Tunnel rockburst type prediction method and system based on multi-dimensional mechanism fusion and medium

Abstract

The invention discloses a tunnel rockburst type prediction method and system based on multi-dimensional mechanism fusion and a medium. Relates to the technical field of tunnel rockburst. Dynamically acquiring multi-dimensional basic information in a tunnel construction process; inverting first proportions of different fracture modes of the micro-seismic event according to the micro-seismic information fusion moment tensor, and performing energy evolution on the acoustic emission information to obtain second proportions of different fracture modes of each acoustic emission event; the dynamic failure weights of different fracture modes are comprehensively obtained; spatial clustering analysis is carried out based on the dynamic damage weight and the multi-dimensional basic information, rockburst type probability models of different rockburst types are constructed, and the probabilities of different rockburst types are predicted; according to the scheme, the rockburst type probability model is constructed in combination with real-time multi-dimensional basic information, dynamic judgment and prediction of rockburst types are achieved, and technical guarantee is provided for deep tunnel construction safety.

Description

Technical Field

[0001] The present invention relates to the technical field of tunnel rock burst, and in particular to a tunnel rock burst type prediction method, system and medium based on multi-dimensional mechanism fusion. Background Art

[0002] Rockbursts are complex and highly destructive dynamic hazards caused by the sudden release of energy from highly stressed rock masses in deep underground projects. Existing rockburst early warning technologies often rely on single microseismic monitoring data, making it difficult to accurately distinguish between strain-type and tectonic rockburst types, leading to a lack of targeted protection measures.

[0003] In addition, traditional methods have the following limitations: (1) The data dimension is single, relying only on the statistics of microseismic events, ignoring key information such as acoustic emission spectrum and structural surface geometry, and unable to fully analyze the rupture mechanism; (2) Static analysis is delayed, and the statistics of damage proportion based on a fixed time window make it difficult to capture the spatiotemporal evolution of rockbursts; (3) The model has poor universality: the existing prediction model is not adaptable enough to complex geological conditions, lacks physical mechanism support, and has weak explanatory power; (4) Passive warning mode: it is impossible to link construction equipment to achieve active prevention and control, and the risk response efficiency is low.

[0004] In recent years, although multi-source data fusion and machine learning methods have been introduced into the field of rockburst prediction, it still faces challenges such as insufficient multi-sensor coordination, insufficient dynamic feature extraction, and lack of physical constraints. Summary of the Invention

[0005] The technical problem to be solved by the present invention is that: traditionally, multi-source data fusion and machine learning methods have been introduced into the field of rockburst prediction, but still face challenges such as insufficient multi-sensor coordination, insufficient dynamic feature extraction, and lack of physical constraints, resulting in defects in the accuracy of rockburst type prediction; the purpose of the present invention is to provide a tunnel rockburst type prediction method, system and medium based on multi-dimensional mechanism fusion, to improve the method on the basis of traditional rockburst prediction technology, to collect multi-dimensional basic information through the integration of multi-source data monitoring system network such as microseismic, acoustic emission, geological radar, etc., to construct a rockburst type probability model, to realize dynamic discrimination and active prevention and control of rockburst type, and to provide technical guarantee for the safety of deep tunnel construction.

[0006] The present invention is achieved through the following technical solutions: This solution provides a tunnel rockburst type prediction method based on multi-dimensional mechanism fusion, including: A multi-source data monitoring system is established in the tunnel construction disturbance zone to dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information, and geological information of the tunnel face area; The first proportions of different rupture modes of microseismic events are inverted by fusion of microseismic information and moment tensor, and the second proportions of different rupture modes of each acoustic emission event are obtained by energy evolution. The dynamic damage weights of different rupture modes are generated by combining the first and second proportions. Based on the spatial cluster analysis of dynamic damage weights and multi-dimensional basic information, a comprehensive judgment model for different rockburst types is obtained; A rockburst type probability model is constructed based on the comprehensive judgment model. The multidimensional basic information of the microseismic event to be predicted is input into the rockburst type probability model to calculate the probabilities of different rockburst types.

[0007] A further optimization scheme is to invert the first proportion of different rupture modes of microseismic events by fusing the moment tensor with microseismic information; including the following method: Obtain the source of the microseismic event and use the moment tensor analysis method to make an equivalent force approximation inference of the source, and obtain the equivalent force matrix containing the moment tensor; The moment tensor in the equivalent force matrix is ​​considered as the interaction of pure double couple, compensating linear vector dipole and isotropic component to establish the correlation relationship R between the moment tensor and the rupture mode; The rupture mode of each microseismic event is determined based on the correlation relationship R and the double couple components of the moment tensor; The proportions of different rupture modes are calculated as the first proportion.

[0008] A further optimization solution is that the correlation relationship R between the moment tensor and the fracture mode includes: ; m i *= M i - tr ( M ) / 3; in, tr ( M ) represents the moment tensor M traces; m i * represents the eigenvalue of the partial tensor; M i Represents the moment tensor M No. i eigenvalues.

[0009] A further optimization scheme is to perform energy evolution on the acoustic emission information to obtain the second proportion of different rupture modes of each acoustic emission event; including the following method: First, the acoustic emission information is subjected to band-pass filtering, and then the acoustic emission information after band-pass filtering is subjected to short-time Fourier transform to obtain the acoustic emission time spectrum; The main frequency component of the acoustic emission spectrum is extracted: different identification frequency peak energy intervals are configured for different rupture modes, and the proportion of identification frequency peaks in different identification frequency peak energy intervals in the acoustic emission spectrum is counted as the second proportion.

[0010] A further optimization scheme is to generate dynamic damage weights of different fracture modes by integrating the first proportion and the second proportion; including the following method: The microseismic information weight is determined based on the inversion confidence in the first ratio inversion process, and the acoustic emission information weight is determined based on the deviation threshold of the energy ratio of the identified frequency peak in the second ratio evolution process; the microseismic information weight and the acoustic emission information weight are combined to obtain the comprehensive weight W 综合 : ; Among them, W MS represents the weight of microseismic information; N MS represents the total number of microseismic events; W AE Represents the weight of acoustic emission information; N AE represents the total number of acoustic emission events; Set a statistical period, calculate the proportion of different rupture modes based on the comprehensive weight within the statistical period, and dynamically calculate the dynamic damage weights of different rupture modes according to the time window.

[0011] A further optimization scheme is to perform spatial cluster analysis based on dynamic damage weights and multi-dimensional basic information to obtain a comprehensive judgment model for different rockburst types; including the following methods: Each microseismic event is considered a microseismic cluster, and the spatial aggregation area of ​​each microseismic cluster is identified. The positional relationship between the microseismic cluster and the structural surface and fault is used as the spatial judgment condition for different rockburst types. The proportion of different rupture modes is determined based on the dynamic damage weight, which serves as the rupture mode judgment condition for different rockburst types. The ratio of rock mass strength to ground stress or the angle between the maximum principal stress direction and the structural plane strike is used as the mechanical judgment condition for different rockburst types.

[0012] A further optimization scheme is to construct a rockburst type probability model based on the comprehensive judgment model; including the following methods: A graph neural network is constructed with each microseismic cluster as a graph node. The edge weights of the graph neural network are determined by the spatial distance between microseismic clusters and whether the microseismic clusters are connected to the structural surface. The probability distribution of the output layer of the graph neural network is: ; Among them, P(k) represents the probability of rockburst type k; e represents the natural base; zk represents the logit value of rockburst type k; j represents the index of rockburst type (from 1 to 3); K represents the total number of rockburst types (3 types: structural type, strain type, and mixed type); zj represents the logit value of index j.

[0013] A further optimization solution is that the rock burst types include: structural rock burst, strain rock burst and mixed rock burst; the rupture modes include: tension rupture, shear rupture and compression rupture.

[0014] This solution also provides a tunnel rockburst type prediction system based on multi-dimensional mechanism fusion, which is used to implement the above-mentioned tunnel rockburst type prediction method based on multi-dimensional mechanism fusion. The system includes: An acquisition module is used to establish a multi-source data monitoring system in the tunnel construction disturbance zone to dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information, and geological information of the tunnel face area; The inversion evolution module is used to invert the first proportion of different rupture modes of microseismic events based on the fusion of microseismic information and moment tensors, and to perform energy evolution on acoustic emission information to obtain the second proportion of different rupture modes of each acoustic emission event. The first and second proportions are combined to generate dynamic damage weights for different rupture modes. The analysis module is used to perform spatial cluster analysis based on dynamic damage weights and multi-dimensional basic information to obtain a comprehensive judgment model for different rockburst types; The calculation module is used to construct a rockburst type probability model based on the comprehensive judgment model, input the multi-dimensional basic information of the microseismic event to be predicted into the rockburst type probability model to calculate the probability of different rockburst types.

[0015] The present solution also provides a computer-readable medium having a computer program stored thereon, and the computer program is executed by a processor to implement the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion as described above.

[0016] Compared with the prior art, the present invention has the following advantages and beneficial effects: 1. The tunnel rockburst type prediction method, system, and medium provided by the present invention, based on multi-dimensional mechanism fusion, improve upon traditional rockburst prediction technology. By integrating a multi-source data monitoring system network, such as microseismic, acoustic emission, and geological radar, to collect multi-dimensional basic information, a rockburst type probability model is constructed based on real-time multi-dimensional basic information. This allows for dynamic identification and prediction of rockburst types, providing technical support for the safety of deep tunnel construction. 2. The present invention provides a tunnel rockburst type prediction method, system, and medium based on multi-dimensional mechanism fusion. This method extracts the proportion of rupture modes through moment tensor inversion (microseismic) and energy evolution analysis (acoustic emission), unifying data from different sensors into the "rupture mechanism" dimension to avoid the bias of simple numerical superposition. 3. The present invention provides a tunnel rockburst type prediction method, system, and medium based on multi-dimensional mechanism fusion. Based on microseismic moment tensor inversion and acoustic emission energy evolution, the dynamic weights of different rupture modes are calculated in real time (e.g., shear rupture accounts for 60% → high rockburst risk). The weights are updated as the construction progresses to reflect the cumulative effect of rock damage. Geological parameters such as rock strength participate in spatial clustering, and spatial cluster analysis is performed based on microseismic event locations and dynamic weights to avoid misjudgment of isolated points and ensure that the prediction results are consistent with the geological conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the following briefly introduces the drawings required for use in the examples. It should be understood that the following drawings only illustrate certain embodiments of the present invention and should not be considered as limiting the scope. A person of ordinary skill in the art can also derive other relevant drawings based on these drawings without inventive effort. In the drawings: Figure 1 Schematic diagram of the flow chart of the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion. DETAILED DESCRIPTION

[0018] 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 in conjunction with examples and drawings. The exemplary embodiments of the present invention and their descriptions are only used to explain the present invention and are not intended to limit the present invention.

[0019] Traditionally, multi-source data fusion and machine learning methods have been introduced into the field of rockburst prediction. However, challenges such as insufficient multi-sensor coordination, inadequate dynamic feature extraction, and lack of physical constraints still lead to deficiencies in rockburst type prediction accuracy. In view of this, this solution provides the following embodiments to address the aforementioned technical issues: Example 1: This example provides a tunnel rockburst type prediction method based on multi-dimensional mechanism fusion, such as Figure 1 As shown, including: Step 1: Establish a multi-source data monitoring system in the tunnel construction disturbance area to dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information, and geological information of the tunnel face area; The multi-source data monitoring system integrates a network of microseismic sensors, acoustic emission sensors, and geological sensors (3D laser scanners) to collect multi-dimensional basic information in real time during tunnel construction. Specifically, the microseismic sensor array consists of at least six triaxial microseismic sensors (covering horizontal, vertical, and vertical positions) evenly distributed along the tunnel cross-section within 30-50 meters behind the tunnel face, ensuring three-dimensional capture of microseismic events across the entire cross-section. Acoustic emission sensors are deployed in areas of potential fracture (such as structural plane intersections and areas of high stress concentration), with four to six sensors spaced every 10 meters. A 3D laser scanner uses a track-mounted geological radar installed behind the tunnel face to scan the geological structure within a 20-meter radius as excavation progresses. With every 20-meter advance, sensors positioned 30 meters ahead of the tunnel face are moved forward in batches to maintain coverage of the construction-disturbed area. A redundant fiber-optic + 5G network is used to transmit microseismic, acoustic emission, and 3D laser scanning information in real time.

[0020] Furthermore, the geological information of the tunnel face area includes fault distribution information, structural surface geometric characteristics, and rock mass mechanical parameters, as shown in Table 1: Table 1 Geological information of the tunnel face

[0021] Step 2: Based on the microseismic information fused with the moment tensor inversion, the first proportion of different rupture modes of microseismic events is obtained, and the energy evolution of the acoustic emission information is performed to obtain the second proportion of different rupture modes of each acoustic emission event; the dynamic damage weights of different rupture modes are generated by combining the first and second proportions; the rupture modes include: tension rupture, shear rupture, and compression rupture.

[0022] The method of inverting the first proportion of different rupture modes of microseismic events based on microseismic information fusion moment tensor inversion includes: The source of the microseismic event is obtained, and the equivalent force approximation of the source is inferred using the moment tensor analysis method to obtain the equivalent force matrix containing the moment tensor; the equivalent force matrix is ​​expressed as: u=GM Where, u The vector representing the far-field displacement, M is the moment tensor, G is the Green function. If the earthquake source is equivalent to a point source in time, the generation of micro-fracture can be regarded as the moment tensor acting on time. t Afterwards, x The displacement generated at . Then the equivalent force matrix is ​​expressed as: ; in, x k Represents the components of the spatial coordinates; x 0 represents the initial location of the earthquake source;t 0 represents the time when the earthquake source event occurred; M ij represents the components of the moment tensor; G ij represents the component of Green's function; u i (x, t) represents the earthquake source as a point source in time, and the generation of micro-fracture is regarded as the moment tensor acting on time. t Afterwards, x The equivalent matrix of the displacement generated at ; In actual field conditions, the generation of micro-fractures is a continuous process, and the generation of displacement is a function change on a time scale. p When the wave is displaced in the far field, we have: ; In the formula: u i Is relative to the i The first movement of the P wave along the axis, r i Represents the radial component from the source to the sensor; r p It represents the lateral component from the source to the sensor; r q It represents the longitudinal component from the source to the sensor; x 0 represents the initial location of the earthquake source; R is the focal distance, ρ is the rock density.

[0023] Expanded to ; in, Indicates the source n The decomposition modes correspond to the Green's function of the b source component under the q field component; q=1,2,3;b=1,2,3;n=1,2,...,N; The moment tensor in the equivalent force matrix is ​​considered as the interaction of pure double couple, compensating linear vector dipole and isotropic component to establish the correlation relationship R between the moment tensor and the rupture mode; ; M DC represents a pure double couple; M CLVD represents the compensated linear vector dipole; M ISO represents the isotropic component; X represents the coefficient of the pure double couple; Y represents the coefficient of the compensated linear vector dipole; Z represents the coefficient of the isotropic component; The correlation relationship R between the moment tensor and the fracture mode includes: ; m i *= M i - tr ( M ) / 3; in, tr ( M ) represents the moment tensor M traces; m i * represents the eigenvalue of the partial tensor; M i Represents the moment tensor M No. i eigenvalues.

[0024] The rupture mode of each microseismic event is determined based on the correlation relationship R and the double couple components of the moment tensor; Results for the correlation R and the double couple components of the moment tensor C DC (representing the shear fracture of the rock mass or the relative dislocation mechanism of the fault) will produce different criteria, and the corresponding failure mode judgments are shown in Table 2: Table 2 Destruction mode judgment Mechanical state <![CDATA[ C DC Value]]> value Fracture mode Pure tension 0% -100% Tensile rupture Shear-Tension 40% -43% Tensile rupture Pure shear 60% -30% Shear fracture Shear-Compression 100% 0% Shear fracture Shear-Compression 60% 30% Compression fracture Shear-Compression 40% 43% Compression fracture Pure compression 0% 100% Compression fracture

[0025] The proportions of different rupture modes are calculated as the first proportion.

[0026] The energy evolution of the acoustic emission information is performed to obtain a second proportion of different rupture modes of each acoustic emission event; including the following method: First, the acoustic emission information is band-pass filtered (20kHz-1MHz) to remove low-frequency noise, and then the acoustic emission information after band-pass filtering is short-time Fourier transformed to obtain the acoustic emission time spectrum; The main frequency component of the acoustic emission spectrum is extracted: different identification frequency peak energy intervals are configured for different rupture modes, and the proportion of identification frequency peaks in different identification frequency peak energy intervals in the acoustic emission spectrum is counted as the second proportion.

[0027] For example, calculate the spectrum energy interval and identify the main frequency peak (such as high frequency peak > 100kHz, low frequency peak < 50kHz).

[0028] Characteristics of tensile rupture include: a high proportion of high-frequency spectrum energy (>70%), corresponding to brittle crack growth; Characteristics of shear fracture include a high proportion of low-frequency spectrum energy (>60%), corresponding to frictional slip.

[0029] The proportion of high-frequency spectrum energy is: ; The proportion of low spectrum energy is: ; If α>0.7, it is marked as tension-dominated; if β>0.6, it is marked as shear-dominated.

[0030] The method of generating dynamic damage weights of different fracture modes by combining the first proportion and the second proportion includes: For microseismic events, when the microseismic sensor detects a waveform amplitude exceeding a threshold, an event recording is triggered. The moment tensor is used to calculate the rupture type (tension, shear, compression) in real time. Each microseismic event is output with a failure mode label and confidence level (e.g., a shear ratio >60% is labeled "shear"). For acoustic emission events, a fast Fourier transform (FFT) is used to extract the dominant frequency energy ratio (high frequency / low frequency). A high-frequency energy ratio >70% is labeled as tension; a low-frequency energy ratio >60% is labeled as shear; and intermediate states are labeled as mixed (compression or tension-shear combination).

[0031] The microseismic information weight is determined based on the inversion confidence in the first ratio inversion process (e.g., when the inversion confidence is >60%, the microseismic information weight is 0.8, which is set based on experience in actual application). The acoustic emission information weight is determined based on the deviation threshold of the energy ratio of the identified frequency peak in the second ratio evolution process (e.g., when α is >0.7, the weight is 0.7, which is set based on experience in actual application). The microseismic information weight and the acoustic emission information weight are combined to obtain the comprehensive weight W. 综合 : ; Among them, W MS represents the weight of microseismic information; N MS represents the total number of microseismic events; W AE Represents the weight of acoustic emission information; N AE represents the total number of acoustic emission events; Set a statistical period, calculate the proportion of different rupture modes based on the comprehensive weight within the statistical period, and dynamically calculate the dynamic damage weights of different rupture modes according to the time window.

[0032] The proportions of different rupture modes are:

[0033]

[0034]

[0035] Among them, P 张拉 represents the proportion of tensile rupture mode; P 剪切 represents the proportion of shear fracture mode; P 压缩Indicates the proportion of compression fracture mode; N 张拉 represents the total number of events in the tensile rupture mode in a rockburst; N 剪切 represents the total number of shear failure mode events in a rockburst; N 压缩 represents the total number of compression failure mode events in a rockburst; N 总 represents the total of all rupture mode events; Mapping the proportion to a weight coefficient, the dynamic destruction weight is:

[0036]

[0037]

[0038] Among them, W 张拉 W represents the dynamic failure weight of the tensile rupture mode; 剪切 W represents the dynamic failure weight of the shear fracture mode; 压缩 Represents the dynamic failure weight of the compression fracture mode; α, β, and γ are the tension experience coefficient, shear experience coefficient, and compression experience coefficient, respectively (e.g., α=1.2 in strain-type rock burst and β=1.5 in tectonic rock burst).

[0039] Step 3: Perform spatial cluster analysis based on dynamic damage weights and multidimensional basic information to obtain a comprehensive judgment model for different rockburst types; the rockburst types include: structural rockburst, strain rockburst, and mixed rockburst; this step specifically includes the following methods: Each microseismic event is considered a microseismic cluster, and the spatial aggregation area of ​​each microseismic cluster is identified. The positional relationship between the microseismic cluster and the structural surface and fault is used as the spatial judgment condition for different rockburst types. Specifically, for each microseismic cluster, the distance D between its geometric center and the nearest structural surface is calculated. If D ≤ 2 m and the cluster density is > 5 events / m³, it is marked as a “structural surface related cluster”.

[0040] Structural surfaces with high roughness (roughness JRC>12) have large friction resistance and are prone to accumulating elastic strain energy, triggering microseismic events of shear fracture; Establish a linear regression model of shear fracture proportion: ;in represents the shear fracture ratio; shear fracture ratio; a represents the first coefficient, b represents the second coefficient; Structural surfaces with steep inclinations (v > 60°) are prone to shear slip, triggering shear failure. Structural surfaces with gentle inclinations (v < 30°) are primarily tensile, triggering tensile failure. The type of rockburst is determined based on roughness and inclination. If the inclination v is > 60° and the shear failure ratio is > 60%, it is considered a tectonic rockburst.

[0041] The density of microseismic clusters in the fault zone is significantly higher than that in the surrounding rock area (density difference > 2 times); shear failure accounts for a high proportion (> 70%) near the strike-slip fault; and tensile failure accounts for a high proportion (> 50%) in the normal fault area. The proportion of different rupture modes is determined based on the dynamic damage weight, which serves as the rupture mode judgment condition for different rockburst types. The ratio of rock mass strength to ground stress SSR or the angle between the maximum principal stress direction σ1 and the structural plane strike is used as the mechanical judgment condition for different rockburst types.

[0042] Specifically, areas with a low SSR (Structural Stress Ratio) of rock mass strength to in-situ stress (SSR < 2) are prone to brittle failure (triggering strain-induced rockbursts). Areas with a high SSR (SSR > 4) experience slow energy accumulation, with shear failure controlled by structural planes predominating (tectonic rockbursts). Regarding in-situ stress direction and microseismic distribution, if the angle between the maximum principal stress direction (σ1) and the strike of the structural plane is less than 30°, the structural plane is likely to be activated, triggering a tectonic rockburst.

[0043] The comprehensive determination results of rockburst type mechanism are shown in Table 3: Table 3 Comprehensive determination of rockburst type and mechanism Rockburst type Space conditions Failure Mode Mechanical conditions Tectonic rock burst Microseismic clusters are distributed along structural surfaces or faults (distance ≤ 2m) Shear ratio>60% <![CDATA[SSR>4 or σ1 direction is nearly parallel to the structural surface]]> strain-type rockburst Microseismic clusters are dispersed and far away from the structural surface Tension ratio>50% <![CDATA[The SSR < 2 or the σ1 direction intersects the strike of the structural plane at a large angle.]]> Mixed rock burst Some microseismic clusters are close to structural surfaces, while others are located in intact rock mass areas. Tension, shear, and compression failures account for >30% \

[0044] Step 4: Construct a rockburst type probability model based on the comprehensive judgment model, input the multi-dimensional basic information of the microseismic event to be predicted into the rockburst type probability model to calculate the probabilities of different rockburst types.

[0045] The method of constructing a rockburst type probability model based on a comprehensive judgment model includes: The rockburst type probability model of this scheme is a physical information graph neural network hybrid driven model, which is obtained by fusing the graph neural network with physical constraints (such as elastic mechanics equations and energy conservation). Its input three-dimensional features are the microseismic event rupture type (tension / shear / compression), energy release rate (dE / dt), and acoustic emission spectrum energy ratio; its edge features are the spatial distance of the microseismic event, structural surface roughness (JRC), and rock mass strength (UCS).

[0046] A graph neural network is constructed with each microseismic cluster as a graph node; the graph node features include failure mode, energy, and spectrum parameters; The edge weights of the graph neural network are determined by the spatial distance between microseismic clusters and whether the microseismic clusters are connected to the structural surface (for example, the closer the spatial distance and the stronger the correlation between the microseismic clusters and the structural surface, the higher the edge weight). The probability distribution of the output layer of the graph neural network is:

[0047] Where P(k) represents the probability of rockburst type k; e represents the natural base; zk represents the logit value of rockburst type k; j represents the index of rockburst type (from 1 to 3); K represents the total number of rockburst types (three types: structural, strain, and mixed).

[0048] Follow the above steps to rebuild the graph data and run the model every 5 minutes to ensure that the prediction results are updated in real time as the construction progresses.

[0049] Finally, based on the calculated probabilities of different rockburst types, construction equipment is linked to achieve active risk prevention and control, and the probabilities of different rockburst types are converted into prevention and control instructions, linking construction equipment to reduce risks.

[0050] Example 2: This example provides a tunnel rockburst type prediction system based on multi-dimensional mechanism fusion, which is used to implement the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion described in Example 1. The system includes: An acquisition module is used to establish a multi-source data monitoring system in the tunnel construction disturbance zone to dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information, and geological information of the tunnel face area; The inversion evolution module is used to invert the first proportion of different rupture modes of microseismic events based on the fusion of microseismic information and moment tensors, and to perform energy evolution on acoustic emission information to obtain the second proportion of different rupture modes of each acoustic emission event. The first and second proportions are combined to generate dynamic damage weights for different rupture modes. The analysis module is used to perform spatial cluster analysis based on dynamic damage weights and multi-dimensional basic information to obtain a comprehensive judgment model for different rockburst types; The calculation module is used to construct a rockburst type probability model based on the comprehensive judgment model, input the multi-dimensional basic information of the microseismic event to be predicted into the rockburst type probability model to calculate the probability of different rockburst types.

[0051] Example 3: This example provides a computer-readable medium having a computer program stored thereon. The computer program is executed by a processor to implement the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion as described in Example 1. Specifically, the following steps are performed: Step 1: Dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information and geological information of the tunnel face area; Step 2: The first proportions of different rupture modes of microseismic events are inverted by fusing the moment tensor with the microseismic information. The second proportions of different rupture modes of each AE event are obtained by energy evolution of the acoustic emission information. The dynamic damage weights of different rupture modes are generated by combining the first and second proportions. Step 3: Perform spatial cluster analysis based on dynamic damage weights and multi-dimensional basic information to obtain a comprehensive judgment model for different rockburst types; Step 4: Construct a rockburst type probability model based on the comprehensive judgment model, input the multi-dimensional basic information of the microseismic event to be predicted into the rockburst type probability model to calculate the probabilities of different rockburst types.

[0052] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A tunnel rockburst type prediction method based on multi-dimensional mechanism fusion, characterized by: include: Build a multi-source data monitoring system in the tunnel construction disturbance area to dynamically collect multi-dimensional basic information during the tunnel construction process; The multi-dimensional basic information includes: microseismic information, acoustic emission information, three-dimensional laser scanning information and geological information of the tunnel face area; The first proportions of different rupture modes of microseismic events are inverted by fusion of microseismic information and moment tensor, and the second proportions of different rupture modes of each acoustic emission event are obtained by energy evolution. The dynamic damage weights of different rupture modes are generated by combining the first and second proportions. Based on the spatial cluster analysis of dynamic damage weights and multi-dimensional basic information, a comprehensive judgment model for different rockburst types is obtained; A rockburst type probability model is constructed based on the comprehensive judgment model. The multidimensional basic information of the microseismic event to be predicted is input into the rockburst type probability model to calculate the probabilities of different rockburst types.

2. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 1 is characterized in that: The first proportion of different rupture modes of microseismic events is inverted by fusing the moment tensor with the microseismic information; Includes methods: Obtain the source of the microseismic event and use the moment tensor analysis method to make an equivalent force approximation inference of the source, and obtain the equivalent force matrix containing the moment tensor; The moment tensor in the equivalent force matrix is ​​considered as the interaction of pure double couple, compensating linear vector dipole and isotropic component to establish the correlation relationship R between the moment tensor and the rupture mode; The rupture mode of each microseismic event is determined based on the correlation relationship R and the double couple components of the moment tensor; The proportions of different rupture modes are calculated as the first proportion.

3. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 2 is characterized in that: The correlation relationship R between the moment tensor and the fracture mode includes: ； m i *= M i - tr ( M ) / 3； in, tr ( M ) represents the moment tensor M traces; m i * represents the eigenvalue of the partial tensor; M i Represents the moment tensor M No. i eigenvalues.

4. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 1 is characterized in that: The energy evolution of the acoustic emission information is performed to obtain a second proportion of different rupture modes of each acoustic emission event; including the following method: First, the acoustic emission information is subjected to band-pass filtering, and then the acoustic emission information after band-pass filtering is subjected to short-time Fourier transform to obtain the acoustic emission time spectrum; The main frequency component of the acoustic emission spectrum is extracted: different identification frequency peak energy intervals are configured for different rupture modes, and the proportion of identification frequency peaks in different identification frequency peak energy intervals in the acoustic emission spectrum is counted as the second proportion.

5. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 4 is characterized in that: The first proportion and the second proportion are combined to generate dynamic damage weights of different rupture modes; Includes methods: The microseismic information weight is determined based on the inversion confidence in the first ratio inversion process, and the acoustic emission information weight is determined based on the deviation threshold of the energy ratio of the identified frequency peak in the second ratio evolution process; the microseismic information weight and the acoustic emission information weight are combined to obtain the comprehensive weight W 综合 : ; Among them, W MS represents the weight of microseismic information; N MS represents the total number of microseismic events; W AE Represents the weight of acoustic emission information; N AE represents the total number of acoustic emission events; Set a statistical period, calculate the proportion of different rupture modes based on the comprehensive weight within the statistical period, and dynamically calculate the dynamic damage weights of different rupture modes according to the time window.

6. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 1 is characterized in that: The spatial cluster analysis based on dynamic damage weight and multi-dimensional basic information is performed to obtain a comprehensive judgment model for different rock burst types; Includes methods: Each microseismic event is considered a microseismic cluster, and the spatial aggregation area of ​​each microseismic cluster is identified. The positional relationship between the microseismic cluster and the structural surface and fault is used as the spatial judgment condition for different rockburst types. The proportion of different rupture modes is determined based on the dynamic damage weight, which serves as the rupture mode judgment condition for different rockburst types. The ratio of rock mass strength to ground stress or the angle between the maximum principal stress direction and the structural plane strike is used as the mechanical judgment condition for different rockburst types.

7. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 6 is characterized in that: The method of constructing a rockburst type probability model based on a comprehensive judgment model includes: A graph neural network is constructed with each microseismic cluster as a graph node. The edge weights of the graph neural network are determined by the spatial distance between microseismic clusters and whether the microseismic clusters are connected to the structural surface. The probability distribution of the output layer of the graph neural network is: ； Among them, P(k) represents the probability of rockburst type k; e represents the natural base; zk represents the logit value of rockburst type k; j represents the index of rockburst type; K represents the total number of rockburst types; zj represents the logit value of index j.

8. The tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to claim 1 is characterized in that: The rock burst types include: structural rock burst, strain rock burst and mixed rock burst; the rupture modes include: tension rupture, shear rupture and compression rupture.

9. The tunnel rockburst type prediction system based on multi-dimensional mechanism fusion is characterized by: A system for implementing the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion according to any one of claims 1 to 8, comprising: An acquisition module is used to establish a multi-source data monitoring system in the tunnel construction disturbance zone to dynamically collect multi-dimensional basic information during the tunnel construction process; the multi-dimensional basic information includes: microseismic information, acoustic emission information, 3D laser scanning information, and geological information of the tunnel face area; The inversion evolution module is used to invert the first proportion of different rupture modes of microseismic events based on the fusion of microseismic information and moment tensors, and to perform energy evolution on acoustic emission information to obtain the second proportion of different rupture modes of each acoustic emission event. The first and second proportions are combined to generate dynamic damage weights for different rupture modes. The analysis module is used to perform spatial cluster analysis based on dynamic damage weights and multi-dimensional basic information to obtain a comprehensive judgment model for different rockburst types; The calculation module is used to construct a rockburst type probability model based on the comprehensive judgment model, input the multi-dimensional basic information of the microseismic event to be predicted into the rockburst type probability model to calculate the probability of different rockburst types.

10. A computer-readable medium having a computer program stored thereon, characterized in that: The computer program is executed by a processor to implement the tunnel rockburst type prediction method based on multi-dimensional mechanism fusion as described in any one of claims 1 to 8.

Images