Near-bit stratum detection method based on while-drilling sound wave foresight

Abstract

The invention discloses a near-bit stratum detection method based on while-drilling sound wave foresight, and belongs to the technical field of stratum detection, and the method comprises the steps: S1, obtaining while-drilling sound wave signals generated by rock breaking of a drill bit in a drilling process, carrying out the time-frequency analysis of the while-drilling sound wave signals, and determining while-drilling sound wave source features through combining theoretical analysis; s2, carrying out independent component analysis on the acoustic wave signals while drilling, combining with ensemble empirical mode decomposition (CEEMD) and wavelet transform, and identifying effective acoustic wave signals while drilling from complex background noise; and S3, establishing an unfavorable geologic body geophysical model comprising a lithologic interface, a fault fracture zone, a karst cave and a water-rich stratum, and adopting a time domain finite difference numerical simulation method to obtain a propagation rule and characteristic parameters of the acoustic wave while drilling in the model. Through multi-algorithm cooperative processing, the while-drilling sound wave signals with high signal-to-noise ratio and obvious characteristics are extracted from complex background noise, and the reliability and precision of signal processing are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of stratum exploration, and more particularly to a near-bit stratum exploration method based on acoustic wave look-ahead while drilling. BACKGROUND

[0002] The measurement-while-drilling technology is a widely used method in oil and gas exploration and geological drilling, which obtains physical signals near the drill bit or drill pipe during drilling to assist in judging the structure and properties of the stratum. As an important means, acoustic wave measurement while drilling can provide acoustic wave information of the stratum in front of the drill bit, so as to realize the pre-judgment of lithology changes and adverse geological bodies (such as fault fracture zone, karst cave, water-rich layer, etc.). However, the existing acoustic wave measurement while drilling method has the following main problems: complex noise is generated by rock breaking of the drill bit during drilling, including mechanical vibration, drilling pressure fluctuation, and drill pipe resonance, etc., so that the acoustic wave signal while drilling appears as low signal-to-noise ratio and unclear time-frequency characteristics at the receiving end. The traditional filtering or simple frequency domain analysis method is difficult to accurately extract the effective acoustic wave signal from the complex background, which affects the accuracy of stratum exploration. The existing technology usually cannot accurately determine the comprehensive characteristic parameters of the acoustic wave source while drilling, such as the main frequency, frequency band width, energy attenuation coefficient, and directivity, etc. by combining theoretical analysis, which leads to the lack of pertinence of signal processing and stratum imaging model, and affects the reliability of the imaging result. The traditional method relies on empirical formula or simple numerical simulation, and it is difficult to fully consider the influence of complex adverse geological bodies such as lithology boundary, fault fracture zone, karst cave, and water-rich layer, etc., and lacks high-precision acoustic wave propagation rule modeling and characteristic parameter extraction means. At the same time, the existing imaging method has limited training and prediction ability for large-scale and multi-type geological bodies, and it is difficult to realize high-precision real-time imaging of the stratum in front of the near-bit. SUMMARY

[0003] The present application aims to provide a near-bit stratum exploration method based on acoustic wave look-ahead while drilling, to solve the problems raised in the background art: complex noise is generated by rock breaking of the drill bit during drilling, including mechanical vibration, drilling pressure fluctuation, and drill pipe resonance, etc., so that the acoustic wave signal while drilling appears as low signal-to-noise ratio and unclear time-frequency characteristics at the receiving end. The traditional filtering or simple frequency domain analysis method is difficult to accurately extract the effective acoustic wave signal from the complex background, which affects the accuracy of stratum exploration. The existing technology usually cannot accurately determine the comprehensive characteristic parameters of the acoustic wave source while drilling, such as the main frequency, frequency band width, energy attenuation coefficient, and directivity, etc. by combining theoretical analysis, which leads to the lack of pertinence of signal processing and stratum imaging model, and affects the reliability of the imaging result. The traditional method relies on empirical formula or simple numerical simulation, and it is difficult to fully consider the influence of complex adverse geological bodies such as lithology boundary, fault fracture zone, karst cave, and water-rich layer, etc., and lacks high-precision acoustic wave propagation rule modeling and characteristic parameter extraction means. At the same time, the existing imaging method has limited training and prediction ability for large-scale and multi-type geological bodies, and it is difficult to realize high-precision real-time imaging of the stratum in front of the near-bit.

[0004] The near-bit formation detection method based on the while-drilling acoustic forward-looking comprises the following steps: S1, acquiring the while-drilling acoustic signal generated by the rock breaking of the drill bit in the drilling process, performing time-frequency analysis on the while-drilling acoustic signal, and determining the while-drilling acoustic source characteristics in combination with theoretical analysis; S2, performing independent component analysis on the while-drilling acoustic signal, and combining with the ensemble empirical mode decomposition (CEEMD) and wavelet transform to identify the effective while-drilling acoustic signal from the complex background noise; S3, establishing a geophysical model of adverse geological bodies including lithological boundary surface, fault fracture zone, karst cave and water-rich formation, and using the time-domain finite difference numerical simulation method to obtain the propagation law and characteristic parameters of the while-drilling acoustic signal in the model; S4, constructing a while-drilling acoustic field characteristic sample library based on the characteristic parameters, and training the sample library by using a deep learning model to obtain an adverse geological body imaging model; S5, inputting the acquired while-drilling acoustic signal into the adverse geological body imaging model to obtain the imaging result of the formation structure in front of the drill bit.

[0005] Preferably, the determination of the while-drilling acoustic source characteristics in S1 comprises the following steps: The time-frequency distribution characteristics of the while-drilling acoustic signal in different time intervals are obtained by using short-time Fourier transform to perform time-frequency decomposition on the while-drilling acoustic signal; The frequency spectrum distribution is fitted and parameter inversion based on the while-drilling acoustic propagation theory, the main frequency component, frequency band width, energy attenuation coefficient and directivity parameter of the while-drilling acoustic source are extracted, so as to obtain a source characteristic parameter set representing the comprehensive characteristics of the while-drilling acoustic source in time domain and frequency domain.

[0006] Preferably, the step of identifying the effective while-drilling acoustic signal from the complex background noise in S2 is as follows: S2.1, performing detrending and normalization processing on the while-drilling acoustic signal, setting a band-pass filter according to the while-drilling acoustic source characteristic parameters (including the main frequency component and the frequency band width) determined in S1, performing band-pass filtering on the while-drilling acoustic signal, then framing and windowing the filtered signal according to a fixed frame length to obtain a frame signal set to be decomposed; S2.2, applying ensemble empirical mode decomposition (CEEMD) to the frame signal, decomposing it into several intrinsic mode functions (IMF) and a residual term, and calculating the instantaneous energy, instantaneous frequency distribution and spectral centroid of each IMF for representing the time-frequency energy concentration; S2.3, performing a discrete wavelet transform on the IMF, extracting different frequency band features according to a multi-scale decomposition mode, and adopting a soft threshold method to denoise the wavelet coefficients, and then reconstructing the denoised IMF through inverse wavelet transform, while calculating the multi-scale energy distribution and energy proportion; S2.4, according to the while-drilling acoustic source feature parameters in S2, and in combination with the IMF features, ranking and screening the IMF according to the energy proportion, the spectral matching degree with the source main frequency band, the instantaneous frequency stability, and the kurtosis, and retaining a plurality of candidate IMF with high matching degree with the while-drilling acoustic source features; S2.5, forming a matrix of the candidate IMF in the multi-channel dimension, inputting an independent component analysis (ICA) algorithm for blind source separation to obtain a plurality of independent components, and respectively calculating the cross-correlation coefficient with the while-drilling acoustic source features, the energy concentration degree of the independent component, the signal-to-noise ratio, and the instantaneous frequency consistency; S2.6, according to the energy proportion of the independent component, the cross-correlation coefficient with the while-drilling acoustic source feature template, the signal-to-noise ratio, and the instantaneous frequency consistency, selecting the independent component meeting the requirements as an effective independent component, and after wavelet domain denoising processing, weighting and superimposing to obtain a reconstructed effective while-drilling acoustic signal; Preferably, when the source feature parameters cannot be obtained, a preset empirical frequency band is used for band-pass filtering.

[0007] Preferably, the specific steps of S3 are as follows: S3.1, establishing a geophysical model of the adverse geological body according to the lithology, fault fracture zone, karst cave, and water-rich layer of the drilled stratum; S3.2, obtaining the propagation law of the while-drilling acoustic wave in the geophysical model by using a numerical simulation method; S3.3, extracting acoustic wave propagation feature parameters including propagation time, amplitude, energy attenuation, and spectral features from the simulation results; S3.4, combining the feature parameters with corresponding geological type information to construct an adverse geological body acoustic field feature sample library for subsequent imaging model training.

[0008] Preferably, the specific steps of S4 are as follows: S4.1, inputting the adverse geological body acoustic field feature sample library into a deep learning model for training the model to identify different types of adverse geological bodies; S4.2, constructing a deep learning model including a deep neural network, a convolutional neural network, or a space-time Transformer, etc., for establishing a mapping relationship between the feature parameters and the geological types; S4.3, defining a loss function and iteratively optimizing the model parameters to minimize the error between the prediction result and the true geological type label. S4.4, after training, an adverse geology body imaging model is obtained, which can input the acoustic features while drilling to predict the formation structure in front of the drill bit to realize formation imaging.

[0009] Preferably, the S5 specific steps are as follows: S5.1, input the effective acoustic features while drilling obtained into the trained adverse geology body imaging model to predict the formation type label in front of the drill bit; S5.2, map the predicted formation type label to a spatial distribution representation to form a three-dimensional or two-dimensional imaging of the formation structure in front of the drill bit; S5.3, set a decision threshold according to the imaging result to mark the key adverse geology body area to realize abnormal formation identification; S5.4, output the formation type prediction result, spatial imaging result and marking result for guiding drilling operation or geological analysis.

[0010] Preferably, the model fitting of the source feature parameters adopts least square fitting, genetic algorithm or particle swarm optimization algorithm, and the parameters include a source main frequency component in the range of 100-5000 Hz, a frequency band width in the range of 50-1000 Hz, an energy attenuation coefficient in the range of 0.1-5 dB / m, and a directivity parameter is an angle range related to the rotation direction of the drill bit.

[0011] Preferably, the frame length of the frame after band-pass filtering and windowing in S2 is 50-500 ms, the sampling rate is 10-100 kHz, the windowing type is Hanning window or Blackman window, the CEEMD decomposition layer number is 5-10 layers, the wavelet denoising adopts Daubechies wavelet, the soft threshold is set according to 0.2-0.5 times of the channel noise energy, and the ICA separation matrix is realized by maximizing mutual information or minimizing high-order moment.

[0012] Preferably, the fault fracture zone, cave and water-rich layer in the adverse geology body geophysical model can be alternatively set as different formation combinations, and the acoustic wave propagation law is simulated by the finite difference time domain method, and the feature parameters include propagation time, amplitude, spectral center frequency, instantaneous frequency and energy attenuation index.

[0013] Compared with the prior art, the advantages of the present application are: (1) The present application combines multi-algorithm collaborative processing such as ensemble empirical mode decomposition (CEEMD), wavelet transform and independent component analysis (ICA) to extract high signal-to-noise ratio and obvious feature acoustic signals while drilling from complex background noise, which improves the reliability and accuracy of signal processing.

[0014] (2) The present application accurately obtains the main frequency component, frequency band width, energy attenuation coefficient and directivity parameter of the while-drilling acoustic wave source through short-time Fourier transform, model fitting and parameter inversion, so that the signal processing and formation imaging are targeted, and the accuracy of formation identification is enhanced.

[0015] (3) The present application realizes accurate characterization of the acoustic wave behavior of complex geological bodies by establishing a geophysical model of adverse geological bodies containing lithological boundaries, fault fracture zones, karst caves and water-rich layers, and extracting acoustic wave propagation characteristic parameters by using time-domain finite difference numerical simulation.

[0016] (4) The present application realizes automatic identification and prediction of different types of adverse geological bodies based on a deep learning model trained based on an acoustic wave field characteristic sample library, can quickly map the while-drilling acoustic wave characteristics to two-dimensional or three-dimensional imaging results of the formation structure in front of the near-bit, and supports real-time decision-making for drilling operations.

[0017] (5) The present application can early warn potential geological risks, optimize drilling strategies, reduce drilling accidents and drill bit wear, and improve the safety and economic benefits of drilling operations by marking key adverse geological body regions and identifying abnormal formations. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The present application is a near-bit formation detection method based on while-drilling acoustic wave forward-looking. DETAILED DESCRIPTION

[0019] The technical solutions of the present application will be described below in conjunction with embodiments, obviously, the described embodiments are only some 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 are within the scope of protection of the present application.

[0020] Please refer to Figure 1 , the near-bit formation detection method based on while-drilling acoustic wave forward-looking includes the following steps: S1, obtaining the while-drilling acoustic wave signal generated by the rock breaking of the drill bit during drilling, performing time-frequency analysis on the while-drilling acoustic wave signal, and determining the while-drilling acoustic wave source characteristics in combination with theoretical analysis; The determination of the while-drilling acoustic wave source characteristics in S1 includes the following steps: The frequency spectrum distribution characteristics of the while-drilling acoustic wave in different time intervals are obtained by performing time-frequency decomposition on the while-drilling acoustic wave signal using short-time Fourier transform; when performing short-time Fourier transform (STFT) time-frequency analysis on the while-drilling acoustic wave signal, the time window length of the STFT is 10-50 ms, the overlap rate is 50%-80%, and the frequency resolution is 5-50 Hz; Based on the theory of acoustic propagation while drilling, the spectrum distribution is model fitted and parameter inverted, the model fitting method includes least square fitting or genetic algorithm optimization to obtain the source characteristic parameters with minimum fitting error; the main frequency component, frequency band width, energy attenuation coefficient and directivity parameter of the acoustic source while drilling are extracted to obtain the source characteristic parameter set representing the comprehensive characteristics of the acoustic source while drilling in time domain and frequency domain.

[0021] The source characteristic parameters of the acoustic source while drilling are extracted from the fitting model, including: main frequency component f0, typically ranging from 500 to 5000 Hz; frequency band width Δf, typically ranging from 200 to 2000 Hz; energy attenuation coefficient α, typically ranging from 0.05 to 0.5 dB / m; directivity parameter θ, typically ranging from ±30°, used to represent the directional energy distribution of the source; S2, independent component analysis is performed on the acoustic signal while drilling, combined with ensemble empirical mode decomposition (CEEMD) and wavelet transform, to identify the effective acoustic signal while drilling from complex background noise; S3, a geophysical model of adverse geologic body including lithology interface, fault fracture zone, cave and water-rich stratum is established, and the propagation law and characteristic parameters of the acoustic wave while drilling in the model are obtained by using time domain finite difference numerical simulation method; S4, based on the characteristic parameters, a characteristic sample library of acoustic wave field of adverse geologic body is constructed, and a deep learning model is used to train the sample library to obtain an imaging model of adverse geologic body; S5, the acquired acoustic signal while drilling is input into the imaging model of adverse geologic body to obtain the imaging result of the structure of the stratum in front of the drill bit.

[0022] The step of identifying the effective acoustic signal while drilling from complex background noise in S2 is as follows: S2.1 signal preprocessing The acoustic signal while drilling is subjected to detrending and normalization processing; The signal is divided into frames according to the frame length , and a Hamming window or a Hanning window is used for windowing, the frame overlap rate is , and the sampling rate is ; A band-pass filter is set according to the source characteristic parameters determined in S1, and the filter range covers half of the main frequency component ± frequency band width.

[0023] S2.2 ensemble empirical mode decomposition (CEEMD) The frame signal is subjected to CEEMD decomposition to obtain a plurality of IMF and residual terms; The noise amplitude of CEEMD is the standard deviation of the original signal times, and the iteration number is times.

[0024] S2.3 Wavelet transform denoising Discrete wavelet transform (multi-scale decomposition of 3-6 layers) is performed on the IMF, and the wavelet basis is Daubechies or Symlets wavelet; The wavelet coefficients are denoised by soft threshold method, and the threshold formula is:

[0025] Wherein: represents the noise standard deviation of the IMF signal; represents the length of the IMF signal.

[0026] Inverse wavelet transform is performed to reconstruct the denoised IMF, and the multi-scale energy distribution and energy proportion are calculated.

[0027] S2.4 IMF screening The IMF is sorted and screened according to the following indicators: Energy proportion:

[0028] Wherein: represents the energy of the th IMF; represents the sum of the energies of all IMFs.

[0029] Spectrum matching degree:

[0030] Wherein: represents the spectrum of the th IMF; represents the spectrum of the source feature template; represents the sum of the frequencies

[0031] Instantaneous frequency stability:

[0032] Wherein: represents the instantaneous frequency standard deviation of the th IMF.

[0033] Kurtosis:

[0034] Wherein: represents the​ The kurtosis of the IMF.

[0035] IMFs that meet the above criteria are selected as candidate IMFs.

[0036] S2.5 Independent Component Analysis (ICA) Construct a matrix of candidate IMFs in multiple dimensions. Input the ICA algorithm for blind source separation and calculate the separation matrix. Iterate and optimize until convergence; Independent components that meet the following criteria are selected as valid independent components: Energy percentage

[0037] Cross-correlation number with source characteristic template

[0038] Signal-to-noise ratio

[0039] Instantaneous frequency consistency

[0040] S2.6 Effective Signal Reconstruction After wavelet domain denoising of the independent components that meet the conditions, they are weighted and superimposed according to their energy proportions to obtain the final effective drilling acoustic signal. :

[0041] in: This represents the reconstructed effective acoustic signal while drilling. Indicates the first Each independent component signal; This represents the weighting coefficient, which is set based on the energy percentage of each independent component. Indicates the first Energy of an independent component; Indicates the number of effective independent components; This represents the wavelet domain denoising function.

[0042] The specific steps of S3 are as follows: S3.1 Establishment of Adverse Geological Body Model Construct a geophysical model of an adverse geological body that includes lithological interfaces, fault fracture zones, karst caves, and water-rich strata; Model parameters include lithological density Elastic modulus Poisson's ratio Fault thickness radius of the karst cave Porosity of aquifers, etc., are used to describe the physical properties of formations.

[0043] Note: These parameters are used to describe the spatial distribution and physical properties of the adverse geological body. The specific values can be set according to the on-site measurement or geological data.

[0044] S3.2 Acoustic wave numerical simulation while drilling The propagation of acoustic waves while drilling in the model is simulated using the finite difference time domain (FDTD) method, and the basic update formula is:

[0045] wherein: represents the displacement of the th grid node at the th time step; represents the time step; represents the density of the th grid node; represents the elastic modulus of the th grid node; represents the spatial derivative operator.

[0046] S3.3 Feature parameter extraction After simulating the propagation field of acoustic waves while drilling, the following feature parameter set is extracted: :

[0047] wherein: represents the propagation time of the peak value of the acoustic wave; represents the peak amplitude; represents the energy attenuation coefficient, wherein is the source amplitude, is the amplitude at the propagation distance ; represents the central frequency of the acoustic wave or the spectral feature parameter.

[0048] S3.4 Construction of acoustic wave field feature sample library The feature parameters extracted in S3.3 are combined with the corresponding geological body type label to form an adverse geological body acoustic wave field feature sample library:

[0049] wherein: represents the adverse geological body acoustic wave field feature sample library; represents the number of samples; represents the feature parameter set of the th sample; represents the geological type label corresponding to the th sample.

[0050] The specific steps of S4 are as follows: S4.1 Sample Input The acoustic field feature sample library of adverse geological bodies constructed in S3.4 Input a deep learning model; in: Indicates the first A set of acoustic wave characteristic parameters for each sample (propagation time, amplitude, energy attenuation, spectral characteristics, etc.). Indicates the first The label of the adverse geological body type corresponding to each sample; This indicates the total number of samples.

[0051] S4.2 Model Structure Utilizing models such as deep neural networks (DNN), convolutional neural networks (CNN), or spatiotemporal Transformers. The model is trained on samples and outputs... To predict the type of geological body:

[0052] in: Represents a deep learning model. It is a set of trainable parameters; Indicates the first The type of geological body predicted for each sample.

[0053] S4.3 Loss Function and Training Optimization Define loss function To measure the deviation between the predicted result and the true label, for example, cross-entropy loss:

[0054] in: Indicates the total number of geological body categories; Indicates the first Each sample in category The actual label (0 or 1) on it; Indicates the model output in the category The predicted probability.

[0055] Update using optimization algorithms such as gradient descent or Adam. until Convergence or reaching the preset number of iterations.

[0056] S4.4 Imaging Model Output After training, an imaging model of the adverse geological body is obtained. :

[0057] Wherein: denotes the trained bad geology body imaging model; denotes the optimized parameters obtained by training; the while-drilling acoustic features can be input to predict the near-bit formation structure and realize formation imaging.

[0058] The specific steps of S5 are as follows: S5.1 Feature input The effective while-drilling acoustic signal feature vector obtained is input into the trained bad geology body imaging model :

[0059] Wherein: denotes the while-drilling acoustic feature parameters, including propagation time, peak amplitude, energy attenuation coefficient, and spectral features; denotes the trained bad geology body imaging model; denotes the near-bit formation type label predicted by the model.

[0060] S5.2 Imaging inference The is converted into a spatial distribution representation of the formation structure in front of the near-bit :

[0061] Wherein: denotes the spatial coordinates on which the formation imaging result is obtained; denotes a function based on model output for spatial interpolation or mapping.

[0062] S5.3 Visualization and analysis The is visualized as a three-dimensional formation model or a two-dimensional profile for determining the position and range of the bad geology body; The threshold value can be combined to mark the key bad geology body:

[0063] Wherein: denotes the marking matrix, 1 represents the bad geology body area, and 0 represents the normal formation; denotes the formation type determination threshold value, which is determined according to experience or model training.

[0064] S5.4 Output and application The , and results are output for guiding drilling operations or geological analysis; Through the output, real-time monitoring and early warning of the formation structure in front of the drill bit can be realized.

[0065] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited to the above-mentioned embodiments, and the above-mentioned embodiments and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A method for near-bit formation detection based on acoustic look-ahead while drilling, characterized in that, The method comprises the following steps: S1, acquiring a while-drilling acoustic signal generated by a drill bit in a drilling process, performing time-frequency analysis on the while-drilling acoustic signal, and determining a while-drilling acoustic source feature in combination with theoretical analysis; S2, performing independent component analysis on the while-drilling acoustic signal, identifying effective while-drilling acoustic signals from complex background noise in combination with ensemble empirical mode decomposition (CEEMD) and wavelet transform; S3, establishing a geophysical model of a bad geological body including a lithology boundary surface, a fault fracture zone, a karst cave and a water-rich stratum, and obtaining a propagation rule and a characteristic parameter of the while-drilling acoustic signal in the model by using a time-domain finite difference numerical simulation method; S4, constructing a while-drilling acoustic field characteristic sample library based on the characteristic parameter, and training the sample library by using a deep learning model to obtain a bad geological body imaging model; S5, inputting the acquired while-drilling acoustic signal into the bad geological body imaging model to obtain an imaging result of a stratum structure in front of the drill bit.

2. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 1, wherein, The step of determining the while-drilling acoustic source feature in S1 comprises the following steps: obtaining frequency spectrum distribution characteristics of the while-drilling acoustic signal in different time intervals by performing time-frequency decomposition on the while-drilling acoustic signal by using a short-time Fourier transform; extracting a main frequency component, a frequency band width, an energy attenuation coefficient and a directivity parameter of the while-drilling acoustic source by performing model fitting and parameter inversion on the frequency spectrum distribution based on a while-drilling acoustic propagation theory, so as to obtain a source feature parameter set representing comprehensive characteristics of the while-drilling acoustic source in the time domain and the frequency domain.

3. The near-bit formation investigation method based on acoustic while drilling look-ahead of claim 2, wherein, The step of identifying the effective while-drilling acoustic signal from the complex background noise in S2 comprises the following steps: S2.1, performing detrend processing and normalization processing on the while-drilling acoustic signal, setting a band-pass filter according to the while-drilling acoustic source feature parameter determined in S1, performing band-pass filtering on the while-drilling acoustic signal, then framing and windowing the filtered signal according to a fixed frame length to obtain a frame signal set to be decomposed; S2.2, applying ensemble empirical mode decomposition (CEEMD) to the frame signal to decompose it into a plurality of intrinsic mode functions (IMFs) and a residual term, and calculating an instantaneous energy, an instantaneous frequency distribution and a frequency spectrum centroid of each IMF for representing time-frequency energy concentration; S2.3, performing discrete wavelet transform on the IMF, extracting different frequency band features according to a multi-scale decomposition manner, performing denoising processing on wavelet coefficients by using a soft threshold method, and reconstructing the denoised IMF by inverse wavelet transform, while calculating a multi-scale energy distribution and an energy proportion; S2.4, sorting and screening the IMFs according to the while-drilling acoustic source feature parameter and the IMF features, and retaining a plurality of candidate IMFs with high matching degrees with the while-drilling acoustic source feature according to indexes such as an energy proportion, a frequency spectrum matching degree with a main frequency band of the source, an instantaneous frequency stability and a kurtosis. S2.5, the candidate IMF is constituted into a matrix in the multi-channel dimension, an independent component analysis (ICA) algorithm is inputted to perform blind source separation, a plurality of independent components are obtained, and a cross-correlation coefficient with the while-drilling acoustic wave source feature in S2, an energy concentration degree of the independent component, a signal-to-noise ratio and an instantaneous frequency consistency are respectively calculated; S2.6, according to the energy proportion of the independent component, the cross-correlation coefficient with the while-drilling acoustic wave source feature template, the signal-to-noise ratio and the instantaneous frequency consistency, an independent component meeting the requirements is selected as an effective independent component, a wavelet domain denoising processing is performed on the effective independent component, and then the effective independent component is weighted and superimposed to obtain a reconstructed effective while-drilling acoustic wave signal.

4. The near-bit formation investigation method based on acoustic while drilling look-ahead of claim 3, wherein, When the source feature parameter cannot be obtained, a preset empirical frequency band is used for band-pass filtering.

5. The near-bit formation investigation method based on acoustic wave while-drilling look-ahead of claim 3, wherein, The specific steps of S3 are as follows: S3.1, a geophysical model of the adverse geologic body is established according to the lithology, fault fracture zone, karst cave and water-rich layer and the like of the drilled stratum; S3.2, in the geophysical model, a numerical simulation method is used to obtain the propagation law of the while-drilling acoustic wave; S3.3, acoustic wave propagation feature parameters are extracted from the simulation result, including propagation time, amplitude, energy attenuation and spectral feature; S3.4, the feature parameters are combined with corresponding geologic type information to construct an adverse geologic body acoustic wave field feature sample library, which is used for subsequent imaging model training.

6. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 5, wherein, The specific steps of S4 are as follows: S4.1, the adverse geologic body acoustic wave field feature sample library is inputted into a deep learning model to train the model to identify different types of adverse geologic bodies; S4.2, a deep learning model is constructed, including a deep neural network, a convolutional neural network or a space-time Transformer, to establish a mapping relationship between the feature parameters and the geologic types; S4.3, a loss function is defined and the model parameters are iteratively optimized to minimize the error between the prediction result and the real geologic type label; S4.4, after the training is completed, an adverse geologic body imaging model is obtained, which can input the while-drilling acoustic wave feature to predict the stratum structure near the front of the drill bit, thereby realizing stratum imaging.

7. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 6, wherein, The specific steps of S5 are as follows: S5.1, the obtained effective while-drilling acoustic wave feature is inputted into the trained adverse geologic body imaging model to predict the stratum type label near the front of the drill bit; S5.2, the predicted stratum type label is mapped to a spatial distribution representation to form a three-dimensional or two-dimensional imaging of the stratum structure near the drill bit; S5.3, according to the imaging result, a determination threshold is set to mark the key adverse geologic body area, thereby realizing abnormal stratum identification; S5.4, the stratum type prediction result, the spatial imaging result and the marking result are outputted to guide the drilling operation or perform geologic analysis.

8. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 2, wherein, The model fitting of the source feature parameter adopts a least square fitting, a genetic algorithm or a particle swarm optimization algorithm, and the parameters include a source main frequency component in a range of 100-5000 Hz, a frequency band width in a range of 50-1000 Hz, an energy attenuation coefficient in a range of 0.1-5 dB / m and a directivity parameter as an angle range related to the rotation direction of the drill bit.

9. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 3, wherein, The frame length of the band-pass filtered frame in S2 is 50-500 ms, the sampling rate is 10-100 kHz, the window type is Hanning window or Blackman window, the decomposition layer number of ensemble empirical mode decomposition is 5-10 layers, the wavelet denoising adopts Daubechies wavelet, the soft threshold is set according to 0.2-0.5 times of the channel noise energy, and the ICA separation matrix is achieved by maximizing mutual information or minimizing high-order moments.

10. The near-bit formation detection method based on acoustic wave while-drilling look-ahead of claim 5, wherein, The fault fracture zone, karst cave and water-rich layer in the geophysical model of the adverse geological body can be alternatively set as different stratum combinations, the sound wave propagation law is simulated by a finite difference method in time domain, and the characteristic parameters include propagation time, amplitude, spectral center frequency, instantaneous frequency and energy attenuation index.

Images