Tectonic coal while-drilling acoustic vibration signal discrimination method and tectonic coal while-drilling acoustic vibration signal discrimination device

Abstract

The invention provides a tectonic coal while-drilling acoustic-vibration signal discrimination method and a tectonic coal while-drilling acoustic-vibration signal discrimination device. In the coal hole drilling process, acoustic vibration signal sources are complex, waveforms received by sensors are actually mixed responses formed by superposition of multiple vibration sources, effective acoustic vibration signals caused by coal body cutting damage are accurately discriminated, and the method is a key basis for establishing a reliable while-drilling tectonic coal identification method. Meanwhile, the coal body damage acoustic vibration signal contains various characteristic parameters, but not all the parameters can effectively reflect the coal body structure difference. Therefore, extraction of the characteristic parameters sensitive to the change of the coal body structure becomes the premise of constructing a high-precision tectonic coal recognition model. The method mainly solves the two technical problems of accurate discrimination and sensitive feature extraction of the coal body while-drilling damage acoustic vibration signals, so that the tectonic coal distribution identification precision and advanced detection capability are remarkably improved, and key technical support is provided for coal and gas outburst prevention and control.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of advanced detection of tectonic coal at a coal mining face, and in particular to a method and device for identifying tectonic coal using acoustic and vibration signals while drilling. BACKGROUND

[0002] Coal mine safety technology, as a core support system in the field of energy exploitation, is widely used in the prevention and control of coal and gas outburst disasters. In related technologies, a tectonic coal detection technology system is constructed through the collaborative operation of coal wall observation, drilling core sampling and logging curve inversion. Specifically, the system covers the entire process from stratum information collection to mechanical property testing, including key links such as acquisition of coal and rock mass physical and mechanical parameters, drilling construction data recording and multi-source information fusion analysis. With the increase of mining depth in high-gas mines, the traditional method has been difficult to meet the dynamic outburst prevention needs, and the existing technology usually uses acoustic and vibration sensors installed on the drilling rig body to collect signals, but is limited by the complexity of drilling conditions, and there are significant deficiencies in signal separation accuracy and feature extraction reliability.

[0003] However, in the existing acoustic and vibration identification method while drilling, the signals collected directly by the drilling rig body contain multiple sources of interference such as idling, rod withdrawal and rod connection, and no signal identification standard based on the failure mechanism of coal and rock mass has been established. Specifically, the time domain amplitude and frequency domain distribution of the acoustic and vibration signals during drilling overlap, and the traditional threshold screening method cannot effectively distinguish between coal cutting failure signals and working condition interference signals. Among them, the differences in physical and mechanical properties between tectonic coal and primary structure coal result in insufficient correlation of acoustic and vibration characteristics, and the existing feature extraction method does not combine physical and mechanical parameters for sensitive feature screening, which may cause feature misselection or model overfitting problems, or lead to a tectonic coal distribution prediction error rate of more than 30% due to insufficient signal purity, thereby affecting the scientific deployment of outburst prevention measures and the stability of coal mine safety production. SUMMARY

[0004] The present application aims to at least partially solve one of the technical problems in the related art.

[0005] To this end, a first object of the present application is to propose a method for identifying tectonic coal acoustic and vibration signals while drilling.

[0006] A second object of the present application is to propose a device for identifying tectonic coal acoustic and vibration signals while drilling.

[0007] To achieve the above object, the first aspect of the present application provides a method for identifying coal while drilling acoustic vibration signals, comprising: S1, obtaining the original acoustic vibration signals of the drilling process through the near-bit acoustic vibration signal acquisition probe, and recording the time stamp and drilling distance data corresponding to the drill pipe idling, drill bit coal breaking, rod retraction and rod connection working conditions; S2, performing time domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signals based on the time stamp, filtering the non-coal breaking condition signals, and then using wavelet packet transform and empirical mode decomposition to separate the remaining signals in the frequency domain to extract the high-frequency breaking signal frequency band; S3, processing the separated high-frequency breaking signal frequency band using a band-pass filter method to obtain pure coal cutting damage acoustic vibration signals; S4, based on the amplitude, dominant frequency, rise time and duration characteristics of the coal cutting damage acoustic vibration signals, using correlation analysis method to screen sensitive features, and constructing a coal identification model through convolutional neural network and recurrent neural network, and adjusting the model parameters by five-fold cross-validation and Bayesian optimization to realize the prediction of coal distribution.

[0008] In an embodiment of the present application, the original acoustic vibration signals of the drilling process are obtained through the near-bit acoustic vibration signal acquisition probe, and the time stamp and drilling distance data corresponding to the drill pipe idling, drill bit coal breaking, rod retraction and rod connection working conditions are recorded, which further comprises: S11, the original acoustic vibration signal acquisition probe is arranged by combining piezoelectric sensors and MEMS microphones, and the collected data is uploaded in real time through a wireless radio frequency transmission unit; S12, the drilling distance data is obtained through joint calibration of a laser ranging module and a drill pipe rotation counter.

[0009] In an embodiment of the present application, the time domain amplitude threshold value analysis and duration threshold value analysis on the original acoustic vibration signals based on the time stamp, and the filtering of the non-coal breaking condition signals, and then using wavelet packet transform and empirical mode decomposition to separate the remaining signals in the frequency domain to extract the high-frequency breaking signal frequency band further comprises: S21, the time domain amplitude threshold value is determined based on the statistical characteristics after excluding the abnormal values of the box plot, specifically taking 1.5 times of the average amplitude of the coal breaking signal as the upper limit of the threshold value; S22, the duration threshold value is set according to 2 times of the average duration of the rod connection process, and the threshold value adjustment mechanism is automatically triggered when the duration of the rod connection signal is less than the preset threshold value.

[0010] In an embodiment of the present application, the processing of the separated high-frequency breaking signal frequency band using a band-pass filter method to obtain pure coal cutting damage acoustic vibration signals further comprises: S31, the center frequency range of the band-pass filter is 2-8 kHz, and the bandwidth is dynamically adjusted according to the dominant frequency distribution of the IMF component after empirical mode decomposition; S32, the processing process uses an adaptive filtering algorithm to calculate the signal spectrum characteristics in real time and update the filtering parameters through a sliding window.

[0011] In an embodiment of the present application, the amplitude, dominant frequency, rise time and duration characteristics of the coal body cutting damage acoustic vibration signal are based on a correlation analysis method to screen sensitive features, and a convolutional neural network and a recurrent neural network are used to construct a tectonic coal recognition model, and five-fold cross-validation and Bayesian optimization are used to adjust the model parameters to realize tectonic coal distribution prediction, and the correlation analysis method is used to determine that the feature is a tectonic coal sensitive feature when the Pearson correlation coefficient and the grey correlation degree are greater than a preset threshold.

[0012] In an embodiment of the present application, further comprising: S5, spatially registering the tectonic coal distribution image of the borehole wall collected by the borehole viewer with the prediction result of the tectonic coal recognition model, and verifying the model prediction accuracy through joint analysis of image features and acoustic vibration features.

[0013] To achieve the above purpose, the second aspect of the present application provides a tectonic coal while-drilling acoustic vibration signal discrimination device, comprising: an acoustic vibration signal acquisition and working condition recording module, used to acquire original acoustic vibration signals of the drilling process through a near-bit acoustic vibration signal acquisition probe, and record the time stamps and drilling distance data corresponding to the drill pipe idle, coal breaking, rod retraction and rod connection working conditions; a time domain analysis and frequency domain feature separation module, used to perform time domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signals based on the time stamps, filter non-coal breaking working condition signals, and then use wavelet packet transform and empirical mode decomposition to separate the frequency domain features of the remaining signals, and extract high-frequency breaking signal frequency bands; a band-pass filtering processing module, used to process the separated high-frequency breaking signal frequency bands using a band-pass filtering method to obtain pure coal body cutting damage acoustic vibration signals; a feature analysis and model construction module, used to screen sensitive features using a correlation analysis method based on the amplitude, dominant frequency, rise time and duration characteristics of the coal body cutting damage acoustic vibration signals, and construct a tectonic coal recognition model through a convolutional neural network and a recurrent neural network, and use five-fold cross-validation and Bayesian optimization to adjust the model parameters to realize tectonic coal distribution prediction.

[0014] In an embodiment of the present application, the original acoustic vibration signal acquisition and working condition recording module is further used to: acquire the original acoustic vibration signals of the drilling process through a combination of piezoelectric sensors and MEMS microphones, and upload the collected data in real time through a wireless radio frequency transmission unit; and acquire the drilling distance data through joint calibration of a laser ranging module and a drill pipe rotation counter.

[0015] In an embodiment of the present application, the time domain analysis and frequency domain feature separation module is further configured to: determine an amplitude threshold based on the statistical features after the outlier rejection of the box plot, specifically, taking 1.5 times of the average amplitude of the coal breaking signal as the upper threshold; set a duration threshold based on 2 times of the average duration of the contact rod process, and automatically trigger a threshold adjustment mechanism when the duration of the contact rod signal is less than the preset threshold.

[0016] In an embodiment of the present application, the band-pass filtering processing module is further configured to: set a band-pass filter with a center frequency range of 2-8 kHz, and dynamically adjust the bandwidth according to the main frequency distribution of the IMF component after the empirical mode decomposition; adopt an adaptive filtering algorithm to calculate the signal spectrum features in real time through a sliding window and update the filtering parameters.

[0017] In an embodiment of the present application, the feature analysis and model construction module is further configured to: in the correlation analysis method, when the Pearson correlation coefficient and the grey correlation degree are both greater than a preset threshold, determine that the feature is a tectonic coal sensitive feature.

[0018] In an embodiment of the present application, further comprising: an image registration and joint analysis module configured to perform spatial registration on the tectonic coal distribution image collected by the borehole viewer and the prediction result of the tectonic coal recognition model, and verify the model prediction accuracy through joint analysis of the image features and the acoustic vibration features.

[0019] The method and device of the embodiments of the present application can accurately identify the acoustic vibration signal generated by the coal cutting damage in the drilling process, effectively improve the recognition accuracy of the tectonic coal distribution, and realize the advanced dynamic detection of the tectonic coal in the mining working face.

[0020] Additional aspects and advantages of the present application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the following description, taken in conjunction with the accompanying drawings, in which: Figure 1 is a flowchart of a tectonic coal while-drilling acoustic vibration signal identification method according to an embodiment of the present application; Figure 2 is a flowchart of the implementation process of a tectonic coal while-drilling acoustic vibration recognition method in a mining working face according to an embodiment of the present application; Figure 3 is a structural schematic diagram of a near-bit acoustic vibration signal acquisition probe according to an embodiment of the present application; Figure 4 is a tectonic coal cutting damage acoustic vibration signal identification flowchart in a drilling process according to an embodiment of the present application; Figure 5 is a flow chart of a coal construction model construction method according to an embodiment of the present application; Figure 6 is a structural schematic diagram of a coal construction while-drilling acoustic vibration signal discrimination device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] It should be noted that the embodiments in the present application and the features in the embodiments can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0023] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present application.

[0024] A coal construction while-drilling acoustic vibration signal discrimination method and device according to an embodiment of the present application will be described below with reference to the accompanying drawings.

[0025] Embodiment 1 Figure 1 is a flow chart of a coal construction while-drilling acoustic vibration signal discrimination method according to an embodiment of the present application, as shown in Figure 1 , comprising: S1, obtaining the original acoustic vibration signal of the drilling process through the near-bit acoustic vibration signal acquisition probe tube, and recording the time stamp and drilling distance data corresponding to the drill pipe idle, drill bit coal breaking, rod retraction and rod connection working conditions.

[0026] Specifically, this step obtains the original acoustic vibration signal of the drilling process through the near-bit acoustic vibration signal acquisition probe tube, and synchronously records the time stamp and drilling distance data corresponding to different working conditions such as drill pipe idle, drill bit coal breaking, rod retraction and rod connection. It is a key data acquisition link to realize the construction of coal while-drilling identification. In some implementations, the probe tube uses a high-sensitivity piezoelectric acoustic vibration sensor, and its sampling frequency is usually set to 10 kHz to 50 kHz to ensure that high-frequency breaking signals and low-frequency friction signals generated during the interaction between the drill bit and the coal rock mass can be captured. The sensor installation position is not more than 1.5 meters away from the drill bit end to reduce the interference of drill pipe vibration on the signal and improve the signal-to-noise ratio and spatial resolution of the signal.

[0027] Specifically, the acquisition system needs to have at least 12-bit ADC precision to ensure that the dynamic range and quantization error of the acoustic vibration signal are controlled within a reasonable range. At the same time, the circuit board built-in the probe needs to support data caching function, usually using FAT32 file system for data storage, the sampling duration can be set to continuous recording or triggered recording according to the drilling construction progress, and the trigger threshold is generally set to 2 times the standard deviation of the amplitude exceeding the background noise mean. Drilling distance data is collected in real time by the encoder or laser ranging device of the drilling machine, with an accuracy requirement of not less than ±1 cm and a sampling frequency of 1 Hz to 5 Hz, to ensure alignment with the timestamp of the acoustic vibration signal.

[0028] Further, this step is suitable for coal mine underground extraction drilling or detection drilling during the construction process, especially in complex geological structure and tectonic coal development area, and has significant engineering practical value. By collecting acoustic vibration signals in real time during drilling and recording working condition information, high-quality and high-timeliness raw data support can be provided for subsequent signal discrimination and feature extraction.

[0029] Specifically, this step realizes high-precision and high-fidelity acquisition of acoustic vibration signals during drilling, and provides a reliable data basis for subsequent construction of a tectonic coal identification model by synchronously recording the time stamp and drilling distance, which is the premise and core input link of data-driven modeling in the entire method.

[0030] Further, S1 includes: S11, the original acoustic vibration signal acquisition probe adopts a combination of piezoelectric sensors and MEMS microphones, and collects data in real time through a wireless radio frequency transmission unit.

[0031] Specifically, in some implementations, the original acoustic vibration signal acquisition probe adopts a combination of piezoelectric sensors and MEMS microphones, and collects data in real time through a wireless radio frequency transmission unit, and the technical implementation principle is based on multi-source heterogeneous signal fusion and low-delay wireless communication technology. The piezoelectric sensor is usually made of high-sensitivity PZT (lead zirconate titanate) material, has a wide frequency response range (generally 0.1 Hz to 10 kHz), and is suitable for capturing high-frequency vibration signals (usually between 2 kHz and 8 kHz) generated by the rupture of coal and rock mass during drilling. The output voltage of the piezoelectric sensor is proportional to the vibration acceleration, which can be directly used for time domain and frequency domain analysis. The MEMS microphone is used to collect acoustic signals, and its working frequency range is usually 20 Hz to 20 kHz, with good signal-to-noise ratio (SNR≥60 dB) and low power consumption characteristics, suitable for capturing low-frequency acoustic vibration signals (usually between 100 Hz and 2 kHz) generated by the friction between the drill bit and the coal and rock mass.

[0032] In terms of specific operation, the probe pipe is internally integrated with multi-channel signal conditioning circuit, including preamplifier, anti-aliasing filter and analog-to-digital converter (ADC), to realize synchronous sampling and digital processing of the output signals of the piezoelectric sensor and the MEMS microphone. The sampling frequency is usually set to 20 kHz to 50 kHz to meet the restoration requirements of high-frequency signals according to the Nyquist sampling theorem. The collected acoustic vibration data are preliminarily processed by an embedded microcontroller, including signal filtering, feature extraction and data packaging, and then uploaded to the ground monitoring system in a low-latency manner through a wireless radio frequency transmission unit (such as ZigBee, LoRa or 2.4 GHz ISM frequency band wireless module) to realize real-time acoustic vibration signal feedback during drilling.

[0033] Specifically, this step is deployed in the drilling process in the coal mine underground in actual application, the probe pipe is installed at the front end of the drill pipe, the distance from the drill bit is not more than 1.5 meters, to ensure the near-field nature and accuracy of signal acquisition. Its technical effect lies in effectively separating the effective acoustic vibration signals generated by the drill bit during drilling, excluding the interference of non-coal body damage signals such as drill pipe idling and rod retraction, providing high-quality and high signal-to-noise ratio original data for subsequent feature extraction and identification model construction of tectonic coal, thereby improving the accuracy and real-time performance of tectonic coal identification, and having important engineering value for the early warning of coal mine outburst disasters.

[0034] S12, the drilling distance data is obtained through joint calibration of the laser ranging module and the drill pipe rotation counter.

[0035] Specifically, the drilling distance data is obtained through joint calibration of the laser ranging module and the drill pipe rotation counter. This step is a key link in the present application to realize the accurate correspondence between the acoustic vibration signal while drilling and the drilling position, and its technical implementation principle is based on the synchronous measurement and fusion processing of the displacement and rotation angle of the drill pipe during drilling.

[0036] Specifically, the laser ranging module is installed at the end of the propulsion mechanism of the drilling machine, for real-time measurement of the linear propulsion distance of the drill pipe. This module uses a high-precision laser ranging sensor (such as SICK LMS511 series), which has a measurement accuracy of ±1 mm and a sampling frequency of 100 Hz, which can meet the real-time monitoring requirements of dynamic displacement during underground drilling. The drill pipe rotation counter is installed on the rotating drive shaft of the drilling machine, for recording the number of rotations of the drill pipe, so as to calculate the actual cutting footage length of the drill bit in the coal and rock mass. The rotation counter usually uses an optical encoder (such as an incremental encoder with a resolution of 1024 PPR), which calculates the rotation angle through pulse signals, and converts it into the actual drilling distance in combination with the drill bit diameter and the cutting feed rate.

[0037] Furthermore, to ensure consistency between the measurement data of the two systems, joint calibration is performed every 5 meters of drilling distance. The calibration process includes: first, obtaining the absolute displacement value of the current drilling through the laser ranging module; second, calculating the theoretical drilling distance based on the number of rotations recorded by the rotary counter and the drill bit cutting parameters (such as a feed per revolution of 0.5 mm); finally, comparing the two, if the deviation exceeds a set threshold (such as ±2 cm), an automatic correction mechanism is triggered to adjust the pulse equivalent of the rotary counter or the sampling compensation coefficient of the laser ranging module to eliminate system errors.

[0038] Specifically, in practical applications, this step is suitable for underground drilling environments in coal mines, especially in high-gas and outburst-prone mines, to establish the spatial correspondence between acoustic and vibration signals during drilling and the coal body structure. Accurate drilling distance data provides reliable spatial coordinate support for subsequent temporal analysis of acoustic and vibration signals and model training, thereby improving the positioning accuracy and model generalization ability of coal structure identification.

[0039] Specifically, this joint calibration method effectively solves the distance error problem caused by drill pipe elastic deformation, slippage, or measurement drift during drilling, ensuring high-precision matching between acoustic and vibration signals and drilling position. It provides a data foundation for real-time, dynamic, and high-resolution identification of structural coal, and significantly improves the reliability and practicality of the drilling identification system.

[0040] S2, based on the timestamp, perform time-domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signal, filter out non-coal breaking condition signals, and use wavelet packet transform and empirical mode decomposition to separate the frequency domain features of the remaining signal to extract the high-frequency breaking signal band.

[0041] Specifically, in some implementations, this step performs time-domain amplitude threshold analysis and duration threshold analysis on the raw acoustic vibration signal based on timestamps, aiming to identify effective coal cutting and damage signals from the complex drilling process. Specifically, firstly, the acquired acoustic vibration signal is precisely mapped to the drilling rig's operating conditions (such as idling, coal breaking, rod retraction, and rod extension) using timestamps, thereby classifying the signal's operating conditions. In the time-domain amplitude threshold analysis, a sliding window method is used to segment the signal, with the window length typically set to 50ms to 200ms and a step size of 25ms to balance real-time performance and signal integrity. When setting the amplitude threshold, the signal amplitude distribution during the coal breaking stage of the drill bit is referenced, and the threshold is typically set to 1.5 to 2 times the background noise amplitude to effectively eliminate low-amplitude interference signals during the idling and rod retraction stages.

[0042] Furthermore, in the duration threshold analysis, transient signals caused by short-duration operations such as rod connection are identified by analyzing the duration characteristics of signal segments. The rod connection process typically lasts less than 1 second, while the signal duration of the coal breaking process is longer, usually more than 2 seconds. Therefore, a duration threshold of 1.2 seconds is set, and signal segments exceeding this threshold are retained as valid signals that may contain information about coal body fracturing.

[0043] Furthermore, after completing the aforementioned time-domain screening, wavelet packet transform (WPT) and empirical mode decomposition (EMD) are used to separate the frequency domain features of the remaining signal. The wavelet packet transform employs the db4 wavelet basis function, with a decomposition level of 5 layers covering the 0-20kHz frequency range to extract the high-frequency fracture signal band (typically 5-15kHz). Empirical mode decomposition is used to adaptively separate the intrinsic mode functions (IMFs) in the signal and, combined with Hilbert transform, obtains instantaneous frequency features, thereby enhancing the ability to identify coal seam fracture characteristics.

[0044] Specifically, this step plays a crucial role in signal preprocessing within the overall technical solution. Through multi-dimensional threshold analysis and frequency domain decomposition methods, it effectively removes interference signals from non-cracked coal working conditions and extracts high-frequency cracking signals that reflect the structural characteristics of the coal body. This provides high-quality data input for subsequent sensitive feature extraction and the construction of tectonic coal identification models, significantly improving the accuracy and reliability of tectonic coal identification.

[0045] Furthermore, S2 includes: S21, the time-domain amplitude threshold is determined based on the statistical characteristics after excluding outliers in the box plot, specifically by taking 1.5 times the average amplitude of the coal breaking signal as the upper limit of the threshold.

[0046] Specifically, in the stage of identifying acoustic and vibration signals of coal cutting and damage during drilling, this invention employs a statistical method based on box plot outlier elimination, combined with amplitude characteristics, to set a reasonable amplitude threshold to filter out non-coal breaking signals. Specifically, the technical principle of this step is based on the amplitude differences of acoustic and vibration signals under different drilling rig conditions (such as idling, rod retraction, rod extension, and coal breaking). Typical amplitude characteristics of coal breaking signals are extracted through statistical analysis, and threshold criteria are established accordingly.

[0047] Specifically, the collected acoustic and vibration signals are first classified and processed according to the drilling rig's operating conditions, and time-domain waveform data for each stage are extracted. For the coal breaking signal, the box plot statistical characteristics of its amplitude are calculated, including the first quartile (Q1), the third quartile (Q3), and the interquartile range (IQR = Q3 - Q1). Outliers exceeding 1.5 times the IQR range are identified and removed through the box plot, thereby obtaining a more robust amplitude mean.

[0048] Further, in the downhole environment of the mining face, the acoustic vibration signal collection probe is installed at the near-bit end of the drill pipe, and the sampling frequency is usually set to 10 kHz to 20 kHz to ensure complete capture of high-frequency breakage signals. By setting an amplitude threshold, low-amplitude background noise can be effectively suppressed, and the signal-to-noise ratio and stability of signal identification can be improved.

[0049] Specifically, this step plays a key role in the entire technical solution and is the basis for subsequent sensitive feature extraction and construction of a coal recognition model. By scientifically setting the amplitude threshold, the discrimination accuracy of coal breaking signals can be significantly improved, providing high-quality input data for constructing a high-precision structural coal recognition model, thereby improving the reliability and practicality of the coal mine safety warning system.

[0050] S22, the duration threshold is set according to 2 times the average duration of the connection rod process, and the threshold adjustment mechanism is automatically triggered when the duration of the connection rod signal is less than the preset threshold.

[0051] Specifically, in the acoustic vibration signal discrimination stage of coal cutting and damage during drilling, the setting of the duration threshold is one of the key technical links. In some implementations, the duration threshold is set according to 2 times the average duration of the connection rod process, and the technical principle is based on statistical analysis and dynamic identification of the time characteristics of acoustic vibration signals under different working conditions during drilling. Specifically, the connection rod process is a typical non-cutting working condition in drilling construction, and its acoustic vibration signal usually exhibits a short, high-frequency, and large-amplitude transient response, which is significantly different from the persistent, low-frequency, and relatively stable acoustic vibration signal generated when the drill bit actually cuts coal and rock. Therefore, by setting a reasonable duration threshold, the connection rod signal and the coal cutting and damage signal can be effectively distinguished, thereby improving the accuracy of signal discrimination.

[0052] Specifically, the system first collects multiple connection rod signal samples during drilling construction, calculates the average duration (denoted as T_avg), and sets the duration threshold to 2 x T_avg. In actual application, if the average duration of the connection rod signal is 0.15 seconds, the threshold is set to 0.3 seconds. When a segment of acoustic vibration signal is detected with a duration less than 0.3 seconds, the system will automatically trigger the threshold adjustment mechanism, recalculate the average duration of the connection rod signal in the current construction section, and update the threshold to the new 2 times value. This mechanism can adapt to the fluctuations in connection rod time caused by different drill models, operation habits, and geological conditions, ensuring the robustness of signal discrimination.

[0053] Further, this step is deployed in the downhole drilling monitoring system in the coal mine, combined with the near-bit acoustic vibration signal collection probe, to process acoustic vibration data in real time during drilling. Through the combination of time window sliding analysis and threshold judgment, the system can dynamically identify and eliminate interference signals during the connection rod stage, and retain acoustic vibration characteristics that truly reflect coal cutting and damage.

[0054] Specifically, the duration threshold setting method effectively improves the discrimination accuracy of coal cutting damage signals, reduces the misjudgment rate, provides a high-quality data basis for subsequent sensitive feature extraction and construction of coal recognition model training, and thus enhances the reliability and practicality of the entire drilling acoustic vibration recognition system.

[0055] S3, using a band-pass filtering method to process the separated high-frequency fracture signal frequency band, obtaining pure coal cutting damage acoustic vibration signals.

[0056] Specifically, in the discrimination stage of coal cutting damage acoustic vibration signals during drilling, using a band-pass filtering method to process the separated high-frequency fracture signal frequency band is a key technical step for extracting effective acoustic vibration features of tectonic coal. Based on the distribution characteristics of acoustic vibration signals in the frequency domain, this step realizes the accurate extraction of coal and rock mass fracture signals by setting a reasonable frequency range, thereby removing non-destructive low-frequency interference signals such as drill pipe idling, rod withdrawal, and rod connection.

[0057] Specifically, the band-pass filtering method uses a digital filter structure, usually an IIR (infinite impulse response) or FIR (finite impulse response) filter, whose design is based on the frequency band characteristics of the fracture signals generated when the drill bit cuts coal and rock. According to experimental data and spectral analysis results, the high-frequency acoustic vibration signals generated during the cutting and destruction of tectonic coal and primary structure coal usually have a frequency band between 2 kHz and 10 kHz, while the signals generated by non-destructive operations such as drill pipe idling and rod withdrawal have a dominant frequency below 0.5 kHz. Therefore, the passband range of the band-pass filter can be set to 2 kHz to 10 kHz, and the stopband is set to below 0.5 kHz and above 12 kHz, respectively, to ensure the preservation of effective signals and the suppression of interference signals.

[0058] Further, to improve filtering accuracy, a window function design method (such as the Hanning window, Hamming window) can be used to optimize the frequency response characteristics of the FIR filter, or a bilinear transformation method can be used to design an IIR filter to achieve a steeper transition band and lower phase distortion. In actual processing, the signal sampling rate is usually set to 20 kHz or higher to meet the Nyquist sampling theorem and ensure complete capture of high-frequency signals. After time-domain reconstruction, the filtered signals can be further used for feature extraction, such as amplitude, dominant frequency, rise time, etc., to provide high-quality input data for subsequent tectonic coal recognition models.

[0059] Specifically, this step plays a role in the overall technical solution, and its processing results directly affect the extraction accuracy of sensitive features and the recognition performance of the model. Through the band-pass filtering method, high-frequency acoustic vibration signals generated by coal fracture can be effectively separated, thereby improving the accuracy and stability of tectonic coal recognition, and providing reliable data support for coal mine safety warning and disaster prevention.

[0060] Further, S3 comprises: S31, the center frequency range of the band-pass filter is 2-8 kHz, and the bandwidth is dynamically adjusted according to the main frequency distribution of the IMF component after empirical mode decomposition.

[0061] Specifically, the center frequency range of the band-pass filter is set to 2-8 kHz, and the technical principle is based on the fact that the acoustic vibration signals excited by the rupture of coal and rock mass during drilling have significant high-frequency components, while the acoustic vibration signals caused by non-destructive operations such as drill pipe idling, rod withdrawal, and rod connection are mainly concentrated in the low frequency band. Therefore, by setting the center frequency range of the band-pass filter, the high-frequency acoustic vibration signals related to the cutting and destruction of the coal body can be effectively extracted, thereby realizing the discrimination of effective signals and the suppression of noise signals.

[0062] Further, the bandwidth of the band-pass filter is not fixed, but is dynamically adjusted according to the main frequency distribution of the IMF (Intrinsic Mode Function) component obtained after empirical mode decomposition (EMD). The EMD method decomposes the original acoustic vibration signal into several IMF components, each of which represents the local characteristics of the signal at different time scales. By calculating the power spectral density (PSD) of each IMF component, the main frequency band related to the coal mass rupture process can be identified, which is usually concentrated in the range of 2-8 kHz. Further, combined with the spectral energy proportion analysis, the concentration interval of the main frequency distribution can be determined, so as to dynamically set the upper and lower limit frequencies of the band-pass filter, for example, the bandwidth can be set to 1-3 kHz, which depends on the frequency band width whose energy proportion in the IMF component exceeds 80%.

[0063] Specifically, this step is usually executed in real time during the drilling process in the well. The acoustic vibration signal is obtained by a high-sensitivity piezoelectric acoustic vibration sensor in the near-bit acquisition probe. The sampling frequency is generally set to 20-50 kHz to ensure complete capture of high-frequency components. The parameter adjustment of the filter can be realized by an embedded signal processing unit, and an adaptive filtering algorithm (such as an adaptive bandwidth adjustment algorithm) is used for online optimization to adapt to the changes in the physical and mechanical properties of different coal and rock masses.

[0064] Specifically, by dynamically adjusting the bandwidth of the band-pass filter, the effective acoustic vibration signals in the cutting and destruction process of the coal body can be more accurately extracted, the signal-to-noise ratio of the tectonic coal recognition and the accuracy of the feature extraction are improved, high-quality data input is provided for subsequent sensitive feature analysis and deep learning model training, thereby improving the prediction performance and engineering applicability of the tectonic coal recognition model.

[0065] S32, the processing process adopts an adaptive filtering algorithm to calculate signal spectrum features in real time through a sliding window and update filtering parameters.

[0066] Specifically, in some implementations, the processing process adopts an adaptive filtering algorithm to calculate signal spectrum features in real time through a sliding window and dynamically update filtering parameters to achieve accurate extraction of coal cutting and damage acoustic vibration signals during drilling. The core of this step is to use adaptive filtering technology to process the collected original acoustic vibration signals online, thereby effectively suppressing interference signals during non-coal body damage processes such as drill rig idling, rod withdrawal, and rod connection, and improving the accuracy of tectonic coal identification.

[0067] Specifically, the method first divides the collected acoustic vibration signals into fixed-length sliding windows according to time sequence (such as a window length of 256 points and a sliding step of 128 points) to ensure signal time domain continuity and real-time spectrum analysis. The signals in each window are analyzed by fast Fourier transform (FFT) or short-time Fourier transform (STFT) to extract key frequency domain features such as dominant frequency, frequency band energy distribution, and spectral entropy. Subsequently, an adaptive filtering algorithm (such as the LMS algorithm or the RLS algorithm) is used to filter the signals, where the parameters of the filter (such as filter order, convergence step μ, error threshold ε, etc.) are dynamically adjusted according to the spectrum features of the current window. For example, in the RLS algorithm, the forgetting factor λ can be set to 0.98 to balance system stability and response speed to new data.

[0068] Further, the length and step of the sliding window need to be optimized according to the sampling frequency (such as 20 kHz) and signal feature variation rate to ensure the resolution and real-time performance of spectrum analysis. The update frequency of the filter is usually set to 10-20 times per second to adapt to the dynamic changes of coal and rock mass damage signals during drilling. In addition, the extraction range of spectrum features is usually concentrated in the 0-5 kHz frequency band, where the energy difference between tectonic coal and primary structure coal in the 2-4 kHz frequency band is particularly significant and can be used as a key discriminant.

[0069] Specifically, this step is mainly used for online processing of acoustic vibration signals while drilling in coal mines, especially in high-gas mines and outburst mines. By collecting signals in real time through a near-bit acoustic vibration probe and combining with drilling process conditions (such as drilling distance, drill bit speed, etc.), dynamic identification of tectonic coal distribution is achieved. This method can be integrated into the drilling construction system to provide real-time warning and decision support for the site.

[0070] Specifically, by combining adaptive filtering with sliding window spectrum analysis, high-frequency characteristic signals generated by coal cutting and damage can be effectively separated, improving the input data quality of the tectonic coal identification model and enhancing the prediction accuracy and stability of the model, thereby providing reliable technical support for safe and efficient coal mining.

[0071] S4, based on the amplitude, dominant frequency, rise time and duration characteristics of the coal body cutting damage acoustic vibration signal, sensitive features are screened by using correlation analysis method, and a tectonic coal recognition model is constructed by convolutional neural network and recurrent neural network, and the model parameters are adjusted by five-fold cross validation and Bayesian optimization to realize tectonic coal distribution prediction.

[0072] Specifically, the present application extracts the amplitude, dominant frequency, rise time and duration of the coal body cutting damage acoustic vibration signal in the drilling process, and uses correlation analysis method to screen sensitive features, and then constructs a tectonic coal recognition model combining convolutional neural network (CNN) and recurrent neural network (RNN) to realize high-precision prediction of tectonic coal distribution. In some implementations, this step first extracts the features of the screened pure acoustic vibration signal, wherein the amplitude feature is usually represented by the signal peak value or root mean square value (RMS), the dominant frequency feature is obtained by fast Fourier transform (FFT) or wavelet transform, the rise time is defined as the time required for the signal to reach the peak value from the baseline, and the duration is the time window of the signal effective energy duration, which is usually set as the duration of the signal amplitude exceeding the set threshold.

[0073] Further, in order to screen out sensitive features with significant distinguishing ability for tectonic coal, Pearson correlation coefficient and grey correlation degree analysis method are used to evaluate the correlation between features and physical and mechanical properties of coal body (such as uniaxial compressive strength, elastic modulus, etc.). In parameter setting, the Pearson correlation coefficient threshold is set to be greater than or equal to 0.85 to ensure that the selected features have a strong linear relationship with the target variable; the grey correlation degree is calculated by the geometric similarity between the feature sequence and the reference sequence, and the features with correlation degree greater than 0.8 are selected as the input variables of the model.

[0074] Specifically, the present application adopts CNN-RNN fusion architecture, in which CNN is used to extract local spatial features of acoustic vibration signal, and RNN (such as LSTM or GRU) is used to capture time dependence of signal. The input layer is the screened multi-dimensional feature vector, and the output layer is the binary classification label (tectonic coal / primary structure coal). In order to improve the generalization ability and prediction accuracy of the model, five-fold cross validation (5-Fold Cross Validation) is used for model training and evaluation to ensure full use of the data set and stability of the model. At the same time, Bayesian optimization (Bayesian Optimization) algorithm is introduced to automatically optimize the model hyperparameters (such as learning rate, batch size, number of hidden layer nodes, type of activation function, etc.) to realize optimal model configuration.

[0075] Specifically, this step plays a core role in the entire technical solution, which effectively improves the accuracy and robustness of tectonic coal recognition through the fusion of feature screening and deep learning model. In practical applications, the model can be deployed in the downhole acoustic vibration signal acquisition system to process acoustic vibration data in real time during drilling, providing dynamic prediction support for tectonic coal distribution in mining faces, and thus providing a scientific basis for coal and gas outburst prevention.

[0076] Further, S4 comprises: S41, in the correlation analysis method, when the Pearson correlation coefficient and the grey correlation degree are both greater than a preset threshold, it is determined that the feature is a tectonic coal sensitive feature.

[0077] Specifically, in the tectonic coal while-drilling acoustic vibration identification method of the present application, the step "when the Pearson correlation coefficient and the grey correlation degree are both greater than a preset threshold, it is determined that the feature is a tectonic coal sensitive feature" is a key link in building a tectonic coal recognition model. This step selects acoustic vibration signal feature parameters that can effectively represent the differences between tectonic coal and primary structure coal through a combination of statistical methods and system engineering.

[0078] Specifically, this step first extracts the amplitude, dominant frequency, rise time, duration and other time-frequency domain features of the coal cutting damage acoustic vibration signal identified, and performs statistical analysis on these features, such as calculating the maximum value, minimum value, mean value, standard deviation, etc. Subsequently, the Pearson correlation coefficient analysis method is used to evaluate the linear correlation between these features and the physical and mechanical properties of the coal body (such as uniaxial compressive strength, shear strength, elastic modulus, etc.), and the grey correlation degree analysis method is introduced to measure the degree of non-linear correlation between the feature parameters and the change in the structure of the coal body. The Pearson correlation coefficient is used to quantify the strength of the linear relationship between the feature and the target variable, and its value range is [-1, 1], when the coefficient is greater than 0.85, it indicates that the two have a high positive correlation; the grey correlation degree is calculated by calculating the geometric proximity between the reference sequence and the comparison sequence to evaluate the consistency of their system behavior, when the correlation degree is greater than 0.85, it means that the feature has a significant response to the change in the structure of the coal body.

[0079] Further, the double threshold values of the Pearson correlation coefficient and the grey correlation degree in this step are both set to 0.85, which is determined based on the statistical analysis results of multiple sets of experimental data, ensuring that the selected features have significant distinguishing ability under different coal body structures. In addition, during the feature screening process, box plot analysis should also be combined to eliminate outliers, in order to improve the stability and reliability of the model input data.

[0080] Specifically, this step is mainly used for feature selection in the construction of the tectonic coal identification model, providing a high-quality input feature set for subsequent ReliefF feature importance ranking and deep learning model training. Through this screening mechanism, noise features unrelated to the coal structure can be effectively excluded, improving the generalization ability and prediction accuracy of the model.

[0081] Specifically, this step ensures that the selected features are representative in terms of physical meaning and technical feasibility through a dual correlation analysis method, thereby significantly improving the accuracy and robustness of the tectonic coal identification model and providing a scientific basis for predicting the outburst danger zone of the coal mining face.

[0082] S5, spatially register the tectonic coal distribution image collected by the borehole viewer with the prediction result of the tectonic coal identification model, and verify the model prediction accuracy through joint analysis of image features and acoustic vibration features.

[0083] Specifically, this step involves spatially registering the tectonic coal distribution image collected by the borehole viewer with the prediction result of the tectonic coal identification model, and verifying the model prediction accuracy through joint analysis of image features and acoustic vibration features. The technical implementation principle is based on multi-source heterogeneous data fusion and spatial information alignment, aiming to improve the accuracy and reliability of tectonic coal identification.

[0084] Specifically, the spatial registration process uses a three-dimensional space mapping method based on image coordinates and borehole trajectory coordinates. The images collected by the borehole viewer are usually a sequence of two-dimensional images taken continuously along the borehole axis, with each image corresponding to a specific borehole depth. To achieve spatial alignment with the model prediction results, each pixel point in the image sequence needs to be mapped to a three-dimensional coordinate system. Typically, borehole trajectory data (such as borehole inclination, azimuth, and depth) and image acquisition timestamps are synchronized to construct an image depth-coordinate mapping table. Optionally, interpolation algorithms (such as linear interpolation or spline interpolation) can be used for fine processing of image depth to ensure that the image and model prediction results have consistent resolution and accuracy in the spatial dimension.

[0085] Further, image feature extraction can use edge detection, texture analysis (such as gray level co-occurrence matrix GLCM), and color histogram methods to identify visual features such as fractures, fractured zones, and coal structure abnormalities. Acoustic vibration features include amplitude, dominant frequency, rise time, and duration, which are extracted through band-pass filtering and spectral analysis during the coal cutting and destruction process. During joint analysis, feature matching algorithms (such as SIFT and SURF) or deep learning-based image-signal fusion models can be used to correlate and verify image features and acoustic vibration features.

[0086] Specifically, the spatial registration error should be controlled within ±5cm, the image resolution is not less than 1024x768 pixels, the acoustic vibration signal sampling frequency is 20kHz, and the main frequency analysis range is 100Hz-5kHz. This step is mainly used for verifying the prediction results of the while-drilling acoustic vibration identification model in practical applications, especially in the drilling operation near the geological structure zone. Through the comparison between the image and the model results, the identification accuracy and robustness of the model can be effectively evaluated, providing data support for subsequent model optimization and field application. The technical value lies in the collaborative verification of multi-source data, improving the field applicability and prediction accuracy of tectonic coal identification.

[0087] The tectonic coal while-drilling acoustic vibration signal identification method of the embodiment of the present application can effectively identify the acoustic vibration signals generated by the cutting and destruction of coal bodies during drilling, significantly improving the accuracy and reliability of tectonic coal identification. Practical application shows that the method can accurately remove the interference of non-coal-breaking conditions such as drill pipe idling and rod connection through multi-condition signal separation and frequency domain feature extraction technology, and successfully extract pure coal cutting and destruction acoustic vibration signals. The convolutional neural network and recurrent neural network fusion model based on sensitive features can greatly improve the identification accuracy of tectonic coal after five-fold cross-validation and Bayesian optimization, achieving high-precision and high-reliability prediction of tectonic coal distribution, and providing strong technical support for accurate judgment of coal mine gas outburst danger.

[0088] Embodiment 2 To achieve the above-mentioned embodiments, an embodiment of the present application further provides a tectonic coal while-drilling acoustic vibration signal identification system. The specific implementation process of the tectonic coal while-drilling acoustic vibration identification method of the mining working face is as follows, including: S101, coal mine underground different intensity coal seam while-drilling acoustic vibration signal acquisition and coal body physical and mechanical property test; S102, cutting and destruction of coal body acoustic vibration signal identification during drilling; S103, sensitive feature extraction of different coal body structure coal while-drilling acoustic vibration signal and tectonic coal identification model construction; S104, tectonic coal while-drilling acoustic vibration identification method verification of the mining working face. The specific implementation scheme is as follows: S101, coal mine underground different intensity coal seam while-drilling acoustic vibration signal acquisition and coal body physical and mechanical property test.

[0089] Specifically, coal blocks and rock blocks of planned construction extraction drilling holes or detection drilling holes of different mines or different coal and rock layers of the same mine are collected, corresponding coal blocks and rock blocks are made into standard columnar samples (diameter 50 cm, height 100 cm), physical and mechanical properties of the corresponding samples are tested, including uniaxial compressive strength, triaxial compressive strength, shear strength, internal friction angle, cohesion and other parameters, and the physical and mechanical properties of the coal and rock are used to distinguish primary structure coal and tectonic coal. According to the test results of the physical and mechanical properties of the coal and rock, N (N≥50) coal and rock layers with significant differences in physical and mechanical properties of the coal and rock are selected, and drilling sound and vibration signals during drilling are collected from the selected N coal and rock layers extraction drilling holes or detection drilling holes. In order to solve the problem that the vibration source is complex and difficult to distinguish when the sound and vibration sensor is installed on the drilling machine in the past, the present application adopts a near-bit sound and vibration signal collection probe (such as Figure 2 shown) to obtain the sound and vibration signals during drilling. The probe mainly includes a sound and vibration sensor for data collection, a circuit board for data storage, a power battery pack and a data transmission unit (such as Figure 2 shown). In order to facilitate the identification of the sound and vibration signals of the coal body cutting and damage during drilling, the corresponding time periods of the drill pipe idling, the drill bit drilling into the coal and rock body, the rod retraction, the rod connection and the drilling distance of the drill pipe corresponding to the time are recorded in detail.

[0090] S102, identification of the sound and vibration signals of the coal body cutting and damage during drilling (such as Figure 3 shown).

[0091] Specifically, the acoustic vibration signal data is classified according to the working conditions of the drilling rig (idling, coal breaking, rod retraction, rod connection). First, the waveform time domain characteristics of acoustic vibration signals of different categories are analyzed, mainly including amplitude and duration. Compared with the acoustic vibration signals of the drill bit breaking coal, the acoustic vibration signals of the drill rod idling and rod retraction process are usually smaller in amplitude, so a reasonable amplitude threshold can be determined by comparing the three types of acoustic vibration signals, and the acoustic vibration signals of the drill rod idling and rod retraction process are filtered to a certain extent. At the same time, the rod connection process is relatively short, and the acoustic vibration signals generated have a short duration, so a reasonable duration threshold can be determined by comparing the coal breaking acoustic vibration signals and the rod connection acoustic vibration signals, and the acoustic vibration signals of the rod connection are filtered to a certain extent. Second, the frequency domain characteristics of the waveform of different categories of acoustic vibration signals are analyzed. The original acoustic vibration waveform signals are converted using wavelet packet transform method and empirical mode decomposition method to obtain different frequency spectrum graphs and different frequency band decomposition waveform graphs of different acoustic vibration signals. The acoustic vibration signals generated by the drill bit breaking coal usually include acoustic vibration signals generated by the friction between the drill bit and the coal rock body and acoustic vibration signals generated by the breaking of the coal rock body. The acoustic vibration signals generated by the friction between the drill bit and the coal rock body are usually low-frequency signals, and the acoustic vibration signals generated by the breaking of the coal rock body are high-frequency signals. The acoustic vibration signals generated by the drill rod idling, rod retraction and rod connection are mainly low-frequency signals, and there is no high-frequency signal of coal rock body breaking. Therefore, by comparing the different frequency spectrum graphs and different frequency band decomposition waveform graphs of the drill rod idling, rod retraction, rod connection and drill bit coal breaking stages, the coal cutting and breaking acoustic vibration signal frequency band of the drilling process can be determined, and the coal cutting and breaking acoustic vibration signals of the drilling process are further obtained using the band-pass filtering method. The overall implementation process is shown in the flowchart Figure 4 .

[0092] S103, extraction of acoustic vibration signal sensitive characteristics of different coal body structure coals and construction of a coal identification model.

[0093] Specifically, based on the identified acoustic vibration signals of the coal cutting and breaking during drilling, the amplitude, dominant frequency, rise time, duration and other characteristics of the acoustic vibration signals of the coal cutting and breaking of different coal body structure coals are extracted, and the above characteristics of the collected same coal seam are statistically analyzed. Box plots of the above characteristics are drawn, and outliers are excluded to obtain statistical characteristics such as maximum value, minimum value and average value of the corresponding characteristics. The correlation between the statistical characteristic values of the acoustic vibration characteristics of the coal cutting and breaking and the physical and mechanical properties of different coal body structure coals is analyzed by using Pearson and gray correlation degree correlation analysis methods, and the characteristics with a correlation coefficient greater than 85% are selected as the acoustic vibration sensitive characteristics of different coal body structure coals, and a database of acoustic vibration sensitive characteristics of different coal body structure coals is established.

[0094] Further, the machine learning method such as ReliefF or decision tree is used to sort the importance of the acoustic vibration sensitive characteristics of different coal body structures while drilling, analyze the influence weight of the sensitive characteristics, further exclude the noise characteristics and obtain important characteristics; the convolutional neural network and recurrent neural network deep learning algorithm are used to construct a drilling acoustic vibration identification structure coal model integrating multi-dimensional parameters. First, the data set is divided into a training set and a test set, the important characteristics obtained are used as an input layer, and the coal body structure is used as an output layer, the model parameters are adjusted by using five-fold cross-validation and Bayesian optimization and the like, and the optimized model is used for coal body structure prediction; based on AUC, ROC curve and the like, the prediction accuracy of the model for a single coal body structure and the comprehensive prediction performance are evaluated, and the effectiveness of the model is verified by further comparison with the machine learning methods such as support vector machine (SVM), random forest (RF) and BP neural network. The overall implementation process of the drilling identification model of structure coal is shown in Figure 5 .

[0095] S104, verification of the drilling acoustic vibration identification method of structure coal in the mining working face.

[0096] Specifically, since the structure coal is mainly distributed near the geological structure belt, the coal seam with the known geological structure belt is selected for verification. By drilling a borehole in the coal seam with the known geological structure belt, the acoustic vibration signals in the drilling process are collected, the established drilling acoustic vibration identification model of structure coal (such as Figure 4 ) is used to judge the distribution position of the structure coal, and the whole section of the borehole is viewed by using a borehole viewer, and the structure coal positions obtained by the borehole viewing and the drilling acoustic vibration identification are compared, thereby verifying the reliability of the drilling acoustic vibration identification method of structure coal in the mining working face proposed in the present application.

[0097] The drilling acoustic vibration signal discrimination system of structure coal proposed in the embodiments of the present application realizes the advanced detection of the distribution position of structure coal in the mining working face by collecting the drilling acoustic vibration signals of the extraction borehole and the detection borehole drilled every day in the high-gas mine and the outburst mine, and plays an important role in the scientific and accurate implementation of the coal and gas outburst prediction and outburst prevention measures.

[0098] Embodiment 3 In order to realize the above-mentioned embodiments, as Figure 6 shown in the embodiments, a structural schematic diagram of a coal drilling acoustic vibration signal discrimination device 10 is further provided, which comprises: An acoustic vibration signal acquisition and working condition recording module 100 is used to acquire the original acoustic vibration signals in the drilling process through a near-bit acoustic vibration signal acquisition probe pipe, and record the time stamp and drilling distance data corresponding to the working conditions of the drill rod idling, the drill bit breaking coal, the rod retraction and the rod connection.

[0099] Specifically, the original acoustic vibration signals of the drilling process are collected by a combination of piezoelectric sensors and MEMS microphones, and the collected data are uploaded in real time through a wireless radio frequency transmission unit; the drilling distance data are obtained through joint calibration of a laser ranging module and a drill pipe rotation counter.

[0100] The time domain analysis and frequency domain feature separation module 200 is configured to perform time domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signals based on the timestamps, filter non-coal-breaking working condition signals, and then perform frequency domain feature separation on the remaining signals using wavelet packet transform and empirical mode decomposition to extract high-frequency breaking signal bands.

[0101] Specifically, the amplitude threshold is determined based on the statistical characteristics of the outliers excluded based on the box plot, specifically, 1.5 times the average amplitude of the coal-breaking signals is taken as the upper limit of the threshold; the duration threshold is set based on 2 times the average duration of the rod connection process, and the threshold adjustment mechanism is automatically triggered when the duration of the rod connection signal is less than the preset threshold.

[0102] The band-pass filtering processing module 300 is configured to process the separated high-frequency breaking signal bands using a band-pass filtering method to obtain pure coal cutting damage acoustic vibration signals.

[0103] Specifically, a band-pass filter with a center frequency range of 2-8 kHz is set, the bandwidth is dynamically adjusted according to the main frequency distribution of the IMF components after empirical mode decomposition; an adaptive filtering algorithm is used to calculate the signal spectrum features in real time through a sliding window and update the filtering parameters.

[0104] The feature analysis and model construction module 400 is configured to filter sensitive features using a correlation analysis method based on the amplitude, main frequency, rise time and duration characteristics of the coal cutting damage acoustic vibration signals, and construct a coal recognition model through convolutional neural networks and recurrent neural networks, and adjust the model parameters using five-fold cross-validation and Bayesian optimization to realize the construction of the coal distribution prediction.

[0105] Specifically, in the correlation analysis method, when the Pearson correlation coefficient and the grey correlation degree are both greater than the preset threshold, the feature is determined to be a sensitive feature of the tectonic coal.

[0106] The structural coal while-drilling acoustic vibration signal discrimination device can effectively discriminate the acoustic vibration signal generated by the cutting and damage of the coal body in the drilling process, and significantly improve the accuracy and reliability of structural coal identification. Practical application shows that the method can accurately identify and eliminate non-coal-breaking working condition interference such as drill pipe idling and rod connection through multi-sensor data fusion and adaptive signal processing technology, and successfully extract pure acoustic vibration signals reflecting the structural change of the coal body. The convolution neural network and recurrent neural network fusion model constructed based on the double correlation degree feature screening mechanism greatly improves the identification accuracy of structural coal after five-fold cross-validation and Bayesian optimization, realizes high-precision and high-reliability prediction of the distribution of structural coal, and provides reliable technical support for dynamic assessment of the risk of gas outburst in coal mines.

[0107] In the description of the present specification, the description referring to the terms "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" and the like means that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In the present specification, the illustrative description of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, different embodiments or examples described in the present specification and the features of different embodiments or examples can be combined and combined by those skilled in the art without contradiction.

[0108] In addition, the terms "first", "second" are only for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined with "first", "second" can explicitly or implicitly include at least one of the features. In the description of the present application, the meaning of "a plurality of" is at least two, for example, two, three, etc., unless otherwise specifically limited.

Claims

1. A method for identifying acoustic vibration signals during coal drilling, characterized in that, include: S1, the raw acoustic and vibration signals of the drilling process are obtained through the near-bit acoustic and vibration signal acquisition probe, and the timestamps and drilling distance data corresponding to the drill rod idling, drill bit coal breaking, rod retraction and rod connection are recorded; S2, based on the timestamp, perform time-domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signal, filter out non-coal breaking working condition signals, and use wavelet packet transform and empirical mode decomposition to separate the frequency domain features of the remaining signal to extract the high-frequency breaking signal frequency band; S3, the high-frequency fracture signal band after separation is processed by bandpass filtering to obtain a pure acoustic vibration signal of coal body cutting and destruction; S4. Based on the amplitude, dominant frequency, rise time and duration characteristics of the acoustic vibration signal of coal cutting and damage, a correlation analysis method is used to screen sensitive features, and a tectonic coal identification model is constructed by using convolutional neural network and recurrent neural network. Five-fold cross-validation and Bayesian optimization are used to adjust the model parameters to achieve tectonic coal distribution prediction.

2. The method as described in claim 1, characterized in that, S1 further includes: S11, the original acoustic and vibration signal acquisition probe is arranged with a combination of piezoelectric sensor and MEMS microphone, and the acquired data is uploaded in real time through a wireless radio frequency transmission unit; S12, the drilling distance data is obtained through joint calibration of the laser ranging module and the drill pipe rotation counter.

3. The method as described in claim 1, characterized in that, S2 further includes: S21, the time-domain amplitude threshold is determined based on the statistical characteristics after the box plot outliers are excluded, specifically, 1.5 times the average amplitude of the coal breaking signal is taken as the upper limit of the threshold; S22, the duration threshold is set based on twice the average duration of the rod connection process, and the threshold adjustment mechanism is automatically triggered when the duration of the rod connection signal is less than the preset threshold.

4. The method as described in claim 1, characterized in that, S3 further includes: S31, the center frequency range of the bandpass filter is 2-8kHz, and the bandwidth is dynamically adjusted according to the main frequency distribution of the IMF components after empirical mode decomposition; S32, the processing adopts an adaptive filtering algorithm, which calculates the signal spectrum characteristics and updates the filtering parameters in real time through a sliding window.

5. The method as described in claim 1, characterized in that, S4 further includes: S41, In the correlation analysis method, when both the Pearson correlation coefficient and the grey relational degree are greater than the preset threshold, the feature is determined to be a coal-sensitive feature.

6. The method as described in claim 1, characterized in that, Also includes: S5. Spatial registration is performed between the borehole wall coal distribution image collected by the borehole inspection instrument and the prediction results of the coal identification model. The prediction accuracy of the model is verified by joint analysis of image features and acoustic vibration features.

7. A device for identifying acoustic vibration signals during coal drilling, characterized in that, include: The acoustic and vibration signal acquisition and working condition recording module is used to acquire the original acoustic and vibration signals during the drilling process through the near-drill bit acoustic and vibration signal acquisition probe, and record the timestamps and drilling distance data corresponding to the working conditions of drill rod idling, drill bit coal breaking, rod retraction and rod connection; The time-domain analysis and frequency-domain feature separation module is used to perform time-domain amplitude threshold analysis and duration threshold analysis on the original acoustic vibration signal based on the timestamp. After filtering out non-coal breaking working condition signals, wavelet packet transform and empirical mode decomposition are used to perform frequency-domain feature separation on the remaining signal to extract the high-frequency breaking signal band. The bandpass filter processing module is used to process the separated high-frequency fracture signal band using the bandpass filter method to obtain a pure acoustic vibration signal of coal body cutting and destruction. The feature analysis and model building module is used to screen sensitive features based on the amplitude, dominant frequency, rise time and duration characteristics of the acoustic vibration signal of coal cutting and damage, and to build a tectonic coal identification model through convolutional neural network and recurrent neural network. The model parameters are adjusted by five-fold cross-validation and Bayesian optimization to achieve the prediction of tectonic coal distribution.

8. The apparatus as claimed in claim 7, characterized in that, The original acoustic and vibration signal acquisition and operating condition recording module is also used for: The raw acoustic and vibration signals during the drilling process are collected by combining piezoelectric sensors and MEMS microphones, and the collected data is uploaded in real time through a wireless radio frequency transmission unit. Drilling distance data is obtained through joint calibration of the laser ranging module and the drill pipe rotation counter.

9. The apparatus as claimed in claim 7, characterized in that, The time-domain analysis and frequency-domain feature separation module is also used for: The amplitude threshold is determined based on the statistical characteristics after outlier exclusion from the box plot. Specifically, 1.5 times the average amplitude of the coal breaking signal is taken as the upper limit of the threshold. The duration threshold is set based on twice the average duration of the relay process, and the threshold adjustment mechanism is automatically triggered when the relay signal duration is less than the preset threshold.

10. The apparatus as claimed in claim 7, characterized in that, The bandpass filtering module is also used for: Set a bandpass filter with a center frequency range of 2-8kHz, and dynamically adjust the bandwidth according to the main frequency distribution of the IMF components after empirical mode decomposition. An adaptive filtering algorithm is adopted to calculate the signal spectrum characteristics and update the filtering parameters in real time through a sliding window.

11. The apparatus as claimed in claim 7, characterized in that, The feature analysis and model building module is also used for: In the correlation analysis method, when both the Pearson correlation coefficient and the grey relational degree are greater than the preset threshold, the feature is determined to be a coal-sensitive feature.

12. The apparatus as claimed in claim 7, characterized in that, Also includes: The image registration and joint analysis module is used to spatially register the images of coal distribution on the borehole wall collected by the borehole sighting instrument with the prediction results of the coal identification model, and to verify the prediction accuracy of the model through joint analysis of image features and acoustic vibration features.

Images