A method for earthquake source location based on Bayesian uncertainty estimation of time difference measurement

Abstract

The present invention provides a source location method based on Bayesian time-difference measurement uncertainty estimation, comprising deploying a microseismic monitoring system in a typical mine, conducting multiple blasting vibration experiments with known source locations, collecting and screening blasting waveform signals, and processing the filtered low-signal-to-noise ratio waveform signals using an improved empirical mode decomposition technique and wavelet packet noise reduction technique; automatically picking time-difference measurement data of different sources based on the processed waveform signals; constructing a Bayesian estimation framework to estimate the uncertainty of the time-difference measurement data, and calculating it using Monte Carlo sampling technology; adding a bias correction term to the time-difference data based on the estimated time-difference measurement uncertainty to perform an unbiased correction, and then performing positioning using a positioning method to obtain high-precision microseismic source location results. The present invention improves the accuracy of time-difference measurement data picking, lays a good foundation for time-difference measurement uncertainty estimation, and achieves high-precision positioning of mine microseismic sources.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional microseismic positioning, and in particular to a method for earthquake source positioning based on Bayesian time difference measurement uncertainty estimation. Background Art

[0002] Microseismic location technology can dynamically invert "spatiotemporal intensity" information, such as the time and spatial location of rock mass ruptures and the energy released by the rupture source. Deciphering the temporal and spatial intensity evolution of microseismic data provides an important basis for rock mass stability assessment and accurate early warning of potential rockbursts. Time-difference measurement data recorded by the network serves as a key input for earthquake source location. The accuracy of its prior distribution directly affects the accuracy of earthquake source location, which is severely limited by the uncertainty of time-difference measurement.

[0003] Most current microseismic location methods are essentially extensions and improvements based on the least squares principle. They assume that time-difference measurement data obey the zero-mean Gaussian independent and identically distributed assumption, without making any additional assumptions or estimates on the time-difference measurement data. In mining engineering environments, abnormal signal interference, channel crosstalk, weak waveform first arrivals, non-line-of-sight propagation, multipath propagation, travel time path delays, or decreased sensor sensitivity can all lead to large differences in measurement uncertainty, and biased, heteroscedastic, and intercorrelated measurement data are common. Furthermore, due to the square amplification effect of the least squares principle, the aforementioned prior distribution errors are more prominent, leading to even greater positioning errors. Therefore, existing traditional microseismic location methods are unable to accurately obtain the location of microseismic sources in complex mining environments.

[0004] However, for mine dynamic disasters, high-precision positioning of mine microseismic sources is the key to achieving accurate early warning of mine rock bursts and ensuring the safe and efficient mining of strategic mineral resources. The inability to accurately locate the microseismic source makes the prevention and control of mine dynamic disasters more difficult. Summary of the Invention

[0005] In order to solve the technical problem that existing microseismic locating methods cannot accurately obtain the microseismic source position under complex mining environment conditions, making the prevention and control of mine dynamic disasters more difficult, the present invention provides a source locating method based on Bayesian time difference measurement uncertainty estimation.

[0006] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0007] A method for earthquake source location based on Bayesian time difference measurement uncertainty estimation includes the following steps:

[0008] S1. Deploy a microseismic monitoring system in a typical mine. The system includes a sensor network and a blasting source with a known source location within the monitoring system. Conduct multiple blasting vibration experiments with known source locations, collect and screen blasting waveform signals, and process the resulting low signal-to-noise ratio waveform signals using improved empirical mode decomposition and wavelet packet noise reduction techniques.

[0009] S2, automatically picking up the time difference measurement data of different earthquake sources based on the processed waveform signal;

[0010] S3. Construct a Bayesian estimation framework to estimate the uncertainty of the time difference measurement data and use Monte Carlo sampling technology to calculate it;

[0011] S4. Based on the estimated uncertainty of the time difference measurement, the bias correction term is added to the time difference data for unbiased correction, and then positioning is performed using the positioning method to obtain high-precision microseismic source positioning results.

[0012] Furthermore, the steps of the empirical mode decomposition technology in step S1 specifically include:

[0013] S11, adding multiple adaptive Gaussian white noise time series of different amplitudes to the original time series of the screened burst waveform signal;

[0014] S12, averaging the upper and lower envelopes of each noisy time series to obtain an interpolation of the noisy signal and the corresponding average time as a new time series;

[0015] S13. Continue adding Gaussian white noise to the new time series, and repeat steps S11 and S12 until the new time series meets the termination condition.

[0016] Furthermore, the termination condition in step S13 is:

[0017] The number of extreme values ​​and zero crossings of the new time series is less than or equal to 1; and

[0018] The average value of the upper and lower envelopes of the new time series at any time is 0.

[0019] Furthermore, the wavelet packet denoising technique in step S1 adopts a fixed threshold technique, specifically:

[0020]

[0021] Where T represents the threshold, σ represents the standard deviation of the noise, and N represents the length of the signal in the time domain.

[0022] Furthermore, the fixed threshold technology adopts a soft threshold function, which is specifically expressed as:

[0023]

[0024] Where S(t) represents the soft threshold function, x represents the decomposition factor, and |·| represents the absolute value.

[0025] Furthermore, the steps of estimating the uncertainty of the time difference measurement data using the Bayesian estimation framework in step S3 are specifically as follows:

[0026] Assuming that the microseismic monitoring system has been in a stable state during the research period, that is, the uncertainty of the time difference measurement of the microseismic monitoring system remains approximately unchanged, the number of sensor pairs involved in the inversion of the time difference measurement uncertainty is set to k, the number of microseismic and blasting events is set to m, the number of time difference measurement data is n, and 2k≤n is satisfied, then the Bayesian estimation framework is expressed as:

[0027]

[0028] Where P(X) is the marginal likelihood, P(θ) is the initial prior probability of the parameter to be estimated, and the default assumption of the traditional positioning method, that is, the zero-mean Gaussian independent and identically distributed, is used as the prior distribution of the parameter θ; the prior probability P(θ) will be further updated by the sample information X measured by the time difference measurement, P(θ|X) is the updated posterior probability, and P(X|θ) is the likelihood function conditioned on the model parameter θ, expressed as:

[0029]

[0030] Among them, C e represents T(θ) cal -T mea The covariance matrix of , T(θ) cal is the calculated value of the time difference data, T mea is the measurement value of the time difference data, and d represents the dimension of the data.

[0031] Furthermore, the Monte Carlo sampling technique in step S3 specifically includes:

[0032] Taking 0 µs as the initial value for iteration, the time difference measurement uncertainty of each sensor is estimated by iterating 100,000 times. During the iteration, one parameter is updated each time, and the update speed follows Gaussian distribution sampling with a mean of 0 and a standard deviation of 0.25. After the parameter is updated, the probability density function P(X|θ') of the new parameter model θ' is calculated, and the acceptance rate α(m'|m) of the new parameter model is expressed as:

[0033]

[0034] When α(m'|m) is greater than or equal to u, the new model m' is accepted; otherwise, the new model will be rejected; each time u∈[0,1] represents a random number in the range of 0-1, which obeys the uniform distribution.

[0035] Compared with the prior art, the present invention provides a method for earthquake source location based on Bayesian time difference measurement uncertainty estimation, which includes two major parts: microseismic waveform signal collection and processing and Bayesian estimation and positioning of microseismic time difference measurement uncertainty. The microseismic waveform signal collection and processing includes using a signal noise reduction and purification method to reduce the noise of the blasting vibration signal, and continuing to use the Akaike information method to automatically pick up the time difference measurement data of the earthquake source; then, based on the time difference measurement data, sensor coordinates, earthquake source location and wave velocity structure information, a Bayesian estimation framework for time difference measurement uncertainty is constructed to obtain an accurate time difference measurement prior distribution. The time difference data is unbiasedly corrected by the bias term to achieve fine positioning of the mine microseismic source. Therefore, the present invention improves the accuracy of time difference measurement data picking, lays a good foundation for time difference measurement uncertainty estimation, solves the problem of solving the probability distribution of non-repeatable microseismic measurement data, accurately obtains the prior distribution of time difference measurement, and achieves high-precision positioning of the microseismic source. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of the earthquake source location method based on Bayesian time difference measurement uncertainty estimation provided by the present invention.

[0037] Figure 2 The present invention provides a granite sample with a known earthquake source. DETAILED DESCRIPTION

[0038] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below with reference to specific illustrations.

[0039] Please refer to Figure 1 As shown, the present invention provides a method for earthquake source location based on Bayesian time difference measurement uncertainty estimation, comprising the following steps:

[0040] S1. Deploy a microseismic monitoring system in a typical mine. The system includes a sensor network and a blasting source with a known source location within the monitoring system. Conduct multiple blasting vibration experiments with known source locations, collect and screen blasting waveform signals, and process the resulting low signal-to-noise ratio waveform signals using improved empirical mode decomposition and wavelet packet noise reduction techniques.

[0041] S2. Automatically picking up the time difference measurement data of different earthquake sources based on the processed waveform signal. Specifically, the existing Akaike information method can be used to automatically pick up the time difference measurement data of the earthquake sources.

[0042] S3. Construct a Bayesian estimation framework to estimate the uncertainty of the time difference measurement data and use Monte Carlo sampling technology to calculate it. The algorithms for estimating the uncertainty of the time difference measurement data and the Monte Carlo sampling technology are written and implemented using Matlab software.

[0043] S4. Based on the estimated uncertainty of the time difference measurement, the bias correction term is added to the time difference data for unbiased correction, and then positioning is performed using the traditional linear positioning method to obtain high-precision microseismic source positioning results.

[0044] As a specific embodiment, the steps of the empirical mode decomposition technology in step S1 specifically include:

[0045] S11, adding multiple adaptive Gaussian white noise time series of different amplitudes to the original time series of the screened burst waveform signal;

[0046] S12, averaging the upper and lower envelopes of each noisy time series to obtain an interpolation of the noisy signal and the corresponding average time as a new time series;

[0047] S13. Continue adding Gaussian white noise to the new time series, and repeat steps S11 and S12 until the new time series meets the termination condition.

[0048] As a specific embodiment, the termination condition in step S13 is:

[0049] The number of extreme values ​​and zero crossings of the new time series is less than or equal to 1; and

[0050] The average value of the upper and lower envelopes of the new time series at any time is 0.

[0051] As a specific embodiment, the wavelet packet denoising technology in step S1 adopts a fixed threshold technology, specifically:

[0052]

[0053] Where T represents the threshold, σ represents the standard deviation of the noise, and N represents the length of the signal in the time domain.

[0054] As a specific embodiment, the fixed threshold technology adopts a soft threshold function, which is specifically expressed as:

[0055]

[0056] Where S(t) represents the soft threshold function, x represents the decomposition factor, and |·| represents the absolute value.

[0057] As a specific embodiment, the step of estimating the uncertainty of the time difference measurement data using the Bayesian estimation framework in step S3 is specifically as follows:

[0058] Assuming that the microseismic monitoring system has been in a stable state during the research period, that is, the uncertainty of the time difference measurement of the microseismic monitoring system remains approximately unchanged, the number of sensor pairs involved in the inversion of the time difference measurement uncertainty is set to k, the number of microseismic and blasting events is set to m, the number of time difference measurement data is n, and 2k≤n is satisfied, then the Bayesian estimation framework is expressed as:

[0059]

[0060] Where P(X) is the marginal likelihood, P(θ) is the initial prior probability of the parameter to be estimated, and the default assumption of the traditional positioning method, that is, the zero-mean Gaussian independent and identically distributed, is used as the prior distribution of the parameter θ; the prior probability P(θ) will be further updated by the sample information X measured by the time difference measurement, P(θ|X) is the updated posterior probability, and P(X|θ) is the likelihood function conditioned on the model parameter θ, expressed as:

[0061]

[0062] Among them, C e represents T(θ) cal -T mea The covariance matrix of , T(θ) cal is the calculated value of the time difference data, T mea is the measurement value of the time difference data, and d represents the dimension of the data.

[0063] As a specific embodiment, the Monte Carlo sampling technique in step S3 specifically includes:

[0064] Taking 0 µs as the initial value for iteration, the time difference measurement uncertainty of each sensor is estimated by iterating 100,000 times. During the iteration, one parameter is updated each time, and the update speed follows Gaussian distribution sampling with a mean of 0 and a standard deviation of 0.25. After the parameter is updated, the probability density function P(X|θ') of the new parameter model θ' is calculated, and the acceptance rate α(m'|m) of the new parameter model is expressed as:

[0065]

[0066] When α(m'|m) is greater than or equal to u, the new model m' is accepted; otherwise, the new model will be rejected; each time u∈[0,1] represents a random number in the range of 0-1, which obeys the uniform distribution.

[0067] As a specific example, please refer to Figure 2As shown, the implementation method is implemented through an indoor lead-breaking experiment. Acoustic emission signals are received using a microseismic waveform acquisition system with a known source on a granite specimen. The acquisition system includes a sensor, preamplifier, signal collector, and signal analyzer. The microseismic timing is simulated by breaking the pencil lead. The fracture angle between the pencil lead and the surface of the granite specimen (where the red cross represents the sensor installation location and the blue dot represents the acoustic emission source) is 30°. Acoustic emission signals are collected using a piezoelectric sensor and amplified by a 40dB preamplifier. The acoustic emission signals are then collected using a signal analyzer and stored on a laptop for further analysis.

[0068] The low signal-to-noise ratio waveform signals obtained by screening are processed using improved empirical mode decomposition and wavelet packet denoising techniques. Time difference measurement data is then extracted based on the processed waveform signals and used to estimate the uncertainty of the time difference measurement. A bias correction term for the estimated time difference measurement uncertainty is added to the time difference data for unbiased correction, and then positioning is performed using a traditional linear positioning method. Taking the acoustic emission source at (25, 0, 50) (unit: mm) as an example, the accuracy before time difference measurement uncertainty estimation is 4.03 mm, while the positioning accuracy after correcting the time difference data is 3.48 mm, demonstrating the feasibility and effectiveness of the Bayesian estimation framework.

[0069] The above embodiments can prove the applicability of the present invention in microseismic positioning, prove the feasibility of the Bayesian estimation framework, and can further improve positioning accuracy.

[0070] Compared with the prior art, the present invention provides a method for earthquake source location based on Bayesian time difference measurement uncertainty estimation, which includes two major parts: microseismic waveform signal collection and processing and Bayesian estimation and positioning of microseismic time difference measurement uncertainty. The microseismic waveform signal collection and processing includes using a signal noise reduction and purification method to reduce the noise of the blasting vibration signal, and continuing to use the Akaike information method to automatically pick up the time difference measurement data of the earthquake source; then, based on the time difference measurement data, sensor coordinates, earthquake source location and wave velocity structure information, a Bayesian estimation framework for time difference measurement uncertainty is constructed to obtain an accurate time difference measurement prior distribution. The time difference data is unbiasedly corrected by the bias term to achieve fine positioning of the mine microseismic source. Therefore, the present invention improves the accuracy of time difference measurement data picking, lays a good foundation for time difference measurement uncertainty estimation, solves the problem of solving the probability distribution of non-repeatable microseismic measurement data, accurately obtains the prior distribution of time difference measurement, and achieves high-precision positioning of the microseismic source.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for earthquake source location based on Bayesian time difference measurement uncertainty estimation, characterized in that: The following steps are involved: S1. Deploy a microseismic monitoring system in a typical mine. The system includes a sensor network and a blasting source with a known source location within the monitoring system. Conduct multiple blasting vibration experiments with known source locations, collect and screen blasting waveform signals, and process the resulting low signal-to-noise ratio waveform signals using improved empirical mode decomposition and wavelet packet noise reduction techniques. S2, automatically picking up the time difference measurement data of different earthquake sources based on the processed waveform signal; S3. Construct a Bayesian estimation framework to estimate the uncertainty of the time difference measurement data and use Monte Carlo sampling technology to calculate it; S4. Based on the estimated uncertainty of the time difference measurement, the bias correction term is added to the time difference data for unbiased correction, and then positioning is performed using the positioning method to obtain high-precision microseismic source positioning results; wherein, The steps of the empirical mode decomposition technology in step S1 specifically include: S11, adding multiple adaptive Gaussian white noise time series of different amplitudes to the original time series of the screened burst waveform signal; S12, averaging the upper and lower envelopes of each noisy time series to obtain an interpolation of the noisy signal and the corresponding average time as a new time series; S13. Continue adding Gaussian white noise to the new time series and repeat steps S11 and S12 until the new time series meets the termination condition; the termination condition is: The number of extreme values ​​and zero crossings of the new time series is less than or equal to 1; and The average value of the upper and lower envelopes of the new time series at any time is 0; The wavelet packet denoising technique in step S1 adopts a fixed threshold technique, specifically: Where T represents the threshold, σ represents the standard deviation of the noise, and N represents the length of the signal in the time domain; The fixed threshold technology adopts a soft threshold function, which is specifically expressed as: Where S(t) represents the soft threshold function, x represents the decomposition factor, and |·| represents the absolute value.

2. The earthquake source location method based on Bayesian time difference measurement uncertainty estimation according to claim 1, characterized in that: The steps of estimating the uncertainty of the time difference measurement data using the Bayesian estimation framework in step S3 are specifically as follows: Assuming that the microseismic monitoring system has been in a stable state during the research period, that is, the uncertainty of the time difference measurement of the microseismic monitoring system remains approximately unchanged, the number of sensor pairs involved in the inversion of the time difference measurement uncertainty is set to k, the number of microseismic and blasting events is set to m, the number of time difference measurement data is n, and 2k≤n is satisfied, then the Bayesian estimation framework is expressed as: Where P(X) is the marginal likelihood, P(θ) is the initial prior probability of the parameter to be estimated, and the default assumption of the traditional positioning method, that is, the zero-mean Gaussian independent and identically distributed, is used as the prior distribution of the parameter θ; the prior probability P(θ) will be further updated by the sample information X measured by the time difference measurement, P(θ|X) is the updated posterior probability, and P(X|θ) is the likelihood function conditioned on the model parameter θ, expressed as: Among them, C e represents T(θ) cal -T mea The covariance matrix of , T(θ) cal is the calculated value of the time difference data, T mea is the measurement value of the time difference data, and d represents the dimension of the data.

3. The earthquake source location method based on Bayesian time difference measurement uncertainty estimation according to claim 2, characterized in that: The Monte Carlo sampling technique in step S3 specifically includes: Taking 0 µs as the initial value for iteration, the time difference measurement uncertainty of each sensor is estimated by iterating 100,000 times. During the iteration, one parameter is updated each time, and the update speed follows Gaussian distribution sampling with a mean of 0 and a standard deviation of 0.

25. After the parameter is updated, the probability density function P(X|θ′) of the new parameter model θ′ is calculated, and the acceptance rate α(m′|m) of the new parameter model is expressed as: When α(m'|m) is greater than or equal to u, the new model m' is accepted; otherwise, the new model will be rejected; each time u∈[0,1] represents a random number in the range of 0-1, which obeys the uniform distribution.

Images