Surface anomaly monitoring data fusion system based on vibration signal processing

Abstract

The present invention proposes a surface anomaly monitoring data fusion system based on vibration signal processing, which relates to the field of earthquake precursor monitoring technology. The system includes: a data acquisition module for real-time monitoring of the surface using a distributed multi-parameter sensor network to obtain time-series monitoring data related to earthquake precursors; a signal processing module for processing the time-series monitoring data using a dual-parameter joint optimization algorithm and outputting separated signal components containing precursor information; a data fusion and state inversion module for fusing the separated signal components using a dual-time-scale analysis framework and outputting multi-scale precursor state estimation results and uncertainty quantification indicators; and a monitoring and early warning module for outputting graded earthquake precursor early warning information through a three-layer anomaly recognition architecture and probabilistic risk assessment. The present invention can achieve high-precision separation and identification of earthquake precursor-related signals and reduce false alarms through an intelligent early warning mechanism.

Description

Technical Field

[0001] The present invention relates to the technical field of earthquake precursor monitoring, and in particular to a surface anomaly monitoring data fusion system based on vibration signal processing. Background Art

[0002] Earthquake precursor observations are a crucial foundation for earthquake prediction research. They identify possible earthquake precursors by monitoring abnormal changes in the geophysical field before an earthquake. Earthquake precursor phenomena involve changes in multiple physical parameters, including seismic activity, topographic deformation, groundwater dynamics, the geomagnetic field, and the geoelectric field. These parameters often exhibit complex spatiotemporal evolution during earthquake development. Establishing a high-precision, multi-parameter, comprehensive earthquake precursor observation system is crucial for improving earthquake prediction capabilities and reducing earthquake disasters.

[0003] Current technologies for earthquake precursor observation primarily include single-parameter monitoring methods, such as traditional seismic monitoring networks, geomagnetic observations, geoelectric observations, and deformation observations, as well as the recently developed multi-parameter integrated observation technology. However, existing technologies generally suffer from the following problems: First, single-parameter observations are easily affected by environmental interference, making it difficult to accurately identify true earthquake precursor signals; second, there is a lack of effective fusion mechanisms between multi-parameter observation data, which cannot fully utilize the complementary information between different parameters; third, existing methods have limited ability to separate different physical sources in mixed signals, resulting in low accuracy in precursor signal identification; finally, traditional methods lack adaptive mechanisms, making it difficult to dynamically optimize according to changes in the actual observation environment.

[0004] Chinese invention patent application number CN202111504790.3 discloses a method for screening and analyzing earthquake precursor features. This method acquires seismic electromagnetic disturbance signals through a monitoring network deployment, employs filtering, feature extraction, and anomaly detection techniques to identify earthquake precursor features, and combines the sliding quartile method and isolation forest algorithm for anomaly detection. Although this method has made some progress in seismic electromagnetic disturbance monitoring, it primarily processes a single type of electromagnetic signal and lacks the ability to comprehensively separate mixed signals of multiple physical parameters. Its anomaly detection mechanism is relatively simple and cannot effectively handle complex signal mixing and multi-source interference. Summary of the Invention

[0005] In view of this, the present invention provides a surface anomaly monitoring data fusion system based on vibration signal processing, which can achieve high-precision separation and identification of earthquake precursor-related signals, improve the accuracy and reliability of multi-scale precursor state estimation, and reduce false alarms through an intelligent early warning mechanism.

[0006] The technical solution of the present invention is achieved as follows:

[0007] The present invention provides a surface anomaly monitoring data fusion system based on vibration signal processing, comprising:

[0008] The data acquisition module is used to monitor the ground surface in real time using a distributed multi-parameter sensor network and output time-series monitoring data related to earthquake precursors based on a dual-frequency sampling strategy;

[0009] The signal processing module is used to process the time series monitoring data using a dual-parameter joint optimization algorithm, achieve adaptive wavefield separation by simultaneously optimizing the unmixing matrix and seismic structure model parameters, and output the separated signal components containing precursor information;

[0010] The data fusion and state inversion module is used to fuse the separated signal components using a dual-time-scale analysis framework and output multi-scale precursor state estimation results and uncertainty quantification indicators;

[0011] The monitoring and early warning module is used to analyze the multi-scale precursor state estimation results using an adaptive threshold mechanism and an early warning fatigue suppression algorithm, and output earthquake precursor graded early warning information through a three-layer anomaly recognition architecture and probabilistic risk assessment.

[0012] Preferably, the dual-parameter joint optimization algorithm includes:

[0013] A1. Preprocess and whiten the time series monitoring data to obtain the precursor signal mixing matrix;

[0014] A2, initializing the unmixing matrix and seismic tectonic model parameters;

[0015] A3. Constructing a joint objective function including independence measure and earthquake physics constraints based on the precursor signal mixing matrix;

[0016] A4. Calculate the gradient of the joint objective function with respect to the unmixing matrix and the seismic tectonic model parameters;

[0017] A5. Use adaptive learning rate to update the unmixing matrix and seismic tectonic model parameters simultaneously;

[0018] A6. Determine whether the convergence criterion is met. If so, the iteration ends. If not, return to step A4 and continue the iteration.

[0019] A7. Output the optimized unmixing matrix to separate the precursor signal mixing matrix to obtain separated signal components containing precursor information.

[0020] Preferably, the joint objective function is:

[0021] ,

[0022] in, is the joint objective function, is the precursor signal after separation, W is the unmixing matrix, is the seismic tectonic model parameter vector, is the theoretical mixing matrix based on the earthquake tectonic model, is the independence measure function, is the adaptive regularization function, is the Frobenius norm.

[0023] Preferably, the calculation formula of the adaptive threshold mechanism is:

[0024] ,

[0025] in, is the adaptive threshold at time t, is the basic threshold, is the uncertainty adjustment coefficient, is the variance of the precursor state estimate, is a time-varying adjustment factor.

[0026] Preferably, the early warning fatigue suppression algorithm includes:

[0027] B1. Obtain the current precursor warning judgment results and historical warning records as input;

[0028] B2. Set the statistical time window to collect historical data of precursor warning within the window;

[0029] B3. Count the number of precursor warnings within the window and the corresponding verification results;

[0030] B4. Calculate the recent precursor warning accuracy and warning frequency indicators;

[0031] B5. Calculate the precursor warning suppression factor based on the ratio of the warning frequency to the threshold and the accuracy deviation;

[0032] B6. Multiply the original precursor warning result by the warning suppression factor to obtain the final warning output after suppression.

[0033] Preferably, the calculation formula of the early warning inhibition factor is:

[0034] ,

[0035] in, As an early warning inhibitor, is the number of recent precursor warnings, is the warning number threshold, is the accuracy rate of recent precursor warning, is the target accuracy.

[0036] Preferably, the dual-frequency sampling strategy includes: using standard frequency sampling of 100-200 Hz in normal conditions, using high-frequency sampling of 500-1000 Hz in precursor abnormal conditions, and automatically switching to high-frequency sampling mode when the amplitude change rate of the precursor signal exceeds a preset threshold.

[0037] Preferably, the dual time scale analysis framework includes: a short-term scale using a 1-10 minute time window to capture the rapid changes of precursor signals, a long-term scale using a 1-24 hour time window to identify trend changes in precursor signals, and a dynamic adjustment of fusion weights based on the quality and uncertainty of precursor signals.

[0038] Preferably, the three-layer anomaly recognition architecture includes: the signal layer identifies the type of precursor anomaly event through template matching, the state layer monitors the changes in the physical properties of earthquake precursors based on statistical process control, and the system layer identifies the propagation mode of precursor anomalies through spatial correlation analysis.

[0039] Preferably, the convergence criteria of the dual-parameter joint optimization algorithm include: (1) the relative change of the objective function is less than a set threshold; (2) the amplitude of the parameter vector change is less than the convergence tolerance; (3) the number of iterations reaches the maximum limit value.

[0040] The present invention has the following beneficial effects compared to the prior art:

[0041] (1) The present invention achieves full-process technical integration from data acquisition, signal processing, data fusion to monitoring and early warning by building a complete surface anomaly monitoring data fusion system, effectively solving technical problems in traditional earthquake precursor observation, such as low signal separation accuracy, poor multi-parameter fusion effect, and low early warning accuracy;

[0042] (2) By simultaneously optimizing the unmixing matrix and seismic tectonic model parameters and introducing seismic physical constraints as regularization terms, the algorithm overcomes the problem of unstable separation results caused by the lack of physical constraints in traditional blind source separation methods, thereby improving the accuracy and physical rationality of separation of mixed precursor signals.

[0043] (3) By setting analysis windows at two different time scales, short-term and long-term, and dynamically adjusting the fusion weight according to the signal quality, the framework can simultaneously capture the rapid changes and long-term trends of earthquake precursor signals, solving the problem that single time scale analysis easily misses important precursor information;

[0044] (4) The adaptive threshold mechanism dynamically adjusts the warning threshold by considering the uncertainty and time-varying characteristics of the precursor state estimation, avoiding the problem of poor adaptability of the fixed threshold method under different observation environments and reducing the false alarms and missed alarms caused by environmental changes;

[0045] (5) By calculating the suppression factor based on the historical warning frequency and accuracy, the warning fatigue suppression algorithm effectively suppresses the high-frequency and low-accuracy warning output, solving the problem of warning fatigue that is easy to occur in traditional warning systems;

[0046] (6) The three-layer anomaly recognition architecture realizes multi-level analysis from single-point anomaly detection to spatial pattern recognition through the hierarchical recognition mechanism of signal layer, state layer and system layer, overcoming the limitation that single-layer anomaly recognition is easily affected by local interference. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0048] Figure 1 Schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0050] like Figure 1 As shown, the present invention provides a surface anomaly monitoring data fusion system based on vibration signal processing, comprising:

[0051] The data acquisition module is used to monitor the ground surface in real time using a distributed multi-parameter sensor network and output time-series monitoring data related to earthquake precursors based on a dual-frequency sampling strategy;

[0052] The signal processing module is used to process the time series monitoring data using a dual-parameter joint optimization algorithm, achieve adaptive wavefield separation by simultaneously optimizing the unmixing matrix and seismic structure model parameters, and output the separated signal components containing precursor information;

[0053] The data fusion and state inversion module is used to fuse the separated signal components using a dual-time-scale analysis framework and output multi-scale precursor state estimation results and uncertainty quantification indicators;

[0054] The monitoring and early warning module is used to analyze the multi-scale precursor state estimation results using an adaptive threshold mechanism and an early warning fatigue suppression algorithm, and output earthquake precursor graded early warning information through a three-layer anomaly recognition architecture and probabilistic risk assessment.

[0055] Specifically, in one embodiment of the present invention, the data acquisition module adopts a distributed multi-parameter sensor network architecture, deploys multiple sensor nodes in the target monitoring area, and each node is equipped with a variety of sensor devices such as a triaxial accelerometer, a tilt angle sensor, and a pore water pressure gauge. The triaxial accelerometer is used to monitor surface micro-motion and vibration signals. Its sampling accuracy is required to reach the microgravity level, and it can capture weak surface movement changes caused by earthquake precursors. The tilt angle sensor is responsible for monitoring surface tilt changes. Its measurement accuracy needs to reach the arc second level, and it is used to detect surface tilt anomalies caused by crustal deformation. The pore water pressure gauge is used to monitor groundwater pressure changes. Its pressure measurement accuracy is required to reach the Pascal level, which can reflect the changing characteristics of groundwater dynamic parameters.

[0056] The sensor network layout adheres to the principle of multi-level spatial sampling, with the spatial configuration of sensor nodes determined by the geological and structural characteristics of the monitoring area and the characteristics of seismic activity. Primary sensor nodes are arranged in a regular grid pattern, with node spacing determined by the target monitoring frequency and seismic wavelength characteristics, generally set at one-quarter to one-half the wavelength corresponding to the primary frequency band of earthquake precursor signals. For key monitoring areas, such as those near active faults and areas with historical earthquake concentrations, the sensor density should be more than doubled to form a denser observation network. The vertical layout of sensor nodes takes into account topographical undulations and geological stratification. In mountainous and hilly areas, sensors are distributed along elevation gradients, with elevation spacing determined by the terrain slope and geological complexity. Sensors in plain areas primarily consider groundwater level fluctuations and the distribution of soft soil layers. Sensor burial depths are determined based on local geological exploration data, generally in the soil layer above the stable bedrock surface. Auxiliary sensor nodes are deployed in boundary areas to monitor the propagation characteristics of earthquake precursor signals at the outer edges of the monitoring area. Spacing between auxiliary nodes can be appropriately relaxed, but signal continuity with the primary monitoring network must be ensured. In areas with special geological structures, such as karst development areas and areas with weak strata, the sensor type and layout depth need to be adjusted according to the specific geological conditions.

[0057] The system adopts a dual-frequency sampling strategy to adaptively sample earthquake precursor related signals. The core of this strategy is to dynamically adjust the sampling frequency according to the signal status. Under normal monitoring conditions, the system adopts a standard sampling frequency. The frequency range of the data acquisition is set to 100-200Hz, which can meet the sampling requirements of conventional earthquake precursor signals. When the system detects that the amplitude change rate of the precursor signal exceeds the preset threshold, it automatically switches to the high-frequency sampling mode. At this time, the sampling frequency Increased to 500-1000Hz. The judgment conditions for sampling mode switching are:

[0058] ,

[0059] in is the current sampled signal amplitude, Indicates the moment, The high-frequency sampling mode can capture the rapid changes of earthquake precursor signals, ensuring that important transient abnormal information is not missed.

[0060] The sensor network utilizes a distributed synchronization acquisition mechanism, with all sensor nodes achieving time synchronization through GPS clock synchronization or network clock synchronization protocols. Synchronization accuracy is required to reach microsecond levels. Each sensor node is equipped with a local data cache capable of storing raw sampled data for a specified period of time. This local cache prevents data loss in the event of network communication anomalies. Data transmission utilizes a combination of wired and wireless methods, with wired transmission used for data backhaul from core nodes and wireless for data collection at edge nodes.

[0061] Specifically, in one embodiment of the present invention, the signal processing module uses a dual-parameter joint optimization algorithm to achieve adaptive wavefield separation. This algorithm extracts the separated signal components containing precursor information by simultaneously optimizing the unmixing matrix and the seismic structure model parameters. The signal processing process is as follows:

[0062] A1. Preprocess and whiten the time series monitoring data to obtain the precursor signal mixing matrix ,in express The observation signal of the sensor channel, where Respectively represent the first, second, and The observation signal of each sensor channel, Represents matrix transpose.

[0063] The whitening transform eliminates the second-order statistical correlation between signals through decorrelation processing. The whitening matrix By performing eigenvalue decomposition on the covariance matrix, we can obtain: ,in is the eigenvalue diagonal matrix, is the eigenvector matrix, Represents matrix transpose. The signal after whitening transformation It has unit covariance characteristics, is the signal after whitening transformation, is the whitening matrix, is the precursor signal mixing matrix.

[0064] A2. Initialize the unmixing matrix and seismic tectonic model parameters.

[0065] The algorithm initialization stage requires setting the unmixing matrix and the seismotectonic model parameter vector The initial value of the unmixing matrix Initialized to the identity matrix or the initial estimate obtained by principal component analysis, the seismic tectonic model parameters Initialization is performed based on prior geological information, including physical parameters such as stratum density, elastic modulus, and medium attenuation coefficient. Seismic tectonic model parameters are used to construct a theoretical mixing matrix. , which describes the theoretical propagation characteristics of signals from different physical sources in the sensor array. The construction of the theoretical mixing matrix is ​​based on the physical mechanism of seismic wave propagation and takes into account the influence of factors such as stratum structure, medium parameters, and geometric attenuation on signal propagation.

[0066] A3. Based on the precursor signal mixing matrix, a joint objective function including independence measure and earthquake physical constraints is constructed. The objective function expression is:

[0067] ,

[0068] in, is the joint objective function, is the precursor signal after separation, is the unmixing matrix, is the seismic tectonic model parameter vector, is the theoretical mixing matrix based on the earthquake tectonic model, is the independence measure function, is the adaptive regularization function, is the Frobenius norm. Independence measure function Negative entropy or mutual information is used as a metric to evaluate the statistical independence of separated signals. Dynamic adjustment is made based on the degree of deviation between the model parameters and the prior reference values. The expression is:

[0069] ,

[0070] in is the adaptive regularization function, is the basic regularization coefficient, To adjust the parameters, are the reference geological model parameters, is the seismic tectonic model parameter vector, is the L2 norm. When the model parameters are close to the reference values, the regularization strength is high; when the parameters deviate from the reference values, the regularization strength gradually decreases, allowing the algorithm to adjust the model parameters based on the observed data.

[0071] A4. Calculate the gradient of the joint objective function with respect to the unmixing matrix and the seismic tectonic model parameters.

[0072] Gradient calculation is the core step of the optimization algorithm, which requires calculating the partial derivatives of the objective function with respect to the unmixing matrix and model parameters. The gradient of the seismic tectonic model parameters is:

[0073] ,

[0074] The gradient of the unmixing matrix is:

[0075] ,

[0076] in, Represents the parameters of the earthquake tectonic model The gradient operator, Represents the unmixing matrix The gradient operator, is the joint objective function, is the unmixing matrix, is the original observed mixed signal vector, is the independence measure function, is the theoretical mixing matrix based on the earthquake tectonic model, represents the matrix transpose, is the adaptive regularization function, is the Frobenius norm. The gradient calculation involves the derivative of the theoretical mixing matrix with respect to the model parameters , the derivative is obtained by numerical differentiation or analytical derivation method, which reflects the influence of the change of geological model parameters on the signal propagation characteristics.

[0077] A5. Adaptive learning rate is used to simultaneously update the unmixing matrix and seismic tectonic model parameters.

[0078] The parameter update adopts the gradient descent method with adaptive learning rate, and updates the unmixing matrix and seismic structure model parameters at the same time. The update formula is:

[0079] ,

[0080] ,

[0081] in, is the parameter vector of the earthquake tectonic model after the k+1th iteration, is the seismic tectonic model parameter vector after the kth iteration, is the joint objective function Parameters The gradient, is the unmixing matrix, is the unmixing matrix after the k+1th iteration, is the unmixing matrix after the kth iteration, is the joint objective function Unmixing Matrix The gradient, and The adaptive learning rate dynamically adjusts based on the gradient magnitude and iteration history. This adaptive learning rate utilizes a variant of the AdaGrad or Adam optimizer, automatically adjusting the step size based on historical parameter update information to improve the convergence stability of the algorithm. To ensure the numerical stability of the unmixing matrix, the unmixing matrix must be orthogonalized after each update to ensure it meets reversibility requirements.

[0082] A6. Determine whether the convergence criterion is met. If so, end the iteration. If not, return to step A4 and continue the iteration.

[0083] The convergence criterion includes three conditions: the relative change of the objective function is less than the set threshold, the change amplitude of the parameter vector is less than the convergence tolerance, and the number of iterations reaches the maximum limit. The relative change criterion of the objective function is: ,in is the relative change threshold, is the joint objective function after the k+1th iteration, is the joint objective function after the kth iteration. The criterion for the parameter vector change amplitude is: and ,in and is the convergence tolerance, is the parameter vector of the earthquake tectonic model after the k+1th iteration, is the seismic tectonic model parameter vector after the kth iteration, is the unmixing matrix after the k+1th iteration, is the unmixing matrix after the kth iteration, is the L2 norm, is the Frobenius norm. When any convergence condition is met, the algorithm terminates the iteration and outputs the optimized unmixing matrix.

[0084] A7. Output the optimized unmixing matrix to separate the precursor signal mixing matrix to obtain separated signal components containing precursor information.

[0085] In the output stage of the algorithm, the optimized unmixing matrix is ​​used to separate the precursor signal mixing matrix to obtain the separated signal components containing the precursor information. Each separated signal component corresponds to a specific physical process, such as surface deformation caused by earthquake precursors, changes in groundwater dynamics, and rock mass stress adjustments. The separated signals are then physically interpreted and quality assessed to identify signal components highly correlated with earthquake precursors. The algorithm also outputs estimates of seismic tectonic model parameters, which reflect the geological structural characteristics and physical properties of the monitored area.

[0086] Specifically, in one embodiment of the present invention, the data fusion and state inversion module adopts a dual-time-scale analysis framework to fuse the separated signal components output by the signal processing module. The framework realizes multi-scale precursor state estimation by setting two analysis windows of different time scales, short-term and long-term. The short-term scale adopts a time window of 1-10 minutes, which is mainly used to capture the rapid change process of earthquake precursor signals and can identify sudden precursor anomaly events and transient signal characteristics. The long-term scale adopts a time window of 1-24 hours to identify the trend change of earthquake precursor signals and can capture the slowly accumulated precursor process and long-term evolution pattern. The analysis windows of the two time scales adopt a sliding window mechanism to ensure the continuity and real-time performance of state estimation.

[0087] Short-term scale analysis uses sliding window statistical methods to extract features and estimate states of separated signal components. Signal components within , Indicates the moment, represents the value of the kth separated signal component at time t, is the short-term time window length, is the time window for short-term analysis. Calculate the statistical eigenvector , including signal amplitude mean, variance, skewness, kurtosis and other statistics. State estimation is achieved through the Bayesian filtering framework, and the state transfer equation is:

[0088] ,

[0089] in is the short-term state vector at time t, is the state transition matrix, is the short-term state vector at time t-1, is the process noise at time t. The observation equation is:

[0090] ,

[0091] in is the statistical eigenvector at time t, is the observation matrix, is the short-term state vector at time t, is the observation noise at time t. The state estimate and covariance matrix are recursively updated by the Kalman filter algorithm to obtain the short-term precursor state estimate and uncertainty quantification .

[0092] Long-term scale analysis uses trend analysis and period decomposition methods to identify the long-term evolution characteristics of precursor signals. The signal components within is the length of the long-term time window, represents the time window for long-term analysis, Represents the moment. Empirical mode decomposition or wavelet transform is used to extract trend components and periodic components. Trend analysis fits the long-term trend of the signal through linear regression or non-parametric regression methods. The trend estimation model is:

[0093] ,

[0094] in, is the trend component value at time t, is the trend parameter, is the residual term. Periodic analysis uses spectrum analysis to identify the periodic components in the signal and extract the main periodic frequency and corresponding amplitude information. Long-term state estimation combines the trend and periodic analysis results to construct the long-term state vector , updating the long-term state estimate through a similar Bayesian filtering framework and its uncertainty .

[0095] The fusion weights are dynamically adjusted according to the quality and uncertainty of the precursor signal. The weight calculation is based on the signal quality index and the confidence of the state estimate. The fusion weights of the short-term and long-term state estimates are:

[0096] ,

[0097] ,

[0098] in, is the fusion weight of the short-term state estimation at time t, is the fusion weight of the long-term state estimation at time t, and are the short-term and long-term signal quality indicators at time t, is the matrix trace operation, and are the uncertainty covariance matrices of the short-term state estimate and the long-term state estimate at time t, respectively. The signal quality index is calculated by comprehensively considering factors such as signal amplitude, signal-to-noise ratio, spectral purity, and time-domain stability. When signal quality is high and uncertainty is low, the corresponding weight increases; conversely, the weight decreases.

[0099] The multi-scale precursor state estimation is obtained by weighted fusion of short-term and long-term state estimates:

[0100] ,

[0101] is the fusion precursor state estimation vector at time t, is the fusion weight of the short-term state estimation at time t, is the fusion weight of the long-term state estimation at time t, is the short-term precursor state estimation vector at time t, is the long-term precursor state estimation vector at time t.

[0102] The uncertainty quantification of the fused state estimate takes into account the uncertainty and correlation of the estimates at two time scales:

[0103] ,

[0104] in, is the uncertainty covariance matrix of the fused state estimate at time t, is the fusion weight of the short-term state estimation at time t, is the fusion weight of the long-term state estimation at time t, and are the uncertainty covariance matrix of the short-term state estimate at time t and the uncertainty covariance matrix of the long-term state estimate, is the covariance matrix between the short-term and long-term state estimates. Uncertainty quantification indicators include the standard deviation, confidence interval and prediction variance of the state estimate.

[0105] The module also implements state inversion, which uses reverse inference to estimate possible physical source parameters from the observed precursor state. State inversion is based on the physical mechanism model of earthquake precursors and establishes a mapping relationship between state estimates and physical source parameters. The inversion process uses an optimization algorithm to solve the nonlinear inversion problem:

[0106] ,

[0107] in is the physical source parameter vector, is the fusion precursor state estimation vector at time t, is the forward physical model, is the regularization term, is the regularization coefficient. The inversion results provide a physical explanation of earthquake precursors, including key information such as the possible source location, intensity, and occurrence time.

[0108] The output results include multi-scale precursor state estimation vectors, corresponding uncertainty quantification indicators, fusion weight sequences and state inversion parameters.

[0109] Specifically, in one embodiment of the present invention, the monitoring and early warning module uses an adaptive threshold mechanism to analyze and process the multi-scale precursor state estimation results output by the data fusion and state inversion module. This mechanism dynamically adjusts the early warning threshold to adapt to the changing characteristics of the precursor signal under different time and environmental conditions. The calculation formula for the adaptive threshold is:

[0110] ,

[0111] in, is the adaptive threshold at time t, is the basic threshold, is the uncertainty adjustment coefficient, is the variance of the precursor state estimate, is a time-varying adjustment factor.

[0112] The time-varying adjustment factor is obtained through historical model learning, and its expression is:

[0113] ,

[0114] Where t represents the time, is the time-varying adjustment factor, is a periodic basis function, including time patterns such as annual cycle, seasonal cycle, and daily cycle. is the total number of periodic basis functions; is the corresponding weight coefficient, and the historical abnormal event data is fitted and estimated by the least squares method: ,in is the basis function matrix, is the historical abnormal event vector, Represents the matrix transpose; the sum of the weights is 1. To avoid system instability caused by frequent threshold adjustments, an exponentially weighted moving average is used to smoothly update the threshold: ;in is the smoothing parameter, is the original calculation threshold value updated for the k+1th time, that is, the current new calculation value, is the adaptive threshold after the k+1th update, is the adaptive threshold for the kth update, that is, the historical smoothing value.

[0115] The warning fatigue suppression algorithm dynamically adjusts the warning sensitivity by statistically analyzing the historical warning frequency and accuracy, effectively solving the warning fatigue problem caused by frequent false alarms. The warning fatigue suppression algorithm includes:

[0116] B1. Obtain the current precursor warning judgment results and historical warning records as input;

[0117] B2. Set the statistical time window to collect historical data of precursor warning within the window;

[0118] B3. Count the number of precursor warnings within the window and the corresponding verification results;

[0119] B4. Calculate the recent precursor warning accuracy and warning frequency indicators;

[0120] B5. Calculate the precursor warning suppression factor based on the ratio of the warning frequency to the threshold and the accuracy deviation;

[0121] B6. Multiply the original precursor warning result by the warning suppression factor to obtain the final warning output after suppression.

[0122] The calculation formula of the early warning inhibition factor is:

[0123] ,

[0124] in, As an early warning inhibitor, is the number of recent precursor warnings, is the warning number threshold, is the accuracy rate of recent precursor warning, which is calculated by counting the verification results of warning events. The target accuracy is used as a benchmark for evaluating the quality of warnings.

[0125] The final warning output is obtained by multiplying the original precursor warning result with the warning suppression factor: , is the minimum warning threshold, For the final warning output, is the original precursor warning result, As an early warning inhibitor, Represents the maximum value function; the warning suppression factor sets a lower limit value to ensure that the system still maintains basic warning function when the suppression mechanism is activated.

[0126] The three-layer anomaly recognition architecture implements multi-level anomaly detection and identification from the signal layer to the system layer. The signal layer identifies the type of precursor anomaly event using template matching. This layer maintains a library of templates for various typical precursor signals, including signal feature templates for earthquake precursors, artificial interference, and environmental noise. Template matching uses correlation coefficients or normalized cross-correlation functions to calculate the similarity between the observed signal and the template signal. Signals exceeding a set threshold are identified as an anomaly of the corresponding type. The state layer monitors changes in the physical properties of earthquake precursors using statistical process control methods. This layer performs statistical analysis on the precursor state estimation results and calculates control chart statistics such as mean, variance, and trend slope. When the statistics exceed control limits, the state is considered an anomaly. The control limits are set based on the statistical distribution characteristics of the parameters under normal conditions and the system's false alarm rate requirements. The system layer identifies the propagation pattern of precursor anomalies through spatial correlation analysis. This layer analyzes the spatial correlation characteristics of precursor signals between multiple sensor nodes and uses correlation matrices and cluster analysis to identify the spatial propagation path and impact range of the anomaly signal.

[0127] The probabilistic risk assessment model establishes a hierarchical risk assessment system and calculates the comprehensive risk level through a multi-indicator weighted scoring method. The risk assessment is based on four core indicators: the signal amplitude change rate reflects the degree of change in the precursor signal strength; the frequency domain characteristic deviation measures the degree of deviation of the signal spectrum characteristics from the normal state; the spatial consistency index evaluates the spatial consistency of the precursor signals at multiple monitoring points; and the historical pattern matching degree evaluates the danger level of the current anomaly through the pattern similarity with historical earthquake precursor events. The comprehensive risk index is calculated through a linear weighted combination:

[0128] ,

[0129] in, is the comprehensive risk index, is the signal amplitude change rate, is the frequency domain characteristic deviation, is the spatial consistency index, is the historical pattern matching degree, The weight coefficients for each indicator are determined based on historical data statistics and expert experience, and can be optimized and adjusted through machine learning methods. The sum of the weights is 1. Risk levels are graded based on the comprehensive risk index and are set into four levels: low risk, medium risk, high risk, and extremely high risk. Each level corresponds to different early warning response strategies and disposal measures.

[0130] The system implements a hierarchical and progressive early warning strategy, including three levels: status warning, event warning, and trend warning. Status warning targets abnormal fluctuations in precursor state parameters. It is triggered when the state estimate exceeds the normal range. It is mainly used to remind monitoring personnel to pay attention to changes in system status. Event warning targets detected abnormal events. It is triggered when the three-layer anomaly recognition architecture identifies a clear precursor abnormal event. The warning information includes key information such as event type, occurrence time, and impact range. Trend warning targets long-term trend changes. It is triggered when the system detects a persistent abnormal trend that conforms to the earthquake precursor evolution pattern. The warning information includes the trend development direction, possible evolution results, and recommended response measures. The warning output uses a standardized format and contains complete information such as timestamp, warning level, affected area, and recommended measures.

[0131] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. The surface anomaly monitoring data fusion system based on vibration signal processing is characterized by: include: The data acquisition module is used to monitor the ground surface in real time using a distributed multi-parameter sensor network and output time-series monitoring data related to earthquake precursors based on a dual-frequency sampling strategy; The signal processing module is used to process the time series monitoring data using a dual-parameter joint optimization algorithm, achieve adaptive wavefield separation by simultaneously optimizing the unmixing matrix and seismic structure model parameters, and output the separated signal components containing precursor information; The data fusion and state inversion module is used to fuse the separated signal components using a dual-time-scale analysis framework and output multi-scale precursor state estimation results and uncertainty quantification indicators; The monitoring and early warning module is used to analyze the multi-scale precursor state estimation results using an adaptive threshold mechanism and an early warning fatigue suppression algorithm, and output earthquake precursor graded early warning information through a three-layer anomaly recognition architecture and probabilistic risk assessment.

2. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The dual-parameter joint optimization algorithm includes: A1. Preprocess and whiten the time series monitoring data to obtain the precursor signal mixing matrix; A2, initializing the unmixing matrix and seismic tectonic model parameters; A3. Constructing a joint objective function including independence measure and earthquake physics constraints based on the precursor signal mixing matrix; A4. Calculate the gradient of the joint objective function with respect to the unmixing matrix and the seismic tectonic model parameters; A5. Use adaptive learning rate to update the unmixing matrix and seismic tectonic model parameters simultaneously; A6. Determine whether the convergence criterion is met. If so, the iteration ends. If not, return to step A4 and continue the iteration. A7. Output the optimized unmixing matrix to separate the precursor signal mixing matrix to obtain separated signal components containing precursor information.

3. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 2 is characterized in that: The joint objective function is: ， in, is the joint objective function, is the precursor signal after separation, W is the unmixing matrix, is the seismic tectonic model parameter vector, is the theoretical mixing matrix based on the earthquake tectonic model, is the independence measure function, is the adaptive regularization function, is the Frobenius norm.

4. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The calculation formula of the adaptive threshold mechanism is: ， in, is the adaptive threshold at time t, is the basic threshold, is the uncertainty adjustment coefficient, is the variance of the precursor state estimate, is a time-varying adjustment factor.

5. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The early warning fatigue suppression algorithm includes: B1. Obtain the current precursor warning judgment results and historical warning records as input; B2. Set the statistical time window to collect historical data of precursor warning within the window; B3. Count the number of precursor warnings within the window and the corresponding verification results; B4. Calculate the recent precursor warning accuracy and warning frequency indicators; B5. Calculate the precursor warning suppression factor based on the ratio of the warning frequency to the threshold and the accuracy deviation; B6. Multiply the original precursor warning result by the warning suppression factor to obtain the final warning output after suppression.

6. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 5 is characterized in that: The calculation formula of the early warning inhibition factor is: ， in, As an early warning inhibitor, is the number of recent precursor warnings, is the warning number threshold, is the accuracy rate of recent precursor warning, is the target accuracy.

7. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The dual-frequency sampling strategy includes: standard frequency sampling of 100-200Hz under normal conditions, high-frequency sampling of 500-1000Hz under abnormal precursor conditions, and automatic switching to high-frequency sampling mode when the amplitude change rate of the precursor signal exceeds the preset threshold.

8. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The dual-time-scale analysis framework includes: a 1-10 minute time window is used to capture the rapid changes of precursor signals in the short term, and a 1-24 hour time window is used to identify the trend changes of precursor signals in the long term. The fusion weight is dynamically adjusted according to the quality and uncertainty of the precursor signals.

9. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 1 is characterized in that: The three-layer anomaly recognition architecture includes: the signal layer identifies the type of precursor anomaly event through template matching, the state layer monitors the changes in the physical properties of earthquake precursors based on statistical process control, and the system layer identifies the propagation mode of precursor anomalies through spatial correlation analysis.

10. The surface anomaly monitoring data fusion system based on vibration signal processing according to claim 2 is characterized in that: The convergence criteria of the dual-parameter joint optimization algorithm include: (1) the relative change of the objective function is less than the set threshold; (2) the amplitude of the parameter vector change is less than the convergence tolerance; (3) the number of iterations reaches the maximum limit.

Images