occurrence, and calculate the source mechanism and energy parameters of microseismic events based on the microseismic waveform data and the three-dimensional anisotropic wave velocity field model, and generate a microseismic event sequence. The intelligent early warning module for rupture risk is connected to the spatiotemporal strong localization module for microseismic events. It is used to extract and fuse the multidimensional physical features of the microseismic event sequence and output graded early warning information.

2. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 1, characterized in that, The dome surface sensor array module includes: The curved surface adaptive coupling unit is used to realize the acoustic coupling between the sensor and the curved rock mass of the dome; The conformal spatial configuration unit is used to construct the spatial geometric layout of the sensor array. Monitoring sections are set at intervals along the longitudinal direction of the tunnel. At least three sensors are arranged at equal angles along the dome circumferential curved surface in each monitoring section to form a dome circumferential array. Auxiliary sensors are arranged on both sides of the same section to form a sidewall auxiliary array. The dome circumferential array and the sidewall auxiliary array of multiple sections together constitute a three-dimensional monitoring network. The multi-parameter synchronous triggering unit is used to provide time synchronization for the entire system, and adopts a combination of time protocol and hardware pulse calibration to achieve full-channel synchronization; An anti-interference signal conditioning unit is used to ensure the transmission fidelity of micro-seismic signals in interference environments.

3. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 2, characterized in that, In the conformal spatial configuration unit, the sensor layout of the dome circumferential array is such that the main response direction of the sensor points to the center of the dome curvature, matching the dominant radiation direction of the dome fracture source; the three-dimensional monitoring network forms a three-dimensional enveloping observation geometry for the rock mass fracture source above the dome, providing 360-degree azimuth coverage in the horizontal direction and depth constraint in the vertical direction.

4. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 1, characterized in that, The dynamic wave velocity field modeling module includes: The standard source excitation unit is used to provide an elastic wave source with known spatial coordinates and excitation time. It adopts a dual-mode scheme that combines passive use of tunneling blasting source and active excitation of artificial source. In passive mode, the excitation time and spatial coordinates of blasting source are obtained synchronously with the blasting control system hardware. In active mode, a controllable impact source is excited as needed at preset measuring points. The multipath wave velocity extraction unit is used to extract equivalent wave velocity values ​​on multi-directional propagation paths from standard source events. For each standard source event, it calculates the path length based on the known source coordinates and the coordinates of each sensor, calculates the propagation time based on the time of earthquake occurrence and the arrival time of P-waves in each channel, calculates the equivalent wave velocity of each path, and performs confidence assessment and weighting on the wave velocity values ​​of each path. The dynamic interpolation unit for wave velocity field is used to extend discrete path wave velocity sampling values ​​into a continuous three-dimensional anisotropic wave velocity field model. It adopts the Gaussian process regression method, introduces surrounding rock stress monitoring data as covariates, constructs a joint interpolation model, and outputs a three-dimensional voxel grid wave velocity field covering the monitoring area. Each grid cell contains the full wave velocity tensor. The model iteration verification unit is used to determine the update timing and applicable validity of the wave velocity field model. It adopts an iterative update mechanism that combines residual analysis, positioning error threshold, and tunneling progress triple criteria. It retains the wave velocity sampling data of the most recently tunneled section in a sliding window manner to realize on-demand and periodic updates of the wave velocity field model.

5. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 4, characterized in that, In the dynamic interpolation unit of the wave velocity field, the joint interpolation model decomposes the wave velocity value into anisotropic components related to the propagation path direction and isotropic background components related to the stress state of the surrounding rock. By using a joint covariance function, the influence of spatial distance, path direction differences, and stress state differences on the wave velocity correlation is considered simultaneously, so that the generated wave velocity field model maintains physical consistency with the measured stress field distribution trend.

6. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 1, characterized in that, The microseismic event spatiotemporal strong localization module includes: The waveform preprocessing and event detection unit is used to identify microseismic event segments from continuous waveform data. It uses frequency band adaptive filtering to process the original waveform into frequency bands and adopts a multi-channel joint judgment mechanism. When a preset number of channels simultaneously meet the triggering conditions, it confirms that a microseismic event has been detected. The multi-channel first arrival precision picking unit is used to extract the first arrival time of P-waves for each channel of the detected microseismic event. It performs optimal frequency band scanning for the event waveform of each channel, selects the dominant frequency band with the highest signal-to-noise ratio, and runs three independent algorithms in parallel: long-short window ratio method, Akaike information criterion method, and wavelet transform modulus maxima method. The first arrival time is obtained by fusing the algorithms through a voting and verification mechanism, and the confidence level of each channel is output. An anisotropic medium positioning unit is used to invert the spatial coordinates and origin time of microseismic events. Based on the three-dimensional anisotropic wave velocity field model, it uses anisotropic ray tracing algorithm to calculate the theoretical travel time, establishes a positioning model with the weighted sum of squares of the difference between the theoretical arrival time and the actual picked-up arrival time of each channel as the objective function, and uses a hybrid strategy combining grid search and particle swarm optimization to solve the problem, outputting the three-dimensional spatial coordinates of the event, the origin time, and the positioning information ellipsoid. The focal mechanism and energy calculation unit is used to calculate the intensity and rupture mechanism parameters of microseismic events. It uses a grid search method to invert the focal mechanism solution based on the first motion polarity of multi-channel P-waves. After instrument response correction and geometric diffusion correction, it uses the Brune model to fit and calculate the seismic moment and moment magnitude. It calculates the radiated energy by integrating the displacement spectrum and calculates the energy-to-moment ratio.

7. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 6, characterized in that, In the anisotropic medium positioning unit, the weight of the objective function is determined based on the confidence level of each channel output by the multi-channel first arrival precision picking unit, and the channel with higher confidence level has a larger weight; the positioning information ellipsoid is used to characterize the positioning uncertainty.

8. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 1, characterized in that, The intelligent early warning module for rupture risk includes: The fracture energy release rate analysis unit is used to identify the unsteady acceleration characteristics of rock mass fracture from the energy dimension. It constructs the energy time series of microseismic event sequences, smooths them into continuous energy release rate curves using kernel density estimation, calculates the energy acceleration index by fitting an exponential function through a sliding window, identifies the acceleration release start point, and outputs the energy release rate value and energy acceleration index. The fractal dimension analysis unit for fracture volume is used to identify the cluster evolution characteristics of microseismic events from a spatial dimension. It uses a sliding window to extract a subset of microseismic events, uses the box counting method to calculate the spatial fractal dimension, tracks the trend of fractal dimension change, and determines the fracture cluster nucleation state when the fractal dimension continuously decreases from a high value to below a set threshold. It outputs the current fractal dimension value and the trend of change. The focal mechanism consistency analysis unit is used to identify the unified evolution characteristics of the focal mechanism of microseismic events from the mechanism dimension. It constructs a focal mechanism consistency index for microseismic events within a sliding window, and quantifies the degree of consistency by calculating the root mean square value of the spatial angle between the focal mechanism of each event and the average focal mechanism within the window. When the consistency index decreases to below a set threshold, it is determined to be in a state of rupture surface penetration, and outputs the consistency index value and the orientation of the dominant rupture surface. The multi-indicator fusion early warning output unit is used to fuse and infer the outputs of the aforementioned three analysis units and output graded early warning information. It adopts a Bayesian network to construct a three-layer inference structure, with energy acceleration index, fractal dimension and its changing trend, and source mechanism consistency index and its changing trend as observation nodes, energy acceleration release state, fracture cluster nucleation state, and fracture surface penetration state as hidden state nodes, and rock mass fracture risk level as target node. It calculates the posterior probability of risk level through Bayesian inference, outputs level three, level two, and level one early warning information, and pushes it to the monitoring terminal and mobile terminal in a multimodal manner.

9. The microseismic monitoring and intelligent early warning system for the risk of rock mass fracturing in tunnel dome as described in claim 8, characterized in that, In the multi-indicator fusion early warning output unit, the third-level early warning corresponds to the activation of only one of the three hidden states, indicating that attention should be increased; the second-level early warning corresponds to the activation of any two of the three hidden states, indicating that high-risk operations should be stopped and reinforcement should be prepared. A Level 1 warning corresponds to the activation of all three hidden states, instructing the immediate evacuation of personnel and the activation of the emergency response plan.

Images