Hierarchical optimization positioning method and device for rockburst micro-seismic source

Abstract

The invention provides a rockburst micro-seismic source hierarchical optimization positioning method and device, and the method comprises the steps: building a mixed integer nonlinear programming MINLP model containing a wave velocity gradient field variable, carrying out the coarse positioning of a seismic source through the MINLP model, and determining the coordinates of the seismic source and the initial value of a wave velocity field; constructing a layered optimization framework, and optimizing the coarse coordinates of the seismic source and the initial value of the wave velocity field by using the layered optimization framework to obtain the optimized seismic source coordinates and the optimized wave velocity field; virtual nodes are inserted into the MINLP model by using the optimized seismic source coordinates and the actual sensor positions, a virtual sensor network is constructed, and virtual node coordinates are determined; and according to the optimized wave velocity field and the virtual node coordinates, performing posterior probability inference on the optimized seismic source coordinates through Bayesian MCMC sampling, and determining a rock burst event three-dimensional positioning result. By means of the scheme, the positioning precision of the heterogeneous rock mass can be improved, and therefore the adaptability and timeliness of engineering scene application are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of geotechnical engineering safety monitoring, and in particular to a method and device for optimizing the layered positioning of rockburst microseismic sources. Background Art

[0002] Rockbursts are dynamic hazards caused by the sudden rupture of highly stressed rock masses in underground engineering projects. The precise location of their earthquake source directly impacts the timeliness of disaster warnings and the effectiveness of prevention and control measures. Microseismic source location technology analyzes elastic wave signals received by sensor arrays to invert the three-dimensional coordinates of the rupture point, providing critical data support for surrounding rock stability assessment, support design optimization, and disaster source tracing.

[0003] Currently, conventional methods for locating rockburst microseismic events include the Geiger method, the simplex method, and intelligent optimization algorithms (such as the particle swarm optimization (PSO) and the mayfly algorithm (MA)). Furthermore, the industry has proposed several specific solutions, including a joint inversion method for the 3D rock velocity structure and microseismic source location that considers lateral inhomogeneity; a comprehensive 3D microseismic source location method for unknown wave velocity systems based on analytical solutions and the log-logistic probability density function; a particle swarm optimization-based layered rock microseismic source location method; and an automatic search algorithm for the shortest source-sensor path in the presence of voids. However, these methods all have limitations to varying degrees. For example, the particle swarm optimization algorithm is prone to premature convergence, and the mayfly algorithm suffers from low computational efficiency in high-dimensional spaces. Furthermore, the condition number of the nonlinear equations for microseismic location is as high as 106, making traditional iterative methods unstable. A single location using the grid search method takes >300 seconds, making it difficult to meet the requirements of real-time engineering monitoring. Therefore, existing methods cannot well meet the requirements of high-precision positioning and support for inhomogeneous wave velocity simple modes. Summary of the Invention

[0004] The present invention provides a method and device for optimizing the stratified positioning of rockburst microseismic sources, so as to solve the problem of inaccurate microseismic source positioning in the prior art and meet the demand for real-time monitoring in a heterogeneous rock mass environment.

[0005] To this end, the present invention provides the following technical solutions:

[0006] The present invention provides a method for optimizing the location of rockburst microseismic sources by stratification, the method comprising:

[0007] Establishing a mixed integer nonlinear programming (MINLP) model including velocity gradient field variables, using the MINLP model to perform coarse earthquake source positioning and determine the earthquake source coordinates and initial values ​​of the velocity field;

[0008] Constructing a hierarchical optimization framework, and optimizing the initial values ​​of the earthquake source rough coordinates and the wave velocity field using the hierarchical optimization framework to obtain optimized earthquake source coordinates and wave velocity field;

[0009] Inserting virtual nodes into the MINLP model using the optimized source coordinates and actual sensor locations to construct a virtual sensor network and determine the virtual node coordinates;

[0010] The optimized source coordinates are subjected to posterior probability inference through Bayesian MCMC sampling according to the optimized wave velocity field and the virtual node coordinates to determine the three-dimensional positioning result of the rock burst event.

[0011] Optionally, the MINLP model is an objective function that takes minimization of arrival residuals under the wave velocity gradient field constraint as the goal, and introduces mixed integer variables to characterize wave velocity field discretization units.

[0012] Optionally, establishing a MINLP model including velocity gradient field variables includes:

[0013] Obtain raw microseismic signals and actual sensor positions;

[0014] Filtering the original microseismic signal and automatically picking up the P-wave arrival time;

[0015] Based on the picked-up P-wave arrival time and the actual sensor position, an objective function including a wave velocity gradient field variable is established.

[0016] Optionally, the using the MINLP model to perform coarse earthquake source positioning and determine the initial values ​​of the earthquake source coordinates and the wave velocity field includes:

[0017] The monitoring area is divided into cubic units with set side lengths to characterize the spatial variability of the wave velocity field;

[0018] The MINLP model is solved to obtain the initial values ​​of the earthquake source coordinates and the wave velocity field.

[0019] Optionally, the hierarchical optimization framework includes: an outer loop and an inner loop; the outer loop performs Delaunay triangulation on the monitoring area according to the current wave velocity field to generate an unstructured grid topology; the inner loop fixes the wave velocity field topology and uses the L-BFGS algorithm to optimize the source coordinates.

[0020] Optionally, the step of optimizing the initial values ​​of the earthquake source rough coordinates and the wave velocity field using the hierarchical optimization framework to obtain optimized earthquake source coordinates and wave velocity field comprises:

[0021] Inputting the coarse coordinates of the earthquake source and the initial values ​​of the velocity field and the sensor position coordinates, as well as the velocity field dynamically updated by the outer cyclic Delaunay triangulation, into the MINLP model for iterative calculation until an iteration termination condition is reached, and obtaining the gradient information of the most recent n iterations before the iteration termination, where n is a positive integer greater than 1;

[0022] The gradient information of the n iterations is calculated by using an L-BFGS algorithm to obtain optimized source coordinates and a wave velocity field.

[0023] Optionally, the method further comprises adding geometric constraints to the virtual sensor network and inserting virtual nodes between adjacent actual sensors to limit the distance between adjacent nodes to no more than twice the average distance between actual sensors.

[0024] Optionally, the posterior probability inference of the optimized source coordinates according to the optimized wave velocity field and the virtual node coordinates through Bayesian MCMC sampling to determine the three-dimensional positioning result of the rock burst event comprises:

[0025] constructing a Bayesian probability model;

[0026] inputting the optimized source coordinates and the wave velocity field, the virtual node coordinates and the P-wave arrival time into the Bayesian probability model to perform MCMC inference to obtain an MCMC sample set, wherein the MCMC sample set comprises posterior samples after removing preheating samples;

[0027] determining the highest posterior density interval covering a specified probability in the MCMC sample set;

[0028] outputting the source coordinates and their confidence corresponding to the samples in the highest posterior density interval.

[0029] The application also provides a rock burst microseismic source layered optimization positioning device, which comprises:

[0030] a coarse positioning module configured to establish a mixed integer nonlinear programming (MINLP) model containing wave velocity gradient field variables, perform source coarse positioning by using the MINLP model, and determine initial values of source coordinates and a wave velocity field;

[0031] an optimization module configured to construct a layered optimization framework, optimize the initial values of the source coarse coordinates and the wave velocity field by using the layered optimization framework, and obtain optimized source coordinates and a wave velocity field;

[0032] a sensor network construction module configured to insert virtual nodes in the MINLP model by using the optimized source coordinates and actual sensor positions, construct a virtual sensor network, and determine virtual node coordinates;

[0033] a positioning module configured to perform posterior probability inference of the optimized source coordinates according to the optimized wave velocity field and the virtual node coordinates through Bayesian MCMC sampling to determine a three-dimensional positioning result of a rock burst event.

[0034] The present invention also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is run by a processor, the steps of the rockburst microseismic source layered optimization positioning method are executed.

[0035] The method and device for optimizing the hierarchical location of rockburst microseismic sources, provided by embodiments of the present invention, are based on mixed-integer nonlinear programming (MINLP) and Bayesian inference. This method achieves coarse location of the source by constructing a MINLP model that includes a velocity gradient field. This is combined with a virtual sensor network to enhance spatial constraints, and ultimately employs Bayesian MCMC sampling to output confidence intervals. This approach can improve the location accuracy of heterogeneous rock masses, thereby enhancing the adaptability and timeliness of engineering applications.

[0036] The solution of the present invention can be applied to high-precision positioning of earthquake sources in underground projects such as tunnels and mines, and is particularly suitable for real-time monitoring needs in heterogeneous rock environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] To more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort.

[0038] Figure 1 This is a flow chart of a method for optimizing the location of rockburst microseismic sources by stratification provided by an embodiment of the present invention;

[0039] Figure 2 The figure is a structural diagram of a device for optimizing the positioning of rockburst microseismic sources in layers according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to enable those skilled in the art to better understand the solutions of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of the present invention.

[0041] The embodiments of the present invention provide a method and device for optimizing the hierarchical positioning of rockburst microseismic sources. Based on MINLP and Bayesian inference, the method and device achieve coarse source positioning by constructing a MINLP model including a velocity gradient field, enhances spatial constraints by combining a virtual sensor network, and finally uses Bayesian MCMC sampling to output confidence intervals.

[0042] like Figure 1 FIG. 1 is a flow chart of a method for optimizing the location of rockburst microseismic sources by stratification according to an embodiment of the present invention, comprising the following steps:

[0043] Step 101: Establish a MINLP model including velocity gradient field variables, use the MINLP model to perform coarse earthquake source positioning, and determine the earthquake source coordinates and initial values ​​of the velocity field.

[0044] The MINLP model is an optimization model that contains both continuous variables and integer variables, and has nonlinear terms in the objective function or constraints.

[0045] The MINLP model of the embodiment of the present invention is an objective function that aims to minimize the arrival time residual under the wave velocity gradient field constraint, and introduces mixed integer variables to represent the wave velocity field discretization unit. The construction process of the MINLP model is as follows:

[0046] First, the original microseismic signal and the actual sensor position are obtained, the original microseismic signal is filtered, and the P-wave arrival time is automatically picked up. Then, based on the picked-up P-wave arrival time and the actual sensor position, an objective function containing the wave velocity gradient field variable is established.

[0047] The raw microseismic signals can be obtained from the corresponding data providers. P-waves are longitudinal seismic waves, also known as expansion and contraction waves. They are elastic waves that emanate from the earthquake source, and their vibration direction is consistent with the propagation direction. P-waves can be captured using existing techniques, which are not limited in this embodiment of the present invention.

[0048] The following objective function can be constructed based on the picked-up P-wave arrival time and the actual sensor position:

[0049]

[0050] in:

[0051] x, y, z are the three-dimensional coordinates of the earthquake source to be solved;

[0052] x i ,y i ,z i is the three-dimensional coordinate of the i-th sensor;

[0053] is the arrival time of the p-wave measured by the i-th sensor;

[0054] v i is the wave velocity at the location of the i-th sensor;

[0055] is the wave velocity field gradient;

[0056] a is a geological structure influencing factor, which can be determined based on the rock mass integrity coefficient in the Engineering Rock Mass Classification Standard GB / T 50218;

[0057] v max is the maximum wave speed;

[0058] v min is the minimum wave speed;

[0059] L is a characteristic scale of a monitoring area;

[0060] λ is a regularization coefficient for suppressing wave speed mutation; λ can be optimized through cross-validation, and the value range is 0.1-1.0.

[0061] The above objective function is the MINLP model.

[0062] When the MINLP model is used for rough positioning of a seismic source to determine initial values of a seismic source coordinate and a wave speed field, the monitoring area can be divided into cubic units with a set side length (for example, the side length L satisfies d≤L / 10) to represent the spatial variability of the wave speed field; then the MINLP model is solved, for example, the Gurobi solver or the COPT solver can be used to solve the MINLP model, and the rough positioning seismic source coordinate (x, y, z) and the initial wave speed field gradient v0 are output.

[0063] When the MINLP model is used for rough positioning of a seismic source to determine initial values of a seismic source coordinate and a wave speed field, for example, the monitoring area can be divided into cubic units with a set side length, and the characteristic scale L is the cubic unit with the set side length (for example, the side length L satisfies d≤L / 10) to represent the spatial variability of the wave speed field; then the MINLP model is solved, for example, the Gurobi solver or the COPT solver can be used to solve the MINLP model, and the rough positioning seismic source coordinate (x, y, z) and the initial wave speed field gradient v0 are output.

[0064] L is a macroscopic scale of a monitoring area, used to describe the spatial range of the entire monitoring area; d is a microscopic scale of the monitoring area, used to determine the discretization accuracy of the wave speed field.

[0065] Step 102, a hierarchical optimization framework is constructed, and the hierarchical optimization framework is used to optimize the initial values of the rough seismic source coordinate and the wave speed field to obtain an optimized seismic source coordinate and wave speed field.

[0066] In the embodiment of the application, the hierarchical optimization framework includes an outer loop and an inner loop. Wherein:

[0067] The outer loop is used to update the current wave speed field v kPerform Delaunay triangulation on the monitoring area to generate an unstructured grid topology. Delaunay triangulation is a triangulation method widely used in fields such as computer graphics, computational geometry, and numerical analysis, which can generate high-quality, well-structured triangular meshes. Delaunay triangulation is based on the definition of Delaunay edges, that is, edges that satisfy the empty circle property. Specifically, suppose there is an edge e whose two endpoints are a and b. If there exists a circle that passes through points a and b, and the circle does not contain any other points in the point set V, then this edge e is called a Delaunay edge. Based on this definition, if all the edges in a triangulation of the point set V are Delaunay edges, then this triangulation is called a Delaunay triangulation.

[0068] The inner-layer circulating fixed velocity field topology uses the L-BFGS (Limited-memory Broyden-Fletcher-Goldfarb-Shanno) algorithm to optimize the source coordinates. The L-BFGS algorithm is a quasi-Newton algorithm widely used in machine learning and optimization problems and is an iterative algorithm for solving unconstrained nonlinear optimization problems.

[0069] Specifically, the initial values ​​of the source rough coordinates and the wave velocity field and the sensor position coordinates, as well as the wave velocity field v dynamically updated by the outer cyclic Delaunay triangulation, can be k The gradient information of the n most recent iterations before the termination of the iteration is obtained, where n is a positive integer greater than 1, such as n=10. The gradient information of the n iterations is then optimized using the L-BFGS algorithm to obtain the optimized source coordinates and wave velocity field.

[0070] In some embodiments, the iteration termination condition is as follows:

[0071]

[0072] Among them, (x k ,y k ,z k ) is the earthquake source coordinate of the kth iteration, v k is the wave velocity field of the kth iteration.

[0073] Step 103: insert virtual nodes into the MINLP model using the optimized earthquake source coordinates and actual sensor positions, construct a virtual sensor network, and determine the virtual node coordinates.

[0074] In some non-limiting embodiments, geometric constraints can be added to the virtual sensor network. For example, virtual nodes can be inserted between adjacent actual sensors to limit the distance between adjacent nodes to no more than twice the average distance between actual sensors. This constraint can further improve computational accuracy, reduce computational costs, and shorten computational time.

[0075] For example, the geometric constraint equation is as follows:

[0076]

[0077] Among them, D avg is the actual average distance between sensors, in meters.

[0078] The specific process of building a virtual sensor network is as follows:

[0079] (1) Inputting the actual sensor position coordinates and the source coordinates optimized in step 102 into the MINLP model;

[0080] (2) Based on the constrained Delaunay triangulation, virtual nodes are inserted between adjacent actual sensors to ensure that the distance between adjacent nodes does not exceed twice the average distance between actual sensors, and the total number of virtual nodes meets N virtual ≥3N real .

[0081] Among them, N virtual To generate the number of virtual nodes, N real is the actual number of microseismic sensor nodes.

[0082] By adding virtual nodes to the MINLP model, a virtual sensor network is obtained. This virtual sensor network can enhance the geometric constraints of the velocity field inversion and provide denser spatial sampling point information for subsequent Bayesian inference.

[0083] Step 104 : performing a posteriori probability inference on the optimized earthquake source coordinates through Bayesian MCMC sampling according to the optimized wave velocity field and the virtual node coordinates to determine a three-dimensional positioning result of the rockburst event.

[0084] Specifically, a Bayesian probability model can be constructed, which includes a prior distribution and a likelihood function, wherein the prior distribution can be constructed using a standard normal distribution and the likelihood function can be directly called.

[0085] The prior distribution is:

[0086] N(μ,σ 2 )

[0087] Where N is Gaussian distribution, μ is the mean value of the earthquake source coordinate after optimization, σ 2is the variance, which can be calculated from the final residual of the optimization process in step 102, and σ is the standard deviation.

[0088] The likelihood function can use the t distribution, which describes the distribution of the standardized sample mean and is mainly used for statistical analysis of small sample data, especially for estimating the mean when the population variance is unknown.

[0089] The optimized focal coordinates, wave velocity field, virtual node coordinates, and P-wave arrival times are then input into a Bayesian probability model. The NUST sampler is then called to perform a posteriori probability inference on the optimized focal coordinates through Bayesian MCMC (Markov Chain Monte Carlo) sampling to determine the three-dimensional location of the rockburst event. For example, MCMC inference can be performed to obtain an MCMC sample set, which includes posterior samples after removing preheated samples. The highest posterior density interval within the MCMC sample set that covers a specified probability is determined, and the focal coordinates and their confidence levels corresponding to the samples within the highest posterior density interval are output.

[0090] MCMC sampling is a Markov chain-based random sampling method used to approximately generate samples from complex probability distributions. It constructs a Markov chain whose stationary distribution is the target distribution and uses iterative sampling to approximate the target distribution.

[0091] Specifically, the optimized focal coordinates are regarded as prior information, and the focal coordinates are subject to the coordinates as the mean, σ 2 is a Gaussian distribution with a variance of

[0092] In this embodiment of the present invention, Bayesian inference can use a t-distribution to construct a likelihood function with ν = 4 degrees of freedom. For example, MCMC inference can be performed using, but not limited to, the NUTS (No-U-Turn Sampler) sampler, with MCMC sampling times ≥ 5000 and warm-up samples ≥ 2000. Warm-up samples refer to initial iteration samples discarded by the algorithm before actual sampling.

[0093] For the obtained MCMC sample set, the highest posterior density (HPD) interval covering a specified probability (such as 95%) can be extracted from it, and the optimized source location coordinates and their confidence ranges can be output.

[0094] The following further illustrates the process of optimizing the location of rockburst microseismic sources by layering using the method of the present invention.

[0095] Assume that a total of 8 microseismic sensors are deployed in a mine tunnel with a blind zone coverage rate of less than 12%. The P-wave arrival time of the sensors is recorded with a time synchronization error of less than 1 μs. Rock fracture events are artificially simulated in the tunnel, and the true 3D coordinates of the 8 simulated fracture events are recorded. The true 3D coordinates of the fracture events and the coordinates of the 8 sensors are shown in Table 1.

[0096] Table 1

[0097]

[0098] Step 1: MINLP-based coarse earthquake source location and velocity field initialization, as follows:

[0099] (1) The STA / LTA algorithm (short time window 5ms, long time window 50ms, threshold = 2.5) is used to automatically pick the P wave arrival time of the original microseismic signal;

[0100] (2) Constructing the objective function:

[0101]

[0102] (3) Model solution:

[0103] Input parameters are L = 300 m, α = 1.8, vmax = 5200 m / s, vmin = 2800 / ms, discrete unit size d = 30 m, satisfying d ≤ L / 10, and calling the Gurobi solver to solve the MINLP model objective function and output the coarse-located source coordinates and wave velocity gradient field.

[0104] Step 2: Source-velocity field coordinated optimization, as follows:

[0105] (1) Dynamic wave velocity field update

[0106] The outer loop is based on the current wave velocity field v k Perform Delaunay triangulation on the monitoring area to generate an unstructured grid topology; further input the coarse positioning source coordinates (x, y, z) in step 1, the actual sensor position coordinates (x i ,y i ,z i ), and the wave velocity field v dynamically updated by the outer cyclic Delaunay triangulation k , stores the gradient information of the last 10 iterations.

[0107] (2) L-BFGS synchronization optimization

[0108] The inner loop calls the L-BFGS algorithm for optimization calculation to achieve synchronous optimization of the coarse-positioned source coordinates and the wave velocity gradient field. The output optimized source location coordinates are shown in Table 2. The synchronous optimization iteration termination condition is:

[0109]

[0110] Table 2

[0111]

[0112]

[0113] Step 3: Virtual sensor network construction, as follows:

[0114] Input the positioning coordinates (x, y, z) of the optimized vibration output in step 2 and the actual sensor position coordinates (x i ,y i ,z i )(see Table 1), build a virtual sensor network.

[0115] Among them, the actual average distance between sensors D is measured avg =37.5m, calculate the spatial geometric constraint equation, which requires that the distance between adjacent nodes should be less than or equal to 75m.

[0116] Specifically, the actual sensor coordinates in Table 1 and the optimized source location coordinates in Table 2 are input to the MINLP model. The actual number of nodes (i.e., the number of microseismic sensors) N real =8, generate the number of virtual nodes N virtual =24;

[0117] Step 4: Bayesian MCMC uncertainty quantification to obtain the MCMC sample set, as follows:

[0118] (1) The optimized hypocenter coordinates obtained in step 2 are regarded as prior information. The hypocenter coordinates obey a Gaussian distribution with the coordinates as the mean and 3σ as the variance (σ is calculated from the coarse positioning residual). Bayesian inference uses t distribution to construct the likelihood function with the degree of freedom ν = 4. The NUTS sampler is called for MCMC inference. The MCMC sampling is preheated 2000 times and adopted 6000 times.

[0119] Step 5: Extract the 95% highest posterior density interval of the MCMC sample set and output the optimized focal coordinate means and 95% HPD intervals corresponding to the eight rupture events, as shown in Table 3.

[0120] Table 3

[0121]

[0122]

[0123] In order to further verify the positioning effect of the rockburst microseismic source layered optimization positioning method of the present invention, Table 4 shows the positioning results of the particle swarm algorithm and the mayfly algorithm.

[0124] Table 4

[0125]

[0126] The above comparison results show that the solution of the present invention can effectively improve the positioning accuracy of microseismic events, further verifying the superior performance of the algorithm.

[0127] Accordingly, the embodiment of the present invention also provides a device for optimizing the positioning of rockburst microseismic sources by stratification, such as Figure 2 The figure shows a structural diagram of the rockburst microseismic source layered optimization positioning device.

[0128] The rockburst microseismic source layered optimization positioning device 200 includes the following modules:

[0129] A coarse positioning module 201 is used to establish a mixed integer nonlinear programming MINLP model including velocity gradient field variables, and use the MINLP model to perform coarse earthquake source positioning to determine the earthquake source coordinates and the initial value of the velocity field;

[0130] An optimization module 202 is configured to construct a hierarchical optimization framework, and utilize the hierarchical optimization framework to optimize the initial values ​​of the earthquake source rough coordinates and the wave velocity field to obtain optimized earthquake source coordinates and wave velocity field;

[0131] A sensor network construction module 203 is configured to insert virtual nodes into the MINLP model using the optimized earthquake source coordinates and actual sensor locations to construct a virtual sensor network and determine the virtual node coordinates;

[0132] The positioning module 204 is configured to perform a posteriori probability inference on the optimized earthquake source coordinates through Bayesian MCMC sampling according to the optimized wave velocity field and the virtual node coordinates, and determine a three-dimensional positioning result of the rockburst event.

[0133] The specific implementation of each of the above modules can refer to the description in the embodiment of the rockburst microseismic source layered optimization positioning method of the present invention, and will not be repeated here.

[0134] The present invention provides a method and device for optimizing the stratified positioning of rockburst microseismic sources, based on mixed integer nonlinear programming and Bayesian inference, to achieve optimized stratified positioning of rockburst microseismic sources, effectively improving the positioning accuracy of heterogeneous rock masses, thereby improving the adaptability and timeliness of engineering scenario applications.

[0135] The positioning error of the solution of the present invention in a heterogeneous rock environment is ≤2.5m, which is more than 60% higher than the accuracy of traditional methods. The single positioning time takes <20 seconds. It can be better applied to high-precision positioning of earthquake sources in underground projects such as tunnels and mines, and meet the real-time monitoring needs in heterogeneous rock environments.

[0136] It should be understood that Figure 1 At least part of the steps may include multiple steps or multiple stages. These steps or stages are not necessarily performed at the same time, but can be performed at different times. The order of execution of these steps or stages is not necessarily one by one, but can be performed in turn or alternately with other steps or at least part of the steps or stages in other steps.

[0137] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided by the present invention can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical memory, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0138] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0139] The embodiments of the present invention are described in detail above. Specific implementation methods are used herein to illustrate the present invention. The description of the above embodiments is only used to help understand the methods and systems of the embodiments of the present invention. They are only embodiments of a part of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work should fall within the scope of protection of the present invention, and the content of this specification should not be understood as limiting the present invention. Therefore, any modifications, equivalent replacements, 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 method for optimizing the location of rockburst microseismic sources by stratification, characterized in that: The method comprises: Establishing a mixed integer nonlinear programming (MINLP) model including velocity gradient field variables, using the MINLP model to perform coarse earthquake source positioning and determine the earthquake source coordinates and initial values ​​of the velocity field; Constructing a hierarchical optimization framework, and optimizing the initial values ​​of the earthquake source rough coordinates and the wave velocity field using the hierarchical optimization framework to obtain optimized earthquake source coordinates and wave velocity field; Inserting virtual nodes into the MINLP model using the optimized source coordinates and actual sensor locations to construct a virtual sensor network and determine the virtual node coordinates; The optimized source coordinates are subjected to posterior probability inference through Bayesian MCMC sampling according to the optimized wave velocity field and the virtual node coordinates to determine the three-dimensional positioning result of the rock burst event.

2. The rockburst microseismic source layered optimization positioning method according to claim 1 is characterized in that: The MINLP model takes the minimization of the arrival time residual under the wave velocity gradient field constraint as the objective function, and introduces mixed integer variables to characterize the wave velocity field discretization unit.

3. The rockburst microseismic source layered optimization positioning method according to claim 2, characterized in that: The establishment of the MINLP model including the velocity gradient field variable includes: Obtain raw microseismic signals and actual sensor positions; Filtering the original microseismic signal and automatically picking up the P-wave arrival time; Based on the picked-up P-wave arrival time and the actual sensor position, an objective function including a wave velocity gradient field variable is established.

4. The method for optimizing the location of rockburst microseismic sources by stratification according to claim 1, characterized in that: The method of using the MINLP model to perform coarse earthquake source positioning and determine the initial values ​​of the earthquake source coordinates and the wave velocity field includes: The monitoring area is divided into cubic units with set side lengths to characterize the spatial variability of the wave velocity field; The MINLP model is solved to obtain the initial values ​​of the earthquake source coordinates and the wave velocity field.

5. The rockburst microseismic source layered optimization positioning method according to claim 1, characterized in that: The hierarchical optimization framework includes: an outer loop and an inner loop; the outer loop performs Delaunay triangulation on the monitoring area according to the current wave velocity field to generate an unstructured grid topology; the inner loop fixes the wave velocity field topology and uses the L-BFGS algorithm to optimize the source coordinates.

6. The method for optimizing the location of rockburst microseismic sources by stratification according to claim 5, characterized in that: The step of optimizing the initial values ​​of the earthquake source rough coordinates and the wave velocity field using the hierarchical optimization framework to obtain the optimized earthquake source coordinates and wave velocity field comprises: Inputting the coarse coordinates of the earthquake source and the initial values ​​of the velocity field and the sensor position coordinates, as well as the velocity field dynamically updated by the outer cyclic Delaunay triangulation, into the MINLP model for iterative calculation until an iteration termination condition is reached, and obtaining the gradient information of the most recent n iterations before the iteration termination, where n is a positive integer greater than 1; The L-BFGS algorithm is used to optimize the gradient information of the n iterations to obtain optimized source coordinates and wave velocity field.

7. The method for optimizing the location of rockburst microseismic sources by stratification according to claim 1, characterized in that: The method further comprises: Geometric constraints are added to the virtual sensor network, and virtual nodes are inserted between adjacent actual sensors to limit the distance between adjacent nodes to no more than twice the average distance between actual sensors.

8. The method for optimizing the location of rockburst microseismic sources by stratification according to claim 3, characterized in that: The performing of posterior probability inference on the optimized earthquake source coordinates by Bayesian MCMC sampling based on the optimized wave velocity field and the virtual node coordinates to determine the three-dimensional positioning result of the rockburst event includes: Constructing Bayesian probability models; Inputting the optimized earthquake source coordinates and wave velocity field, the virtual node coordinates, and the P-wave arrival time into the Bayesian probability model, performing MCMC inference, and obtaining an MCMC sample set, wherein the MCMC sample set includes posterior samples after removing preheating samples; Determine the highest posterior density interval in the MCMC sample set that covers the specified probability; Output the source coordinates and confidence levels of the samples within the highest posterior density interval.

9. A device for optimizing the location of rockburst microseismic sources by layering, characterized in that: The device comprises: A coarse positioning module is used to establish a mixed integer nonlinear programming MINLP model including wave velocity gradient field variables, and use the MINLP model to perform coarse earthquake source positioning to determine the earthquake source coordinates and the initial value of the wave velocity field; An optimization module is used to construct a hierarchical optimization framework, and optimize the initial values ​​of the earthquake source rough coordinates and the wave velocity field using the hierarchical optimization framework to obtain optimized earthquake source coordinates and wave velocity field; A sensor network construction module is used to insert virtual nodes into the MINLP model using the optimized source coordinates and actual sensor locations to construct a virtual sensor network and determine the virtual node coordinates; A positioning module is used to perform posterior probability inference on the optimized source coordinates through Bayesian MCMC sampling based on the optimized wave velocity field and the virtual node coordinates to determine the three-dimensional positioning result of the rock burst event.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method for optimizing the location of rockburst microseismic sources by layer are executed according to any one of claims 1 to 8.

Images