Microseismic signal analysis sensor arrangement method and system, and computer device

Abstract

The invention discloses a microseismic signal analysis sensor arrangement method and system and a computer device. The method comprises the following steps of: constructing a Port-Hamiltonian neural network model containing a Hamilton quantity and a dissipation item, wherein the Port-Hamiltonian neural network model comprises the Hamilton quantity and the dissipation item; historical micro-seismic data of the micro-seismic monitoring area are collected and preprocessed; based on the neural network model, the geological parameters and the preprocessed historical micro-seismic data, learning the dynamic characteristics of each monitoring unit of the micro-seismic monitoring system, and outputting dynamic parameters such as the inherent frequency, the damping ratio and the modal shape of the system; constructing a data-driven Min-Max MPC controller based on the kinetic parameters, wherein the controller generates a sensor arrangement strategy according to the monitoring precision and the resource consumption; and calculating the optimal arrangement position of the sensor according to the arrangement strategy. The invention further provides a system for implementing the method. According to the method, the Port-Hamiltonian neural network model is introduced, and the Hamilton quantity and the dissipation characteristics of the system are combined, so that the dynamic characteristics of the micro-seismic monitoring system are accurately described.

Description

Technical Field

[0001] The present application relates to the field of microseismic monitoring technology, and in particular to a method and system for arranging microseismic signal analysis sensors and a computer device. Background Art

[0002] In the field of underground engineering and energy development, microseismic monitoring is an important means to evaluate the stability of underground rock masses and predict geological disasters. Existing microseismic monitoring systems are mainly based on empirical methods or simple geometric optimization for sensor layout, which have the following problems: 1. Traditional methods fail to fully consider the dynamic characteristics of microseismic monitoring systems, resulting in insufficient monitoring accuracy; 2. Existing layout strategies lack an in-depth understanding of the energy transfer and dissipation characteristics of the system and cannot accurately capture the propagation characteristics of microseismic signals. Summary of the invention

[0003] In view of this, the present application provides a method and system for arranging microseismic signal analysis sensors, and a computer device, which solves the problems of low sensor arrangement efficiency and insufficient monitoring accuracy in the prior art.

[0004] The present application provides a method for arranging microseismic signal analysis sensors, including:

[0005] Constructing a Port-Hamiltonian neural network model, wherein the Port-Hamiltonian neural network model includes a Hamiltonian quantity and a dissipation term of a microseismic monitoring system;

[0006] Collecting historical microseismic data in a microseismic monitoring area and preprocessing the historical microseismic data;

[0007] Based on the Port-Hamiltonian neural network model, the geological parameters and the pre-processed historical microseismic data, the dynamic characteristics of each monitoring unit of the microseismic monitoring system are learned, and the dynamic parameters including the natural frequency, damping ratio and modal shape of the system are output;

[0008] constructing a data-driven Min-Max MPC controller based on the dynamic parameters, wherein the Min-Max MPC controller generates a sensor deployment strategy according to monitoring accuracy and resource consumption;

[0009] According to the sensor placement strategy, the optimal placement position of the sensor is calculated.

[0010] In the above method, constructing the Port-Hamiltonian neural network model includes: defining the Hamilton quantity of the microseismic monitoring system, the Hamilton quantity includes a kinetic energy term and a potential energy term, wherein the kinetic energy term describes the motion characteristics of the microseismic monitoring system through a modal mass matrix, and the potential energy term describes the deformation characteristics of the microseismic monitoring system through a stiffness matrix; constructing a Port-Hamiltonian neural network architecture corresponding to the Hamilton quantity, including an energy conservation layer, a port interaction layer and a dissipation layer, wherein the energy conservation layer describes the energy change of the microseismic monitoring system based on the kinetic energy term and the potential energy term, the port interaction layer characterizes the energy exchange between the microseismic monitoring system and the outside world, and the dissipation layer describes the energy loss of the microseismic monitoring system through a dissipation term; setting physical constraints in the neural network architecture, the physical constraints include the continuity conditions of displacement and velocity at the boundary, the constitutive equation constraints of the stress-strain relationship, and the constraint that the total energy of the microseismic monitoring system decreases monotonically over time.

[0011] In the above method, preprocessing the historical microseismic data includes: establishing a multi-scale geological model based on the geological parameters, the multi-scale geological model including regional-scale structural units and local-scale rock structure; constructing a microseismic event classification model to classify microseismic events into tensional, shear and mixed types according to the source mechanism; and performing spatiotemporal clustering analysis on the historical microseismic data based on the multi-scale geological model and the microseismic event classification model.

[0012] In the above method, based on the Port-Hamiltonian neural network model, the geological parameters and the pre-processed historical microseismic data, the dynamic characteristics of each monitoring unit of the microseismic monitoring system are learned, and the dynamic parameters including the natural frequency, damping ratio and modal shape of the system are output, including: inputting the geological parameters and the pre-processed historical microseismic data into the Port-Hamiltonian neural network model, calculating the response of the microseismic monitoring system through forward propagation, and obtaining a predicted value; calculating the loss function between the predicted value and the actual value, the loss function including a data fitting loss term, a physical constraint loss term and a regularization term; optimizing the parameters of the Port-Hamiltonian neural network model through back propagation; and extracting the dynamic parameters of each monitoring unit of the microseismic monitoring system based on the optimized Port-Hamiltonian neural network model.

[0013] In the above method, constructing a data-driven Min-Max MPC controller based on the kinetic parameters includes: inputting the kinetic parameters into a prediction model to construct a system state transfer equation; defining a control objective function based on the state transfer equation, wherein the control objective function is used to balance monitoring accuracy and resource consumption; setting constraints for the control objective function, wherein the constraints include a sensor quantity restriction and a layout position constraint; constructing an uncertain set of system models based on historical data, and combining the state transfer equation and the control objective function to construct a data-driven Min-Max MPC controller.

[0014] In the above method, the control objective function includes: a monitoring performance index item, which measures the monitoring accuracy of the sensor layout through signal quality and coverage range; a control cost item, which quantifies the cost of sensor layout through resource consumption; a change rate penalty item, which limits the rate of change of the control quantity by adjusting the amplitude; an energy efficiency index item, which measures the energy utilization efficiency of the layout scheme through the monitoring effect under unit energy consumption; and a layout uniformity index item, which characterizes the uniformity of coverage of the monitoring area through the standard deviation of the spatial distribution of sensors.

[0015] In the above method, the constraints include: a total number of sensors constraint, which limits the upper limit of the number of sensors that can be deployed; a deployment location constraint, which specifies the feasible deployment area of ​​each sensor; a signal quality constraint, which requires that the signal strength at the monitoring point is higher than a preset threshold; an energy consumption constraint, which limits the maximum power consumption of the system; and a communication link constraint, which specifies the maximum communication distance between sensors.

[0016] In the above method, calculating the optimal layout position of the sensor according to the sensor layout strategy includes: constructing a seismic wave propagation velocity model, the seismic wave propagation velocity model includes a P-wave velocity field and an S-wave velocity field; calculating the acoustic wave impedance distribution of the microseismic monitoring area based on the seismic wave propagation velocity model; determining the position of the key monitoring point according to the acoustic wave impedance distribution; combining the key monitoring point position and the sensor layout strategy, using a mixed integer programming method to solve the optimal layout position.

[0017] The embodiment of the present application also provides a microseismic signal analysis sensor arrangement system, including: a model construction module, used to construct a Port-Hamiltonian neural network model, the Port-Hamiltonian neural network model includes the Hamilton quantity and dissipation term of the microseismic monitoring system; a data acquisition and preprocessing module, used to collect geological parameters and historical microseismic data of the microseismic monitoring area, and preprocess the historical microseismic data; a dynamic characteristic learning module, used to learn the dynamic characteristics of each monitoring unit of the microseismic monitoring system based on the Port-Hamiltonian neural network model, the geological parameters and the preprocessed historical microseismic data, and output dynamic parameters including the system natural frequency, damping ratio and modal shape; a controller construction module, used to construct a data-driven Min-Max MPC controller based on the dynamic parameters, and the Min-Max MPC controller generates a sensor arrangement strategy according to the monitoring accuracy and resource consumption; an arrangement position calculation module, used to calculate the optimal arrangement position of the sensor according to the sensor arrangement strategy.

[0018] An embodiment of the present application also provides a computer device, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the above-mentioned method of arranging microseismic signal analysis sensors.

[0019] An embodiment of the present application also provides a computer-readable storage medium, which stores computer instructions, and the computer instructions are used to enable a computer to execute the above-mentioned method of arranging microseismic signal analysis sensors.

[0020] An embodiment of the present application also provides a computer program product, including computer instructions, characterized in that when the computer instructions are executed by a processor, the steps of the above-mentioned method for arranging microseismic signal analysis sensors are implemented.

[0021] This application has the following technical effects:

[0022] 1. By introducing the Port-Hamiltonian neural network model and combining the Hamiltonian and dissipative characteristics of the system, the dynamic characteristics of the microseismic monitoring system are accurately described, and the scientific nature of the sensor layout is improved;

[0023] 2. The data-driven Min-Max MPC controller is used to achieve multi-objective optimization of monitoring accuracy and resource consumption, improving system operation efficiency;

[0024] 3. Based on multi-scale geological models and microseismic event classification, in-depth mining of historical data is achieved, which improves the pertinence of sensor layout schemes;

[0025] 4. By setting up a complete constraint system, the feasibility and reliability of the sensor layout plan are ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the disclosed embodiments in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 A schematic diagram of a flow chart of a method for arranging sensors for microseismic signal analysis provided in an embodiment of the present application;

[0028] Figure 2 This is a structural diagram of the microseismic signal analysis sensor arrangement system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0029] In order to make the purpose, technical solutions and advantages of the present application more clear, the present application is further described in detail below in conjunction with embodiments.

[0030] Embodiment 1:

[0031] like Figure 1 As shown, the arrangement method of microseismic signal analysis sensors provided in the embodiment of the present application is as follows:

[0032] S1: Construct Port-Hamiltonian neural network model;

[0033] In step S1, the Port-Hamiltonian framework is an energy-based modeling method that can effectively describe the dynamic characteristics of complex physical systems. In this embodiment, the neural network model constructed by the framework can accurately describe the energy transfer and dissipation process of the microseismic monitoring system.

[0034] Specifically, the Hamiltonian of the microseismic monitoring system (hereinafter referred to as the system) is first defined, which consists of kinetic energy terms and potential energy terms. The kinetic energy term describes the motion characteristics of the system through the modal mass matrix, which contains information such as mass distribution and inertial characteristics; the potential energy term describes the deformation characteristics of the system through the stiffness matrix, reflecting the elastic potential energy stored in the system during the deformation process.

[0035] Among them, the kinetic energy term describes the motion characteristics of the system through the modal mass matrix M, which can be expressed as:

[0036]

[0037] Among them, q is the generalized coordinate vector of the system. The potential energy term describes the deformation characteristics of the system through the stiffness matrix K, which can be expressed as:

[0038]

[0039] Next, a neural network architecture is constructed based on the defined Hamilton quantity. The network architecture contains three key layers: energy conservation layer, port interaction layer, and dissipation layer. The energy conservation layer is the first layer. Based on the kinetic energy term and potential energy term of the Hamilton quantity, the energy change of the system is described through the Symplectic structure to ensure the conservation and transfer of energy within the system. The port interaction layer is the second layer. The graph neural network structure is used to establish the energy exchange relationship between the internal units of the system and between the system and the external environment, including the influence of external excitation, boundary conditions, etc. The dissipation layer is the third layer. The Rayleigh dissipation function (i.e., dissipation term) is introduced to describe the energy loss process of the system.

[0040] Finally, in order to ensure the physical rationality of the model, a series of physical constraints need to be set. First, the continuity conditions of displacement and velocity must be met at the system boundary to ensure the continuous change of physical quantities in space; second, the stress-strain relationship needs to meet the constraints of the constitutive equation to ensure the correct description of the mechanical properties of the material; finally, considering the irreversibility of energy dissipation, the total energy of the system is required to decrease monotonically over time, which reflects the thermodynamic characteristics of the system.

[0041] In summary, step S1 specifically includes:

[0042] S1.1: Define Hamiltonian quantities for microseismic monitoring systems;

[0043] S1.1.1: The kinetic energy term of the motion characteristics of the microseismic monitoring system is described by the modal mass matrix;

[0044] S1.1.2: Potential energy terms describing the deformation characteristics of the microseismic monitoring system through the stiffness matrix;

[0045] S1.2: Construct a Port-Hamiltonian neural network architecture including an energy conservation layer, a port interaction layer, and a dissipation layer;

[0046] S1.2.1: Energy conservation layer based on kinetic energy terms and potential energy terms to describe the energy changes of microseismic monitoring systems;

[0047] S1.2.2: Port interaction layer that characterizes the energy exchange between the microseismic monitoring system and the outside world;

[0048] S1.2.3: The dissipative layer that describes the energy loss of the microseismic monitoring system through the dissipative term;

[0049] S1.3: Set physical constraints;

[0050] S1.3.1: Continuity conditions for displacements and velocities at boundaries;

[0051] S1.3.2: Constitutive equation constraints on stress-strain relationships;

[0052] S1.3.3: The constraint that the total energy of the microseismic monitoring system decreases monotonically over time;

[0053] S2: Acquisition and preprocessing of historical microseismic data;

[0054] In step S2, historical microseismic data are collected and preprocessed. The sensor layout scheme of the microseismic monitoring system needs to fully consider the geological characteristics of the monitoring area and the laws of historical microseismic activity. To this end, a multi-scale geological model is first established based on geological parameters. The model describes the geological characteristics of the monitoring area from two levels: regional scale and local scale. At the regional scale, a three-dimensional geometric model including major structural units such as faults and folds is constructed; at the local scale, the structural characteristics of the rock strata are described in detail, including lithology distribution, bedding occurrence, and joint and fissure development. For example, in an application in a certain mining area, the regional scale model describes a main fault with a strike of NE45° and a dip of 75° and its related secondary fault system, while the local scale model describes in detail the rock strata structure within 10 meters around the working face, including the spatial distribution of roof sandstone, coal seam, and floor mudstone.

[0055] On this basis, a microseismic event classification model is constructed. According to different focal mechanisms, microseismic events are divided into three categories: tension, shear, and mixed. Among them, tension events are mainly related to rock stratum tensile fractures, characterized by a large volume expansion component; shear events are mostly related to fault slip, with obvious shear failure characteristics; mixed events have both tension and shear characteristics. The classification process uses a deep learning method based on waveform features to extract the spectral features, waveform features, and polarity features of microseismic signals, and establish a focal mechanism identification model.

[0056] Subsequently, based on the multi-scale geological model and microseismic event classification results, a spatiotemporal clustering analysis was performed on the historical microseismic data. In the time dimension, the dynamic time warping algorithm was used to identify the evolution pattern of the microseismic sequence; in the spatial dimension, the density clustering algorithm was combined to analyze the spatial distribution characteristics of microseismic events. For example, through spatiotemporal clustering analysis, it can be found that during the advancement of the mining face, tensile microseismic events mostly occurred within 20-30 meters in front of the working face, and showed obvious clustering characteristics when the working face advancement speed increased. Spatiotemporal clustering analysis reveals the distribution patterns of microseismic events in the time and space dimensions, which helps to identify high-risk areas and optimize sensor layout plans.

[0057] In summary, step S2 specifically includes:

[0058] S2.1: Build a multi-scale geological model;

[0059] S2.1.1: Regional scale structural units;

[0060] S2.1.2: Local-scale rock structure;

[0061] S2.2: Construct a microseismic event classification model;

[0062] S2.3: Perform spatiotemporal cluster analysis;

[0063] S3: Learning the dynamic characteristics of each monitoring unit of the microseismic monitoring system;

[0064] After completing the construction of the Port-Hamiltonian neural network model and preprocessing of historical data, it is necessary to learn the dynamic characteristics of the system in a data-driven way. The specific implementation process is as follows:

[0065] First, the geological parameters and preprocessed historical microseismic data are input into the Port-Hamiltonian neural network model, and the dynamic response of the microseismic monitoring system is calculated through forward propagation. During the forward propagation process, the state evolution of the system follows the Port-Hamiltonian equation:

[0066]

[0067] Among them, q and p represent generalized position and momentum respectively, H is Hamiltonian, R is dissipation matrix, G is input matrix, and u is external excitation. By solving this equation, the response prediction value of each monitoring unit of the system is obtained.

[0068] Next, the loss function between the predicted value and the actual observed value is calculated. This loss function contains three items: the data fitting loss term is used to measure the deviation between the model predicted value and the observed value, in the form of weighted mean square error; the physical constraint loss term ensures that the model prediction results meet the physical constraints, including energy conservation constraints, boundary condition constraints, etc.; the regularization term is used to prevent the model from overfitting, using a combination of L1 and L2 regularization. Specifically, the loss function expression is:

[0069] L=αL data +βL physics +γL reg

[0070] Among them, L data L is the data fitting loss term, which measures the deviation between the model prediction value and the actual observation value; physics is the physical constraint loss term, ensuring that the model meets the physical constraint conditions; L reg is a regularization term, combined with L1 and L2 regularization to prevent overfitting of the model; α is the weight coefficient of the data fitting loss term, which is used to control the degree of fit of the model to the actual observed data; β is the weight coefficient of the physical constraint loss term, which is used to control the degree of compliance of the model with physical laws; γ is the weight coefficient of the regularization term, which is used to control the complexity of the model. For example, in a practical application, it was verified that the model performance is optimal when α=0.6, β=0.3, and γ=0.1.

[0071] Subsequently, the parameters of the Port-Hamiltonian neural network model are optimized by the back-propagation algorithm. During the back-propagation process, special attention should be paid to maintaining the physical rationality of the model. For example, by introducing the Symplectic gradient descent method, the Hamiltonian structure of the system is ensured during the optimization process; the positive definiteness of the mass matrix and stiffness matrix is ​​guaranteed by the projection algorithm.

[0072] Finally, the dynamic parameters of the system are extracted based on the optimized model, including natural frequency, damping ratio and modal shape. These parameters reflect the essential dynamic characteristics of the microseismic monitoring system and provide a theoretical basis for the subsequent optimization of sensor layout. For example, the natural frequency spectrum of the system can be obtained through eigenvalue analysis, and the vibration modes of each order can be obtained through modal analysis. This information directly guides the spatial layout of the sensor.

[0073] In one embodiment, the above method specifically includes:

[0074] S3.1: Calculate the response of the microseismic monitoring system through forward propagation and obtain the predicted value;

[0075] S3.2: Calculate the loss function;

[0076] S3.2.1: Data fitting loss term;

[0077] S3.2.2: Physical constraint loss term;

[0078] S3.2.3: Regularization term;

[0079] S3.3: Optimize the parameters of the neural network model by back-propagation;

[0080] S3.4: Extract kinetic parameters;

[0081] Finally, key dynamic parameters are extracted from the optimized model, including characteristic quantities such as natural frequency, damping ratio and mode shape. These parameters directly reflect the dynamic characteristics of the system and provide an important basis for subsequent controller design.

[0082] S4: Build a data-driven Min-Max MPC controller;

[0083] In step S4, based on the extracted dynamic parameters, a controller is designed to generate an optimal sensor deployment strategy. The controller can balance monitoring accuracy and resource consumption while taking into account system uncertainty.

[0084] First, the dynamic parameters are input into the prediction model to construct the system state transfer equation. The discrete state space model of the system can be expressed as:

[0085] x k+1 =f(x k ,u k )+w k

[0086] Among them, x k represents the system state at time k (including the displacement and velocity of each monitoring point), u k represents the control input (i.e., the sensor layout), ω k represents the system uncertainty. The state transfer equation f(·) is given by the Port-Hamiltonian framework:

[0087]

[0088] Among them, J is an antisymmetric matrix representing the energy conservation term, R is a symmetric positive definite matrix representing the dissipation term, and Δt is the sampling interval.

[0089] This equation describes the evolution of the system state over time, including the system's dynamic characteristics and the influence of external excitation. The state transition equation adopts a discrete form, which is convenient for numerical solution and real-time control.

[0090] Next, define the control objective function, which needs to comprehensively consider multiple performance indicators. Specifically, the monitoring performance indicator item evaluates the information acquisition ability of the sensor layout through the condition number of the Fisher information matrix; the control cost item considers resource consumption such as the number of sensors and installation costs; the change rate penalty item limits the drastic changes in the sensor position to ensure the stability of the solution; the energy efficiency indicator item evaluates the monitoring effect under unit energy consumption; the layout uniformity indicator item ensures uniform coverage of the monitoring area. The overall objective function expression is:

[0091]

[0092] Among them, N p is the predicted time domain length; J perf(xk) To monitor the performance indicators, the Fisher information matrix is ​​used to evaluate the information acquisition capability of the sensor layout; cost(uk) To control the cost, consider resource consumption such as the number of sensors and installation cost; J rate(Δuk) is the penalty term for the rate of change, which limits the drastic change of the sensor position; J energy(zk,uk) is the energy efficiency index item, which evaluates the monitoring effect under unit energy consumption; J uniform(uk) To arrange uniformity indicators and ensure uniform coverage of the monitoring area; i (i=1,2,3,4,5) is the weight coefficient of each item.

[0093] Based on the objective function, a series of constraints are set to ensure the feasibility of the solution. The total number of sensors constraint limits the maximum number of sensors that can be deployed; the layout location constraint takes into account factors such as geological conditions and engineering feasibility, and stipulates the feasible installation area for each sensor; the signal quality constraint requires that the signal-to-noise ratio of each monitoring point is not lower than the preset threshold; the energy consumption constraint ensures that the system power consumption does not exceed the power supply capacity; the communication link constraint ensures the connectivity of the sensor network. These constraints can be expressed as:

[0094]

[0095] u i ∈Ω i ,i=1,2,…,n

[0096] SNR(x k )≥SNR min

[0097] P(u k )≤P max

[0098]

[0099] Among them, Ω irepresents the feasible layout area of ​​the i-th sensor, SNR(·) represents the signal-to-noise ratio function, P(·) represents the power consumption function, d ij represents the communication distance between sensors i and j.

[0100] Finally, the uncertainty set W of the system model is constructed based on historical data, and the data-driven Min-Max MPC controller is constructed by combining the state transition equation and the control objective function. The controller solves the following cross-point problem:

[0101]

[0102] By solving the optimization problem, a robust sensor placement strategy is obtained which can guarantee performance in the worst case.

[0103] Based on the above, it can be concluded that S4 specifically includes:

[0104] S4.1: Construct the system state transition equation;

[0105] S4.2: Define the control objective function;

[0106] S4.2.1: Monitoring performance indicators;

[0107] S4.2.2: Control cost items;

[0108] S4.2.3: Rate of change penalty term;

[0109] S4.2.4: Energy efficiency indicators;

[0110] S4.2.5: Arrangement uniformity index items;

[0111] S4.3: Set constraints;

[0112] S4.3.1: Constraint on the total number of sensors;

[0113] S4.3.2: Arrangement location constraints;

[0114] S4.3.3: Signal quality constraints;

[0115] S4.3.4: Energy consumption constraints;

[0116] S4.3.5: Communication link constraints;

[0117] S4.4: Construct Min-Max MPC controller;

[0118] Finally, a data-driven Min-Max MPC controller is constructed to obtain the optimal control strategy by solving the cross-point problem. This controller can ensure the robustness of control performance while considering system uncertainties.

[0119] S5: Calculate the optimal arrangement position of the sensor;

[0120] In step S5, based on the sensor placement strategy generated by the Min-Max MPC controller and in combination with the seismic wave propagation characteristics, the specific installation locations of the sensors are determined.

[0121] First, a seismic wave propagation velocity model is constructed. This model includes the P-wave velocity field and the S-wave velocity field, and describes the propagation law of seismic waves through the wave equation in the layered medium:

[0122]

[0123] Where ρ is the medium density, λ and μ are Lamé constants, and u is the displacement vector. For layered media, the wave velocity parameters of each layer can be obtained through acoustic logging data. For example, in an application in a mining area, the detailed velocity structure from the surface to a depth of -800 meters was obtained through logging data, where the P-wave velocity of the surface sandstone was about 3500m / s, and the P-wave velocity of the coal seam was about 2200m / s.

[0124] Then, the acoustic impedance distribution of the monitoring area is calculated based on the wave velocity model. The acoustic impedance Z is defined as the product of density and wave velocity:

[0125] Z=ρv

[0126] Where v is the velocity of seismic waves. At the interface of the medium, the sudden change of acoustic impedance will cause the reflection and transmission of waves, affecting the propagation of signals. By analyzing the distribution of acoustic impedance, favorable channels and unfavorable areas for signal propagation can be identified.

[0127] Next, the key monitoring points are determined based on the acoustic impedance distribution. The selection of key monitoring points needs to consider the following factors: First, more monitoring points should be arranged in areas with frequent microseismic activity; second, areas with large acoustic impedance gradients need to be densely arranged to accurately capture waveform characteristics; third, key structural parts (such as fault intersections and stress concentration areas) should be the focus of monitoring. Specifically, the importance of monitoring at each location can be quantified through the following evaluation function:

[0128] I(x)=w1f activity (x)+w2f impedance (x)+w3f structure (x)

[0129] Among them, f activity represents the microseismic activity index, f impedance Represents the characteristic index of acoustic impedance, f structure represents the construction importance index, w i is the weight coefficient of each indicator.

[0130] Finally, combined with the key monitoring point locations and sensor layout strategies, the mixed integer programming method is used to solve the optimal layout location. The mathematical model of the optimization problem can be expressed as:

[0131]

[0132] y ij ∈{0,1}

[0133] Among them, x i Represents the position coordinates of the i-th sensor, y ij To arrange the decision variables, c ij is the monitoring effect score of location i on monitoring point j, d(·,·) represents the spatial distance function, and Ω is the feasible layout area. By solving this optimization problem, the optimal sensor layout position that meets various constraints can be obtained.

[0134] In one embodiment, S5 further includes:

[0135] S5.1: Construct a seismic wave propagation velocity model;

[0136] S5.2: Calculate the acoustic impedance distribution;

[0137] S5.3: Determine the locations of key monitoring points;

[0138] S5.4: Find the optimal layout location.

[0139] The construction of the seismic wave propagation velocity model is the basis for optimizing the sensor layout. The model needs to consider the propagation characteristics of P waves and S waves in different media and describe the propagation law of seismic waves through wave equations. The model parameters can be obtained through field test data, such as acoustic logging data.

[0140] The calculation of acoustic impedance distribution reflects the energy change characteristics of seismic waves during propagation. By analyzing the spatial distribution of acoustic impedance, favorable and unfavorable areas for signal propagation can be identified, providing an important reference for sensor layout.

[0141] The determination of key monitoring points requires comprehensive consideration of multiple factors. Each potential monitoring point is scored using an evaluation function, taking into account factors such as the frequency of microseismic activity, acoustic impedance characteristics, and structural location importance. Locations with higher scores will be prioritized for deployment.

[0142] Finally, the mixed integer programming method is used to solve the optimal layout position. This optimization process needs to consider both discrete variables (number of sensors) and continuous variables (layout coordinates) at the same time, and obtain the optimal solution that meets various constraints through iterative calculation, and finally determine the specific installation location of each sensor.

[0143] The microseismic signal analysis sensor layout method in this embodiment accurately describes the system dynamic characteristics through the Port-Hamiltonian neural network model, uses the Min-Max MPC control strategy to achieve a balance between monitoring effect and resource consumption, and determines the optimal layout position in combination with the seismic wave propagation characteristics, providing a scientific basis for the efficient deployment of the microseismic monitoring system. This method has been applied in many engineering practices and achieved good results. For example, in an application in a certain mining area, compared with the traditional layout method, the microseismic positioning accuracy of this method was improved by 30%, and the system energy consumption was reduced by 25%, which fully verified the practical value of this method.

[0144] Example 2

[0145] Based on the above method, this embodiment also provides a microseismic signal analysis sensor arrangement system. Figure 2 As shown, the following is a detailed description of each functional module of the system:

[0146] The model construction module 10 is used to construct the Port-Hamiltonian neural network model. This module adopts a modular design concept and mainly includes the following sub-units: the Hamilton quantity definition unit is responsible for constructing the energy description of the system, including the kinetic energy calculation unit based on the modal mass matrix and the potential energy calculation unit based on the stiffness matrix; the neural network architecture unit is responsible for realizing the construction of the energy conservation layer, the port interaction layer and the dissipation layer; the physical constraint setting unit is responsible for realizing the embedding of physical constraints such as boundary conditions and constitutive relations. The output of the model construction module is a complete Port-Hamiltonian neural network model structure.

[0147] The data acquisition and preprocessing module 20 is responsible for collecting and processing microseismic monitoring data. The module first obtains the original data collected by the on-site microseismic monitoring instrument through the data acquisition interface unit, including waveform data, trigger time, sampling parameters and other information. Then the data preprocessing unit performs basic processing such as filtering, noise reduction, and normalization. The geological model construction unit establishes a multi-scale geological model based on geological exploration data such as drilling and logging. The event classification unit uses a deep learning method to classify microseismic events, and the spatiotemporal analysis unit completes cluster analysis.

[0148] The dynamic characteristics learning module 30 is the core computing unit of the system. This module includes: the model training unit is responsible for the parameter optimization of the Port-Hamiltonian neural network, in which the loss function calculation subunit realizes the calculation of data fitting loss, physical constraint loss and regularization term, and the parameter update subunit adopts the improved stochastic gradient descent algorithm to realize the optimization of network parameters; the dynamic parameter extraction unit is responsible for extracting characteristic parameters such as the natural frequency, damping ratio and modal shape of the system from the trained model.

[0149] The controller construction module 40 is used to generate the sensor deployment strategy. The module first constructs the system state transfer equation by the prediction model unit, then designs the multi-objective optimization function by the objective function construction unit, and the constraint setting unit is responsible for realizing the mathematical description of various constraint conditions. The controller optimization unit uses a robust optimization algorithm to solve the Min-Max MPC control problem and outputs the optimized sensor deployment strategy.

[0150] The layout position calculation module 50 is responsible for determining the specific installation position of the sensor. This module includes: a wave velocity model construction unit responsible for establishing the P-wave and S-wave velocity field models; an impedance analysis unit calculating the acoustic wave impedance distribution; a key point selection unit determining important monitoring positions based on comprehensive evaluation; and a position optimization unit using a mixed integer programming algorithm to calculate the final layout plan.

[0151] In addition, the system also includes a human-computer interaction interface module, which provides functions such as parameter configuration, operation monitoring and result display. This module adopts a layered design, including: the data display layer is responsible for visualizing the system operation status and calculation results; the parameter configuration layer provides a configuration interface for model parameters, control parameters, etc.; the task management layer realizes the creation, execution and management of calculation tasks.

[0152] The above functional modules are connected through standardized data interfaces to form a complete system architecture. The data interaction between modules adopts a unified data format and communication protocol to ensure the stability and reliability of system operation. The system also reserves a function expansion interface to support the flexible access of new algorithms and new functions.

[0153] In actual applications, users can configure system parameters and start the calculation process through the human-computer interaction interface. The system will automatically complete the entire process from data processing to solution generation and output a detailed calculation report. The modular design of the system makes it well maintainable and scalable, and can adapt to the application requirements of different engineering scenarios.

Claims

1. A method for arranging microseismic signal analysis sensors, characterized in that: include: Constructing a Port-Hamiltonian neural network model, wherein the Port-Hamiltonian neural network model includes a Hamiltonian quantity and a dissipation term of a microseismic monitoring system; Collecting historical microseismic data in the microseismic monitoring area and preprocessing the historical microseismic data; Based on the Port-Hamiltonian neural network model, the geological parameters and the pre-processed historical microseismic data, the dynamic characteristics of each monitoring unit of the microseismic monitoring system are learned, and the dynamic parameters including the natural frequency, damping ratio and modal shape of the system are output; constructing a data-driven Min-Max MPC controller based on the dynamic parameters, wherein the Min-Max MPC controller generates a sensor deployment strategy according to monitoring accuracy and resource consumption; According to the sensor placement strategy, the optimal placement position of the sensor is calculated.

2. The method according to claim 1, characterized in that Constructing the Port-Hamiltonian neural network model includes: Define Hamiltonian quantities of the microseismic monitoring system, wherein the Hamiltonian quantities include kinetic energy terms and potential energy terms, wherein the kinetic energy terms describe the motion characteristics of the microseismic monitoring system through a modal mass matrix, and the potential energy terms describe the deformation characteristics of the microseismic monitoring system through a stiffness matrix; Constructing a Port-Hamiltonian neural network architecture corresponding to the Hamiltonian quantity, including an energy conservation layer, a port interaction layer and a dissipation layer, wherein the energy conservation layer describes the energy change of the microseismic monitoring system based on the kinetic energy term and the potential energy term, the port interaction layer characterizes the energy exchange between the microseismic monitoring system and the outside world, and the dissipation layer describes the energy loss of the microseismic monitoring system through the dissipation term; Physical constraints are set in the neural network architecture, including continuity conditions of displacement and velocity at the boundary, constitutive equation constraints of stress-strain relationship, and constraints that the total energy of the microseismic monitoring system decreases monotonically over time.

3. The method according to claim 1, characterized in that Preprocessing the historical microseismic data includes: Based on the geological parameters, a multi-scale geological model is established, wherein the multi-scale geological model includes regional-scale structural units and local-scale rock formation structures; A microseismic event classification model was constructed to classify microseismic events into tension, shear and mixed types according to the focal mechanism. Based on the multi-scale geological model and the microseismic event classification model, a spatiotemporal clustering analysis is performed on historical microseismic data.

4. The method according to claim 1, characterized in that Based on the Port-Hamiltonian neural network model, the geological parameters and the pre-processed historical microseismic data, the dynamic characteristics of each monitoring unit of the microseismic monitoring system are learned, and the dynamic parameters including the natural frequency, damping ratio and modal shape of the system are output, including: Inputting the geological parameters and the pre-processed historical microseismic data into the Port-Hamiltonian neural network model, calculating the response of the microseismic monitoring system through forward propagation, and obtaining a predicted value; Calculating a loss function between the predicted value and the actual value, wherein the loss function includes a data fitting loss term, a physical constraint loss term, and a regularization term; Optimizing the parameters of the Port-Hamiltonian neural network model by back propagation; The dynamic parameters of each monitoring unit of the microseismic monitoring system are extracted based on the optimized Port-Hamiltonian neural network model.

5. The method according to claim 1, characterized in that Building a data-driven Min-Max MPC controller based on the dynamic parameters includes: Inputting the kinetic parameters into a prediction model to construct a system state transfer equation; Based on the state transfer equation, a control objective function is defined, where the control objective function is used to balance monitoring accuracy and resource consumption; Setting constraints for the control objective function, wherein the constraints include a sensor quantity restriction and a layout position constraint; An uncertain set of system models is constructed based on historical data, and a data-driven Min-Max MPC controller is constructed by combining the state transfer equation and the control objective function.

6. The method according to claim 5, characterized in that The control objective function includes: Monitoring performance indicators, wherein the monitoring performance indicators measure the monitoring accuracy of the sensor arrangement through signal quality and coverage; a control cost term, the control cost term quantifying the cost of the sensor deployment through resource consumption; A change rate penalty term, wherein the change rate penalty term limits the change rate of the control amount by adjusting the amplitude; An energy efficiency index item, wherein the energy efficiency index item measures the energy utilization efficiency of the layout scheme through monitoring results under unit energy consumption; The arrangement uniformity index item characterizes the uniformity of the monitoring area coverage through the standard deviation of the sensor spatial distribution.

7. The method according to claim 5, characterized in that The constraints include: A total number of sensors constraint, wherein the total number of sensors constraint defines an upper limit on the number of sensors that can be deployed; A placement position constraint, wherein the placement position constraint specifies a feasible placement area for each sensor; A signal quality constraint, wherein the signal quality constraint requires that the signal strength at the monitoring point is higher than a preset threshold; an energy consumption constraint, the energy consumption constraint defining a maximum power consumption of the system; A communication link constraint specifies a maximum communication distance between sensors.

8. The method according to claim 1, characterized in that: Calculating the optimal arrangement position of the sensor according to the sensor arrangement strategy includes: Constructing a seismic wave propagation velocity model, wherein the seismic wave propagation velocity model includes a P-wave velocity field and an S-wave velocity field; Calculating the acoustic impedance distribution of the microseismic monitoring area based on the seismic wave propagation velocity model; Determining the location of key monitoring points according to the acoustic wave impedance distribution; Combining the key monitoring point locations and the sensor layout strategy, a mixed integer programming method is used to solve the optimal layout location.

9. A microseismic signal analysis sensor arrangement system, characterized in that: include: A model building module, used to build a Port-Hamiltonian neural network model, wherein the Port-Hamiltonian neural network model includes a Hamilton quantity and a dissipation term of a microseismic monitoring system; A data acquisition and preprocessing module, used to acquire geological parameters and historical microseismic data in the microseismic monitoring area, and preprocess the historical microseismic data; A dynamic characteristic learning module is used to learn the dynamic characteristics of each monitoring unit of the microseismic monitoring system based on the Port-Hamiltonian neural network model, the geological parameters and the pre-processed historical microseismic data, and output dynamic parameters including the natural frequency, damping ratio and modal shape of the system; A controller building module, for building a data-driven Min-Max MPC controller based on the kinetic parameters, wherein the Min-Max MPC controller generates a sensor deployment strategy according to monitoring accuracy and resource consumption; The arrangement position calculation module is used to calculate the optimal arrangement position of the sensor according to the sensor arrangement strategy.

10. A computer device, characterized in that: The computer device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method for arranging microseismic signal analysis sensors as described in any one of claims 1 to 8.

Images