Method and system for automatically cleaning seismic data during excavation

Abstract

The invention discloses an automatic cleaning method and system for digging seismic data, and the method comprises the steps: collecting seismic data, carrying out the processing of the seismic data through first processing, and determining an optimal reference channel; and performing data evaluation on the seismic data by adopting the second processing and the optimal reference channel to obtain a quality score, and determining a data cleaning threshold according to the quality score. The method comprises the following steps of: performing data exploration, determining an optimal reference trace, performing mutual dry seismic interference by using all trace data and the optimal reference trace, extracting an abstracted contour of a generated mutual dry seismic interference profile, analyzing whether energy peaks with absolute advantages exist in the abstracted contour trace by trace, and dividing the trace number of the energy peaks with the absolute advantages by the total trace number to obtain a seismic interference result. And obtaining a quality evaluation result. The cleaning threshold of the high-quality data is determined through the seismic data, and the high-quality data is cleaned by using the threshold. And then false high-quality data in the selected high-quality data are independently removed, so that the technical problem of inaccurate result of the seismic data cleaning method in the prior art is solved.

Description

Technical Field

[0001] The invention belongs to the field of seismic data cleaning with mining machinery as the seismic source, and relates to a data cleaning method, in particular to an automatic cleaning method and system for seismic data during excavation. Background Art

[0002] When using a tunnel boring machine as the seismic source for advanced detection of the head and side walls of a coal mine tunnel, seismic data is collected continuously for 24 hours, which will generate a huge amount of data, of which high-quality data accounts for a very small proportion. If data cleaning is not performed, poor-quality data will cause pollution, which will not only waste computing resources and take a long time, but also fail to obtain good results. Therefore, data cleaning is necessary. Since seismic data is collected continuously for 24 hours and the data volume is huge, this data cleaning method must be completely separated from human intervention. Any cleaning method that requires human intervention will be exhausted in the face of an endless stream of data, causing the idea of ​​real-time processing and real-time imaging to fail.

[0003] There are currently two ways to automatically clean data. The first is to use the fact that the belt conveyor is running during tunneling, and the entire tunnel has strong vibrations. By scanning the amplitude of the data channel, the amplitude threshold for judging whether the tunneling machine is working is determined, and the working period and rest period of the tunneling machine are divided accordingly. However, when the tunneling machine is working, it does not always forcefully cut the coal wall. There are times when it digs the floating coal on the ground. At this time, the belt conveyor is still running, and the amplitude of the data channel is also large, but the corresponding data quality is not good. Therefore, this method cannot guarantee that the quality of the selected data must be excellent. The second method is to select a channel as a reference channel, correlate the reference channel signal with the remaining channels in the time domain, and judge the quality of the data by analyzing the correlation results. Although this method has certain effects, it has three shortcomings: 1. Due to time domain correlation, the time consumption increases exponentially with respect to the data length, and long data blocks will lead to a significant increase in time consumption; 2. It is unable to eliminate high-correlation data caused by large vibration signals such as microseismicity that are not caused by tunnel boring machines; 3. The entire process requires manual determination of the best reference track, manual determination of data cleaning standards, and manual inspection of whether there is high-correlation data caused by sources other than tunnel boring machines, which cannot meet the requirements of automatic data processing.

[0004] In summary, the existing methods for cleaning seismic data during excavation have the problems of inaccurate results, time-consuming calculations, frequent manual intervention, and inability to achieve unattended automated cleaning. Summary of the invention

[0005] In view of the shortcomings of the prior art, the purpose of the present invention is to provide an automatic cleaning method and system for seismic data during excavation, so as to solve the technical problem of inaccurate results of seismic data cleaning methods in the prior art.

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

[0007] An automatic cleaning method for seismic data during excavation is based on a seismic data acquisition system, wherein the seismic data acquisition system includes a plurality of geophones and signal transmission cables, wherein the geophones are installed on coal wall anchors; the seismic data is a vibration signal simultaneously acquired by a plurality of geophones when a tunneling machine cuts a coal wall and causes vibration, and a vibration signal acquired by a geophone is called a data channel, and specifically includes the following steps:

[0008] Step 1, collecting seismic data, processing the seismic data using a first process, and determining an optimal reference trace;

[0009] The first process includes process A' and process B';

[0010] The input of the process A' is a two-dimensional original data segment r and the track number n of the data track closest to the roadheader, and the output is a pulse intercept segment No. 1;

[0011] The input of the process B' is the main frequency F of the direct wave when the first pulse cutting section and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r;

[0012] Step 2, using the second processing and the best reference trace determined in step 1 to evaluate the seismic data, obtain a quality score, and determine a data cleaning threshold according to the quality score;

[0013] The second process includes process A" and process B";

[0014] The input of the process A" is a plurality of two-dimensional original data segments R and the track number n of the best reference track determined in step 1, and the output is a plurality of No. 2 pulse interception segments;

[0015] The input of the process B" is a plurality of No. 2 pulse interception sections and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R;

[0016] Step 3, using the data cleaning threshold obtained in step 2 to clean the seismic data, obtain high-quality data segments, and remove false high-quality data in the high-quality data segments based on process B'';

[0017] The input of the process B'' is the two-dimensional original data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

[0018] Step 1 specifically includes the following steps:

[0019] Step 1.1, after laying the survey line and successfully collecting and uploading the data, select the seismic data of the first day when the tunnel boring machine has tunneling activities;

[0020] Step 1.2, slicing the seismic data obtained in step 1.1 using a first preset time to obtain a plurality of two-dimensional original data segments r, sequentially sending the plurality of two-dimensional original data segments r and the channel number n of the data channel closest to the tunnel boring machine into process A', and outputting a first pulse interception segment;

[0021] Step 1.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the No. 1 pulse interception section ω , and the pulse interception section No. 1 obtained in step 1.2 and the main frequency F of the direct wave when the roadheader cuts the coal wall are respectively ω Input process B' to obtain the quality scores of all two-dimensional raw data segments r;

[0022] Step 1.4, selecting a data segment corresponding to the two-dimensional original data segment r whose quality score is greater than a first threshold;

[0023] Step 1.5, using the data segments obtained in step 1.4, taking each data track as a reference track, searching for the best reference track among all tracks, and obtaining a plurality of candidate best reference tracks;

[0024] Step 1.6, using the following formula to calculate ref(b) of the best reference track to be selected obtained in step 1.5, and taking the best reference track to be selected corresponding to its maximum value as the best reference track;

[0025]

[0026] in:

[0027] cov(b,j) represents the covariance between the best reference track b and data track j;

[0028] cov(j,j) represents the variance of the jth data track;

[0029] N represents the number of all data channels;

[0030] ref(b) represents the sum of squares of correlation coefficients between the candidate best reference trace b and all candidate best reference traces.

[0031] Step 2 specifically includes the following steps:

[0032] Step 2.1, slicing and segmenting the seismic data using a second preset time to obtain a plurality of two-dimensional original data segments R;

[0033] Step 2.2, sequentially sending the multiple two-dimensional original data segments R obtained in step 2.1 and the channel number n of the best reference channel determined in step 1 to process A'' to obtain the second pulse interception segment corresponding to each of the multiple two-dimensional original data segments R;

[0034] Step 2.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the second pulse interception section ω , and sequentially compare the multiple No. 2 pulse interception segments obtained in step 2.2 with the main frequency F of the direct wave when the roadheader cuts the coal wall ω Input data flow B” to obtain the quality scores of all two-dimensional raw data segments R, that is, to obtain the quality score curve of seismic data;

[0035] Step 2.4: Determine the data cleaning threshold based on the quality score curve obtained in step 2.3.

[0036] The data cleaning threshold determined according to the quality score curve obtained in step 2.3 is specifically:

[0037] Step 2.4.1: For the two-dimensional original data segments R with a quality score lower than P, they are directly cleaned; for the two-dimensional original data segments R with a quality score higher than Q, they are determined to be high-quality data segments; for the two-dimensional original data segments R with a quality score falling within the range of [P, Q], they are determined to be optional data;

[0038] In step 2.4.2, the optional data in the range [P, Q] are sorted from large to small according to the quality score, and the quality scores in the top 80% are taken as the data cleaning threshold.

[0039] Step three specifically includes the following steps:

[0040] Step 3.1, according to the data cleaning threshold obtained in step 2, remove the data segments in the seismic data whose quality scores are lower than the data cleaning threshold to obtain high-quality data segments;

[0041] Step 3.2, remove the false high-quality data in the high-quality data segment obtained in step 3.1.

[0042] The method of removing false high-quality data from the high-quality data segment is specifically:

[0043] Step 3.2.1: Obtain the main frequency F of the microseismic event on the high-quality data segment m , the two-dimensional original data segment r corresponding to the selected high-quality data segment and the main frequency F of the microseismic event m Input process B'', output the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment;

[0044] Step 3.2.2, determine whether the quality score obtained in step 3.2.1 exceeds the second threshold. If so, the two-dimensional original data segment r is false high-quality data caused by a large vibration signal not caused by tunnel boring machine activity, and set its quality score to -1 and directly discard it.

[0045] Process A' specifically includes the following processes:

[0046] Q1.1, according to the data channel number n closest to the tunnel boring machine, find the data channel corresponding to the channel number n from the input two-dimensional data segment r, and save it as the reference channel r(n);

[0047] Q1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N];

[0048] Q2.3, the length of the middle part of the time domain mutual interference result obtained in Q1.2 is T SEG The data is of length T SEG The two-dimensional output data, namely the No. 1 pulse intercept segment;

[0049] Process A" specifically includes the following processes:

[0050] G1.1, according to the track number n of the best reference track determined in step 1, find the corresponding data track from the input two-dimensional data segment R, and save it as the reference track r(n);

[0051] G1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N];

[0052] G1.3, the length of the middle part of the time domain mutual interference result obtained in process 1.2 is T SEG The data is of length T SEG The two-dimensional output data is the second pulse intercept segment.

[0053] Process B' specifically includes the following processes:

[0054] Q3.1, extract the contour features of the intercepted segment of pulse No. 1 and abstract it to obtain m data abstract contours of each data channel;

[0055] Q3.2, analyze whether the m data abstract profiles of a data channel obtained in Q3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in its data abstract profile. If there is no absolutely dominant energy peak, then this data channel is set to zero;

[0056] Q3.3, perform the operation of Q3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the first pulse interception segment;

[0057] The value of the ratio is between [0, 1];

[0058] Process B" specifically includes the following processes:

[0059] G3.1, extract the contour features of the second pulse segment and abstract it to obtain m data abstract contours of each data channel;

[0060] G 3.2, analyzing whether the m data abstract profiles of a data channel obtained in G 3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in the data abstract profile; if there is no absolutely dominant energy peak, then this data channel is set to zero;

[0061] G3.3, perform the operation of G3.2 on all data channels, calculate the ratio of the number of non-zero data channels to the total number of data channels, and the ratio is the quality score of the second pulse interception segment;

[0062] The value of the ratio is between [0, 1];

[0063] Process B"' specifically includes the following processes:

[0064] W3.1, extract contour features from the two-dimensional original data segment r corresponding to the high-quality data segment, and abstract it to obtain m data abstract contours for each data channel;

[0065] W3.2, analyze whether the m data abstract profiles of a data channel obtained in W3.1 have an energy peak that is absolutely dominant. If there is an energy peak that is absolutely dominant, then only the energy peak is retained in its data abstract profile; if there is no energy peak that is absolutely dominant, then this data channel is set to zero;

[0066] W3.3, perform the operation of W3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment;

[0067] The value of the ratio is between [0, 1].

[0068] 9. The method for automatic cleaning of seismic data during excavation as claimed in claim 1, characterized in that the process C specifically includes the following processes:

[0069] Process 1.1, transform the reference channel r(n) into the frequency domain, denoted as G(r n,s )W(s), where G(rn,s ) is the Green's function from source s to detector n, W(s) is the frequency domain transform of the source signal;

[0070] Process 1.2, transform the data channel r(i) into the frequency domain, denoted as G(r i,s )W(s), where G(r i,s ) is the Green's function from source s to detector i, W(s) is the frequency domain transform of the source signal;

[0071] Process 1.3, according to the cross-correlation operation in the time domain is equivalent to the conjugate multiplication of the two spectra in the frequency domain, the frequency domain transform C(n,i) of the time domain cross-correlation function of the reference channel r(n) and the data channel r(i) is obtained;

[0072] C(n,i)=W(s)G(r n,s )W * (s)G * (r i ,s)=|W(s)| 2 G(r n,s )G * (r i ,s)

[0073] in:

[0074] W(s) is the frequency domain transform of the source function;

[0075] W * (s) is the complex conjugate of the frequency domain transform of the source function;

[0076] |W(s)| 2 =W(s)W * (s) is the frequency domain conjugate product of the autocorrelation function of the source function, which represents the influence of the source function on the cross-correlation function between the reference channel r(n) and the data channel r(i) in the above formula;

[0077] G * (r i , s) is the complex conjugate of the Green's function from source s to receiver i;

[0078] Process 1.4, according to the following formula, the expression H(n,i) in the frequency domain of the cross-correlation result between the reference channel r(n) and the data channel r(i) in the time domain is obtained after removing the source effect and normalizing;

[0079]

[0080] in:

[0081] |C(n,n)| is the modulus of the frequency domain transform of the autocorrelation function of the reference channel r(n);

[0082] |C(i,i)| is the modulus of the frequency domain transform of the autocorrelation function of data channel r(i);

[0083] Process 1.5, perform inverse Fourier transform on H(n,i) obtained in process 1.4 to obtain the cross-correlation result h(n,i) in the time domain after removing the source effect and normalizing the signals r(n) and r(i);

[0084]

[0085] In G3.1, Q3.1 or W3.1, the specific operation of obtaining m data abstract profiles for each data channel is: ω The length is an interval, and the first pulse interception segment, the second pulse interception segment or the high-quality data segment with a length of T is cut into m segments;

[0086]

[0087] T ω =1 / F ω

[0088] in:

[0089] T ω Indicates the length of time;

[0090] The ceil function means rounding up to an integer, that is, the number at the end is less than 2T ω The length part is made into a separate section.

[0091] An automatic cleaning system for seismic data while drilling, based on the automatic cleaning method for seismic data while drilling, comprises a data acquisition unit, a data quality evaluation unit and a data cleaning unit connected in sequence;

[0092] The data acquisition unit is used to process the seismic data using a first process to determine an optimal reference trace;

[0093] The first process includes process A' and process B';

[0094] The input of the process A' is a two-dimensional original data segment r and the track number n of the data track closest to the roadheader, and the output is a pulse intercept segment No. 1;

[0095] The input of the process B' is the main frequency F of the direct wave when the first pulse cutting section and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r;

[0096] The data quality evaluation unit is used to evaluate the seismic data using the second processing and the best reference trace to obtain a quality score, and determine a data cleaning threshold according to the quality score;

[0097] The second process includes process A" and process B";

[0098] The input of the process A" is a plurality of two-dimensional original data segments R and the track number n of the best reference track, and the output is a plurality of second pulse interception segments;

[0099] The input of the process B" is a plurality of No. 2 pulse interception sections and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R;

[0100] The data cleaning unit is used to clean the seismic data using the obtained data cleaning threshold to obtain high-quality data segments, and to eliminate false high-quality data in the high-quality data segments based on process B'';

[0101] The input of the process B'' is the two-dimensional original data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

[0102] Compared with the prior art, the present invention has the following beneficial technical effects:

[0103] (I) In the present invention, the best reference channel is first determined through data investigation, and then all channel data are used to perform mutual interference seismic interference with the best reference channel. The abstract outline of the generated mutual interference seismic interference profile is extracted, and the abstract outline is analyzed channel by channel to see if there is an absolutely dominant energy peak. The number of channels with an absolutely dominant energy peak is divided by the total number of channels to obtain the quality evaluation result. Furthermore, the cleaning threshold of high-quality data is determined through seismic data, and this threshold is used to clean out high-quality data that can be sent to the next step of the process. Before actually sending it to the next step of the process, the false high-quality data caused by large vibration signals caused by non-tunneling machine activities such as microseismic are separately eliminated from these selected high-quality data, thereby solving the technical problem of inaccurate results of seismic data cleaning methods in the prior art.

[0104] (II) In the present invention, the greater the vibration caused by the tunnel boring machine cutting the coal wall, the stronger the ability to suppress other interferences in the entire arrangement, and the higher the data quality. Therefore, after the mutual interference seismic interference processing, there are more channels with energy peaks that are absolutely dominant. Therefore, it is very reasonable to score the quality of the data by dividing the number of channels with energy peaks that are absolutely dominant by the total number of channels (for false high-quality data caused by microseismic events, etc., special cleaning is done). On this basis, by automatically determining the best reference channel, automatically determining the data cleaning standard, and automatically cleaning false high-quality data, the entire process is run, and the process adopted greatly reduces the need for manual intervention, and automatically cleans out high-quality and real high-quality data from massive on-dig seismic data, laying the first step for realizing fully automatic processing of on-dig seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] Figure 1 The automatic data cleaning flow chart of the present invention;

[0106] Figure 2 This is the data quality evaluation result of the seismic data of a coal mine in the past 10 hours divided into 1-minute segments in this embodiment;

[0107] Figure 3 The seismic interference result of the excavation seismic data in this embodiment;

[0108] Figure 4 This is the evaluation result after abstract outline in this embodiment.

[0109] The specific contents of the present invention are further explained in detail below in conjunction with embodiments. DETAILED DESCRIPTION

[0110] It should be noted that, unless otherwise specified, all components in the present invention are components known in the art.

[0111] Specific embodiments of the present invention are given below. It should be noted that the present invention is not limited to the following specific embodiments, and all equivalent changes made on the basis of the technical solution of this application fall within the protection scope of the present invention.

[0112] The present invention provides an automatic cleaning method for seismic data during excavation, which is based on a seismic data acquisition system. The seismic data acquisition system includes a plurality of geophones and signal transmission cables. The geophones are installed on coal wall anchors. The seismic data is a vibration signal collected simultaneously by a plurality of geophones when a tunneling machine cuts a coal wall and causes vibration. The vibration signal collected by one geophone is called a data channel. The method specifically includes the following steps:

[0113] Step 1, collecting seismic data, processing the seismic data using a first process, and determining an optimal reference trace;

[0114] The first process includes process A' and process B';

[0115] The input of process A' is a two-dimensional raw data segment r and the track number n of the data track closest to the roadheader, and the output is the No. 1 pulse intercept segment;

[0116] The input of process B' is the main frequency F of the direct wave when the No. 1 pulse is cut and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r;

[0117] Step 2, using the second processing and the best reference trace determined in step 1 to evaluate the seismic data, obtain a quality score, and determine a data cleaning threshold according to the quality score;

[0118] The second process includes process A” and process B”;

[0119] The input of process A" is a plurality of two-dimensional original data segments R and the track number n of the best reference track determined in step 1, and the output is a plurality of No. 2 pulse interception segments;

[0120] The input of process B is multiple No. 2 pulse interception segments and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R;

[0121] Step 3, using the data cleaning threshold obtained in step 2 to clean the seismic data, obtain high-quality data segments, and remove false high-quality data in the high-quality data segments based on process B'';

[0122] The input of process B'' is the two-dimensional raw data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

[0123] In this scheme, the geophone is arranged behind the tunnel boring machine, about 40 to 50 meters away from the tunnel boring machine (the geophone cannot be installed in this range). The geophone is installed on the coal wall anchor to receive the vibration signal from the coal seam. The larger vibrations that may be transmitted from the coal seam are: first, microseismic events caused by roof rupture; second, vibrations caused by the tunnel boring machine cutting the coal wall; third, vibrations caused by the belt conveyor transporting coal. Among them, only the vibration caused by the tunnel boring machine cutting the coal wall is the vibration used in seismic exploration during excavation;

[0124] Seismic data is two-dimensional data related to time and detector number, and data channel is one-dimensional data related to time;

[0125] The roadheader has the following three working states, among which the state of actively cutting the coal wall is the seismic data required in this scheme. Therefore, the quality score is used to select the data segment when the roadheader is in the state of actively cutting the coal wall in the seismic data.

[0126] (1) Complete shutdown state. At this time, the tunnel boring machine and belt conveyor are not working, the tunnel is relatively quiet, and there is no effective signal.

[0127] (2) Coal-digging state: At this time, the cutting head of the tunnel boring machine digs loose the floating coal on the surface, and the shovel board sends the floating coal to the belt conveyor and transports it to the surface. Both the tunnel boring machine and the belt conveyor are working, but there is a high-quality signal only when the cutting head cuts the hard coal wall. There is no high-quality signal at other times.

[0128] (3) Actively cutting the coal wall. At this time, the cutting head of the tunnel boring machine presses against the coal wall and actively cuts it, generating a large number of high-quality signals.

[0129] First, the best reference channel is determined through data investigation. Then, all channel data are used to perform mutual seismic interference with the best reference channel. The abstract outline of the generated mutual seismic interference profile is extracted. The abstract outline is analyzed channel by channel to see if there is an absolutely dominant energy peak. The number of channels with an absolutely dominant energy peak is divided by the total number of channels to obtain the quality evaluation result. Furthermore, the cleaning threshold of high-quality data is determined through seismic data, and this threshold is used to clean out high-quality data that can be sent to the next step of the process. Before actually sending it to the next step of the process, the false high-quality data caused by large vibration signals caused by non-tunneling machine activities such as microseismic activities are separately eliminated from these selected high-quality data, solving the technical problem of inaccurate results of seismic data cleaning methods in the prior art.

[0130] The greater the vibration caused by the tunnel boring machine cutting the coal wall, the stronger the ability to suppress other interference in the entire arrangement, and the higher the data quality. Therefore, after the mutual interference seismic interference processing, there are more channels with absolutely dominant energy peaks. Therefore, it is very reasonable to score the data quality by dividing the number of channels with absolutely dominant energy peaks by the total number of channels (for false high-quality data caused by microseismic events, etc., special cleaning is done). On this basis, by automatically determining the best reference channel, automatically determining the data cleaning standard, and automatically cleaning false high-quality data, the entire process is run, and the process adopted greatly reduces the need for manual intervention, and automatically cleans out high-quality and real high-quality data from massive on-dig seismic data, laying the first step for the fully automatic processing of on-dig seismic data.

[0131] Step 1 specifically includes the following steps:

[0132] Step 1.1, after laying the survey line and successfully collecting and uploading the data, select the seismic data of the first day when the tunnel boring machine has tunneling activities;

[0133] Step 1.2, slicing the seismic data obtained in step 1.1 using a first preset time to obtain a plurality of two-dimensional original data segments r, sequentially sending the plurality of two-dimensional original data segments r and the channel number n of the data channel closest to the tunnel boring machine into process A', and outputting a first pulse interception segment;

[0134] Step 1.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the No. 1 pulse interception section ω , and the pulse interception section No. 1 obtained in step 1.2 and the main frequency F of the direct wave when the roadheader cuts the coal wall are respectively ω Input process B' to obtain the quality scores of all two-dimensional raw data segments r;

[0135] Step 1.4, selecting a data segment corresponding to the two-dimensional original data segment r whose quality score is greater than a first threshold;

[0136] Step 1.5, using the data segments obtained in step 1.4, taking each data track as a reference track, searching for the best reference track among all tracks, and obtaining a plurality of candidate best reference tracks;

[0137] Step 1.6, using the following formula to calculate ref(b) of the best reference track to be selected obtained in step 1.5, and taking the best reference track to be selected corresponding to its maximum value as the best reference track;

[0138]

[0139] in:

[0140] cov(b,j) represents the covariance between the best reference track b and data track j;

[0141] cov(j,j) represents the variance of the jth data track;

[0142] N represents the number of all data channels;

[0143] ref(b) represents the sum of squares of correlation coefficients between the candidate best reference trace b and all candidate best reference traces.

[0144] In the above technical solution, the first preset time is five minutes, and the first threshold value can be selected as 0.5.

[0145] Step 2 specifically includes the following steps:

[0146] Step 2.1, slicing and segmenting the seismic data using a second preset time to obtain a plurality of two-dimensional original data segments R;

[0147] Step 2.2, sequentially sending the multiple two-dimensional original data segments R obtained in step 2.1 and the channel number n of the best reference channel determined in step 1 to process A'' to obtain the second pulse interception segment corresponding to each of the multiple two-dimensional original data segments R;

[0148] Step 2.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the second pulse interception section ω , and sequentially compare the multiple No. 2 pulse interception segments obtained in step 2.2 with the main frequency F of the direct wave when the roadheader cuts the coal wall ω Input data flow B” to obtain the quality scores of all two-dimensional raw data segments R, that is, to obtain the quality score curve of seismic data;

[0149] Step 2.4: Determine the data cleaning threshold based on the quality score curve obtained in step 2.3.

[0150] In the above technical solution, the second preset time ranges from 1 to 10 minutes.

[0151] According to the quality score curve obtained in step 2.3, the data cleaning threshold is determined as follows:

[0152] Step 2.4.1: For the two-dimensional original data segments R with a quality score lower than P, they are directly cleaned; for the two-dimensional original data segments R with a quality score higher than Q, they are determined to be high-quality data segments; for the two-dimensional original data segments R with a quality score falling within the range of [P, Q], they are determined to be optional data;

[0153] In step 2.4.2, the optional data in the range [P, Q] are sorted from large to small according to the quality score, and the quality scores in the top 80% are taken as the data cleaning threshold.

[0154] In the above technical solution, [P, Q] takes the value of [0.3, 0.7];

[0155] The working state of the roadheader corresponding to the two-dimensional original data segment R with a quality score lower than P is: completely stopped state; the working state of the roadheader corresponding to the two-dimensional original data segment R with a quality score higher than Q is: actively cutting the coal wall state; the working state of the roadheader corresponding to the two-dimensional original data segment R with a quality score falling within the range of [P, Q] is: coal digging state.

[0156] Step three specifically includes the following steps:

[0157] Step 3.1, according to the data cleaning threshold obtained in step 2, remove the data segments in the seismic data whose quality scores are lower than the data cleaning threshold to obtain high-quality data segments;

[0158] Step 3.2, remove the false high-quality data in the high-quality data segment obtained in step 3.1.

[0159] The specific steps to remove false high-quality data from the high-quality data segment are:

[0160] Step 3.2.1: Obtain the main frequency F of the microseismic event on the high-quality data segment m , the two-dimensional original data segment r corresponding to the selected high-quality data segment and the main frequency F of the microseismic event m Input process B'', output the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment;

[0161] Step 3.2.2, determine whether the quality score obtained in step 3.2.1 exceeds the second threshold. If so, the two-dimensional original data segment r is false high-quality data caused by a large vibration signal not caused by tunnel boring machine activity, and set its quality score to -1 and directly discard it.

[0162] In the above technical solution, the second threshold can be reduced to 0.7.

[0163] Process A' specifically includes the following processes:

[0164] Q1.1, according to the data channel number n closest to the tunnel boring machine, find the data channel corresponding to the channel number n from the input two-dimensional data segment r, and save it as the reference channel r(n);

[0165] Q1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N];

[0166] Q2.3, the length of the middle part of the time domain mutual interference result obtained in Q1.2 is T SEG The data is of length T SEG The two-dimensional output data, namely the No. 1 pulse intercept segment;

[0167] Process A" specifically includes the following processes:

[0168] G1.1, according to the track number n of the best reference track determined in step 1, find the corresponding data track from the input two-dimensional data segment R, and save it as the reference track r(n);

[0169] G1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N];

[0170] G1.3, the length of the middle part of the time domain mutual interference result obtained in process 1.2 is T SEG The data is of length T SEG The two-dimensional output data is the second pulse intercept segment.

[0171] Process B' specifically includes the following processes:

[0172] Q3.1, extract the contour features of the intercepted segment of pulse No. 1 and abstract it to obtain m data abstract contours of each data channel;

[0173] Q3.2, analyze whether the m data abstract profiles of a data channel obtained in Q3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in its data abstract profile. If there is no absolutely dominant energy peak, then this data channel is set to zero;

[0174] Q3.3, perform the operation of Q3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the first pulse interception segment;

[0175] The value of the ratio is between [0, 1];

[0176] Process B" specifically includes the following processes:

[0177] G3.1, extract the contour features of the second pulse segment and abstract it to obtain m data abstract contours of each data channel;

[0178] G 3.2, analyzing whether the m data abstract profiles of a data channel obtained in G 3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in the data abstract profile; if there is no absolutely dominant energy peak, then this data channel is set to zero;

[0179] G3.3, perform the operation of G3.2 on all data channels, calculate the ratio of the number of non-zero data channels to the total number of data channels, and the ratio is the quality score of the second pulse interception segment;

[0180] The value of the ratio is between [0, 1];

[0181] Process B"' specifically includes the following processes:

[0182] W3.1, extract contour features from the two-dimensional original data segment r corresponding to the high-quality data segment, and abstract it to obtain m data abstract contours for each data channel;

[0183] W3.2, analyze whether the m data abstract profiles of a data channel obtained in W3.1 have an energy peak that is absolutely dominant. If there is an energy peak that is absolutely dominant, then only the energy peak is retained in its data abstract profile; if there is no energy peak that is absolutely dominant, then this data channel is set to zero;

[0184] W3.3, perform the operation of W3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment;

[0185] The value of the ratio is between [0, 1].

[0186] In the above technical solution, in process B' and process B", the input pulse interception segment, F ω It is the main frequency of the direct wave during the period when the roadheader cuts the coal wall. The purpose is to evaluate the data quality. In process B”, the input is the high-quality data segment, F ω It is the dominant frequency of microseismic events, and its purpose is to eliminate false high-quality data.

[0187] In the above 3.2, the segment length used for segmentation is 2T ω, which is twice the time length of the wavelength of the strong energy main frequency. Therefore, the strong energy main frequency wave part in the data profile can be summarized and described by 3-6 abstract sample points. The m data abstract profiles of a data channel obtained in 3.2 are sorted from large to small, and the first 6 abstract sample points are used to estimate the main amplitude of the strong energy main frequency wave. There is no strong energy main frequency wave amplitude outside of 3 times the number of sample points (3*6=18) (the direct wave or microseismic signal propagates through the surrounding rock without dispersion effect, and its wave velocity is relatively stable, which means that its wave train length will not extend too long, so this estimate is very reasonable). Take the 19th-ranked value as the measurement standard, and set the value of m abstract points that is less than or equal to 2 times this value to zero (this threshold value is for example, and can be slightly increased or decreased). If there is no data greater than 2 times this value in the entire abstract data channel, it means that there is no energy peak that is absolutely dominant in this channel, that is, there is definitely no direct wave, and the entire abstract data channel is set to zero.

[0188] Process C specifically includes the following processes:

[0189] Process 1.1, transform the reference channel r(n) into the frequency domain, denoted as G(r n,s )W(s), where G(r n,s ) is the Green's function from source s to detector n, W(s) is the frequency domain transform of the source signal;

[0190] Process 1.2, transform the data channel r(i) into the frequency domain, denoted as G(r i,s )W(s), where G(r i,s ) is the Green's function from source s to detector i, W(s) is the frequency domain transform of the source signal;

[0191] Process 1.3, according to the cross-correlation operation in the time domain is equivalent to the conjugate multiplication of the two spectra in the frequency domain, the frequency domain transform C(n,i) of the time domain cross-correlation function of the reference channel r(n) and the data channel r(i) is obtained;

[0192] C(n,i)=W(s)G(r n,s )W * (s)G * (r i ,s)=|W(s)| 2 G(r n,s )G * (r i ,s)

[0193] in:

[0194] W(s) is the frequency domain transform of the source function;

[0195] W * (s) is the complex conjugate of the frequency domain transform of the source function;

[0196] |W(s)| 2 =W(s)W * (s) is the frequency domain conjugate product of the autocorrelation function of the source function, which represents the influence of the source function on the cross-correlation function between the reference channel r(n) and the data channel r(i) in the above formula;

[0197] G * (r i , s) is the complex conjugate of the Green's function from source s to receiver i;

[0198] Process 1.4, according to the following formula, the expression H(n,i) in the frequency domain of the cross-correlation result between the reference channel r(n) and the data channel r(i) in the time domain is obtained after removing the source effect and normalizing;

[0199]

[0200] in:

[0201] |C(n,n)| is the modulus of the frequency domain transform of the autocorrelation function of the reference channel r(n);

[0202] |C(i,i)| is the modulus of the frequency domain transform of the autocorrelation function of data channel r(i);

[0203] Process 1.5, perform inverse Fourier transform on H(n,i) obtained in process 1.4 to obtain the cross-correlation result h(n,i) in the time domain after removing the source effect and normalizing the signals r(n) and r(i);

[0204]

[0205] In G3.1, Q3.1 or W3.1, the specific operation of obtaining m data abstract profiles for each data channel is: ω The length is an interval, and the first pulse interception segment, the second pulse interception segment or the high-quality data segment with a length of T is cut into m segments;

[0206]

[0207] T ω =1 / F ω

[0208] in:

[0209] T ω Indicates the length of time;

[0210] The ceil function means rounding up to an integer, that is, the number at the end is less than 2T ω The length part is made into a separate section.

[0211] In the above technical solution, the maximum absolute value of the data in the input data segment is used to replace the data in the current segment, and the two-dimensional original data segment r with a length of T is abstracted into a two-dimensional data abstract outline with a length of m.

[0212] An automatic cleaning system for seismic data while drilling, based on an automatic cleaning method for seismic data while drilling, comprises a data acquisition unit, a data quality evaluation unit and a data cleaning unit connected in sequence;

[0213] The data acquisition unit is used to process the seismic data using the first process to determine the best reference track;

[0214] The first process includes process A' and process B';

[0215] The input of process A' is a two-dimensional raw data segment r and the track number n of the data track closest to the roadheader, and the output is the No. 1 pulse intercept segment;

[0216] The input of process B' is the main frequency F of the direct wave when the No. 1 pulse is cut and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r;

[0217] The data quality evaluation unit is used to evaluate the seismic data using the second processing and the best reference trace to obtain a quality score, and determine a data cleaning threshold according to the quality score;

[0218] The second process includes process A” and process B”;

[0219] The input of process A” is a plurality of two-dimensional raw data segments R and the track number n of the best reference track, and the output is a plurality of No. 2 pulse intercept segments;

[0220] The input of process B is multiple No. 2 pulse interception segments and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R;

[0221] The data cleaning unit is used to clean the seismic data using the obtained data cleaning threshold to obtain high-quality data segments, and to eliminate false high-quality data in the high-quality data segments based on process B'';

[0222] The input of process B'' is the two-dimensional raw data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

[0223] Example:

[0224] The data quality of a coal mine seismic data collected over the past 10 hours was evaluated in 1-minute segments. The evaluation results are shown in the figure. Figure 2The evaluation results clearly show the time when the roadheader is actively cutting the coal wall and the time when it is resting.

[0225] Take a section of excavation seismic data, obtain the mutual interference seismic interference profile, extract the contour features, and then analyze each channel to see if there is an absolutely dominant energy peak. The result is as follows: Figure 3-4 . Figure 3 It is the result of the simultaneous excavation seismic data being processed by mutual seismic interference. Figure 4 yes Figure 3 After abstraction and contouring, the weak energy clusters are set to zero and the dominant energy clusters are retained after evaluation. As can be seen from the figure, the present invention is relatively robust and can effectively identify channels without relevant energy peaks, making a sufficiently reliable evaluation of the quality of the data block.

Claims

1. An automatic cleaning method for seismic data during excavation, based on a seismic data acquisition system, wherein the seismic data acquisition system comprises a plurality of geophones and signal transmission cables, wherein the geophones are installed on coal wall anchors; the seismic data is a vibration signal collected simultaneously by a plurality of geophones when a tunnel boring machine cuts a coal wall and causes vibration, and a vibration signal collected by a geophone is called a data channel, wherein the seismic data is characterized in that: The specific steps include: Step 1, collecting seismic data, processing the seismic data using a first process, and determining an optimal reference trace; The first process includes process A' and process B'; The input of the process A' is a two-dimensional original data segment r and the track number n of the data track closest to the roadheader, and the output is a pulse intercept segment No. 1; The input of the process B' is the main frequency F of the direct wave when the first pulse cutting section and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r; Step 2, using the second processing and the best reference trace determined in step 1 to evaluate the seismic data, obtain a quality score, and determine a data cleaning threshold according to the quality score; The second process includes process A" and process B"; The input of the process A" is a plurality of two-dimensional original data segments R and the track number n of the best reference track determined in step 1, and the output is a plurality of No. 2 pulse interception segments; The input of the process B" is a plurality of No. 2 pulse interception sections and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R; Step 3, using the data cleaning threshold obtained in step 2 to clean the seismic data, obtain high-quality data segments, and remove false high-quality data in the high-quality data segments based on process B''; The input of the process B'' is the two-dimensional original data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

2. The method for automatic cleaning of seismic data during excavation as claimed in claim 1, characterized in that: Step 1 specifically includes the following steps: Step 1.1, after laying the survey line and successfully collecting and uploading the data, select the seismic data of the first day when the tunnel boring machine has tunneling activities; Step 1.2, slicing the seismic data obtained in step 1.1 using a first preset time to obtain a plurality of two-dimensional original data segments r, sequentially sending the plurality of two-dimensional original data segments r and the channel number n of the data channel closest to the tunnel boring machine into process A', and outputting a first pulse interception segment; Step 1.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the No. 1 pulse interception section ω , and the pulse interception section No. 1 obtained in step 1.2 and the main frequency F of the direct wave when the roadheader cuts the coal wall are respectively ω Input process B' to obtain the quality scores of all two-dimensional raw data segments r; Step 1.4, selecting a data segment corresponding to the two-dimensional original data segment r whose quality score is greater than a first threshold; Step 1.5, using the data segments obtained in step 1.4, taking each data track as a reference track, searching for the best reference track among all tracks, and obtaining a plurality of candidate best reference tracks; Step 1.6, using the following formula to calculate ref(b) of the best reference track to be selected obtained in step 1.5, and taking the best reference track to be selected corresponding to its maximum value as the best reference track; in: cov(b,j) represents the covariance between the best reference track b and data track j; cov(j,j) represents the variance of the jth data track; N represents the number of all data channels; ref(b) represents the sum of squares of correlation coefficients between the candidate best reference trace b and all candidate best reference traces.

3. The method for automatic cleaning of seismic data during excavation as claimed in claim 1, characterized in that: Step 2 specifically includes the following steps: Step 2.1, slicing and segmenting the seismic data using a second preset time to obtain a plurality of two-dimensional original data segments R; Step 2.2, sequentially sending the multiple two-dimensional original data segments R obtained in step 2.1 and the channel number n of the best reference channel determined in step 1 to process A'' to obtain the second pulse interception segment corresponding to each of the multiple two-dimensional original data segments R; Step 2.3, obtain the main frequency F of the direct wave when the tunnel boring machine cuts the coal wall on the second pulse interception section ω , and sequentially compare the multiple No. 2 pulse interception segments obtained in step 2.2 with the main frequency F of the direct wave when the roadheader cuts the coal wall ω Input data flow B", and obtain the quality scores of all two-dimensional raw data segments R, that is, obtain the quality score curve of seismic data; Step 2.4: Determine the data cleaning threshold based on the quality score curve obtained in step 2.

3.

4. The method for automatic cleaning of seismic data during excavation as claimed in claim 3, characterized in that: The data cleaning threshold determined according to the quality score curve obtained in step 2.3 is specifically: Step 2.4.1: For the two-dimensional original data segments R with a quality score lower than P, they are directly cleaned; for the two-dimensional original data segments R with a quality score higher than Q, they are determined to be high-quality data segments; for the two-dimensional original data segments R with a quality score falling within the range of [P, Q], they are determined to be optional data; In step 2.4.2, the optional data in the range [P, Q] are sorted from large to small according to the quality score, and the quality scores in the top 80% are taken as the data cleaning threshold.

5. The method for automatic cleaning of seismic data while drilling as claimed in claim 1, characterized in that: Step three specifically includes the following steps: Step 3.1, according to the data cleaning threshold obtained in step 2, remove the data segments in the seismic data whose quality scores are lower than the data cleaning threshold to obtain high-quality data segments; Step 3.2, remove the false high-quality data in the high-quality data segment obtained in step 3.

1.

6. The method for automatic cleaning of seismic data during excavation as claimed in claim 5, characterized in that: The method of removing false high-quality data from the high-quality data segment is specifically: Step 3.2.1: Obtain the main frequency F of the microseismic event on the high-quality data segment m , the two-dimensional original data segment r corresponding to the selected high-quality data segment and the main frequency F of the microseismic event m Input process B'', output the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment; Step 3.2.2, determine whether the quality score obtained in step 3.2.1 exceeds the second threshold. If so, the two-dimensional original data segment r is false high-quality data caused by a large vibration signal not caused by tunnel boring machine activity, and set its quality score to -1 and directly discard it.

7. The method for automatic cleaning of seismic data during excavation as claimed in claim 1, characterized in that: Process A' specifically includes the following processes: Q1.1, according to the data channel number n closest to the tunnel boring machine, find the data channel corresponding to the channel number n from the input two-dimensional data segment r, and save it as the reference channel r(n); Q1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N]; Q2.3, the length of the middle part of the time domain mutual interference result obtained in Q1.2 is T SEG The data is of length T SEG The two-dimensional output data, namely the No. 1 pulse intercept segment; Process A" specifically includes the following processes: G1.1, according to the track number n of the best reference track determined in step 1, find the corresponding data track from the input two-dimensional data segment R, and save it as the reference track r(n); G1.2, perform the operation of process C on the data channel r(i) in the input two-dimensional data segment r and the reference channel r(n) respectively, and obtain the normalized time domain mutual interference result of the reference channel r(n) and the data channel r(i), i∈[1,N]; G1.3, the length of the middle part of the time domain mutual interference result obtained in process 1.2 is T SEG The data is of length T SEG The two-dimensional output data is the second pulse intercept segment.

8. The method for automatic cleaning of seismic data during excavation as claimed in claim 1, characterized in that: Process B' specifically includes the following processes: Q3.1, extract the contour features of the intercepted segment of pulse No. 1 and abstract it to obtain m data abstract contours of each data channel; Q3.2, analyze whether the m data abstract profiles of a data channel obtained in Q3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in its data abstract profile. If there is no absolutely dominant energy peak, then this data channel is set to zero; Q3.3, perform the operation of Q3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the first pulse interception segment; The value of the ratio is between [0, 1]; Process B" specifically includes the following processes: G3.1, extract the contour features of the second pulse segment and abstract it to obtain m data abstract contours of each data channel; G 3.2, analyzing whether the m data abstract profiles of a data channel obtained in G 3.1 have an absolutely dominant energy peak. If there is an absolutely dominant energy peak, then only the energy peak is retained in the data abstract profile; if there is no absolutely dominant energy peak, then this data channel is set to zero; G3.3, perform the operation of G3.2 on all data channels, calculate the ratio of the number of non-zero data channels to the total number of data channels, and the ratio is the quality score of the second pulse interception segment; The value of the ratio is between [0, 1]; Process B"'Specific The process includes: W3.1, extract contour features from the two-dimensional original data segment r corresponding to the high-quality data segment, and abstract it to obtain m data abstract contours for each data channel; W3.2, analyze whether the m data abstract profiles of a data channel obtained in W3.1 have an energy peak that is absolutely dominant. If there is an energy peak that is absolutely dominant, then only the energy peak is retained in its data abstract profile; if there is no energy peak that is absolutely dominant, then this data channel is set to zero; W3.3, perform the operation of W3.2 on all data channels, and calculate the ratio of the number of non-zero data channels to the total number of data channels, which is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment; The value of the ratio is between [0, 1].

9. The method for automatic cleaning of seismic data while drilling as claimed in claim 1, characterized in that: The process C specifically includes the following process: Process 1.1, transform the reference channel r(n) into the frequency domain, denoted as G(r n,s )W(s), where G(r n,s ) is the Green's function from source s to detector n, W(s) is the frequency domain transform of the source signal; Process 1.2, transform the data channel r(i) into the frequency domain, denoted as G(r i,s )W(s), where G(r i,s ) is the Green's function from source s to detector i, W(s) is the frequency domain transform of the source signal; Process 1.3, according to the cross-correlation operation in the time domain is equivalent to the conjugate multiplication of the two spectra in the frequency domain, the frequency domain transform C(n,i) of the time domain cross-correlation function of the reference channel r(n) and the data channel r(i) is obtained; C(n,i)=W(s)G(r n,s )W * (s)G * (r i ,s)=|W(s)| 2 G(r n,s )G * (r i ,s) in: W(s) is the frequency domain transform of the source function; W * (s) is the complex conjugate of the frequency domain transform of the source function; |W(s)| 2 =W(s)W * (s) is the frequency domain conjugate product of the autocorrelation function of the source function, which represents the influence of the source function on the cross-correlation function between the reference channel r(n) and the data channel r(i) in the above formula; G * (r i , s) is the complex conjugate of the Green's function from source s to receiver i; Process 1.4, according to the following formula, the expression H(n,i) in the frequency domain of the cross-correlation result between the reference channel r(n) and the data channel r(i) in the time domain is obtained after removing the source effect and normalizing; in: |C(n,n)| is the modulus of the frequency domain transform of the autocorrelation function of the reference channel r(n); |C(i,i)| is the modulus of the frequency domain transform of the autocorrelation function of data channel r(i); Process 1.5, perform inverse Fourier transform on H(n,i) obtained in process 1.4 to obtain the cross-correlation result h(n,i) in the time domain after removing the source effect and normalizing the signals r(n) and r(i); 10. The method for automatic cleaning of seismic data during excavation as claimed in claim 8, characterized in that: In G3.1, Q3.1 or W3.1, the specific operation of obtaining m data abstract profiles for each data channel is: ω The length is an interval, and the first pulse interception segment, the second pulse interception segment or the high-quality data segment with a length of T is cut into m segments; T ω =1 / F ω in: T ω Indicates the length of time; The ceil function means rounding up to an integer, that is, the number at the end is less than 2T ω The length part is made into a separate section.

11. An automatic cleaning system for seismic data while drilling, based on the automatic cleaning method for seismic data while drilling as claimed in any one of claims 1 to 10, comprising a data acquisition unit, a data quality evaluation unit and a data cleaning unit connected in sequence; The data acquisition unit is used to process the seismic data using a first process to determine an optimal reference trace; The first process includes process A' and process B'; The input of the process A' is a two-dimensional original data segment r and the track number n of the data track closest to the roadheader, and the output is a pulse intercept segment No. 1; The input of the process B' is the main frequency F of the direct wave when the first pulse cutting section and the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment r; The data quality evaluation unit is used to evaluate the seismic data using the second processing and the best reference trace to obtain a quality score, and determine a data cleaning threshold according to the quality score; The second process includes process A" and process B"; The input of the process A" is a plurality of two-dimensional original data segments R and the track number n of the best reference track, and the output is a plurality of second pulse interception segments; The input of the process B" is a plurality of No. 2 pulse interception sections and the main frequency F of the direct wave when the roadheader cuts the coal wall. ω , the output is the quality score of the two-dimensional original data segment R; The data cleaning unit is used to clean the seismic data using the obtained data cleaning threshold to obtain high-quality data segments, and to eliminate false high-quality data in the high-quality data segments based on process B''; The input of the process B'' is the two-dimensional original data segment r corresponding to the high-quality data segment and the main frequency F of the microseismic event m , the output is the quality score of the two-dimensional original data segment r corresponding to the high-quality data segment.

Images