A method for fusing the time series characteristics of microseismic activity parameters

Abstract

The present invention discloses a method for fusing time series features of microseismic activity parameters, including: First, obtaining microseismic monitoring data; Second, obtaining b-value fusion feature values based on the WFTBlock module in WFTNet; Third, obtaining energy index fusion feature values based on the WFTBlock module in WFTNet; Fourth, obtaining Schmidt number fusion feature values based on the WFTBlock module in WFTNet; Fifth, obtaining comprehensive fusion feature values of microseismic activity parameters. The method of the present invention has simple steps and reasonable design. The original sequences of b-value, energy index, and Schmidt number are subjected to feature extraction and fusion through the WFTBlock module in WFTNet, and then the comprehensive fusion feature values of microseismic activity parameters are obtained by averaging the fusion feature values of each parameter, which is convenient for accurately predicting rock bursts through the comprehensive fusion feature values of microseismic activity parameters in the follow-up.

Description

Technical Field

[0001] The present invention belongs to the technical field of microseismic activity feature fusion, and particularly relates to a method for fusing time series features of microseismic activity parameters. Background Technique

[0002] In recent years, the prediction of rock burst disasters based on the characterization of microseismic activity parameters has provided valuable theoretical and practical experience for improving the accuracy of rock burst early warning. However, due to the complexity of microseismic activity parameters themselves in the process of data collection, processing and analysis, the spatio-temporal characteristics of the rock burst events they represent are often relatively vague and uncertain, resulting in the current rock burst disaster prediction based on microseismic activity parameters still mainly relying on pure data-driven deep learning methods. Although these data-driven prediction methods can use a large amount of historical data for training, they have a strong dependence on the potential feature patterns of the data. The prediction methods based on deep learning directly predict based on activity parameters such as b-value, energy index and Schmidt number, without performing feature extraction and fusion on microseismic activity parameters, and cannot understand the local periodic characteristics and global periodicity of the data.

[0003] Therefore, there is currently a lack of a reasonably designed method for fusing time series features of microseismic activity parameters. The original sequences of b-value, energy index and Schmidt number are subjected to feature extraction and fusion through the WFTBlock module in WFTNet to obtain the fused feature values of b-value, energy index and Schmidt number, and then the fused feature values of each parameter are averaged to obtain the comprehensive fused feature value of microseismic activity parameters, which is convenient for accurately predicting rock bursts through the comprehensive fused feature value of microseismic activity parameters in the follow-up. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a method for fusing time series features of microseismic activity parameters in view of the above deficiencies in the prior art. The method steps are simple and reasonably designed. The original sequences of b-value, energy index and Schmidt number are subjected to feature extraction and fusion through the WFTBlock module in WFTNet to obtain the fused feature values of b-value, energy index and Schmidt number, and then the fused feature values of each parameter are averaged to obtain the comprehensive fused feature value of microseismic activity parameters, which is convenient for accurately predicting rock bursts through the comprehensive fused feature value of microseismic activity parameters in the follow-up.

[0005] To solve the above technical problem, the technical solution adopted by the present invention is: A method for fusing time series features of microseismic activity parameters, characterized in that the method includes the following steps:

[0006] Step 1: Obtain microseismic monitoring data:

[0007] Step 101: Use a microseismic monitoring system to monitor the mine area to be monitored, and obtain the microseismic events within a unit time, and the microseismic magnitudes and energies corresponding to each microseismic event; among them, the mine area to be monitored is a mine roadway or a goaf;

[0008] Step 102: Use a computer to obtain the b-value, energy index, and Schmidt number based on the microseismic events within a unit time and the microseismic magnitudes and energies corresponding to each microseismic event;

[0009] Step 103: Repeat Step 101 to Step 102 multiple times to obtain the original b-value sequence, energy index original sequence, and Schmidt number original sequence for n unit times; where n is a positive integer;

[0010] Step Two: Obtain the b-value fusion eigenvalue based on the WFTBlock module in WFTNet:

[0011] Step 201: Set the block length and divide the original b-value sequence into multiple b-value sequence blocks;

[0012] Step 202: Use a computer to input each b-value sequence block into the WFTBlock module in WFTNet to obtain the period weighting coefficient α, Fourier transform eigenvalue and wavelet transform eigenvalue

[0013] Step 203: Use a computer to obtain the b-value fusion eigenvalue of each b-value sequence block according to the formula where is the preset weight of the wavelet transform corresponding to the b-value sequence block, is the preset weight of the Fourier transform corresponding to the b-value sequence block,

[0014] Step Three: Obtain the energy index fusion eigenvalue based on the WFTBlock module in WFTNet;

[0015] According to the method in Step Two, input the original energy index sequence to obtain the energy index fusion eigenvalue of each energy index sequence block

[0016] Step Four: Obtain the Schmidt number fusion eigenvalue based on the WFTBlock module in WFTNet;

[0017] According to the method in Step Two, input the original Schmidt number sequence to obtain the Schmidt number fusion eigenvalue of each Schmidt number sequence block

[0018] ​Step 5. Obtain the comprehensive fusion eigenvalue of microseismic activity parameters:

[0019] Step 501. Perform average processing on the b-value fusion eigenvalue energy index fusion eigenvalue Schmidt number fusion eigenvalue of the corresponding block to obtain the comprehensive fusion eigenvalue F of microseismic activity parameters for each block;

[0020] Step 502. Denote the ratio of the comprehensive fusion eigenvalue F of microseismic activity parameters of each block to the block length as the comprehensive fusion eigenvalue of microseismic activity parameters per unit time.

[0021] The above method for fusing time series characteristics of microseismic activity parameters is characterized in that: in Step 202, the specific process is as follows:

[0022] Step 2021. Input any b-value sequence block into the WFTBlock module in WFTNet using a computer; wherein, the WFTBlock module includes a feature extraction stage and a feature fusion stage, and the feature extraction stage includes a Fourier transform branch and a wavelet transform branch;

[0023] Step 2022. Process any b-value sequence block through the Fourier transform branch of the feature extraction stage in the WFTBlock module using a computer to obtain each frequency and the corresponding amplitude;

[0024] Step 2023. Process any b-value sequence block through the wavelet transform branch of the feature extraction stage in the WFTBlock module using a computer to obtain a time-frequency diagram;

[0025] Step 2024. Pass each frequency and the corresponding amplitude through the feature fusion stage in the WFTBlock module to obtain the Fourier transform eigenvalue Pass the time-frequency diagram through the feature fusion stage in the WFTBlock module to obtain the wavelet transform eigenvalue

[0026] The above method for fusing time series characteristics of microseismic activity parameters is characterized in that: in Step 2022, the specific process is as follows:

[0027] Step A. Arrange each frequency in descending order to obtain the first m frequencies and the amplitudes corresponding to the first m frequencies; wherein, the amplitude corresponding to the i-th frequency is denoted as A i ; i is a positive integer, 1 ≤ i ≤ m; m is a positive integer;

[0028] Step B. Use a computer to calculate according to the formula to obtain the period weighting coefficient α; wherein, Represents the maximum value of the squared amplitude corresponding to the first m frequencies.

[0029] The present invention has the following advantages compared with the prior art:

[0030] 1. The method steps of the present invention are simple and reasonably designed, solving the problem of directly performing intelligent prediction of rock bursts based on multiple variables currently.

[0031] 2. The present invention respectively performs feature extraction and fusion on each microseismic activity parameter based on the WFTBlock module in WFTNet, obtaining the b-value fusion eigenvalue, energy index fusion eigenvalue, and Schmidt number fusion eigenvalue. Through the feature extraction of each microseismic activity parameter, the local periodic features and global periodic features of the data can be captured simultaneously, and the period weighting coefficient adaptively balances these features, further improving the fusion performance of the data set with various features.

[0032] 3. The present invention performs average processing on the b-value fusion eigenvalue, exponential fusion eigenvalue, and Schmidt number fusion eigenvalue to obtain the comprehensive fusion eigenvalue of the microseismic activity parameter, and converts it into the comprehensive fusion eigenvalue of the microseismic activity parameter per unit time, so as to facilitate subsequent rock burst prediction and early warning based on the comprehensive fusion eigenvalue of the microseismic activity parameter per unit time, improving the prediction performance.

[0033] In summary, the method steps of the present invention are simple and reasonably designed. The original b-value sequence, original energy index sequence, and original Schmidt number sequence are subjected to feature extraction and fusion through the WFTBlock module in WFTNet to obtain the b-value fusion eigenvalue, energy index fusion eigenvalue, and Schmidt number fusion eigenvalue, and then the average processing of the fusion eigenvalues of each parameter is performed to obtain the comprehensive fusion eigenvalue of the microseismic activity parameter, which is convenient for subsequent accurate prediction of rock bursts through the comprehensive fusion eigenvalue of the microseismic activity parameter.

[0034] Next, through the drawings and embodiments, the technical solutions of the present invention will be further described in detail. Description of the Drawings

[0035] Figure 1 Is the method flow block diagram of the present invention. Detailed Embodiments

[0036] As Figure 1 shown, the method for fusing the time series features of the microseismic activity parameters of the present invention includes the following steps:

[0037] Step 1. Obtain microseismic monitoring data:

[0038] Step 101: Use a microseismic monitoring system to monitor the mine area to be monitored, and obtain microseismic events within a unit time, as well as the microseismic magnitude and energy corresponding to each microseismic event; among them, the mine area to be monitored is a mine roadway or a goaf.

[0039] Step 102: Use a computer to obtain the b-value, energy index, and Schmidt number based on the microseismic events within a unit time and the microseismic magnitude and energy corresponding to each microseismic event.

[0040] Step 103: Repeat Step 101 to Step 102 multiple times to obtain the original b-value sequence, energy index original sequence, and Schmidt number original sequence for n unit times; where n is a positive integer.

[0041] Step Two: Obtain the b-value fusion eigenvalue based on the WFTBlock module in WFTNet:

[0042] Step 201: Set the block length and divide the original b-value sequence into multiple b-value sequence blocks.

[0043] Step 202: Use a computer to input each b-value sequence block into the WFTBlock module in WFTNet to obtain the period weighting coefficient α, Fourier transform eigenvalue and wavelet transform eigenvalue

[0044] Step 203: Use a computer to obtain the b-value fusion eigenvalue of each b-value sequence block according to the formula where where is the preset weight of the wavelet transform corresponding to the b-value sequence block, is the preset weight of the Fourier transform corresponding to the b-value sequence block,

[0045] Step Three: Obtain the energy index fusion eigenvalue based on the WFTBlock module in WFTNet;

[0046] According to the method in Step Two, input the original energy index sequence to obtain the energy index fusion eigenvalue of each energy index sequence block

[0047] Step Four: Obtain the Schmidt number fusion eigenvalue based on the WFTBlock module in WFTNet;

[0048] According to the method in Step Two, input the original Schmidt number sequence to obtain the Schmidt number fusion eigenvalue of each Schmidt number sequence block

[0049] Step Five: Obtain the comprehensive fusion eigenvalue of microseismic activity parameters:

[0050] Step 501: Fuse the b-value eigenfeatures of the corresponding block Energy index fused eigenfeatures Schmidt number fused eigenfeatures Perform averaging processing to obtain the comprehensive fused eigenfeatures F of the microseismic activity parameters of each block;

[0051] Step 502: Denote the ratio of the comprehensive fused eigenfeatures F of the microseismic activity parameters of each block to the block length as the comprehensive fused eigenfeatures of the microseismic activity parameters per unit time.

[0052] In this embodiment, step 202 is as follows:

[0053] Step 2021: Input any b-value sequence block into the WFTBlock module in WFTNet using a computer; wherein, the WFTBlock module includes a feature extraction stage and a feature fusion stage, and the feature extraction stage includes a Fourier transform branch and a wavelet transform branch;

[0054] Step 2022: Process any b-value sequence block through the Fourier transform branch of the feature extraction stage in the WFTBlock module using a computer to obtain each frequency and the corresponding amplitude;

[0055] Step 2023: Process any b-value sequence block through the wavelet transform branch of the feature extraction stage in the WFTBlock module using a computer to obtain a time-frequency diagram;

[0056] Step 2024: Pass each frequency and the corresponding amplitude through the feature fusion stage in the WFTBlock module to obtain Fourier transform eigenfeatures Pass the time-frequency diagram through the feature fusion stage in the WFTBlock module to obtain wavelet transform eigenfeatures

[0057] In this embodiment, step 2022 is as follows:

[0058] Step A: Arrange each frequency in descending order to obtain the first m frequencies and the amplitudes corresponding to the first m frequencies; wherein, the amplitude corresponding to the i-th frequency is denoted as A i ; i is a positive integer, 1 ≤ i ≤ m; m is a positive integer;

[0059] Step B: Use a computer to calculate according to the formula to obtain the period weighting coefficient α; wherein, represents the maximum value of the squares of the amplitudes corresponding to the first m frequencies.

[0060] In this embodiment, it should be noted that the acquisition of the b value, energy index, and Schmidt number can refer to the parameter acquisition in the patent CN202410667273.5, a multi-variable intelligent prediction method for mine rock bursts.

[0061] In this embodiment, it should be noted that the value of m is 5.

[0062] In this embodiment, it should be noted that the value of n is not less than 100.

[0063] In this embodiment, it should be noted that and take values from 0 to 1, and the weights will represent the importance of local periodic features and global periodic features in calculating the fused feature values. and the values can be adjusted according to actual requirements.

[0064] In this embodiment, it should be noted that since the block contains multiple unit timestamps, these periodic features need to be transformed into comprehensive fused feature values corresponding to each unit timestamp for subsequent time series analysis. To simplify this process, the averaging method of period-weighted features is adopted. The comprehensive fused feature value of the microseismic activity parameters of the block and the average feature value of the block length are assigned to all unit timestamps within the block.

[0065] In this embodiment, it should be noted that the block length is 3, that is, one block represents the overall features of 3 days.

[0066] In this embodiment, the unit time is one day, which can be adjusted according to actual requirements.

[0067] In summary, the method of the present invention has simple steps and reasonable design. The original sequences of the b value, energy index, and Schmidt number are subjected to feature extraction and fusion through the WFTBlock module in WFTNet to obtain the fused feature values of the b value, energy index, and Schmidt number. Then, the average processing of the fused feature values of each parameter is performed to obtain the comprehensive fused feature value of the microseismic activity parameters, which is convenient for accurately predicting rock bursts through the comprehensive fused feature value of the microseismic activity parameters subsequently.

[0068] The above is only a preferred embodiment of the present invention, and does not impose any limitation on the present invention. Any simple modification, change, and equivalent structural change made to the above embodiments according to the technical essence of the present invention still fall within the protection scope of the technical solution of the present invention.

Claims

1. A method for fusing the time series characteristics of microseismic activity parameters, characterized in that The method includes the following steps: Step 1: Obtain microseismic monitoring data: Step 101: Use a microseismic monitoring system to monitor the mine area to be monitored, and obtain microseismic events within a unit time, and the microseismic magnitude and energy corresponding to each microseismic event; wherein, the mine area to be monitored is a mine roadway or a goaf; Step 102: Use a computer to obtain the b-value, energy index, and Schmidt number based on the microseismic events within a unit time and the microseismic magnitude and energy corresponding to each microseismic event; Step 103: Repeat Step 101 to Step 102 multiple times to obtain the original sequences of b-values, energy indices, and Schmidt numbers for n unit times; wherein, n is a positive integer; Step 2: Obtain the b-value fusion eigenvalue based on the WFTBlock module in WFTNet: Step 201: Set the block length and divide the original b-value sequence into multiple b-value sequence blocks; Step 202: Use a computer to input each b-value sequence block into the WFTBlock module in WFTNet, and obtain the periodic weighting coefficient α, Fourier transform eigenvalue, and wavelet transform eigenvalue Step 203: Use a computer to calculate according to the formula to obtain the b-value fusion eigenvalues of each b-value sequence block where is the preset weight of wavelet transform corresponding to the b-value sequence block, is the preset weight of Fourier transform corresponding to the b-value sequence block, Step 3: Obtain the energy index fusion eigenvalue based on the WFTBlock module in WFTNet; According to the method in Step 2, input the original sequence of energy indices to obtain the energy index fusion eigenvalues of each energy index sequence block Step 4: Obtain the Schmidt number fusion eigenvalue based on the WFTBlock module in WFTNet; According to the method in Step 2, input the original Schmidt number sequence to obtain the Schmidt number fusion eigenvalues of each Schmidt number sequence block Step 5: Obtain the comprehensive fusion eigenvalue of microseismic activity parameters: Step 501: Fuse the b-value eigenvalue of the corresponding block Energy index fusion eigenvalue Schmidt number fusion eigenvalue Perform averaging processing to obtain the comprehensive fusion eigenvalue F of the microseismic activity parameters of each block; Step 502: Denote the ratio of the comprehensive fusion eigenvalue F of the microseismic activity parameters of each block to the block length as the comprehensive fusion eigenvalue of the microseismic activity parameters per unit time.

2. The microseismic activity parameter time series feature fusion method according to claim 1, wherein: Step 202, the specific process is as follows: Step 2021: Use a computer to input any b-value sequence block into the WFTBlock module in WFTNet; wherein, the WFTBlock module includes a feature extraction stage and a feature fusion stage, and the feature extraction stage includes a Fourier transform branch and a wavelet transform branch; Step 2022: Use a computer to process any b-value sequence block through the Fourier transform branch of the feature extraction stage in the WFTBlock module to obtain each frequency and the corresponding amplitude; Step 2023: Use a computer to process any b-value sequence block through the wavelet transform branch of the feature extraction stage in the WFTBlock module to obtain a time-frequency diagram; Step 2024: Pass each frequency and its corresponding amplitude through the feature fusion stage in the WFTBlock module to obtain the Fourier transform eigenvalue Pass the time-frequency diagram through the feature fusion stage in the WFTBlock module to obtain the wavelet transform eigenvalue 3. A method for fusing time series characteristics of microseismic activity parameters according to claim 1, characterized in that: Step 2022, the specific process is as follows: Step A: Arrange each frequency in descending order, and obtain the first m frequencies and the amplitudes corresponding to the first m frequencies; where the amplitude corresponding to the i-th frequency is denoted as A i ; i is a positive integer, 1 ≤ i ≤ m; m is a positive integer; Step B. Use a computer to calculate according to the formula to obtain the period weighting coefficient α; where represents the maximum value of the squared amplitudes corresponding to the first m frequencies.

Images