A multi-source monitoring data adaptive fusion and parameter correction method and system

Abstract

The application provides a kind of multi-source monitoring data adaptive fusion and parameter correction method and system, it is related to multi-source data collaborative processing technical field, method includes collecting multi-source monitoring data and pre-processing, constructs initial weighted linear basic fusion model, core parameter is divided into azimuth correction parameter and time service correction parameter, based on variance decomposition analysis error sensitive source and set correction priority, for time service parameter, scene is used to adopt travel time stability analysis or background noise cross-correlation superposition algorithm, combined with event trigger mechanism dynamic correction clock difference, cross-validation collaborative verification correction result, system includes data collection and basic fusion model construction, parameter division and priority setting, azimuth parameter adaptive correction, time service parameter adaptive correction and collaborative verification and system integration module, the application improves the fusion precision and reliability of monitoring data.

Description

Technical Field

[0001] This invention relates to the field of multi-source data collaborative processing technology, and in particular to an adaptive fusion and parameter correction method and system for multi-source monitoring data. Background Technology

[0002] With the development of the Internet of Things (IoT) and sensing technology, the fusion processing of multi-source heterogeneous monitoring data has become a core technical means to improve the perception accuracy of the system, and has been widely used in many fields such as disaster early warning, environmental monitoring, and industrial IoT. Earthquake monitoring, as a core task of geophysical observation, directly affects the reliability of earthquake early warning, focal mechanism inversion, and crustal structure imaging based on its data quality. With the dense deployment of monitoring networks, it has become possible to acquire multi-source monitoring data, such as active P-wave data from air guns, teleseismic P-wave data, and background noise data. Data fusion can effectively improve the spatiotemporal resolution and anti-interference capability of monitoring. However, parameter errors introduced during the acquisition, transmission, and processing of multi-source data, such as azimuth deviation and time service asynchrony, have become key bottlenecks restricting the accuracy of fusion. From the perspective of parameter error sources, azimuth parameters are mainly affected by sensor installation deviations at stations, local interference in the geomagnetic field, and multipath effects of wave propagation, leading to distortion in the direction of arrival estimation. Time service parameters are affected by clock hardware drift, network synchronization delays, and external interference, resulting in clock errors and frequency drift, which disrupt the synchronization of event timing. Existing data fusion methods often employ static weighted models, such as fixed-weight linear fusion, which fail to dynamically adjust weights based on data quality. Parameter correction is typically performed in isolation, lacking analysis of error-sensitive sources and prioritization. For example, when fusing active airgun data with teleseismic data, fixed weights cannot adapt to fluctuations in the signal-to-noise ratio, easily introducing fusion biases in low-quality data segments. Furthermore, traditional parameter correction methods often ignore the coupling effect between parameters, resulting in a lack of collaborative verification of correction results and the accumulation of errors in the fusion model. While correction methods based on traditional optimization algorithms, such as least-squares fitting, can handle some parameter errors, they fail to achieve adaptive fusion for the heterogeneous characteristics of multi-source data. For instance, azimuth correction relies on polarization analysis or travel time inversion, but the algorithm selection lacks adaptability to data scenarios such as airgun signal coverage and non-coverage areas, leading to correction failure at remote stations. In time service correction, travel time stability analysis is applicable to airgun data but cannot cover all stations, while background noise cross-correlation methods are greatly affected by environmental noise. Existing systems lack dynamic triggering mechanisms to switch correction strategies. In addition, the verification process after multi-source data fusion is weak, often relying on manual quality control, making real-time error identification and feedback difficult.

[0003] In summary, existing multi-source monitoring data fusion and parameter correction methods have significant shortcomings in terms of adaptive weight adjustment for data quality, collaborative correction of parameter errors, and full-scene coverage. Therefore, there is an urgent need for a new adaptive fusion and parameter correction method and system for multi-source monitoring data. Summary of the Invention

[0004] The purpose of this invention is to provide an adaptive fusion and parameter correction method and system for multi-source monitoring data, in order to solve the problems existing in the prior art, such as the inability of static weighted fusion models to dynamically adjust weights according to data quality, the isolated parameter correction process lacking error-sensitive source analysis and priority setting, the difficulty of traditional correction algorithms adapting to the heterogeneous characteristics of multi-source data leading to insufficient full-scene coverage, and the weak verification of correction results relying on manual quality control, making it difficult to achieve real-time error identification and feedback. The specific technical solution is as follows:

[0005] This invention provides an adaptive fusion and parameter correction method for multi-source monitoring data, comprising:

[0006] Step 1: Collect multi-source monitoring data, preprocess it, and construct a basic fusion model for multi-source monitoring data;

[0007] Step 2: Divide the core parameters to be corrected in the multi-source monitoring data fusion model into azimuth correction parameters and time service correction parameters, analyze the error sensitivity sources of azimuth correction parameters and time service correction parameters, and set correction priorities based on sensitivity analysis;

[0008] Step 3: Embed multiple azimuth estimation algorithms into the multi-source monitoring data fusion model, construct an adaptive azimuth parameter fusion correction framework, estimate azimuth parameters in parallel using multiple algorithms, and introduce an adaptive weight adjustment mechanism for weighted fusion to output the azimuth correction result.

[0009] Step 4: Embed time service parameter estimation algorithms in the multi-source monitoring data fusion model according to different scenarios, construct an adaptive fusion correction framework for time service parameters, adopt differentiated correction strategies for stations covered by the active source signal of the air gun and stations not covered, introduce an event triggering mechanism to dynamically activate the correction process, and output the time service correction results.

[0010] Step 5: Perform collaborative verification based on the azimuth correction results and time service correction results, and feed the correction parameters back to the multi-source monitoring data fusion model to update the model parameters.

[0011] Further, the multi-source monitoring data mentioned in step 1 includes air gun active source P-wave data, teleseismic P-wave data, and background noise data; the preprocessing includes quality control, outlier removal, and standardization of the original multi-source monitoring data; the basic fusion model of the multi-source monitoring data is a weighted linear fusion model, the output of which is the output of the basic fusion model. The data sources include air gun active source P-wave data, teleseismic P-wave data, and background noise data. The initial weight coefficients are set based on the observation accuracy of each data source and satisfy the weight and constraint of 1.

[0012] Furthermore, the quality control in step 1 includes using a screening mechanism based on the signal-to-noise ratio threshold to remove data segments with a signal-to-noise ratio below 3dB for the active source P-wave data of the air gun, identifying and removing records severely affected by local interference through polarization analysis for the teleseismic P-wave data, and removing outliers for the background noise data using the probability density function estimation method; the standardization process normalizes each source data to the same numerical range.

[0013] Further, the azimuth correction parameter mentioned in step 2 is defined as the deflection angle deviation between the station sensor coordinate system and the geographic coordinate system. Its error sensitive sources include sensor installation deviation, local geomagnetic field interference, and multipath effect. The time service correction parameter is defined as the deviation between the station clock and standard time, including clock error and frequency drift. Its error sensitive sources include clock hardware drift, network synchronization delay, and external interference. The sensitivity analysis adopts the variance decomposition method to calculate the partial derivatives of the azimuth parameter change and the time service parameter change with the variance of the fusion model output. The priority determination rule is that if the sensitivity of the azimuth parameter is higher than that of the time service parameter, the azimuth correction takes priority, and vice versa.

[0014] Furthermore, the various azimuth estimation algorithms mentioned in step 3 include a polarization characteristic extraction algorithm, a particle motion analysis algorithm, and a signal-to-noise ratio weighted multi-event inversion algorithm. The polarization characteristic extraction algorithm calculates the principal direction of the polarization ellipsoid based on the vector waveform of the P-wave data from the active air gun source or the remote earthquake P-wave data, obtains the azimuth estimate, and evaluates the degree of polarization to determine the reliability of the estimate. The particle motion analysis algorithm extracts the motion direction through principal component analysis based on the particle motion trajectory of the background noise data or P-wave data, obtains the azimuth estimate, and calculates the confidence level. The signal-to-noise ratio weighted multi-event inversion algorithm estimates the azimuth based on the travel time inversion of multiple remote earthquake events, and its weight is determined by the event signal-to-noise ratio. The adaptive weight adjustment mechanism is to weight and fuse the estimates of multiple algorithms, and define the azimuth correction result after fusion as the weighted sum of the estimates of each algorithm. The weights are calculated based on the real-time confidence index and historical error variance of each algorithm through an objective function that minimizes the error. The weights are updated once every time window to adapt to changes in data quality.

[0015] Furthermore, the scenario-specific embedded time service parameter estimation algorithm in step 4 includes: a station embedded travel time stability analysis algorithm for stations covered by the air gun active source signal, which uses the precise excitation time of the air gun active source P-wave data as a time reference and estimates the time service parameter deviation by comparing the residual between the theoretical travel time and the observed travel time; a background noise cross-correlation superposition algorithm for stations not covered by the air gun active source signal, which estimates the clock error based on the time offset of the peak value of the cross-correlation function of the background noise data; a travel time stability analysis algorithm that defines the travel time residual sequence and estimates the time service parameter deviation by least squares fitting, with the weights determined by the signal-to-noise ratio of the air gun event; and a background noise cross-correlation superposition algorithm that extracts continuous recording segments from the preprocessed background noise data, calculates the noise cross-correlation function between station pairs, and superimposes multiple time windows to enhance the signal, estimating the clock error by comparing the peak offset of linear regression with the theoretical travel time difference between station pairs.

[0016] Furthermore, the triggering conditions of the event triggering mechanism described in step 4 include a data quality threshold, a time period threshold, and an external event trigger. The data quality threshold triggers recalibration when the average signal-to-noise ratio of the air gun active source P-wave data is lower than 5 dB or the power spectral density of the background noise data is abnormal. The time period threshold automatically triggers calibration at fixed intervals. The external event trigger triggers immediate calibration when a teleseismic event or a station maintenance event is detected.

[0017] Furthermore, the collaborative verification in step 5 includes cross-validation of azimuth parameters and cross-validation of time service parameters. Cross-validation of azimuth parameters uses the geometric constraint relationship between multiple source data to compare the corrected direction of arrival with the reference value. Cross-validation of time service parameters calculates a consistency index by comparing the clock errors estimated by different algorithms. If the consistency index is higher than the threshold, the verification is successful.

[0018] Furthermore, in step 5, after feeding the correction parameters back to the multi-source monitoring data basic fusion model, the improvement in fusion accuracy is evaluated by the variance reduction rate. A variance reduction rate greater than 0.5 is considered as effective correction.

[0019] This invention also discloses an adaptive fusion and parameter correction system for multi-source monitoring data, used to implement the method described above. The system includes:

[0020] The data collection and basic fusion model construction module is used to collect multi-source monitoring data in earthquake monitoring scenarios, preprocess the multi-source monitoring data, and construct a multi-source monitoring data basic fusion model based on the preprocessed multi-source monitoring data.

[0021] The parameter division and priority setting module is used to divide the core parameters to be corrected in the multi-source monitoring data basic fusion model into azimuth correction parameters and time service correction parameters, analyze error sensitive sources, and set correction priorities based on sensitivity analysis.

[0022] The azimuth parameter adaptive correction module is used to embed multiple azimuth estimation algorithms into the multi-source monitoring data basic fusion model, construct an azimuth parameter adaptive fusion correction framework, estimate azimuth parameters in parallel through multiple algorithms, introduce an adaptive weight adjustment mechanism for weighted fusion, and output azimuth correction results.

[0023] The time service parameter adaptive correction module is used to embed time service parameter estimation algorithms into the multi-source monitoring data basic fusion model according to different scenarios, construct a time service parameter adaptive fusion correction framework, adopt differentiated correction strategies for different stations, introduce an event triggering mechanism to dynamically activate the correction process, and output time service correction results.

[0024] The collaborative verification and system integration module is used to perform collaborative verification based on the azimuth correction results and the time service correction results, and feeds back the correction parameters to the multi-source monitoring data basic fusion model to update the model parameters. Finally, the entire correction framework is integrated into the automated operation platform.

[0025] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, performs the steps of the method described herein.

[0026] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method.

[0027] The beneficial effects of this invention are as follows: The adaptive fusion and parameter correction method and system for multi-source monitoring data provided by this invention significantly improves the processing accuracy and reliability of earthquake monitoring data by constructing a basic fusion model and intelligently dividing correction priorities; the method decomposes the core parameter correction into two parallel paths: azimuth angle and time service, and integrates multi-source algorithms such as polarization analysis, travel time inversion, and noise cross-correlation for adaptive fusion correction, effectively overcoming the limitations of a single data source or algorithm; by introducing a confidence-based dynamic weight adjustment and event triggering mechanism, it can adapt to the dynamic changes in data quality, ensuring stable and high-precision correction results under different monitoring scenarios; at the same time, it utilizes cross-validation and integration mechanisms to achieve closed-loop optimization and quality control of correction parameters, thereby greatly improving the accuracy and stability of multi-source data fusion results, providing a higher quality data foundation for earthquake monitoring and early warning, and enhancing the system's real-time performance, robustness, and automation level.

[0028] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0029] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0030] Figure 1 This is a schematic diagram illustrating the steps of an adaptive fusion and parameter correction method for multi-source monitoring data according to the present invention;

[0031] Figure 2 This is a schematic diagram of the structure of an adaptive fusion and parameter correction system for multi-source monitoring data according to the present invention. Detailed Implementation

[0032] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0033] In an embodiment of the present invention, an adaptive fusion and parameter correction method for multi-source monitoring data is provided. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0034] Step 1: Collect multi-source monitoring data in earthquake monitoring scenarios, and construct a multi-source monitoring data fusion model based on the multi-source monitoring data.

[0035] Specifically, the multi-source monitoring data includes airgun active source P-wave data, teleseismic P-wave data, and background noise data. The airgun active source P-wave data is generated by an airgun source system deployed in the monitoring area and received by a seismic detector array. Its data characteristics include P-wave first arrival time, amplitude spectrum, and travel time residual. The teleseismic P-wave data originates from distant seismic events (epicentral distance greater than 1000 km) and is recorded by broadband seismic stations. Key parameters include P-wave relative intensity, frequency content, and waveform correlation. The background noise data is obtained by continuously recording environmental seismic background signals in the monitoring area. Its main characteristics are environmental noise power spectral density and spatial correlation function. The collection of the multi-source monitoring data must be carried out under a unified spatiotemporal reference, with a time alignment accuracy better than 0.01 seconds, and spatial coordinates unified to the geocentric coordinate system to eliminate systematic errors introduced by differences in spatiotemporal references.

[0036] In the data preprocessing stage, the raw multi-source monitoring data first undergoes quality control and outlier removal: for the active P-wave data from the air gun, a screening mechanism based on the signal-to-noise ratio threshold is used to remove data segments with a signal-to-noise ratio below 3dB; for the teleseismic P-wave data, polarization analysis is used to identify and remove records severely affected by local interference; for the background noise data, probability density function estimation is used to remove outliers caused by instrument transient malfunctions or human activity. Subsequently, the data that passes quality control is standardized to normalize the data from each source to the same numerical range (e.g., the [-1,1] interval) to eliminate the influence of dimensions.

[0037] Based on the preprocessed multi-source monitoring data, a basic fusion model for the multi-source monitoring data is constructed. This model is an initial weighted linear fusion framework, and its expression is as follows: M base D represents the output of the basic fusion model. i These represent the preprocessed air gun active source P-wave data (i=1), teleseismic P-wave data (i=2), and background noise data (i=3), respectively. For the initial weighting coefficients of each data source, during the model initialization phase, default values ​​are set based on the theoretical observation accuracy of each data source (e.g., for P-wave data from an air gun active source, due to the high controllability of the signal, the initial weights are set to a certain value). Set to 0.5, teleseismic P-wave data Set to 0.3, background noise data (Set to 0.2), and satisfy the weight and constraints. This basic fusion model provides an initial data field and a weight optimization benchmark for the adaptive parameter correction in subsequent steps.

[0038] Step 2: Divide the core parameters to be corrected in the multi-source monitoring data basic fusion model into two categories: azimuth correction parameters and time service correction parameters, and clarify the error sensitivity sources and correction priorities of the two types of parameters;

[0039] Specifically, based on the multi-source monitoring data fusion model constructed in step 1, the core parameters to be corrected in the model are divided into two categories: azimuth correction parameters and time service correction parameters. The error sensitivity sources and correction priorities for these two types of parameters are also identified. It should be noted that the parameter division is based on their impact mechanism on the quality of seismic monitoring data fusion: azimuth parameters mainly affect the positioning accuracy of wave propagation direction, while time service parameters mainly affect the synchronization of event travel time calculations.

[0040] Specifically, the azimuth correction parameter is defined as the deflection angle deviation Δθ between the station sensor coordinate system and the geographic coordinate system, and its error sensitive sources include:

[0041] Sensor installation deviation: Due to mechanical installation errors during station deployment, the azimuth angle deviates.

[0042] Local interference from the geomagnetic field: Geomagnetic anomalies can cause distortion in polarization analysis and amplify azimuth errors;

[0043] Multipath effect: Multipath scattering caused by seismic waves propagating in complex geological structures leads to instability in polarization direction estimation.

[0044] Time service correction parameters are defined as the deviation Δt between the station clock and the standard time, including clock error and frequency drift. The sources of error sensitivity include:

[0045] Clock hardware drift: Accumulated clock error caused by temperature changes or crystal aging;

[0046] Network synchronization delay: During data transmission, network jitter causes inaccurate timestamps;

[0047] External interference, such as electromagnetic interference or power fluctuations, can affect clock stability.

[0048] The setting of correction priority is based on the parameter values ​​of the fusion model output M. base Sensitivity analysis was performed using variance decomposition to calculate the contribution rate of each variance to determine priority. The total variance of the fusion model output was defined as: ,in Let be the variance of the i-th class of data. To calculate the covariance, the effects of azimuth parameter variation δθ and time service parameter variation δt on the covariance are calculated using perturbation analysis. partial derivative S θ and S t The priority determination rule is: if S θ Greater than S t If the azimuth error is high, then azimuth correction takes priority; otherwise, time service correction takes priority. In actual earthquake monitoring scenarios, because azimuth error has a more sensitive impact on positioning accuracy, azimuth correction is usually set as a high priority, and time service correction as a secondary priority. This division provides target guidance for the specific correction framework in steps 3 and S4.

[0049] S3, In the basic fusion model of the multi-source monitoring data, a polarization characteristic extraction algorithm, a particle motion analysis algorithm, and a signal-to-noise ratio weighted multi-event inversion algorithm are embedded to construct an adaptive azimuth parameter fusion correction framework. This framework uses the M output from the basic fusion model as the basis for the azimuth parameter fusion correction. base Using this as input, azimuth parameters are estimated in parallel through multiple algorithms, and an adaptive weight adjustment mechanism is introduced to optimize the fusion result, achieving high-precision correction. Specifically, the construction of the adaptive azimuth parameter fusion correction framework includes the following steps:

[0050] S301, based on the vector waveform (three-component data) of the air gun active source P-wave data or teleseismic P-wave data, calculate the principal direction of the polarization ellipsoid, and define the polarization degree P and the estimated azimuth angle θ1 as: , , where λ max ,λ mid ,λ min Here, eigenvalues ​​of the covariance matrix are listed in descending order, and UE and UN represent the amplitudes of the east-west and north-south components, respectively. The degree of polarization P is used to assess the reliability of the estimate; P > 0.5 indicates a valid estimate. Based on the particle trajectory from background noise data or P-wave data, the motion direction is extracted using principal component analysis (PCA). The azimuth estimate θ2 is defined as the angle between the direction of the first principal component and true north, and its confidence level C2 is expressed as the variance ratio explained by the principal components. λ1 is the largest eigenvalue. For multiple seismic events (such as teleseismic events), the azimuth is estimated using the travel-time inversion method. The estimated azimuth value θ3 is defined as the inversion result, and its weight is determined by the event signal-to-noise ratio (SNR) using the following formula: Where m is the number of events. and SNR represents the observed and theoretical travel times, respectively. j Let be the signal-to-noise ratio of the j-th event.

[0051] S302 introduces an adaptive weight adjustment mechanism to perform weighted fusion of estimates from multiple algorithms. The azimuth correction result θ after fusion is defined. corrected for: ;where θ n The estimated value for the nth algorithm (n=1,2,3 correspond to polarization characteristic extraction, particle motion analysis, and signal-to-noise ratio weighted inversion, respectively), w n For adaptive weights, and satisfying The weights are calculated based on the real-time confidence metrics of each algorithm (such as polarization degree P, explained variance ratio C2, and inversion residuals) and historical error variance. Specifically, the weights w n The expression is obtained by optimizing the objective function to minimize the error, and the formula is: , where σ train,i Let σ represent the historical training error variance of the i-th algorithm. ij Let represent the error covariance between the i-th and j-th algorithms, and n be the algorithm index (n=1,2,3). The weights are updated every time window (e.g., 1 hour) to adapt to changes in data quality. In a specific embodiment, the objective function is defined as the mean squared error of the fusion estimate: Where J represents the objective function value, E represents the mathematical expectation operator, and θ true The true azimuth angle (which can be calibrated using reference stations), θ correctedThis represents the corrected azimuth estimate. The weights w are calculated using a function constructed using the Lagrange multiplier method. n The specific process is as follows: Define the constraints, and the weights must satisfy the normalization constraint. Where n=1,2,3 correspond to the polarization characteristic extraction algorithm, the particle motion analysis algorithm, and the signal-to-noise ratio weighted multi-event inversion algorithm, respectively; construct the Lagrangian function. Where J is the objective function for minimizing the error. λ Let L be the Lagrange multiplier; to find the extrema, take the partial derivatives of the Lagrange function L with respect to w1, w2, w3 and λ respectively, and set the partial derivatives to 0 to obtain the system of equations. (n=1,2,3) and Solving the system of equations: By solving the above linear system of equations, we obtain the analytical solution of the adaptive weight wn that satisfies the constraints and minimizes the objective function J, ensuring that the weight allocation conforms to the principle of minimizing error.

[0052] To control the convergence accuracy of the azimuth parameter, the convergence condition is set as follows: the change in the correction result of consecutive iterations is less than a threshold. The convergence criterion is defined as follows: Where ϵ = 0.1° is the preset threshold, and k is the number of iterations. If convergence is not achieved, return to S301 for re-estimation until the accuracy requirement is met. This framework outputs a high-precision azimuth correction result θ. corrected This provides the foundation for collaborative verification in step 5.

[0053] Step 4: For stations covered by the active air gun signal, a travel time stability analysis algorithm is embedded in the multi-source monitoring data fusion model. The precise excitation time and travel time stability of the active air gun P-wave data are used to evaluate and correct the deviation of the time service parameters. For stations not covered by the active air gun signal, a background noise cross-correlation superposition algorithm is embedded to estimate the clock difference in the time service parameters based on the background noise data. At the same time, an event triggering mechanism is introduced to trigger the time service parameter correction process in the corresponding scenario, so as to achieve full-scenario coverage correction of the time service parameters.

[0054] Specifically, the adaptive fusion correction framework for time service parameters is constructed based on the time service correction parameters defined in step 2 (defined as the deviation Δt between the station clock and the standard time, including clock error and frequency drift), and differentiated algorithms are adopted for different data coverage scenarios of different stations. The scenario-based correction strategy is dynamically activated by an event-triggered mechanism to ensure the real-time nature and adaptability of the correction process.

[0055] First, for stations covered by the active air gun source signal (i.e., stations deployed within the effective excitation range of the air gun source system and capable of reliably receiving P-wave data from the active air gun source), a travel time stability analysis algorithm is embedded in the multi-source monitoring data fusion model. This algorithm uses the precise excitation time of the air gun active source P-wave data (recorded by a high-precision clock of the air gun source system, with an accuracy better than 0.001 seconds) as the time reference, and estimates the time service parameter deviation by comparing the residuals between the theoretical travel time and the observed travel time. The travel time residual sequence is defined as: ,in For the P-wave observation travel time of the k-th airgun event at the station, The theoretical travel time is calculated based on station and source locations (using a one-dimensional velocity model such as the AK135 model). The estimated value of the time service parameter deviation Δt is obtained by least-squares fitting of the residual sequence: where the weight w k The signal-to-noise ratio (SNR) of an air gun incident determines w k = SNR k / ∑SNR k This ensures that high signal-to-noise ratio events dominate the correction process. Simultaneously, a time-stability index σΔT (standard deviation of time-stability residuals) is introduced to assess correction reliability: if σΔT < 0.01 seconds, the estimation is considered valid; otherwise, a re-acquisition or data filtering process is triggered.

[0056] Secondly, for stations not covered by the active source signal of the air gun (such as remote areas or air gun excitation blind zones), a background noise cross-correlation superposition algorithm is embedded. This algorithm estimates the clock difference by analyzing the time offset of the cross-correlation function peak based on the cross-correlation function of the background noise data. The specific steps include: extracting continuous recording segments from the preprocessed background noise data, calculating the noise cross-correlation function (NCF) between station pairs, and superimposing the NCF of multiple time windows to enhance the signal. The time offset δτ of the cross-correlation peak is defined as:

[0057] Where δτ represents the peak time offset (in seconds), τ represents the time delay variable, and NCF m (τ) represents the noise cross-correlation function value of the m-th time window, M represents the total number of time windows (usually M≥24, corresponding to data of more than 24 hours), argmax τ This represents the value of τ that maximizes the function value. The estimated value of the clock difference Δt is obtained by using the linear regression peak offset and the theoretical travel time difference between the station pairs: , where Δt ref The known clock bias of the reference station (calibrated via GPS synchronization) is used. To improve robustness, only cross-correlation peaks with a signal-to-noise ratio higher than 10 are used, and anomalous offsets are removed.

[0058] The event triggering mechanism dynamically activates the above correction process, and the triggering conditions include:

[0059] Data quality threshold: When the average signal-to-noise ratio of the air gun active source P-wave data is lower than 5dB, or the power spectral density of the background noise data is abnormal (deviating from the historical mean by 3 times the standard deviation), recalibration is triggered.

[0060] Time period threshold: Automatically triggers correction at fixed intervals (e.g., 24 hours) to address clock drift.

[0061] External event trigger: When a teleseismic event (magnitude > 5) or a station maintenance event (such as hardware restart) is detected, immediate correction is triggered.

[0062] The time service correction result Δt is output through adaptive weight fusion. corrected The weights are assigned based on the confidence index of each algorithm: the weights w for the time-travel stability algorithm. travel The weight w of the background noise algorithm is proportional to the reciprocal of the time-stability index. noise The signal-to-noise ratio is proportional to the peak value of the cross-correlation. The fusion formula is: , where Δt corrected The time service correction result after fusion (unit: seconds), Δt travel Δt represents the time service parameter deviation value estimated by the time-stability algorithm. noise w represents the clock difference value estimated using the background noise cross-correlation algorithm. travel and w noise These are the weights of the corresponding algorithms, and they satisfy the weights and constraints w. travel + w noise =1. If data is missing for a certain scenario, the weight is automatically adjusted to 0, and the algorithm dominates for the other scenario. The framework runs iteratively until convergence, and the convergence condition is that the difference between two consecutive correction results is less than a threshold.

[0063] Step 5: Based on the azimuth correction result output by S3 and the time service correction result output by S4, the azimuth parameter and time service parameter are cross-validated to complete the collaborative verification of the two types of parameter correction results. The adaptive fusion correction framework for azimuth parameters, the adaptive fusion correction framework for time service parameters, and the collaborative verification mechanism are integrated into the automated operation platform and embedded into the existing seismic network operation and maintenance system to realize real-time adaptive fusion of multi-source monitoring data, automatic identification of errors in correction parameters, and visual quality control feedback of correction results.

[0064] Specifically, the collaborative verification and integration mechanism aims to ensure the accuracy and consistency of the parameter correction results output from steps 3 and S4, and to achieve operationalization through system integration. First, a cross-validation method is used to collaboratively verify the azimuth parameter and time service parameter: Azimuth parameter cross-validation: The correction results are verified using the geometric constraints between multi-source data. For example, based on the corrected azimuth of the airgun active source P-wave data, the direction of arrival (Backazimuth) is calculated and compared with the polarization analysis results of the teleseismic P-wave data, defining the verification residual δθ = θ. corrected -θ reference , where the reference value θ reference The result can be obtained through an independent calibration method (such as GPS orientation calibration). If the absolute value of the residual is less than the threshold, the verification is successful; otherwise, the S3 framework is triggered for recalibration.

[0065] Cross-validation of time service parameters: Validation is performed by comparing clock errors estimated by different algorithms with relative clock errors between stations. For example, a consistency index is calculated between clock errors estimated using an air gun timekeeping stability algorithm and clock errors estimated using a background noise cross-correlation algorithm. If C Δt If the value is > 0.9, the verification is successful; otherwise, arbitration correction will be performed by combining the network synchronization data between stations (such as NTP protocol logs).

[0066] After successful verification, the correction parameters are fed back to the multi-source monitoring data fusion model constructed in step 1 to update the model weights and parameters, thereby improving the quality of data fusion. The model output is M. base The accuracy is assessed by the variance reduction rate: ,in and These are the variances of the fused model outputs before and after correction, respectively. A VRR > 0.5 is considered a valid correction.

[0067] Optionally, the entire parameter calibration framework is finally integrated into an automated operation platform. This platform, developed based on a microservice architecture, is embedded in existing seismic network operation and maintenance systems (such as Earthworm or SeisComP) to achieve the following functions: Real-time adaptive fusion: The platform continuously receives multi-source monitoring data streams, automatically triggers the S1-S5 process, and outputs the fused results after calibration at short time windows; Automatic error identification: By monitoring calibration residuals and confidence indices, abnormal errors (such as sensor failures or clock jumps) are identified in real time, and alarm events are generated; Visualized quality control feedback: A web interface is provided to visualize calibration parameters and fusion results, including spatial distribution maps of azimuth calibration results, clock difference time series curves, and data quality heatmaps. Users can interactively adjust parameter thresholds and export quality control reports.

[0068] During integration, the platform and the seismic network system use standard protocols for their data interfaces to ensure compatibility. All algorithm modules are deployed in a containerized manner, supporting horizontal scaling to handle high-concurrency data streams. Through these mechanisms, this invention automates the entire process of multi-source monitoring data acquisition, fusion, correction, and verification, significantly improving the reliability and accuracy of earthquake monitoring.

[0069] This embodiment provides an adaptive fusion and parameter correction method for multi-source monitoring data. By constructing a basic fusion model and intelligently prioritizing corrections, it improves the processing accuracy and reliability of earthquake monitoring data. The method decomposes the core parameter correction into two parallel paths: azimuth and time service. It embeds multi-source algorithms such as polarization analysis, travel time inversion, and noise cross-correlation for adaptive fusion correction, effectively overcoming the limitations of single data sources or algorithms. By introducing a confidence-based dynamic weight adjustment and event triggering mechanism, the method can adapt to dynamic changes in data quality, ensuring stable and high-precision correction results under different monitoring scenarios. Finally, through cross-validation and integration mechanisms, closed-loop optimization and quality control of correction parameters are achieved, ultimately improving the accuracy and stability of multi-source data fusion results and providing a higher-quality data foundation for earthquake monitoring and early warning.

[0070] Accordingly, such as Figure 2 As shown, based on an adaptive fusion and parameter correction method for multi-source monitoring data, this embodiment of the invention also provides an adaptive fusion and parameter correction system for multi-source monitoring data, which implements the adaptive fusion and parameter correction method for multi-source monitoring data of this embodiment of the invention. The system includes:

[0071] The system comprises: a data collection and basic fusion model construction module 401; a parameter partitioning and priority setting module 402; an azimuth parameter adaptive correction module 403; a time service parameter adaptive correction module 404; and a collaborative verification and system integration module 405.

[0072] The data collection and basic fusion model construction module 401 is used to collect multi-source monitoring data in the earthquake monitoring scenario, including air gun active source P-wave data, teleseismic P-wave data and background noise data; to perform quality control, outlier removal and standardization preprocessing on the multi-source monitoring data; and to construct a basic fusion model of multi-source monitoring data based on the preprocessed data.

[0073] The parameter partitioning and priority setting module 402 is used to partition the core parameters to be corrected in the multi-source monitoring data basic fusion model into azimuth correction parameters and time service correction parameters; analyze the error sensitivity sources of the azimuth correction parameters, including sensor installation deviation, local geomagnetic field interference, and multipath effect, and the error sensitivity sources of the time service correction parameters, including clock hardware drift, network synchronization delay, and external interference; perform sensitivity analysis based on the variance decomposition method, calculate the partial derivatives of the azimuth parameter changes and time service parameter changes with the output variance of the fusion model, and set the correction priority. If the sensitivity of the azimuth parameter is higher than that of the time service parameter, then the azimuth correction takes priority, and vice versa.

[0074] The azimuth parameter adaptive correction module 403 is used to embed polarization characteristic extraction algorithm, particle motion analysis algorithm, and signal-to-noise ratio weighted multi-event inversion algorithm into the multi-source monitoring data basic fusion model to construct an adaptive azimuth parameter fusion correction framework; it estimates the azimuth parameter in parallel through multiple algorithms, including estimates based on the principal direction of the polarization ellipsoid, estimates based on the particle motion trajectory, and estimates based on travel time inversion; it introduces an adaptive weight adjustment mechanism, calculates weights based on the real-time confidence index and historical error variance of each algorithm, performs weighted fusion of the estimates from multiple algorithms, and outputs the azimuth correction result; it sets convergence conditions to ensure that the correction accuracy meets the threshold requirements.

[0075] The time service parameter adaptive correction module 404 is used to construct an adaptive fusion correction framework for time service parameters based on different scenarios. For stations covered by the active air gun signal, a travel time stability analysis algorithm is embedded to estimate the time service parameter deviation using the precise excitation time of the active air gun P-wave data. For stations not covered by the active air gun signal, a background noise cross-correlation superposition algorithm is embedded to estimate the clock error based on the peak offset of the cross-correlation function of the background noise data. An event triggering mechanism is introduced to dynamically activate the correction process based on data quality thresholds, time period thresholds, and external event triggers. The estimated values ​​from different algorithms are fused through adaptive weighting to output the time service correction result.

[0076] The collaborative verification and system integration module 405 is used to perform collaborative verification of azimuth correction results and time service correction results using cross-validation methods, including cross-validation of azimuth parameters and cross-validation of time service parameters, to ensure the consistency of correction results; feeds back the correction parameters to the multi-source monitoring data basic fusion model, updates the model weights and parameters, and evaluates the improvement in fusion accuracy through variance reduction rate; integrates the entire correction framework into the automated operation platform and embeds it into the existing seismic network operation and maintenance system to realize real-time adaptive fusion of multi-source monitoring data, automatic error identification, and visualized quality control feedback.

[0077] The adaptive fusion and parameter correction system for multi-source monitoring data provided in this application embodiment can implement the steps and processes of the adaptive fusion and parameter correction method for multi-source monitoring data provided in any of the above embodiments, and achieve the same technical effect, which will not be described in detail here.

[0078] This embodiment provides an adaptive fusion and parameter correction system for multi-source monitoring data. It modularizes and automates the complex processes of the aforementioned methods, achieving efficient operation and maintenance throughout the entire process from data access to correction application. Through the collaborative work of five modules, the system realizes automatic preprocessing of multi-source data, basic fusion, intelligent parameter partitioning, and path-specific adaptive correction of azimuth and time service parameters, greatly reducing manual intervention and improving processing efficiency. The collaborative verification and system integration module ensures the reliability of the correction results and seamlessly embeds the entire framework into the existing seismic network system, possessing strong engineering practical value. The system can perform real-time data stream processing, automatic error identification, and visualized quality control feedback, significantly reducing operation and maintenance costs and providing earthquake monitoring personnel with intuitive and reliable data and decision support, thereby improving the automated operation level of the seismic network and the usability of monitoring data.

[0079] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. An adaptive fusion and parameter correction method for multi-source monitoring data, characterized in that, include: Step 1: Collect multi-source monitoring data, preprocess it, and construct a basic fusion model for multi-source monitoring data; Step 2: Divide the core parameters to be corrected in the multi-source monitoring data fusion model into azimuth correction parameters and time service correction parameters, analyze the error sensitivity sources of azimuth correction parameters and time service correction parameters, and set correction priorities based on sensitivity analysis; Step 3: Embed multiple azimuth estimation algorithms into the multi-source monitoring data fusion model, construct an adaptive azimuth parameter fusion correction framework, estimate azimuth parameters in parallel using multiple algorithms, and introduce an adaptive weight adjustment mechanism for weighted fusion to output the azimuth correction result. Step 4: Embed time service parameter estimation algorithms in the multi-source monitoring data fusion model according to different scenarios, construct an adaptive fusion correction framework for time service parameters, adopt differentiated correction strategies for stations covered by the active source signal of the air gun and stations not covered, introduce an event triggering mechanism to dynamically activate the correction process, and output the time service correction results. Step 5: Perform collaborative verification based on the azimuth correction results and time service correction results, and feed the correction parameters back to the multi-source monitoring data fusion model to update the model parameters.

2. The method as described in claim 1, characterized in that, The multi-source monitoring data mentioned in step 1 includes air gun active source P-wave data, teleseismic P-wave data, and background noise data; the preprocessing includes quality control, outlier removal, and standardization of the original multi-source monitoring data; the basic fusion model of the multi-source monitoring data is a weighted linear fusion model, the output of which is the output of the basic fusion model. The data sources include air gun active source P-wave data, teleseismic P-wave data, and background noise data. The initial weight coefficients are set based on the observation accuracy of each data source and satisfy the weight and constraint of 1.

3. The method as described in claim 2, characterized in that, The quality control described in step 1 includes using a screening mechanism based on the signal-to-noise ratio threshold to remove data segments with a signal-to-noise ratio below 3dB for the active source P-wave data of the air gun; using polarization analysis to identify and remove records severely affected by local interference for the teleseismic P-wave data; and using probability density function estimation to remove outliers for the background noise data. The standardization process normalizes each source data to the same numerical range.

4. The method as described in claim 1, characterized in that, The azimuth correction parameter mentioned in step 2 is defined as the deflection angle deviation between the station sensor coordinate system and the geographic coordinate system. Its error sensitive sources include sensor installation deviation, local interference of the geomagnetic field, and multipath effect. The time service correction parameter is defined as the deviation between the station clock and the standard time, including clock difference and frequency drift. Its error sensitive sources include clock hardware drift, network synchronization delay, and external interference. Sensitivity analysis employs variance decomposition to calculate the partial derivatives of the azimuth parameter changes and time service parameter changes with the variance of the fusion model output. The priority determination rule is that if the sensitivity of the azimuth parameter is higher than that of the time service parameter, then azimuth correction takes priority; otherwise, time service correction takes priority.

5. The method as described in claim 1, characterized in that, The various azimuth estimation algorithms mentioned in step 3 include a polarization characteristic extraction algorithm, a particle motion analysis algorithm, and a signal-to-noise ratio (SNR) weighted multi-event inversion algorithm. The polarization characteristic extraction algorithm calculates the principal direction of the polarization ellipsoid based on the vector waveform of P-wave data from an active air gun source or teleseismic P-wave data, obtains the azimuth estimate, and evaluates the degree of polarization to determine the reliability of the estimate. The particle motion analysis algorithm extracts the motion direction based on the particle motion trajectory of background noise data or P-wave data through principal component analysis, obtains the azimuth estimate, and calculates the confidence level. The SNR weighted multi-event inversion algorithm estimates the azimuth based on the travel time inversion of multiple teleseismic events, with its weights determined by the event SNR. The adaptive weight adjustment mechanism weights and fuses the estimates from multiple algorithms, defining the fused azimuth correction result as the weighted sum of the estimates from each algorithm. The weights are calculated based on the real-time confidence index and historical error variance of each algorithm through an error minimization objective function, and the weights are updated every time window to adapt to changes in data quality.

6. The method as described in claim 1, characterized in that, The scenario-based embedded time service parameter estimation algorithm in step 4 includes a station embedded travel time stability analysis algorithm for air gun active source signal coverage. It uses the precise excitation time of the air gun active source P-wave data as a time reference and estimates the time service parameter deviation by comparing the residual between the theoretical travel time and the observed travel time. For stations not covered by the active source signal of the air gun, a background noise cross-correlation superposition algorithm is embedded. The clock error is estimated based on the time offset of the peak value of the cross-correlation function of the background noise data. The travel time stability analysis algorithm defines the travel time residual sequence and estimates the time service parameter deviation by least squares fitting. The weights are determined by the signal-to-noise ratio of the air gun event. The background noise cross-correlation superposition algorithm extracts continuous recording segments from the preprocessed background noise data, calculates the noise cross-correlation function between station pairs, and superimposes multiple time windows to enhance the signal. The clock error is estimated by linear regression of the peak offset and the theoretical travel time difference between station pairs.

7. The method as described in claim 1, characterized in that, The triggering conditions for the event triggering mechanism described in step 4 include a data quality threshold, a time period threshold, and an external event trigger. The data quality threshold triggers recalibration when the average signal-to-noise ratio of the P-wave data from the active air gun source is lower than 5 dB or the power spectral density of the background noise data is abnormal. The time period threshold automatically triggers calibration at fixed intervals. The external event trigger triggers immediate calibration when a teleseismic event or a station maintenance event is detected.

8. The method as described in claim 1, characterized in that, The collaborative verification described in step 5 includes cross-validation of azimuth parameters and cross-validation of time service parameters. Cross-validation of azimuth parameters uses the geometric constraints between multiple source data to compare the corrected direction of arrival with the reference value. Cross-validation of time service parameters calculates a consistency index by comparing the clock errors estimated by different algorithms. If the consistency index is higher than the threshold, the verification is successful.

9. The method as described in claim 1, characterized in that, In step 5, after the correction parameters are fed back to the multi-source monitoring data basic fusion model, the improvement in fusion accuracy is evaluated by the variance reduction rate. A variance reduction rate greater than 0.5 is considered as effective correction.

10. An adaptive fusion and parameter correction system for multi-source monitoring data, used to implement the method described in any one of claims 1-9, characterized in that, The system includes: The data collection and basic fusion model construction module is used to collect multi-source monitoring data in earthquake monitoring scenarios, preprocess the multi-source monitoring data, and construct a multi-source monitoring data basic fusion model based on the preprocessed multi-source monitoring data. The parameter division and priority setting module is used to divide the core parameters to be corrected in the multi-source monitoring data basic fusion model into azimuth correction parameters and time service correction parameters, analyze error sensitive sources, and set correction priorities based on sensitivity analysis. The azimuth parameter adaptive correction module is used to embed multiple azimuth estimation algorithms into the multi-source monitoring data basic fusion model, construct an azimuth parameter adaptive fusion correction framework, estimate azimuth parameters in parallel through multiple algorithms, introduce an adaptive weight adjustment mechanism for weighted fusion, and output azimuth correction results. The time service parameter adaptive correction module is used to embed time service parameter estimation algorithms into the multi-source monitoring data basic fusion model according to different scenarios, construct a time service parameter adaptive fusion correction framework, adopt differentiated correction strategies for different stations, introduce an event triggering mechanism to dynamically activate the correction process, and output time service correction results. The collaborative verification and system integration module is used to perform collaborative verification based on the azimuth correction results and the time service correction results, and feeds back the correction parameters to the multi-source monitoring data basic fusion model to update the model parameters. Finally, the entire correction framework is integrated into the automated operation platform.

Images