Acoustic remote detection device and detection method suitable for horizontal borehole

Abstract

An acoustic remote detection device and detection method suitable for a horizontal borehole. The detection method comprises: first, acquiring a reflected acoustic signal; subsequently, using a signal preprocessing module to process the reflected acoustic signal to obtain a sequence of reflected acoustic signal segments; then, performing waveform feature extraction on the sequence of the reflected acoustic signal segments to obtain a sequence of reflected acoustic signal segment waveform feature vectors; next, using a positional encoding and self-correlation highlighting module to process the sequence of the reflected acoustic signal segment waveform feature vectors to obtain an important content-highlighted reflected acoustic signal waveform feature vector; and finally, determining an identification result of a geological body on the basis of the important content-highlighted reflected acoustic signal waveform feature vector.

Description

A sonic remote detection device and detection method suitable for horizontal boreholes Technical Field

[0001] This disclosure relates to the field of acoustic long-range detection, specifically to an acoustic long-range detection device and detection method suitable for horizontal boreholes. Background Technology

[0002] In-hole testing techniques help identify potential geological risks outside the borehole by probing geological information near the borehole wall. Currently, most tests are limited to the borehole wall or within 3 meters outside the borehole wall, such as lateral resistivity probing (2.5 meters) and tube wave probing (1-3 meters). To more accurately probe geological structures outside the borehole, acoustic long-range sounding technology has been developed. This technology images the formation around the well by analyzing the energy of sound waves emitted from a sound source in the well and the reflected energy. This technology has made progress in oil and gas reservoir prediction and fracture detection, for example, in applications in the Dagang Oilfield and the Tarim Basin.

[0003] However, in the field of engineering exploration, this technology has not yet been effectively applied due to factors such as the small size of engineering exploration boreholes and the complexity of testing horizontal and vertical boreholes. More importantly, the wave field reflection characteristics of geological bodies related to engineering geology, such as goafs, karst, and faults, differ from those of oil drilling, requiring targeted research. In other words, in terms of data processing, a suitable acoustic long-range detection method for horizontal boreholes has not yet been developed.

[0004] Therefore, there is a need for a sonic remote detection device and detection method suitable for horizontal boreholes. Summary of the Invention

[0005] This disclosure was made in consideration of the above problems. One object of this disclosure is to provide an acoustic remote detection device and method suitable for horizontal boreholes.

[0006] Embodiments of this disclosure provide a method for acoustic long-range detection suitable for horizontal boreholes, comprising:

[0007] Acquire the reflected sound wave signal;

[0008] The reflected sound wave signal is processed using a signal preprocessing module to obtain a sequence of reflected sound wave signal segments;

[0009] Waveform feature extraction is performed on the sequence of the reflected acoustic wave signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments;

[0010] The sequence of waveform feature vectors of the reflected acoustic signal segment is processed using a position encoding and autocorrelation enhancement module to obtain waveform feature vectors that highlight important content of the reflected acoustic signal; and

[0011] Based on the aforementioned key information, the feature vector of the reflected acoustic wave signal waveform is used to determine the identification result of the geological body.

[0012] For example, according to an embodiment of the present disclosure, a method for acoustic long-range detection suitable for horizontal boreholes includes processing the reflected acoustic signal using a signal preprocessing module to obtain a sequence of reflected acoustic signal segments, including:

[0013] The reflected sound wave signal is denoised to obtain a denoised reflected sound wave signal; and

[0014] The noise-reduced reflected sound wave signal is segmented into signal segments to obtain a sequence of reflected sound wave signal segments.

[0015] For example, according to an embodiment of the present disclosure, a method for acoustic long-range detection suitable for horizontal boreholes includes extracting waveform features from a sequence of reflected acoustic signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic signal segments, comprising:

[0016] Each reflected acoustic wave signal segment in the sequence of reflected acoustic wave signal segments is processed by a waveform feature extractor based on a convolutional neural network model to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments.

[0017] For example, according to an embodiment of the present disclosure, a method for long-range acoustic detection in horizontal boreholes, wherein a position encoding and autocorrelation enhancement module is used to process a sequence of waveform feature vectors of the reflected acoustic signal segments to obtain waveform feature vectors of the reflected acoustic signal that highlight important content, including:

[0018] Position encoding and labeling are performed on the sequence of waveform feature vectors of the reflected acoustic signal segment to obtain a sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels; and

[0019] The sequence of waveform feature vectors of the reflected acoustic signal segments containing location labels is processed through an autocorrelation signal feature saliency fusion network to obtain the waveform feature vectors of the reflected acoustic signal that highlight the important content.

[0020] For example, according to an embodiment of the present disclosure, a method for long-range acoustic detection in horizontal boreholes includes position encoding and labeling of a sequence of waveform feature vectors of reflected acoustic signal segments to obtain a sequence of waveform feature vectors of reflected acoustic signal segments containing position labels, comprising:

[0021] Position encoding is performed on each waveform feature vector of the reflected acoustic signal segment in the sequence of waveform feature vectors to obtain a sequence of position-coded feature vectors; and

[0022] The sequence of waveform feature vectors of the reflected acoustic signal segment and the sequence of position-coded feature vectors are fused to obtain the sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels.

[0023] For example, according to an embodiment of the present disclosure, a method for long-range acoustic detection in horizontal boreholes includes processing a sequence of waveform feature vectors of reflected acoustic signal segments containing location markers through an autocorrelation signal feature saliency fusion network to obtain the important content-highlighting reflected acoustic signal waveform feature vectors, including:

[0024] The sequence of waveform feature vectors of the reflected acoustic wave signal segments containing location annotations is processed using the following autocorrelation signal feature saliency fusion formula to obtain the waveform feature vector of the reflected acoustic wave signal highlighting the important content; wherein, the autocorrelation signal feature saliency fusion formula is:

[0025] Among them, e i Let W be the scoring coefficient for the i-th attention. i Let i be the weight coefficient matrix. Let B be the i-th weight coefficient vector. i Let h be the i-th bias vector, and let Selu(·) denote the Selu activation function. i V1 is the i-th waveform feature vector of the reflected acoustic signal segment containing position labels in the sequence of waveform feature vectors containing position labels, where λ and α are weighting coefficients with different values, x represents the value input to the Selu activation function, softmax(·) represents the softmax activation function, t is the length of the sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels, and V1 is the waveform feature vector of the important content highlighting the reflected acoustic signal.

[0026] For example, according to embodiments of the present disclosure, a method for acoustic long-range detection applicable to horizontal boreholes, wherein determining the identification result of a geological body based on the waveform feature vector of the reflected acoustic signal highlighting the important content includes:

[0027] The important content is highlighted by the reflected acoustic signal waveform feature vector, which is then passed through a classifier-based geological body identifier to obtain the identification result, which is used to represent the type label of the geological body.

[0028] For example, the acoustic remote detection method for horizontal boreholes according to embodiments of this disclosure further includes a training step: training the waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the geological body identifier based on the classifier.

[0029] For example, in the acoustic long-range detection method applicable to horizontal boreholes according to embodiments of this disclosure, the training step includes:

[0030] Acquire training data, which includes training reflected acoustic signals and the true values ​​of geological body type labels;

[0031] The training reflected sound wave signal is subjected to signal denoising to obtain the training denoised reflected sound wave signal;

[0032] The trained, denoised, reflected sound wave signal is segmented into signal segments to obtain a sequence of trained reflected sound wave signal segments;

[0033] Each training reflected sound wave signal segment in the sequence of training reflected sound wave signal segments is passed through the waveform feature extractor based on the convolutional neural network model to obtain a sequence of waveform feature vectors for the training reflected sound wave signal segments.

[0034] Position encoding is performed on each waveform feature vector of the training reflected acoustic wave signal segment in the sequence of waveform feature vectors to obtain a sequence of training position encoded feature vectors;

[0035] The sequence of waveform feature vectors of the trained reflected acoustic signal segment and the sequence of waveform feature vectors of the trained position encoding are fused to obtain a sequence of waveform feature vectors of the trained reflected acoustic signal segment containing position labels;

[0036] The sequence of waveform feature vectors of the training reflected acoustic signal segments containing location annotations is passed through the autocorrelation signal feature saliency fusion network to obtain waveform feature vectors of the training important content highlighting reflected acoustic signal segments.

[0037] The training content, highlighting the waveform feature vector of the reflected acoustic signal, is passed through the classifier-based geological body identifier to obtain the classification loss function value; and

[0038] The waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the geological body identifier based on the classifier are trained using the classification loss function value. In each iteration of the training, the waveform feature vector of the reflected sound wave signal, which is an important part of the training, is corrected.

[0039] Embodiments of this disclosure also provide an acoustic remote detection device suitable for horizontal boreholes, comprising:

[0040] The signal acquisition module is used to acquire reflected sound wave signals;

[0041] A preprocessing module is used to process the reflected acoustic wave signal using a signal preprocessing module to obtain a sequence of reflected acoustic wave signal segments;

[0042] The waveform feature extraction module is used to extract waveform features from the sequence of the reflected acoustic wave signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments.

[0043] The explicit processing module is used to process the sequence of waveform feature vectors of the reflected acoustic wave signal segment using the position encoding and autocorrelation explicitation module to obtain important content highlighting waveform feature vectors of the reflected acoustic wave signal; and

[0044] The identification and analysis module is used to determine the identification result of the geological body based on the waveform feature vector of the reflected sound wave signal highlighted by the important content.

[0045] According to embodiments of the present disclosure, an acoustic long-range detection device and method suitable for horizontal boreholes are based on deep learning algorithms and acoustic long-range detection technology. The device performs noise reduction and geological feature extraction on reflected acoustic signals, and introduces position coding to provide richer spatial information for the acoustic signals. Then, a self-attention mechanism is used to improve and optimize the spatial distribution features and overall structural expression of geological features. In this way, a geological body identifier is used to transform this complex spatial distribution geological feature information into intuitive geological body type labels, simplifying the interpretation and analysis process of geological body identification. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings of the embodiments of this disclosure will be briefly described below. Obviously, the drawings described below only relate to some embodiments of this disclosure and are not intended to limit this disclosure.

[0047] Figure 1 shows a schematic diagram of the application architecture of the acoustic remote detection method applicable to horizontal boreholes in an embodiment of this disclosure;

[0048] Figure 2 shows a flowchart of a acoustic remote detection method applicable to horizontal boreholes in an embodiment of this disclosure;

[0049] Figure 3 shows a flowchart of sub-step S520 of the acoustic remote detection method applicable to horizontal boreholes in an embodiment of this disclosure;

[0050] Figure 4 shows a flowchart of sub-step S540 of the acoustic remote detection method applicable to horizontal boreholes in an embodiment of this disclosure;

[0051] Figure 5 shows a schematic diagram of the acoustic remote detection device suitable for horizontal drilling in an embodiment of this disclosure.

[0052] Figure 6 illustrates an application scenario of the acoustic remote detection method suitable for horizontal boreholes according to embodiments of this disclosure; and

[0053] Figure 7 shows a schematic diagram of a storage medium according to an embodiment of the present disclosure. Detailed Implementation

[0054] The technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are also within the scope of protection of this disclosure.

[0055] The terminology used in this specification is that which is currently widely used in the art in consideration of the functionality of this disclosure; however, these terms may vary depending on the intent, precedent, or new technology of those skilled in the art. Furthermore, specific terms may be chosen by the applicant, and in such cases, their detailed meanings will be described in the detailed description of this disclosure. Therefore, the terminology used in this specification should not be construed as simple names, but rather based on the meaning of the terms and the overall description of this disclosure.

[0056] While this disclosure makes various references to certain modules of systems according to embodiments of this disclosure, any number of different modules may be used and run on user terminals and / or servers. The modules described are merely illustrative, and different aspects of the systems and methods may use different modules.

[0057] This disclosure uses flowcharts to illustrate the operations performed by a system according to embodiments of this disclosure. It should be understood that the preceding or following operations are not necessarily performed in exact order. Instead, various steps can be processed in reverse order or simultaneously, as needed. Furthermore, other operations can be added to these processes, or one or more steps can be removed from them.

[0058] Figure 1 shows a schematic diagram of the application architecture of the acoustic remote detection method applicable to horizontal drilling in this embodiment of the present disclosure, including server 100 and terminal device 200.

[0059] Terminal device 200 and server 100 can be connected via the Internet to enable communication between them. Optionally, the Internet described above uses standard communication technologies and / or protocols. The Internet is typically the Internet, but can also be any network, including but not limited to any combination of Local Area Network (LAN), Metropolitan Area Network (MAN), Wide Area Network (WAN), mobile, wired or wireless networks, private networks or virtual private networks. In some embodiments, technologies and / or formats including Hyper Text Markup Language (HTML), Extensible Markup Language (XML), etc., are used to represent data exchanged over the network. Furthermore, conventional encryption technologies such as Secure Socket Layer (SSL), Transport Layer Security (TLS), Virtual Private Network (VPN), and Internet Protocol Security (IPsec) can be used to encrypt all or some links. In other embodiments, customized and / or dedicated data communication technologies can be used to replace or supplement the aforementioned data communication technologies.

[0060] Server 100 can provide various network services to terminal device 200. Server 100 can be a single server, a server cluster consisting of several servers, or a cloud computing center. Specifically, server 100 may include processor 110 (Center Processing Unit, CPU), memory 120, input device 130, and output device 140, etc. Input device 130 may include keyboard, mouse, touch screen, etc., and output device 140 may include display device, such as liquid crystal display (LCD), cathode ray tube (CRT), etc.

[0061] The memory 120 may include a read-only memory (ROM) and a random access memory (RAM), and provides the processor 110 with program instructions and data stored in the memory 120. In this embodiment of the present disclosure, the memory 120 may be used to store a program for a sonic remote detection method applicable to horizontal drilling in this embodiment of the present disclosure.

[0062] The processor 110 executes the steps of any of the acoustic remote detection methods applicable to horizontal boreholes in this disclosure according to the program instructions stored in the memory 120 by calling the program instructions.

[0063] Furthermore, the application architecture diagrams in this disclosure are for the purpose of more clearly illustrating the technical solutions in this disclosure and do not constitute a limitation on the technical solutions provided in this disclosure. Of course, the technical solutions provided in this disclosure are also applicable to similar problems for other application architectures and business applications.

[0064] The following examples or embodiments illustrate a non-limiting description of an acoustic remote detection method suitable for horizontal boreholes provided according to at least one embodiment of the present disclosure. As described below, different features in these specific examples or embodiments can be combined with each other without conflict to obtain new examples or embodiments, all of which are also within the scope of protection of the present disclosure.

[0065] To address the aforementioned technical issues, the technical concept of this application is based on deep learning algorithms and acoustic long-range detection technology. It performs noise reduction and geological feature extraction on reflected acoustic signals, introduces position coding to provide richer spatial information for acoustic signals, and then uses a self-attention mechanism to improve and optimize the spatial distribution characteristics and overall structural expression of geological features. In this way, a geological body identifier is used to transform this complex spatial distribution geological feature information into intuitive geological body type labels, simplifying the interpretation and analysis process of geological body identification.

[0066] Based on this, Figure 2 shows a flowchart of the acoustic remote detection method applicable to horizontal boreholes in an embodiment of this disclosure. For example, the acoustic remote detection method applicable to horizontal boreholes can be executed by a server, which can be server 100 shown in Figure 1. As shown in Figure 2, the acoustic remote detection method applicable to horizontal boreholes according to an embodiment of this disclosure includes the following steps: S510, acquiring reflected acoustic signals; S520, processing the reflected acoustic signals using a signal preprocessing module to obtain a sequence of reflected acoustic signal segments; S530, extracting waveform features from the sequence of reflected acoustic signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic signal segments; S540, processing the sequence of waveform feature vectors for the reflected acoustic signal segments using a position encoding and autocorrelation enhancement module to obtain waveform feature vectors for highlighting important content; and S550, determining the identification result of geological bodies based on the waveform feature vectors for highlighting important content.

[0067] Specifically, in the technical solution of this application, the first step is to acquire the reflected acoustic wave signal. It should be understood that in long-range acoustic detection technology, by radiating a sound field from a sound source in the well to the surrounding strata, the sound wave energy can penetrate the well wall, interact with the underground geological structure, and then be reflected back. That is, the reflected acoustic wave signal carries information about the underground geological structure, such as characteristics of rock strata, lithology, fractures, and pores. By acquiring and further analyzing the reflected acoustic wave signal, geological structure information can be obtained, thereby enabling the identification of geological bodies.

[0068] During the actual acquisition of reflected acoustic signals, various interference factors can occur, such as equipment noise, multiple reflections caused by changes in the geological medium, and wellbore reflections. These noises can contaminate the reflected acoustic signals, affecting the subsequent extraction and analysis of geological features. In this application's technical solution, the reflected acoustic signals are denoised to obtain denoised reflected acoustic signals. More specifically, in the embodiments of this application, frequency domain filtering can be used to denoise the reflected acoustic signals. Next, the denoised reflected acoustic signals are segmented into signal segments to obtain a sequence of reflected acoustic signal segments. Here, segmenting the denoised reflected acoustic signals, which contain a relatively long time period, into shorter reflected acoustic signal segments facilitates subsequent model analysis of each segment, thereby mitigating the situation where local detail information is ignored to some extent.

[0069] Accordingly, in step S520, as shown in FIG3, the signal preprocessing module is used to process the reflected sound wave signal to obtain a sequence of reflected sound wave signal segments, including: S521, performing signal denoising on the reflected sound wave signal to obtain a denoised reflected sound wave signal; and S522, performing signal segmentation on the denoised reflected sound wave signal to obtain a sequence of reflected sound wave signal segments.

[0070] Afterwards, each reflected sound wave signal segment in the sequence is passed through a waveform feature extractor based on a convolutional neural network (CNN) model to obtain a sequence of waveform feature vectors for the reflected sound wave signal segments. A CNN is a deep learning model primarily used to process data with a grid structure, such as images and audio. Specifically, a CNN contains multiple convolutional and pooling layers, progressively extracting abstract features from the data through multiple convolutional processes and reducing the dimensionality of the features through pooling operations. In the technical solution of this application, the waveform feature extractor constructed using a CNN model can effectively learn the waveform features of each reflected sound wave signal segment, such as frequency, amplitude, and waveform shape, thereby characterizing the features of underground geological structures.

[0071] Accordingly, in step S530, waveform feature extraction is performed on the sequence of reflected sound wave signal segments to obtain a sequence of waveform feature vectors for the reflected sound wave signal segments, including: passing each reflected sound wave signal segment in the sequence of reflected sound wave signal segments through a waveform feature extractor based on a convolutional neural network model to obtain a sequence of waveform feature vectors for the reflected sound wave signal segments.

[0072] Subsequently, positional encoding is performed on each waveform feature vector of the reflected acoustic signal segment in the sequence of waveform feature vectors to obtain a sequence of position-coded feature vectors. Here, the purpose of positional encoding is to provide the model with positional information about each waveform feature vector of the reflected acoustic signal segment in the sequence, helping the model better understand the correlation between reflected signal waveform features at different positions in the sequence. Typically, during positional encoding, functions are used to assign a unique position vector to each position in the sequence. For example, triangular positional encoding, absolute positional encoding, or relative positional encoding can be used to process the waveform feature vectors of each reflected acoustic signal segment. Further, the sequence of waveform feature vectors of the reflected acoustic signal segment and the sequence of position-coded feature vectors are fused to obtain a sequence of waveform feature vectors of the reflected acoustic signal segment containing positional annotations. In this way, the geological feature information contained in the original data can be taken into account, and the location information can be fully utilized, thereby improving the model's spatial perception of the distribution of geological features and avoiding confusion of various local geological feature information during feature extraction and visualization.

[0073] The sequence of waveform feature vectors of the reflected acoustic signal segments containing location annotations is then processed through an autocorrelation signal feature saliency fusion network to obtain waveform feature vectors of the reflected acoustic signal segments that highlight important content. That is, the autocorrelation signal feature saliency fusion network is used to obtain the full-time-domain autocorrelation of the sequence of waveform feature vectors of the reflected acoustic signal segments containing location annotations, capturing long-term dependencies and temporal dynamic features in the data. The essence of the autocorrelation signal feature saliency fusion network is a weighted probability distribution mechanism, that is, assigning greater weight to important content and reducing the weight to other content. This mechanism focuses on finding useful information in the input data that is significantly related to the current data, discovering the autocorrelation between the waveform feature vectors of various reflected acoustic signal segments containing location annotations. Here, the autocorrelation signal feature saliency fusion network can learn the dependencies and correlations between the waveform features of different positions and different segments of the reflected acoustic signal segments in the sequence data containing the waveform feature vectors of the reflected acoustic signal segments containing location annotations, and perform autocorrelation modeling on the entire sequence, thereby gaining a more comprehensive understanding of the dynamic fluctuation characteristics and spatial structure distribution in the reflected acoustic signal data.

[0074] Accordingly, in step S540, as shown in FIG4, the sequence of waveform feature vectors of the reflected sound wave signal segment is processed by the position encoding and autocorrelation enhancement module to obtain the waveform feature vector of the reflected sound wave signal highlighting important content, including: S541, performing position encoding and annotation on the sequence of waveform feature vectors of the reflected sound wave signal segment to obtain a sequence of waveform feature vectors of the reflected sound wave signal segment containing position annotations; and S542, passing the sequence of waveform feature vectors of the reflected sound wave signal segment containing position annotations through an autocorrelation signal feature saliency fusion network to obtain the waveform feature vector of the reflected sound wave signal highlighting important content.

[0075] In step S541, the sequence of waveform feature vectors of the reflected acoustic wave signal segment is position-encoded and labeled to obtain a sequence of waveform feature vectors of the reflected acoustic wave signal segment containing position labels. This includes: performing position encoding on each waveform feature vector of the reflected acoustic wave signal segment in the sequence of waveform feature vectors of the reflected acoustic wave signal segment to obtain a sequence of position-encoded feature vectors; and fusing the sequence of waveform feature vectors of the reflected acoustic wave signal segment and the sequence of position-encoded feature vectors to obtain the sequence of waveform feature vectors of the reflected acoustic wave signal segment containing position labels.

[0076] In step S542, the sequence of waveform feature vectors of the reflected acoustic signal segments containing location markers is processed through an autocorrelation signal feature saliency fusion network to obtain the waveform feature vector of the important content-highlighting reflected acoustic signal. This includes processing the sequence of waveform feature vectors of the reflected acoustic signal segments containing location markers using the following autocorrelation signal feature saliency fusion formula to obtain the waveform feature vector of the important content-highlighting reflected acoustic signal. The autocorrelation signal feature saliency fusion formula is as follows:

[0077] Among them, e i Let W be the scoring coefficient for the i-th attention. i Let i be the weight coefficient matrix. Let B be the i-th weight coefficient vector. i Let h be the i-th bias vector, and let Selu(·) denote the Selu activation function. i V1 is the i-th waveform feature vector of the reflected acoustic signal segment containing position labels in the sequence of waveform feature vectors containing position labels, where λ and α are weighting coefficients with different values, x represents the value input to the Selu activation function, softmax(·) represents the softmax activation function, t is the length of the sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels, and V1 is the waveform feature vector of the important content highlighting the reflected acoustic signal.

[0078] Subsequently, the waveform feature vector of the highlighted reflected acoustic signal is passed through a classifier-based geological body identifier to obtain the identification result, which is used to represent the type label of the geological body.

[0079] Accordingly, in step S550, determining the identification result of the geological body based on the waveform feature vector of the highlighted acoustic wave signal includes: passing the waveform feature vector of the highlighted acoustic wave signal through a classifier-based geological body identifier to obtain the identification result, which is used to represent the type label of the geological body.

[0080] Specifically, the waveform feature vector of the highlighted acoustic wave signal is passed through a classifier-based geological body identifier to obtain the identification result, which is used to represent the type label of the geological body. This includes: using the fully connected layer of the classifier-based geological body identifier to fully connect and encode the waveform feature vector of the highlighted acoustic wave signal to obtain an encoded classification feature vector; and inputting the encoded classification feature vector into the Softmax classification function of the classifier-based geological body identifier to obtain the identification result.

[0081] As you can understand, the role of a classifier is to learn classification rules and classifiers using given categories and known training data, and then classify (or predict) unknown data. Logistic regression and SVM are commonly used to solve binary classification problems. For multi-class classification problems, logistic regression or SVM can also be used, but multiple binary classifications are needed to form the multi-class classification. However, this is prone to errors and inefficient. A commonly used multi-class classification method is the Softmax classification function.

[0082] In the above-described technical solution, the sequence of waveform feature vectors of the reflected acoustic signal segment expresses the image semantic features of the reflected acoustic signal in the local time domain. Furthermore, after fusing positional coding features to obtain a sequence of waveform feature vectors of the reflected acoustic signal segment containing positional annotations, and then strengthening the sequence of waveform feature vectors of the reflected acoustic signal segment containing positional annotations through an autocorrelation signal feature saliency fusion network based on the autocorrelation global feature distribution of the combined image semantic-positional coding features in the local time domain, the expression complexity of the fusion of local time domain differences in the waveform feature vectors of the reflected acoustic signal segment in the global time domain will also exist. This causes local distribution differences in the waveform feature vectors of the reflected acoustic signal segment, thus affecting the classification iteration effect.

[0083] Based on this, in a preferred embodiment of this application, the important content highlighted reflected acoustic wave signal waveform feature vector is passed through a classifier-based geological body identifier to obtain the recognition result, including the following steps: calculating the self-mean matrix and autovariance matrix of the important content highlighted reflected acoustic wave signal waveform feature vector, wherein the value at position (i,j) of the self-mean matrix is ​​the mean of the eigenvalues ​​at positions i and j of the important content highlighted reflected acoustic wave signal waveform feature vector, and the value at position (i,j) of the autovariance matrix is ​​the variance of the eigenvalues ​​at positions i and j of the important content highlighted reflected acoustic wave signal waveform feature vector; and performing a moment shift operation between the transpose of the important content highlighted reflected acoustic wave signal waveform feature vector (which serves as a row feature vector) and the self-mean matrix. The matrix multiplication is performed to obtain a first intermediate vector, and the autovariance matrix is ​​multiplied by the feature vector of the reflected acoustic signal waveform highlighting the important content to obtain a second intermediate vector; the point sum of the transposes of the first intermediate vector and the second intermediate vector is calculated to obtain a third intermediate vector; the matrix product of the mean matrix and the autovariance matrix is ​​calculated, and the transpose of the feature vector of the reflected acoustic signal waveform highlighting the important content is multiplied by the matrix product to obtain a fourth intermediate vector; the point sum of the transposes of the third intermediate vector and the fourth intermediate vector is calculated to obtain an optimized feature vector of the reflected acoustic signal waveform highlighting the important content; the optimized feature vector of the reflected acoustic signal waveform highlighting the important content is passed through the classifier-based geological body identifier to obtain the recognition result.

[0084] Based on this, the self-mean matrix and self-variance matrix of the group aggregation statistical evaluation used for the granular local distribution of the feature vector of the reflected acoustic wave signal waveform highlighting the important content are used as the retrieval and response distribution enhancement of the feature vector of the reflected acoustic wave signal waveform highlighting the important content. A referenceless distribution retrieval response framework based on the open domain of the feature distribution of the reflected acoustic wave signal waveform highlighting the important content is constructed. By superimposing responses, the redundancy of the distribution response caused by the local overflow features of the feature vector of the reflected acoustic wave signal waveform highlighting the important content is avoided. This achieves the fidelity constraint of the response self-aggregation statistical correlation of the optimized feature vector of the reflected acoustic wave signal waveform highlighting the important content to the classification target domain, and improves the accuracy of the recognition results obtained by the geological body recognizer based on the classifier of the optimized feature vector of the reflected acoustic wave signal waveform highlighting the important content.

[0085] Furthermore, in the technical solution disclosed herein, the acoustic remote detection method applicable to horizontal boreholes further includes a training step: training the waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the geological body identifier based on the classifier.

[0086] In one example, the training step includes: acquiring training data, which includes training reflected acoustic wave signals and true values ​​of geological body type labels; performing signal denoising on the training reflected acoustic wave signals to obtain denoised reflected acoustic wave signals; segmenting the denoised reflected acoustic wave signals into signal segments to obtain a sequence of training reflected acoustic wave signal segments; passing each training reflected acoustic wave signal segment in the sequence of training reflected acoustic wave signal segments through the waveform feature extractor based on the convolutional neural network model to obtain a sequence of waveform feature vectors for training reflected acoustic wave signal segments; performing position encoding on each waveform feature vector in the sequence of waveform feature vectors for training reflected acoustic wave signal segments to obtain a sequence of position-encoded feature vectors for training; and fusing the waveform feature vectors of the training reflected acoustic wave signal segments. The sequence of shape feature vectors and the sequence of training position encoded feature vectors are used to obtain a sequence of training reflected acoustic signal segment waveform feature vectors containing position labels; the sequence of training reflected acoustic signal segment waveform feature vectors containing position labels is passed through the autocorrelation signal feature saliency fusion network to obtain training important content highlighted reflected acoustic signal waveform feature vectors; the training important content highlighted reflected acoustic signal waveform feature vectors are passed through the classifier-based geological body recognizer to obtain a classification loss function value; and the classification loss function value is used to train the waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the classifier-based geological body recognizer, wherein, in each iteration of the training, the training important content highlighted reflected acoustic signal waveform feature vectors are corrected.

[0087] In particular, considering that the feature vector of the reflected acoustic wave signal waveform highlighted by the training content has rich geometric details in the high-dimensional feature space, there is a distortion in the class representation when it is classified by the classifier to map to the class probability domain. That is, it affects the class representation of the feature vector of the reflected acoustic wave signal waveform highlighted by the training content for the classifier's predetermined class, thereby reducing the accuracy of the recognition result.

[0088] Based on this, the applicant of this application corrects the waveform feature vector of the reflected acoustic wave signal highlighting the important training content each time the waveform feature vector of the trained important content is classified and iterated by the classifier-based geological body identifier.

[0089] Accordingly, in one example, in each iteration of the training, the waveform feature vector of the highlighted reflected acoustic signal is corrected using the following correction formula to obtain the corrected waveform feature vector of the highlighted reflected acoustic signal; wherein, the correction formula is:

[0090] Where p is the class probability value obtained by the classifier through the feature vector V of the reflected sound wave signal waveform highlighted by the training content. This refers to the i-th eigenvalue of the waveform feature vector V of the reflected sound wave signal, which is a key aspect of the training process. V is the mean of all eigenvalues ​​of the waveform feature vector V of the reflected acoustic wave signal highlighted by the training content. ||V||1 represents the first norm of the waveform feature vector V of the reflected acoustic wave signal highlighted by the training content, and α is a weight hyperparameter. exp(·) represents the exponential operation of the value, which means calculating the natural exponential function value raised to the power of the value. v is a weight hyperparameter. i ′ It is the i-th feature value of the waveform feature vector of the reflected sound wave signal, which is an important part of the corrected training.

[0091] Specifically, while preserving the geometric details of the high-dimensional feature manifold under the class probability mapping of the reflected acoustic wave signal waveform feature vector V, the feature value itself is edited in terms of shape attribute relative to the feature set of the reflected acoustic wave signal waveform feature vector V through a class probability distortion consultation. This allows the class probability distortion mapping of the reflected acoustic wave signal waveform feature vector V to the latent class space feature representation. Then, consultation fusion is performed by supplementing the basic low-rank constrained latent inversion representation to bridge the gap between the edited consultation representation and the original geometric details. This improves the class representation of the high-dimensional features of the reflected acoustic wave signal waveform feature vector V, thereby improving the accuracy of the recognition results.

[0092] Based on the above embodiments, referring to Figure 5, a structural schematic diagram of an acoustic long-range detection device 800 suitable for horizontal boreholes according to an embodiment of this disclosure is provided. The acoustic long-range detection device 800 suitable for horizontal boreholes includes: a signal acquisition module 810 for acquiring reflected acoustic signals; a preprocessing module 820 for processing the reflected acoustic signals using the signal preprocessing module to obtain a sequence of reflected acoustic signal segments; a waveform feature extraction module 830 for extracting waveform features from the sequence of reflected acoustic signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic signal segments; a visualization processing module 840 for processing the sequence of waveform feature vectors for the reflected acoustic signal segments using a position encoding and autocorrelation visualization module to obtain waveform feature vectors for highlighting important content in the reflected acoustic signals; and an identification analysis module 850 for determining the identification result of geological bodies based on the waveform feature vectors for highlighting important content in the reflected acoustic signals.

[0093] Here, those skilled in the art will understand that the specific functions and operations of each module in the above-described acoustic remote detection device 800 for horizontal boreholes have been described in detail in the description of the acoustic remote detection method for horizontal boreholes with reference to Figures 2 to 4, and therefore, their repeated description will be omitted.

[0094] Figure 6 illustrates an application scenario of the acoustic remote detection method for horizontal boreholes according to an embodiment of the present disclosure. As shown in Figure 6, in this application scenario, firstly, a reflected acoustic signal (e.g., D as shown in Figure 6) is acquired. Then, the reflected acoustic signal is input to a server (e.g., S as shown in Figure 6) equipped with an acoustic remote detection algorithm for horizontal boreholes. The server can process the reflected acoustic signal using the acoustic remote detection algorithm for horizontal boreholes to obtain an identification result for a type label representing a geological body.

[0095] Based on the above embodiments, this disclosure also provides an electronic device with another exemplary implementation. In some possible implementations, the electronic device in this disclosure may include a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, can implement the steps of the acoustic remote detection method applicable to horizontal drilling in the above embodiments.

[0096] For example, taking the server 100 in FIG1 of this disclosure as an example, the processor in the electronic device is the processor 110 in the server 100, and the memory in the electronic device is the memory 120 in the server 100.

[0097] Embodiments of this disclosure also provide a computer-readable storage medium. FIG7 shows a schematic diagram of a computer-readable storage medium 1000 according to an embodiment of this disclosure. As shown in FIG7, the computer-readable storage medium 1000 stores computer-executable instructions 1001. When the computer-executable instructions 1001 are executed by a processor, an acoustic remote detection method suitable for horizontal drilling according to an embodiment of this disclosure, as described with reference to the above figures, can be executed. The computer-readable storage medium includes, but is not limited to, volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory, etc. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.

[0098] Embodiments of this disclosure also provide a computer program product or computer program including computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform an acoustic remote sensing method suitable for horizontal borehole drilling according to embodiments of this disclosure.

[0099] Furthermore, this disclosure also provides a small-diameter (<75mm) horizontal borehole acoustic radial long-range detection device suitable for engineering exploration. It employs a cableless storage hardware architecture, making it applicable to long-distance (>2km) horizontal exploration boreholes. The small-diameter horizontal borehole acoustic radial long-range detection device mainly consists of a transmitting part (source) and a receiving part. The transmitting part uses magnetostrictive technology to excite sound waves, and the receiving part uses a 12-array receiver. The main unit is integrated into the probe, enabling automatic data acquisition and storage without the need for cables. The probe is connected to the drill rod of a horizontal drilling rig, and the horizontal borehole is advanced for automatic data acquisition and storage.

[0100] Those skilled in the art will understand that the contents disclosed herein can be varied and modified in many ways. For example, the various devices or components described above can be implemented in hardware, or in software, firmware, or a combination of some or all of the three.

[0101] Furthermore, while this disclosure makes various references to certain units in systems according to embodiments of this disclosure, any number of different units may be used and run on clients and / or servers. The units described are merely illustrative, and different aspects of the systems and methods may use different units.

[0102] Those skilled in the art will understand that all or part of the steps in the above methods can be implemented by a program instructing related hardware, and the program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk. Optionally, all or part of the steps in the above embodiments can also be implemented using one or more integrated circuits. Accordingly, each module / unit in the above embodiments can be implemented in hardware or as a software functional module. This disclosure is not limited to any particular combination of hardware and software.

[0103] Unless otherwise defined, all terms used herein (including technical and scientific terms) shall have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure pertains. It should also be understood that terms such as those defined in a common dictionary shall be interpreted as having a meaning consistent with their meaning in the context of the relevant art, and not as having an idealized or highly formalized meaning, unless expressly defined herein.

[0104] The foregoing description is intended to illustrate the present disclosure and should not be construed as limiting it. While several exemplary embodiments of the present disclosure have been described, those skilled in the art will readily understand that many modifications may be made to the exemplary embodiments without departing from the novel teachings and advantages of the present disclosure. Therefore, all such modifications are intended to be included within the scope of the present disclosure as defined by the claims. It should be understood that the foregoing description is intended to illustrate the present disclosure and should not be construed as limiting it to the specific embodiments disclosed, and modifications to the disclosed embodiments and other embodiments are intended to be included within the scope of the appended claims. The present disclosure is defined by the claims and their equivalents.

Claims

1. A method for long-range acoustic detection suitable for horizontal boreholes, characterized in that, include: Acquire the reflected sound wave signal; The reflected sound wave signal is processed using a signal preprocessing module to obtain a sequence of reflected sound wave signal segments; Waveform feature extraction is performed on the sequence of the reflected acoustic wave signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments; The sequence of waveform feature vectors of the reflected acoustic wave signal segment is processed using a position encoding and autocorrelation enhancement module to obtain waveform feature vectors of the reflected acoustic wave signal that highlight important content. as well as Based on the aforementioned key information, the feature vector of the reflected acoustic wave signal waveform is used to determine the identification result of the geological body.

2. The acoustic long-range detection method for horizontal boreholes according to claim 1, characterized in that, The reflected acoustic wave signal is processed using a signal preprocessing module to obtain a sequence of reflected acoustic wave signal segments, including: The reflected sound wave signal is denoised to obtain a denoised reflected sound wave signal; and The noise-reduced reflected sound wave signal is segmented into signal segments to obtain a sequence of reflected sound wave signal segments.

3. The acoustic long-range detection method for horizontal boreholes according to claim 2, characterized in that, The sequence of reflected acoustic wave signal segments is subjected to waveform feature extraction to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments, including: Each reflected acoustic wave signal segment in the sequence of reflected acoustic wave signal segments is processed by a waveform feature extractor based on a convolutional neural network model to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments.

4. The acoustic long-range detection method for horizontal boreholes according to claim 3, characterized in that, The sequence of waveform feature vectors of the reflected acoustic wave signal segment is processed using a position encoding and autocorrelation enhancement module to obtain important content highlighting of the reflected acoustic wave signal waveform feature vectors, including: Position encoding and labeling are performed on the sequence of waveform feature vectors of the reflected acoustic signal segment to obtain a sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels; and The sequence of waveform feature vectors of the reflected acoustic signal segments containing location labels is processed through an autocorrelation signal feature saliency fusion network to obtain the waveform feature vectors of the reflected acoustic signal that highlight the important content.

5. The acoustic long-range detection method for horizontal boreholes according to claim 4, characterized in that, Position encoding and annotation are performed on the sequence of waveform feature vectors of the reflected acoustic signal segment to obtain a sequence of waveform feature vectors of the reflected acoustic signal segment containing position annotations, including: Position encoding is performed on each waveform feature vector of the reflected acoustic signal segment in the sequence of waveform feature vectors to obtain a sequence of position-coded feature vectors; and The sequence of waveform feature vectors of the reflected acoustic signal segment and the sequence of position-coded feature vectors are fused to obtain the sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels.

6. The acoustic long-range detection method for horizontal boreholes according to claim 5, characterized in that, The sequence of waveform feature vectors of the reflected acoustic signal segments containing location annotations is processed through an autocorrelation signal feature saliency fusion network to obtain the waveform feature vectors of the reflected acoustic signal segments that highlight the important content, including: The sequence of waveform feature vectors of the reflected acoustic wave signal segments containing location annotations is processed using the following autocorrelation signal feature saliency fusion formula to obtain the waveform feature vector of the reflected acoustic wave signal highlighting the important content; wherein, the autocorrelation signal feature saliency fusion formula is: Among them, e i Let W be the scoring coefficient for the i-th attention. i Let i be the weight coefficient matrix. Let B be the i-th weight coefficient vector. i Let h be the i-th bias vector, and let Selu(·) denote the Selu activation function. i V1 is the i-th waveform feature vector of the reflected acoustic signal segment containing position labels in the sequence of waveform feature vectors containing position labels, where λ and α are weighting coefficients with different values, x represents the value input to the Selu activation function, softmax(·) represents the softmax activation function, t is the length of the sequence of waveform feature vectors of the reflected acoustic signal segment containing position labels, and V1 is the waveform feature vector of the important content highlighting the reflected acoustic signal.

7. The acoustic long-range detection method for horizontal boreholes according to claim 6, characterized in that, Based on the aforementioned key information, the feature vector of the reflected acoustic wave signal waveform is highlighted to determine the identification result of the geological body, including: The important content is highlighted by the reflected acoustic signal waveform feature vector, which is then passed through a classifier-based geological body identifier to obtain the identification result, which is used to represent the type label of the geological body.

8. The acoustic long-range detection method for horizontal boreholes according to claim 7, characterized in that, It also includes a training step: training the waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the geological body identifier based on the classifier.

9. The acoustic long-range detection method for horizontal boreholes according to claim 8, characterized in that, The training steps include: Acquire training data, which includes training reflected acoustic signals and the true values ​​of geological body type labels; The training reflected sound wave signal is subjected to signal denoising to obtain the training denoised reflected sound wave signal; The trained, denoised, reflected sound wave signal is segmented into signal segments to obtain a sequence of trained reflected sound wave signal segments; Each training reflected sound wave signal segment in the sequence of training reflected sound wave signal segments is passed through the waveform feature extractor based on the convolutional neural network model to obtain a sequence of waveform feature vectors for the training reflected sound wave signal segments. Position encoding is performed on each waveform feature vector of the training reflected acoustic wave signal segment in the sequence of waveform feature vectors to obtain a sequence of training position encoded feature vectors; The sequence of waveform feature vectors of the trained reflected acoustic signal segment and the sequence of waveform feature vectors of the trained position encoding are fused to obtain a sequence of waveform feature vectors of the trained reflected acoustic signal segment containing position labels; The sequence of waveform feature vectors of the training reflected acoustic signal segments containing location annotations is passed through the autocorrelation signal feature saliency fusion network to obtain waveform feature vectors of the training important content highlighting reflected acoustic signal segments. The training content, highlighting the waveform feature vector of the reflected acoustic signal, is passed through the classifier-based geological body identifier to obtain the classification loss function value; and The waveform feature extractor based on the convolutional neural network model, the autocorrelation signal feature saliency fusion network, and the geological body identifier based on the classifier are trained using the classification loss function value. In each iteration of the training, the waveform feature vector of the reflected sound wave signal, which is an important part of the training, is corrected.

10. A long-range acoustic detection device suitable for horizontal boreholes, characterized in that, include: The signal acquisition module is used to acquire reflected sound wave signals; A preprocessing module is used to process the reflected acoustic wave signal using a signal preprocessing module to obtain a sequence of reflected acoustic wave signal segments; The waveform feature extraction module is used to extract waveform features from the sequence of the reflected acoustic wave signal segments to obtain a sequence of waveform feature vectors for the reflected acoustic wave signal segments. The explicit processing module is used to process the sequence of waveform feature vectors of the reflected acoustic wave signal segment using the position encoding and autocorrelation explicit processing module to obtain important content highlighting waveform feature vectors of the reflected acoustic wave signal. as well as The identification and analysis module is used to determine the identification result of the geological body based on the waveform feature vector of the reflected sound wave signal highlighted by the important content.

Images