Micro-seismic monitoring and rockburst early warning method and device

Abstract

The embodiment of the invention provides a micro-seismic monitoring and rockburst early warning method and device, and the method comprises the steps: deploying a sensor network to collect multi-source monitoring data which comprises acceleration data, speed data and strain data; constructing a dynamic system model represented by a Koopman operator based on historical multi-source monitoring data; combining the dynamic system model with real-time multi-source monitoring data, and performing state estimation through an improved Kalman filtering framework; based on the state estimation result, establishing a multi-scale prediction model for prediction by using a Koopman spectrum analysis method, and outputting a prediction result containing the frequency of the micro-seismic event, the vibration energy level and a Koopman spectrum characteristic index; a prediction result is corrected in real time through a tool variable method and a self-adaptive control method; and respectively comparing the microseismic event frequency, the vibration energy level and the Koopman spectrum characteristic index in the corrected prediction result with corresponding early warning thresholds, and outputting early warning information of different grades according to comparison results.

Description

Technical Field

[0001] The present application relates to the technical field of rock mechanics monitoring, and in particular to a method and device for microseismic monitoring and rock burst warning. Background Art

[0002] In underground engineering and mining, rockburst is a sudden dynamic destruction phenomenon of rock mass, which may cause major safety accidents and economic losses. When the deep rock mass is under high ground stress, the elastic strain energy accumulated inside the rock mass will be released instantly due to stress redistribution caused by mining activities, resulting in sudden and explosive destruction of the rock mass. According to statistics, the frequency and intensity of rockbursts increase exponentially with the increase of mining depth. In deep well mining below 3,000 meters, about 80% of safety accidents are related to rockbursts.

[0003] There are two methods for rock burst warning in the prior art. The first is monitoring through sensors, and the second is prediction through mechanical models. The inventors found that the existing rock burst warning methods mainly have the following problems:

[0004] 1. The traditional single sensor monitoring method is difficult to fully reflect the dynamic characteristics of the rock mass, resulting in incomplete monitoring information;

[0005] 2. Existing early warning models are mostly based on empirical statistics or simplified mechanical models, and fail to fully consider the nonlinear dynamic characteristics of the rock failure process. Summary of the invention

[0006] In view of this, the present application provides a method and device for microseismic monitoring and rockburst warning, which solves the problems in the prior art of incomplete monitoring information, overly simplified warning models, low prediction accuracy and lack of dynamic correction capabilities of rockburst warning systems.

[0007] The present application provides a method for microseismic monitoring and rockburst early warning, including:

[0008] Deploy a sensor network to collect multi-source monitoring data, wherein the multi-source monitoring data includes acceleration data, velocity data, and strain data;

[0009] Construct a dynamic system model represented by the Koopman operator based on historical multi-source monitoring data;

[0010] The dynamic system model is combined with real-time multi-source monitoring data to perform state estimation through an improved Kalman filter framework;

[0011] Based on the state estimation results, a multi-scale prediction model is established using the Koopman spectrum analysis method to perform prediction, and a prediction result including microseismic event frequency, vibration energy level and Koopman spectrum characteristic index is output;

[0012] Through the instrumental variable method and adaptive control method, the forecast results are corrected in real time;

[0013] The microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction result are respectively compared with the corresponding warning thresholds, and different levels of warning information are output according to the comparison results.

[0014] Optionally, the deploying the sensor network includes:

[0015] Divide the monitoring sub-areas according to the geological structural characteristics of the monitoring area;

[0016] Based on the sound wave propagation path and attenuation characteristics, calculate the minimum number of sensors required for each monitoring sub-area;

[0017] The spatial coordinates of each sensor are determined by an optimization algorithm, so that the monitoring ranges of adjacent sensors have an overlap of 10% to 30%, and the area of ​​the monitoring blind spot is less than 5% of the total monitoring area;

[0018] The physical arrangement of the sensor array is completed according to the determined spatial coordinates.

[0019] Optionally, constructing a dynamic system model represented by a Koopman operator based on historical multi-source monitoring data includes:

[0020] The extended dynamic mode decomposition method is used to select the basis function space including polynomial basis and radial basis function to construct the initial Koopman eigenfunction;

[0021] Performing eigenvalue decomposition on the initial Koopman eigenfunction to obtain optimized Koopman eigenfunction and eigenvalue, and constructing a Koopman spectrum analysis framework;

[0022] The time arrival boundary is determined based on the Koopman spectrum analysis framework, and the time arrival boundary is used as a warning time window to limit the time range of state estimation and prediction.

[0023] Optionally, combining the dynamic system model with real-time multi-source monitoring data to perform state estimation includes:

[0024] Based on the spectral characteristics of the Koopman operator, an improved Kalman filter state observer is constructed;

[0025] Using a sliding time window for online state estimation, the sliding time window length is 3 to 5 times the characteristic time scale;

[0026] Based on the state estimation result, a multi-scale prediction model is established by using the Koopman spectrum analysis method to perform prediction, including:

[0027] According to the state estimation result, a fast mode and a slow mode are identified using a Koopman spectrum analysis method;

[0028] Construct a multi-scale prediction model that includes both fast and slow modes;

[0029] The multi-scale prediction model is used to predict the frequency of microseismic events, the vibration energy level and the Koopman spectrum characteristic index to obtain prediction results.

[0030] Optionally, constructing an improved Kalman filter state observer based on the spectral characteristics of the Koopman operator includes:

[0031] Based on the linear combination of the acquired Koopman eigenfunctions, the state vector expression is established;

[0032] Based on the spectral characteristics of the Koopman operator and the state vector expression, a state transfer matrix is ​​constructed to describe the evolution of the system state over time;

[0033] Determining an observation matrix according to the configuration of the sensor and the state transfer matrix, so as to establish a mapping relationship between observation values ​​and state variables;

[0034] The basic structure of the Kalman filter is defined by the state vector expression, the state transfer matrix and the observation matrix;

[0035] An adaptive covariance matrix update strategy is constructed to simultaneously adjust the system noise covariance matrix and the observation noise covariance matrix of the Kalman filter through an innovative sequence to achieve dynamic optimization of state estimation.

[0036] Optionally, the real-time correction of the prediction results by the instrumental variable method and the adaptive control method includes:

[0037] Design bias correction mechanisms based on instrumental variables that satisfy the condition of zero covariance with measurement noise;

[0038] Building a model-referenced adaptive controller;

[0039] According to the bias correction mechanism of the instrumental variables and the model reference adaptive controller, the forecast results are corrected in real time.

[0040] Optionally, the design is based on a bias correction mechanism of instrumental variables, including:

[0041] Constructing an instrumental variable matrix, wherein the absolute value of the correlation coefficient between the instrumental variable matrix and the measurement noise is less than 0.1;

[0042] Constructing an unbiased estimator based on the instrumental variable matrix to calculate the estimated values ​​of the model parameters of the multi-scale prediction model;

[0043] An exponential weighting strategy is used to update the weight coefficients of the model parameters, and the weight coefficients decay exponentially over time;

[0044] Establish a layered protection mechanism including data quality inspection layer, model diagnosis layer and correction control layer;

[0045] Online parameter estimation is performed based on a sliding time window, and the window length is dynamically adjusted according to the estimation accuracy, wherein the adjustment basis of the window length is the convergence speed and stability index of the parameter estimation.

[0046] Optionally, the constructing a model reference adaptive controller comprises:

[0047] Constructing a reference model based on ideal dynamic characteristics, wherein the reference model includes an ideal system matrix and an input matrix;

[0048] According to the state equation of the reference model and the state equation of the multi-scale prediction model, a parameter adaptation law satisfying the Lyapunov stability condition is constructed;

[0049] According to the parameter adaptation law and the state of the multi-scale prediction model, an adaptive gain matrix and a positive definite matrix are constructed.

[0050] Optionally, according to the bias correction mechanism of the instrumental variable and the model reference adaptive controller, the prediction result is corrected in real time, including:

[0051] Constructing a composite predictor including a nominal prediction model, an instrumental variable bias estimator, and a model reference adaptive controller; the instrumental variable bias estimator is constructed based on a bias correction mechanism of the instrumental variable;

[0052] Based on the composite predictor, a tracking error is calculated, and the prediction result is corrected in real time according to the tracking error.

[0053] Optionally, the microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction result are respectively compared with corresponding warning thresholds, and different levels of warning information are output according to the comparison results, including:

[0054] Set up multi-level warning thresholds, where the frequency of microseismic events, vibration energy level, and Koopman spectrum characteristic indicators each have four levels of thresholds;

[0055] When all indicators are below the first-level threshold, it is judged as a low-risk level;

[0056] When any indicator exceeds the first-level threshold but none exceeds the second-level threshold, it is judged as a medium risk level;

[0057] When any indicator exceeds the second-level threshold but none exceeds the third-level threshold, it is judged as a high-risk level;

[0058] When any indicator exceeds the third-level threshold, it is judged as an extremely high risk level;

[0059] Generate early warning information including risk assessment and treatment recommendations.

[0060] Among them, the warning is upgraded according to the first condition, and the warning is downgraded according to the second condition. The first condition is that the indicator exceeds the higher level threshold for a first time, and the second condition is triggered when the indicator is lower than the current level threshold for a second time;

[0061] Send warning information to the preset receiving terminal.

[0062] The present application also provides a device for microseismic monitoring and rock burst warning, including:

[0063] A data acquisition module, used to collect multi-source monitoring data through a sensor network, wherein the multi-source monitoring data includes acceleration data, velocity data and strain data;

[0064] The dynamic modeling module is used to construct a dynamic system model represented by the Koopman operator based on historical multi-source monitoring data;

[0065] A state estimation module, used to combine the dynamic system model with real-time multi-source monitoring data and perform state estimation through an improved Kalman filter framework;

[0066] A prediction module is used to establish a multi-scale prediction model based on the state estimation result by using Koopman spectrum analysis to perform prediction, and output prediction results including microseismic event frequency, vibration energy level and Koopman spectrum characteristic indicators;

[0067] The result correction module is used to make real-time corrections to the prediction results through the instrumental variable method and the adaptive control method;

[0068] The early warning analysis module is used to compare the microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction results with the corresponding early warning thresholds, and output different levels of early warning information according to the comparison results.

[0069] The present application also provides a computer device, the computer device comprising:

[0070] at least one processor; and,

[0071] a memory communicatively connected to the at least one processor; wherein,

[0072] The memory stores instructions that can be executed 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 above-mentioned microseismic monitoring and rockburst early warning method.

[0073] 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 microseismic monitoring and rockburst early warning method.

[0074] The 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 microseismic monitoring and rockburst early warning method are implemented.

[0075] This application has the following technical effects:

[0076] 1. By deploying a multi-source sensor network, multi-dimensional monitoring data including acceleration, velocity and strain are collected to fully reflect the dynamic characteristics and failure process of the rock mass;

[0077] 2. The Koopman operator theory is used to construct a nonlinear dynamic system model, and the improved Kalman filter framework is used for state estimation to improve the accuracy of system modeling;

[0078] 3. Real-time correction of prediction results is achieved based on instrumental variable method and adaptive control strategy, which effectively reduces the impact of measurement noise and model error;

[0079] 4. Establish a multi-level early warning mechanism, output early warning information based on the comprehensive judgment of multiple indicators, and improve the reliability and timeliness of the early warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0080] 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.

[0081] Figure 1 It is an overall flow chart of the microseismic monitoring and rockburst early warning method in the embodiment of the present application.

[0082] Figure 2 The diagram is a schematic diagram of the sensor network arrangement in an embodiment of the present application.

[0083] Figure 3 4 is a flowchart of data processing and state estimation in an embodiment of the present application.

[0084] Figure 4 This is a flowchart of warning level determination in an embodiment of the present application.

[0085] Figure 5 4 is a structural diagram of the microseismic monitoring and rockburst early warning device in an embodiment of the present application. DETAILED DESCRIPTION

[0086] like Figure 1 As shown, the embodiment of the present application provides a microseismic monitoring and rockburst early warning method, including:

[0087] S1: Deploy sensor networks to collect multi-source monitoring data, including acceleration data, velocity data, and strain data;

[0088] In the specific implementation of this application, it is first necessary to deploy a sensor network to collect multi-source monitoring data. Due to the complexity of rock mass structure and the diversity of destruction processes, a single type of sensor is difficult to fully reflect the dynamic characteristics of rock mass. Therefore, this application adopts a multi-source sensor network, such as Figure 2 As shown, acceleration data, velocity data and strain data are collected simultaneously to obtain comprehensive information on the deformation and failure process of the rock mass.

[0089] Exemplarily, S1 specifically includes:

[0090] S1.1: Divide the monitoring sub-areas according to the geological structure characteristics of the monitoring area;

[0091] Before deploying the sensor network, the monitoring area needs to be reasonably divided. Based on the geological structure characteristics of the monitoring area, such as fault distribution, rock formation orientation, and the degree of development of joints and fissures, the entire monitoring area is divided into several monitoring sub-areas. This division method can ensure that the monitoring layout is adapted to the geological conditions and improve the pertinence and effectiveness of monitoring.

[0092] S1.2: Calculate the minimum number of sensors required for each monitoring sub-area based on the sound wave propagation path and attenuation characteristics;

[0093] For each monitoring sub-area, the number of sensors required needs to be scientifically determined. This application considers the propagation characteristics of sound waves in rock mass, including the attenuation law and propagation path of the waves, and calculates the minimum number of sensors required for each monitoring sub-area. This calculation method can both ensure monitoring coverage and avoid cost waste caused by redundant sensor numbers.

[0094] S1.3: Use optimization algorithms to determine the spatial coordinates of each sensor so that the monitoring ranges of adjacent sensors have a 10% to 30% overlap and the area of ​​the monitoring blind spot is less than 5% of the total monitoring area;

[0095] After determining the number of sensors, an optimization algorithm is used to determine the specific installation location of each sensor. The optimization goal of the algorithm is to make the monitoring range of adjacent sensors overlap by 10% to 30%, while ensuring that the monitoring blind spot area is less than 5% of the total monitoring area. Appropriate overlap can provide data verification and backup, while a smaller blind spot area ensures the continuity and integrity of monitoring.

[0096] S1.4: completing the physical arrangement of the sensor array according to the determined spatial coordinates;

[0097] According to the spatial coordinates determined by optimization, the actual installation of the sensor array is completed.

[0098] S2: Construct a dynamic system model represented by the Koopman operator based on historical multi-source monitoring data;

[0099] In the microseismic monitoring system, building an accurate dynamic model is the key to achieving early warning. Based on the monitoring data, this application uses the Koopman operator theory to build a dynamic model of the rock system. Unlike the traditional modeling method based on physical equations, the Koopman operator method can convert nonlinear systems into linear forms. This conversion makes complex rock dynamic behavior easier to analyze and predict.

[0100] Exemplarily, S2 specifically includes:

[0101] S2.1: Using the extended dynamic mode decomposition method, a basis function space including polynomial basis and radial basis function is selected to construct the initial Koopman eigenfunction;

[0102] First, select a suitable basis function space, which includes two types: polynomial basis functions and radial basis functions. Polynomial basis functions are used to capture the overall change trend of the system and can express the smooth change of the system state. Radial basis functions are specifically used to characterize local characteristics and mutation behaviors, and are particularly suitable for describing small changes before rockbursts. Through the combination of these two types of basis functions, the initial Koopman eigenfunction can be constructed, laying the foundation for the subsequent eigenvalue decomposition.

[0103] S2.2: Perform eigenvalue decomposition on the initial Koopman eigenfunction, obtain the optimized Koopman eigenfunction and eigenvalue, and construct a Koopman spectrum analysis framework;

[0104] The core of eigenvalue decomposition of the initially constructed Koopman eigenfunction is to extract the eigenfunction and eigenvalue that can best describe the dynamic characteristics of the system through numerical calculation methods. The optimized Koopman eigenfunction has a stronger physical meaning and can reflect the essential characteristics of the system. At the same time, through the distribution characteristics of the eigenvalues, the main dynamic modes of the system can be identified and a complete Koopman spectrum analysis framework can be constructed.

[0105] In addition, the framework can not only describe the dynamic characteristics of the system, but also identify the characteristic time scale of the system. Based on this analysis framework, the appropriate time arrival boundary can be determined and used as a warning time window to limit the time range of subsequent state estimation and prediction.

[0106] S2.3: Determine the time arrival boundary based on the Koopman spectrum analysis framework, and use the time arrival boundary as the warning time window to limit the time range of state estimation and prediction;

[0107] The time arrival boundary defines the effective prediction time range of the early warning system, which is determined based on the decay characteristics of the eigenvalue and the characteristic time scale of the system. Through rigorous mathematical analysis, it can be guaranteed that the prediction within this time window has reliable accuracy. The determination of this time boundary is crucial for the practical application of the early warning system, which must ensure sufficient early warning lead time and the reliability of the prediction results.

[0108] S3: Combine the dynamic system model with real-time multi-source monitoring data and perform state estimation through an improved Kalman filter framework;

[0109] like Figure 3 As shown, in order to achieve accurate estimation of the state of the rock system, the present application combines the constructed dynamic system model with multi-source monitoring data acquired in real time and processes them through an improved Kalman filter framework.

[0110] Exemplarily, S3 specifically includes:

[0111] S3.1: Based on the spectral characteristics of the Koopman operator, an improved Kalman filter state observer is constructed;

[0112] This application realizes state estimation by constructing an improved Kalman filter framework, which is a multi-level, progressive process. Specifically, based on the spectral characteristics of the Koopman operator, constructing an improved Kalman filter state observer includes:

[0113] S3.1.1: Establish the state vector expression based on the acquired Koopman eigenfunction;

[0114] Based on the acquired Koopman eigenfunction, a state vector expression is established, which maps the system state to a high-dimensional observation space, so that the behavior of the nonlinear system can be described by linear equations. The state vector contains all key monitoring variables, such as stress, strain, energy and other physical quantities, as well as their time derivatives.

[0115] S3.1.2: Based on the spectral characteristics of the Koopman operator and the state vector expression, construct a state transfer matrix to describe the evolution of the system state over time;

[0116] Based on the spectral characteristics of the Koopman operator, a state transfer matrix is ​​constructed. This matrix describes how the system state evolves over time, and contains information about fast dynamic processes (such as microseismic events) and slow dynamic processes (such as stress accumulation). Through spectral analysis methods, the multi-scale dynamic characteristics of the system can be accurately captured.

[0117] S3.1.3: Determine the observation matrix based on the sensor configuration and state transfer matrix to establish a mapping relationship between observation values ​​and state variables;

[0118] The determined measurement matrix establishes the mapping relationship between sensor measurements and system state variables. Considering the spatial distribution and measurement characteristics of sensors, the measurement matrix needs to reflect the measurement mechanism of different types of sensors (acceleration, velocity, strain, etc.).

[0119] S3.1.4: Define the basic structure of the Kalman filter through the state vector expression, state transfer matrix and observation matrix;

[0120] S3.1.5: Construct an adaptive covariance matrix update strategy to simultaneously adjust the system noise covariance matrix and the observation noise covariance matrix of the Kalman filter through an innovative sequence to achieve dynamic optimization of state estimation;

[0121] After defining the basic structure, by analyzing the statistical characteristics of the innovation sequence (observation residuals), the covariance matrix of the process noise and the observation noise can be adjusted simultaneously, so that the Kalman filter can adapt to the dynamic changes of the environment and improve the accuracy of state estimation.

[0122] S3.2: Use a sliding time window to perform online state estimation, where the length of the sliding time window is 3 to 5 times the characteristic time scale;

[0123] The choice of window length is closely related to the characteristic time scale of the system and is usually taken as 3 to 5 times the characteristic time scale.

[0124] The selection of window length is based on the characteristic time scale of the system. A longer time window is conducive to capturing the long-term change trend of the system, while a shorter time window is conducive to quickly responding to emergencies. In this application, the length of the sliding time window is 3 to 5 times the characteristic time scale, which can not only ensure the stability of the estimation, but also reflect the changes in the system state in a timely manner. In addition, the length of the sliding time window can also be dynamically adjusted according to the stability of the estimation results to achieve a balance between accuracy and real-time performance.

[0125] S4: Based on the state estimation results, a multi-scale prediction model is established using the Koopman spectrum analysis method for prediction;

[0126] Exemplarily, S4 specifically includes:

[0127] S4.1: Based on the state estimation results, the fast mode and slow mode are identified using the Koopman spectrum analysis method;

[0128] Based on the results of state estimation, this application uses the Koopman spectrum analysis method to identify the fast and slow modes in the system. The fast mode usually corresponds to the transient response of the rock mass, such as the sudden process of microseismic events; while the slow mode reflects the gradual process of the rock mass, such as stress accumulation and energy release.

[0129] S4.2: Construct a multi-scale prediction model including fast and slow modes;

[0130] After obtaining these modal information, a multi-scale prediction model including fast mode and slow mode is constructed. This model can simultaneously capture the system behavior on different time scales and improve the comprehensiveness and accuracy of the prediction.

[0131] S4.3: Use the multi-scale prediction model to predict the frequency of microseismic events, vibration energy level and Koopman spectrum characteristic indicators to obtain prediction results;

[0132] Based on the constructed multi-scale prediction model, three key prediction indicators are calculated, among which:

[0133] The frequency of microseismic events is calculated using a method based on time series analysis. The frequency is determined by counting the number of times the amplitude of the microseismic signal exceeds a specific threshold per unit time. During the calculation process, an adaptive threshold strategy is used, and the threshold is dynamically adjusted according to the background noise level to improve the accuracy of event identification.

[0134] The vibration energy level is calculated based on the velocity and acceleration data collected by multi-source sensors. First, the raw data is bandpass filtered to remove high-frequency noise and low-frequency drift, and then the energy characteristics of the signal are calculated in the time domain and frequency domain respectively. The time domain characteristics include the root mean square value and the peak factor, and the frequency domain characteristics include the energy spectrum density and the main frequency band energy ratio. Through the weighted combination of these characteristics, a comprehensive vibration energy level index is obtained.

[0135] The Koopman spectrum characteristic index is calculated using the eigenvalues ​​and eigenfunctions in the dynamic model. Specifically, by analyzing the real and imaginary parts of the Koopman eigenvalue, the instability index of the system can be obtained; by calculating the norm change rate of the eigenfunction, the degree to which the system state deviates from the equilibrium point can be measured. These calculations are performed within a predefined time window, and the final characteristic index is obtained through exponential smoothing.

[0136] S5: The prediction results are corrected in real time through the instrumental variable method and the adaptive control method; this application combines the instrumental variable method and the adaptive control method to achieve a significant improvement in prediction accuracy.

[0137] Exemplarily, S5 specifically includes:

[0138] S5.1: Design bias correction mechanisms based on instrumental variables:

[0139] Specifically, the design of a bias correction mechanism based on instrumental variables includes:

[0140] S5.1.1: construct an instrumental variable matrix, the absolute value of the correlation coefficient between the instrumental variable matrix and the measurement noise is less than 0.1, which can effectively reduce the impact of the measurement noise on the estimation result, thus ensuring the unbiasedness of the estimation result;

[0141] S5.1.2: Construct an unbiased estimator based on the instrumental variable matrix and calculate the model parameter estimates of the multi-scale prediction model;

[0142] S5.1.3: An exponential weighting strategy is used to update the weight coefficients of the model parameters. The weight coefficients decay exponentially over time, that is, new observation data have a larger weight, while the weight of historical data decays exponentially over time, thereby highlighting the importance of recent data, being able to quickly adapt to the dynamic changes of the rock mass state, while maintaining the necessary historical information, and improving the model's ability to adapt to the dynamic changes of the system;

[0143] S5.1.4: Establish a layered protection mechanism including a data quality inspection layer, a model diagnosis layer, and a correction control layer. The data quality inspection layer performs outlier detection and data preprocessing, the model diagnosis layer evaluates model performance and prediction reliability in real time, and the correction control layer dynamically adjusts the correction strategy based on the results of the first two layers. This layered protection mechanism can ensure the accuracy of prediction results from multiple levels.

[0144] S5.1.5: Perform online parameter estimation based on a sliding time window, and dynamically adjust the window length according to the estimation accuracy, wherein the adjustment basis of the window length is the convergence speed and stability index of the parameter estimation, which not only ensures the real-time nature of the estimation, but also maintains the necessary stability.

[0145] S5.2: Build a model reference adaptive controller:

[0146] Specifically, building a model reference adaptive controller includes:

[0147] S5.2.1: Construct a reference model based on ideal dynamic characteristics, wherein the reference model includes an ideal system matrix and an input matrix, which is used to describe the expected dynamic response characteristics of the system and represents the dynamic response characteristics of the system under optimal conditions.

[0148] S5.2.2: Based on the state equations of the reference model and the state equations of the multi-scale prediction model, a parameter adaptation law that satisfies the Lyapunov stability condition is constructed. This parameter adaptation law ensures the stability of the control system and can adapt to the slow changes of system parameters.

[0149] S5.2.3: According to the parameter adaptation law and the state of the multi-scale prediction model, the adaptive gain matrix and the positive definite matrix are constructed. The design of these matrices directly affects the dynamic performance and stability of the system.

[0150] S5.3: Construct a composite predictor that includes a nominal prediction model, an instrumental variable bias estimator, and a model reference adaptive controller; the nominal prediction model provides a baseline prediction value, the instrumental variable bias estimator is built based on the aforementioned instrumental variable bias correction mechanism to eliminate system bias, and the model reference adaptive controller is responsible for compensating for the dynamic error of the system. This composite predictor can handle both the deterministic part and the random disturbance part of the system, significantly improving the accuracy of the prediction.

[0151] S5.4: Based on the composite predictor, the tracking error is calculated, and the prediction result is corrected in real time according to the tracking error.

[0152] S6: comparing the microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction results with the corresponding warning thresholds, and outputting warning information of different levels according to the comparison results;

[0153] Exemplarily, S6 specifically includes:

[0154] S6.1: Set a multi-level warning threshold, where the frequency of microseismic events, the vibration energy level and the Koopman spectrum characteristic index each have four levels of thresholds;

[0155] This multi-indicator, multi-level early warning mechanism can comprehensively assess the risk of rockburst. Specifically, the thresholds set for each monitoring indicator are determined based on a combination of historical data statistics and expert experience, and take into account the response characteristics of the rock mass under different geological conditions.

[0156] S6.2: Continuously compare the real-time corrected prediction results with the preset thresholds, and determine the risk level based on the comparison results.

[0157] When all indicators are below the first-level threshold, it is judged as a low risk level, indicating that the rock mass is in a stable state, and only routine monitoring is required;

[0158] When any indicator exceeds the first-level threshold but none exceeds the second-level threshold, it is judged as a medium risk level. In this case, it is necessary to increase the monitoring frequency and pay special attention to key areas;

[0159] When any indicator exceeds the second-level threshold but none exceeds the third-level threshold, it is judged as a high-risk level. At this time, the emergency plan needs to be activated and corresponding preventive measures need to be taken, such as strengthening support and restricting personnel entry;

[0160] When any indicator exceeds the third-level threshold, it is judged as an extremely high risk level, indicating that the possibility of a rock burst is very high and emergency risk avoidance measures need to be taken immediately. If necessary, work activities in the relevant area should be suspended.

[0161] S6.3: Generate early warning information including risk assessment and disposal suggestions, and send the early warning information to the preset receiving terminal.

[0162] The generation of early warning information adopts a combination of templates and intelligence. For different risk levels, early warning information can be automatically generated, including risk assessment results, current indicator values, forecast trend analysis, and corresponding disposal suggestions. Disposal suggestions include engineering and technical measures, safety management measures, and emergency response measures, which are convenient for on-site personnel to quickly understand and implement. Early warning information is sent to the preset receiving terminal through multiple channels, including but not limited to the monitoring center display, mobile phone text messages, mobile terminal APP, etc., to ensure that relevant personnel can obtain early warning information in a timely manner. At the same time, complete early warning records can also be saved for later analysis and system optimization.

[0163] In addition, if Figure 4, S6 may also include: S6.4: upgrade the warning according to the first condition, and downgrade the warning according to the second condition, the first condition is that the indicator exceeds the higher level threshold for the first time, and the second condition is triggered when the indicator is lower than the current level threshold for the second time. The warning level is updated by adopting a dynamic adjustment mechanism to avoid false alarms or missed alarms caused by instantaneous fluctuations. When the indicator value exceeds the higher level threshold and lasts for the first time, the warning upgrade is triggered; when the indicator value is lower than the current level threshold and lasts for the second time, the warning downgrade is triggered. The first time and the second time here can be set according to the actual situation of the specific project. Usually the first time is shorter than the second time. This setting embodies the principle of "fast rise and slow fall", which is conducive to the stability and reliability of the warning.

[0164] Taking a deep metal mine as an example, the actual operation process of early warning is explained:

[0165] The mine has a stope at the -800m level where mining is ongoing. The following indicators have been monitored:

[0166] Microseismic event frequency: gradually increased from the normal 5 times / day to 15 times / day (first level threshold: 12 times / day)

[0167] Vibration energy level: From the baseline value of 2000J to 5000J (first level threshold: 4000J)

[0168] Koopman spectrum feature index: increased from 0.3 to 0.45 (first level threshold: 0.4)

[0169] When the vibration energy level exceeds 4000J for the first time, a medium risk warning is triggered. The specific operations are as follows:

[0170] Generate warning information: "The vibration energy level of No. 3 stope has exceeded the first-level warning threshold, the current value is 5000J, and it is recommended to:

[0171] Increase monitoring frequency to once an hour

[0172] Check support system integrity

[0173] Restrict non-essential personnel from entering the area."

[0174] In the following 24 hours, the frequency of microseismic events also exceeded the threshold, maintaining a medium risk warning level.

[0175] After 72 hours, if all indicators return to normal levels and remain so for 48 hours, they will be automatically downgraded to a low-risk level.

[0176] This example demonstrates that this application can detect risks in a timely manner, dynamically adjust warning levels, and provide specific disposal suggestions, achieving accurate warning and proactive prevention.

[0177] like Figure 5 As shown, the embodiment of the present application also provides a device for microseismic monitoring and rock burst warning, including:

[0178] A data acquisition module 10, used to collect multi-source monitoring data through a sensor network, wherein the multi-source monitoring data includes acceleration data, velocity data and strain data;

[0179] A dynamic modeling module 20, for constructing a dynamic system model represented by a Koopman operator based on historical multi-source monitoring data;

[0180] A state estimation module 30, for combining the dynamic system model with real-time multi-source monitoring data to perform state estimation through an improved Kalman filter framework;

[0181] A prediction module 40 is used to establish a multi-scale prediction model based on the state estimation result by using Koopman spectrum analysis to perform prediction, and output a prediction result including microseismic event frequency, vibration energy level and Koopman spectrum characteristic index;

[0182] A result correction module 50, used to correct the prediction results in real time by using the instrumental variable method and the adaptive control method;

[0183] The early warning analysis module 60 is used to compare the microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction results with the corresponding early warning thresholds, and output different levels of early warning information according to the comparison results.

[0184] The present application also provides a computer device, the computer device comprising:

[0185] at least one processor; and,

[0186] a memory communicatively connected to the at least one processor; wherein,

[0187] The memory stores instructions that can be executed 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 above-mentioned microseismic monitoring and rockburst early warning method.

[0188] 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 microseismic monitoring and rockburst early warning method.

[0189] The 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 microseismic monitoring and rockburst early warning method are implemented.

[0190] Through the implementation of the above-mentioned early warning mechanism, this application can not only timely discover potential rock burst risks, but also provide specific technical support and decision-making basis for risk prevention and control. This mechanism fully considers the suddenness and harmfulness of rock burst disasters, and through multi-level early warning, dynamic adjustment and timely response measures, a complete early warning and protection system is constructed, which has important practical significance for ensuring the safe production of underground engineering and mines.

Claims

1. A method for microseismic monitoring and rockburst early warning, characterized in that: include: Deploy a sensor network to collect multi-source monitoring data, wherein the multi-source monitoring data includes acceleration data, velocity data, and strain data; Construct a dynamic system model represented by the Koopman operator based on historical multi-source monitoring data; The dynamic system model is combined with real-time multi-source monitoring data to perform state estimation through an improved Kalman filter framework; Based on the state estimation results, a multi-scale prediction model is established using the Koopman spectrum analysis method to perform prediction, and a prediction result including microseismic event frequency, vibration energy level and Koopman spectrum characteristic index is output; Through the instrumental variable method and adaptive control method, the forecast results are corrected in real time; The microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction result are respectively compared with the corresponding warning thresholds, and different levels of warning information are output according to the comparison results.

2. The method according to claim 1, characterized in that The deployment of the sensor network comprises: Divide the monitoring sub-areas according to the geological structural characteristics of the monitoring area; Based on the sound wave propagation path and attenuation characteristics, calculate the minimum number of sensors required for each monitoring sub-area; The spatial coordinates of each sensor are determined by an optimization algorithm, so that the monitoring ranges of adjacent sensors have an overlap of 10% to 30%, and the area of ​​the monitoring blind spot is less than 5% of the total monitoring area; The physical arrangement of the sensor array is completed according to the determined spatial coordinates.

3. The method according to claim 1, characterized in that The dynamic system model represented by the Koopman operator based on historical multi-source monitoring data includes: The extended dynamic mode decomposition method is used to select the basis function space including polynomial basis and radial basis function to construct the initial Koopman eigenfunction; Performing eigenvalue decomposition on the initial Koopman eigenfunction to obtain optimized Koopman eigenfunction and eigenvalue, and constructing a Koopman spectrum analysis framework; The time arrival boundary is determined based on the Koopman spectrum analysis framework, and the time arrival boundary is used as a warning time window to limit the time range of state estimation and prediction.

4. The method according to claim 3, characterized in that The combining of the dynamic system model with real-time multi-source monitoring data to perform state estimation comprises: Based on the spectral characteristics of the Koopman operator, an improved Kalman filter state observer is constructed; Using a sliding time window for online state estimation, the sliding time window length is 3 to 5 times the characteristic time scale; Based on the state estimation result, a multi-scale prediction model is established by using the Koopman spectrum analysis method to perform prediction, including: According to the state estimation result, a fast mode and a slow mode are identified using a Koopman spectrum analysis method; Construct a multi-scale prediction model that includes both fast and slow modes; The multi-scale prediction model is used to predict the frequency of microseismic events, the vibration energy level and the Koopman spectrum characteristic index to obtain prediction results.

5. The method according to claim 4, characterized in that The method of constructing an improved Kalman filter state observer based on the spectral characteristics of the Koopman operator includes: Based on the linear combination of the acquired Koopman eigenfunctions, the state vector expression is established; Based on the spectral characteristics of the Koopman operator and the state vector expression, a state transfer matrix is ​​constructed to describe the evolution of the system state over time; Determining an observation matrix according to the configuration of the sensor and the state transfer matrix, so as to establish a mapping relationship between observation values ​​and state variables; The basic structure of the Kalman filter is defined by the state vector expression, the state transfer matrix and the observation matrix; An adaptive covariance matrix update strategy is constructed to simultaneously adjust the system noise covariance matrix and the observation noise covariance matrix of the Kalman filter through an innovative sequence to achieve dynamic optimization of state estimation.

6. The method according to claim 1, characterized in that The real-time correction of the prediction results by the instrumental variable method and the adaptive control method includes: Design bias correction mechanisms based on instrumental variables that satisfy the condition of zero covariance with measurement noise; Constructing a model-referenced adaptive controller; According to the bias correction mechanism of the instrumental variables and the model reference adaptive controller, the forecast results are corrected in real time.

7. The method according to claim 6, characterized in that The design is based on a bias correction mechanism of instrumental variables, including: Constructing an instrumental variable matrix, wherein the absolute value of the correlation coefficient between the instrumental variable matrix and the measurement noise is less than 0.1; Constructing an unbiased estimator based on the instrumental variable matrix to calculate the estimated values ​​of the model parameters of the multi-scale prediction model; An exponential weighting strategy is used to update the weight coefficients of the model parameters, and the weight coefficients decay exponentially over time; Establish a layered protection mechanism including data quality inspection layer, model diagnosis layer and correction control layer; Online parameter estimation is performed based on a sliding time window, and the window length is dynamically adjusted according to the estimation accuracy, wherein the adjustment basis of the window length is the convergence speed and stability index of the parameter estimation.

8. The method according to claim 7, characterized in that The constructing of the model reference adaptive controller comprises: Constructing a reference model based on ideal dynamic characteristics, wherein the reference model includes an ideal system matrix and an input matrix; According to the state equation of the reference model and the state equation of the multi-scale prediction model, a parameter adaptation law satisfying the Lyapunov stability condition is constructed; According to the parameter adaptation law and the state of the multi-scale prediction model, an adaptive gain matrix and a positive definite matrix are constructed.

9. The method according to claim 8, characterized in that According to the bias correction mechanism of the instrumental variable and the model reference adaptive controller, the forecast results are corrected in real time, including: Constructing a composite predictor including a nominal prediction model, an instrumental variable bias estimator, and a model reference adaptive controller; the instrumental variable bias estimator is constructed based on a bias correction mechanism of the instrumental variable; Based on the composite predictor, a tracking error is calculated, and the prediction result is corrected in real time according to the tracking error.

10. A device for microseismic monitoring and rock burst warning, characterized in that: include: A data acquisition module, used to collect multi-source monitoring data through a sensor network, wherein the multi-source monitoring data includes acceleration data, velocity data and strain data; The dynamic modeling module is used to construct a dynamic system model represented by the Koopman operator based on historical multi-source monitoring data; A state estimation module, used to combine the dynamic system model with real-time multi-source monitoring data and perform state estimation through an improved Kalman filter framework; A prediction module is used to establish a multi-scale prediction model based on the state estimation result by using Koopman spectrum analysis to perform prediction, and output prediction results including microseismic event frequency, vibration energy level and Koopman spectrum characteristic indicators; The result correction module is used to make real-time corrections to the prediction results through the instrumental variable method and the adaptive control method; The early warning analysis module is used to compare the microseismic event frequency, vibration energy level and Koopman spectrum characteristic index in the corrected prediction result with the corresponding early warning threshold value, and output different levels of early warning information according to the comparison results.

Images