While-drilling surrounding rock classification and identification method and device, medium and equipment

Abstract

The invention discloses a while-drilling surrounding rock classification and identification method and device, a medium and equipment, and relates to the technical field of geological prospecting, the method comprises the following steps: collecting surrounding rock data of different while-drilling surrounding rock categories, and establishing an initial model introducing a dynamic preprocessor and a contrast learning engine; training the initial model by using the surrounding rock data; until the initial model meets a preset training ending condition, determining the initial model meeting the training ending condition as a while-drilling surrounding rock classification and identification model; inputting the obtained to-be-classified while-drilling surrounding rock data into a while-drilling surrounding rock classification and recognition model for processing to obtain classification information corresponding to the to-be-classified while-drilling surrounding rock data; efficient and accurate analysis of surrounding rock data is realized, and the speed and accuracy of geological exploration are remarkably improved; according to the method, the dependence on artificial experience is reduced, the cost and the workload are reduced, the generalization ability and the adaptability of the model are enhanced, and complex and diversified geological conditions can be processed.

Description

Technical Field

[0001] The present invention relates to the field of geological prospecting technology, and more specifically, to a method, device, medium and equipment for classifying and identifying surrounding rocks while drilling. Background Art

[0002] In the traditional process of while-drilling surrounding rock classification, it is usually dependent on the experience of geologists and intuitive analysis of core and logging data; however, this method is inefficient and subject to human subjective judgment. With the development of computer technology and machine learning, it is possible to apply automated methods for while-drilling surrounding rock classification and identification, and it is expected to improve accuracy, consistency and efficiency. Summary of the invention

[0003] In view of the deficiencies in the prior art, the object of the present invention is to provide a method, device, medium and equipment for classifying and identifying surrounding rocks while drilling, so as to achieve accurate classification of surrounding rocks while drilling.

[0004] The above technical objectives of the present invention are achieved through the following technical solutions: In a first aspect, the present application provides a method for classification and identification of surrounding rocks while drilling, comprising the following specific steps: Collect surrounding rock data of different types of surrounding rock while drilling, and establish an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data; The initial model is trained using surrounding rock data; the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn the feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels; Until the initial model meets the preset training end condition, the initial model that meets the training end condition is determined as the while-drilling surrounding rock classification and identification model; The acquired drilling surrounding rock data to be classified are input into the drilling surrounding rock classification and identification model for processing to obtain classification information corresponding to the drilling surrounding rock data to be classified.

[0005] Based on the above technical solution, the present invention can also be improved as follows.

[0006] Furthermore, the above initial model includes an input and preprocessing module, a self-supervised learning module, a meta-learning module, a feature fusion and classification module and an output module, the preprocessing module includes a dynamic preprocessor, and the self-supervised learning module includes a contrastive learning engine.

[0007] Furthermore, the above-mentioned input and preprocessing module also includes a multimodal sensor, wherein: The multimodal sensor is used to receive and integrate surrounding rock data or surrounding rock data to be classified while drilling. The output of the multimodal sensor is as follows: ; , ; In the formula, represents the output of a multimodal perceptron; represents a nonlinear activation function; represents the weight matrix; is the bias vector; It is a multimodal feature representation of data, representing a single modality feature The result of weighted summation; Representing modal characteristics The importance weight of Indicates the total number of modes; Indicates The original input data of the modality; For the The encoder function corresponding to the modality; The dynamic preprocessor is used to select the optimal processing strategy according to the data characteristics and preprocess the integrated data. The preprocessing includes feature extraction and noise reduction. The output of the dynamic preprocessor is as follows: ; In the formula, Represents the output of the dynamic preprocessor; represents the original data matrix; represents a multimodal perceptron; is an adaptive enhancement parameter; represents an adaptive unit; represents the feature weight vector; represents feature selection and recombination; represents adversarial deep autoencoder; represents the nonlinear activation function in the autoencoder, Indicates normalization processing.

[0008] Furthermore, the above comparative learning engine learns the feature representations to distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels, which is achieved by the following steps: The input data is randomly and reasonably transformed to form positive sample pairs and negative sample pairs, wherein the positive sample pair includes a first enhanced sample and a second enhanced sample obtained by different transformation methods, and the negative sample pair includes a third sample and the first enhanced sample or the second enhanced sample, and the original data in the positive sample pair and the original data in the negative sample pair are surrounding rock data of different types of surrounding rock while drilling; The preset contrast loss function is used to calculate the contrast loss between positive sample pairs and negative sample pairs, so as to learn the feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels.

[0009] Furthermore, the above contrast loss function is specifically: ,in: ; , ; In the formula, represents the contrast loss, represents a hyperparameter; represents the multi-view contrast loss, represents the local structure preserving term; represents the temperature parameter, represents the cosine similarity function, Indicates the number of data sources or perspectives, Represents a single view or data source, including well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, geochemical data; represents a positive sample pair, represents a negative sample pair, Indicates the batch size; express, represents the degree matrix, Represents the feature representation matrix of all samples after passing through the encoder, Represents matrix transpose.

[0010] Furthermore, the feature fusion and classification module is used to effectively fuse the features of different data sources and compress them into a compact representation, specifically: ,in: ; In the formula, represents the fused representation, represents the activation function, represents a gating unit; represents the gating parameter; Indicates The weight coefficient of the mode; Indicates The characteristic transfer function of the mode; represents the bias term; Represents the feature conversion function Parameter set; represents the scoring function; Indicates The feature representation of the modality, Indicates the number of different modes.

[0011] Furthermore, the output module is used to output classification information, and the classification information is specifically: , ;in: , ; In the formula, Represents classification information, which is the category label output by the surrounding rock classification and identification model while drilling; represents the number of category labels, Represents the labels of each category obtained after multi-category confidence weighting and context-dependent classification steps The enhancement score, represents the classification confidence score, Represents a collection of adjacent categories The sum of all class probabilities in , For a neighboring category The probability of Indicates A set of adjacent categories related to a class, Represents the adjustment parameter used to balance the direct probability and related categories of impact; Represents another adjustment parameter to balance the original score and contextual information The weight of .

[0012] In a second aspect, the present application provides a device for classifying and identifying surrounding rocks while drilling, which is applied to a method for classifying and identifying surrounding rocks while drilling in any one of the first aspects, comprising: Model building module, which is used to collect surrounding rock data of different types of surrounding rock while drilling and build an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data; The model training module is used to train the initial model using the surrounding rock data; the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn the feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels; A training end module is used to determine the initial model that meets the preset training end condition as the surrounding rock classification and identification model while drilling until the initial model meets the preset training end condition; The classification and identification module is used to input the acquired drilling surrounding rock data to be classified into the drilling surrounding rock classification and identification model for processing, so as to obtain the classification information corresponding to the drilling surrounding rock data to be classified.

[0013] In a third aspect, the present application provides a non-transitory computer-readable storage medium, characterized in that the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable a computer to execute any one of the methods in the first aspect.

[0014] In a fourth aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements any one of the methods in the first aspect when executing the computer program.

[0015] Compared with the prior art, the present invention has at least the following beneficial effects: In the present application, firstly, an initial model is constructed which introduces a dynamic preprocessor and a comparative learning engine, wherein the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn the feature representations that distinguish different while-drilling surrounding rock categories when the surrounding rock data has no category labels; secondly, the constructed initial model is trained with surrounding rock data of different while-drilling surrounding rock categories, so that the model can identify the corresponding categories from the outputs such as logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, geochemical data, etc.; finally, after the initial model training is completed, a while-drilling surrounding rock classification and identification model can be obtained, and the acquired while-drilling surrounding rock data to be classified can be input into the while-drilling surrounding rock classification and identification model for processing, thereby obtaining the classification information corresponding to the while-drilling surrounding rock data to be classified.

[0016] In this application, by introducing the method of surrounding rock classification and identification while drilling, efficient and accurate analysis of surrounding rock data is achieved, which significantly improves the speed and accuracy of geological exploration; this method not only reduces the dependence on manual experience, reduces costs and workload, but also enhances the generalization ability and adaptability of the model, and can handle complex and diverse geological conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings: Figure 1 A method flow chart of a classification and identification method in an embodiment of the present invention; Figure 2 Schematic diagram of a model for classification and identification of surrounding rocks while drilling in an embodiment of the present invention; Figure 3 Schematic diagram of an input and preprocessing module in an embodiment of the present invention; Figure 4 Schematic diagram of a self-supervised learning module in an embodiment of the present invention; Figure 5 Schematic diagram of a meta-learning module in an embodiment of the present invention; Figure 6 Schematic diagram of a feature fusion and classification module in an embodiment of the present invention; Figure 7 Schematic diagram of a cross-domain attention network in an embodiment of the present invention; Figure 8 is a schematic diagram of an output module in an embodiment of the present invention; Fig. 9 Schematic diagram of the connection of the classification and identification device in the embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0019] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0020] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0021] In the description of the embodiments of the present invention, "plurality" means at least 2.

[0022] Embodiment 1: In order to solve the problems that the traditional classification of surrounding rocks while drilling usually relies on the experience of geologists and intuitive analysis of core and logging data, resulting in low efficiency and subject to human subjective judgment, with the development of computer technology and machine learning, and in order to improve the accuracy, consistency and efficiency of the classification of surrounding rocks while drilling, this embodiment provides a method for classification and identification of surrounding rocks while drilling, such as Figure 1 As shown, the following specific steps are included: S1, collects surrounding rock data of different types of surrounding rock while drilling, and establishes an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data.

[0023] Among them, the above-mentioned logging data include resistivity logging data, density logging data and natural gamma-ray logging data; core sample analysis data include porosity and mineral composition; seismic reflection data include reflection amplitude and phase data; surrounding rock thickness and depth data include top and bottom plate burial depth data of each layer and layer thickness change data; geochemical data include stable isotope ratios (such as isotope ratios of elements such as carbon, oxygen, and sulfur) and trace element concentrations.

[0024] For the initial model that introduces the dynamic preprocessor and the contrastive learning engine, the multimodal perceptron can be used to receive and integrate the surrounding rock data; the dynamic preprocessor can be used to automatically select the best preprocessing method according to the data characteristics, and use the deep autoencoder to perform feature extraction and noise reduction on the integrated surrounding rock data.

[0025] Optionally, the above initial model is as follows Figure 2 As shown, it includes an input and preprocessing module, a self-supervised learning module, a meta-learning module, a feature fusion and classification module and an output module. The preprocessing module includes a dynamic preprocessor, and the self-supervised learning module includes a contrastive learning engine.

[0026] Among them, the above-mentioned input and preprocessing module also includes a multimodal sensor, such as Figure 3 The figure shows a schematic diagram of the input and preprocessing module, which is further described below.

[0027] The multimodal sensor is used to receive and integrate surrounding rock data or surrounding rock data to be classified while drilling. The output of the multimodal sensor is specifically: ; , ; In the formula, represents the output of a multimodal perceptron; Represents a nonlinear activation function, such as ReLU, Sigmoid; Represents the weight matrix, used to connect multimodal feature representations and the output of the multimodal perceptron y ; is the bias vector; It is a multimodal feature representation of data, representing a single modality feature The result of weighted summation; Representing modal characteristics The importance weight of Indicates the total number of modes; Indicates The original input data of the modality; For the The encoder function corresponding to the modality.

[0028] Specifically, the perception process of the multimodal sensor for surrounding rock data is described as follows: For each type of surrounding rock data (e.g., well logging data, core sample analysis data, seismic reflection data, etc.), first, the corresponding encoder multimodal perceptron function The original input data Convert to a single modality feature representation , which ensures that each type of data is converted into a uniform, processable form. Perform weighted summation to obtain multimodal feature representation , where the importance of each mode is expressed by the weight Finally, through the nonlinear activation function Multimodal feature representation Transform and combine the weight matrix and the bias vector b Generate final output y The present application integrates different types of surrounding rock data using a multimodal sensor, thereby improving the accuracy and robustness of the while-drilling surrounding rock classification and identification model for while-drilling surrounding rock classification and identification.

[0029] Furthermore, the above-mentioned dynamic preprocessor is used to select the optimal processing strategy according to the data characteristics and preprocess the integrated data. The preprocessing includes feature extraction processing and noise reduction processing. The output of the dynamic preprocessor is specifically: ; In the formula, Represents the output of the dynamic preprocessor; Represents the original data matrix, where each row represents a sample point and each column represents an observation feature (such as well logging data, core images, seismic reflection data, etc.); represents a multimodal perceptron; For adaptive enhancement parameters, intelligently select data enhancement strategies suitable for the current sample set (such as rotation, scaling, brightness adjustment, etc.) according to the statistical characteristics of the input data; Represents an adaptive unit, based on Perform a series of random but reasonable transformations on the input data to increase data diversity and help the model generalize better; Represents the feature weight vector, which is obtained by evaluating the importance of each feature through explanatory analysis methods (such as Shapley value or LIME); Represents feature selection and recombination, according to Filter out the most valuable feature subsets and reorganize these features to discover potential feature interaction patterns; Represents adversarial deep autoencoders. Through adversarial training (introducing the idea of ​​generative adversarial networks (GANs)), the autoencoders can learn more robust feature representations in adversarial environments to prevent overfitting. Represents the nonlinear activation function in the autoencoder, such as ReLU; It represents the standardization process, which standardizes the features processed by the above steps to ensure that all features have the same scale range.

[0030] Specifically, the process of dynamic preprocessing of the integrated surrounding rock data by the dynamic preprocessor can be divided into the following steps: First, the multimodal perceptron intelligently selects a data augmentation strategy suitable for the current sample set based on the statistical characteristics of the input data, such as rotation, scaling, brightness adjustment, etc., in order to improve the quality and diversity of the data. Next, the adaptive unit selects the data augmentation strategy based on the adaptive enhancement parameters. A series of random but reasonable transformations are performed on the input data. Then, the feature weight vector , using explanatory analysis methods to evaluate the importance of each feature. Based on the evaluation results, the most valuable feature subsets are screened out, and these features are reorganized to discover potential feature interaction patterns. This step aims to extract the feature combinations that contribute most to model predictions. Subsequently, through adversarial training (introducing the idea of ​​generative adversarial networks GAN), the autoencoder learns more robust feature representations in an adversarial environment. This training method can prevent overfitting and ensure that the model can still maintain good performance when facing complex and changing data. Finally, the features processed by the above steps are standardized to ensure that all features have the same scale range, so as to avoid some features dominating the model's learning process due to excessive numerical ranges, thereby affecting the fairness and accuracy of the model.

[0031] In the above, the dynamic preprocessor integrates multimodal surrounding rock data, intelligently selects the optimal data enhancement strategy, dynamically adjusts feature selection and reorganization, and uses adversarial deep autoencoders for robust feature extraction and noise reduction, and finally outputs a standardized and high-quality feature matrix. This process can not only automatically optimize the preprocessing steps to significantly improve the quality of surrounding rock data and the generalization ability of the model, but also enhance the understanding and recognition accuracy of complex geological features, laying a solid foundation for subsequent while-drilling surrounding rock classification and identification tasks.

[0032] Optionally, the comparative learning engine learns feature representations to distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels, which is achieved by the following steps: S11, randomly and reasonably transform the input data to form positive sample pairs and negative sample pairs, the positive sample pair includes a first enhanced sample and a second enhanced sample obtained by different transformation methods, the negative sample pair includes a third sample, and the first enhanced sample or the second enhanced sample, and the original data in the positive sample pair and the original data in the negative sample pair are surrounding rock data of different while-drilling surrounding rock categories.

[0033] S12, using a preset contrast loss function to calculate the contrast loss between the positive sample pair and the negative sample pair, so as to learn the feature representation for distinguishing different types of surrounding rock while drilling when the surrounding rock data has no category labels.

[0034] The above steps S11-S12 are implemented by the contrastive learning engine in the self-supervised learning module, such as Figure 4 As shown, the self-supervised learning module may include a contrastive learning engine and a predictive modeling unit, wherein the contrastive learning engine is used to construct positive sample pairs (similar surrounding rock samples while drilling) and negative sample pairs (different surrounding rock samples while drilling), so that the surrounding rock classification and identification model while drilling can learn useful representations from the surrounding rock data without labels. The predictive modeling unit is used to train the model to predict missing data parts or future time series values, so as to enhance the model's understanding of the internal structure of the surrounding rock data. For example, in seismic reflection data, the predictive modeling unit can be used to predict the changing trend of underground structures.

[0035] Furthermore, the contrastive learning engine includes an encoder, a projection head, a data enhancement unit, a positive sample pair and a negative sample pair construction unit, and a contrastive loss calculation unit connected in sequence, such as Figure 4 As shown, the encoder uses a convolutional neural network (CNN) or a fully connected network (FCN) to convert the dynamically preprocessed surrounding rock data into a higher-level feature representation, aiming to capture the essential attributes of the input data and remove noise, and retain information that is critical to distinguishing different samples. The projection head consists of two fully connected layers, and an activation function (such as ReLU) is set between the two fully connected layers to introduce nonlinearity. The projection head is used to perform a nonlinear transformation on the higher-level feature representation output by the encoder and map it to another feature space. This step helps to enhance the learning ability of the model, making similar data points closer in the new space, while dissimilar data points are farther away.

[0036] In order to construct positive pairs, the data enhancement unit needs to perform a series of random but reasonable transformations on the input data. For example, for core images, rotation, cropping, color jittering, etc. can be used to generate different views. For another example, noise addition and smoothing can be used for logging curves. By adopting a series of enhancement strategies, the data enhancement unit can ensure that even different observations of the same data point can generate rich learning signals, thereby helping the model to learn more generalizable representations.

[0037] Specifically, the positive sample pair and negative sample pair construction unit is used to create positive sample pairs (i.e., different views from the same data point) for different views generated by the data enhancement unit. For example, given an original sample x, two enhanced samples x1 and x2 are generated by two different enhancement methods (such as rotation and cropping), then (x1, x2) constitutes a positive sample pair; and samples are selected from different categories as negative sample pairs (i.e., views from different data points). For example, given an original sample x and its enhanced sample x1, as well as original samples y from different categories, then (x1, y) constitutes a negative sample pair. Positive sample pairs are used to guide the model to learn to make similar samples close to each other in the feature space, while negative sample pairs encourage the model to push dissimilar samples away in the feature space.

[0038] Optionally, the above contrast loss function is specifically: ,in: ; , ; In the formula, represents the contrast loss, represents a hyperparameter used to balance the importance of multi-view contrast loss and local structure preservation term; represents the multi-view contrast loss, represents the local structure preserving term; represents the temperature parameter, represents the cosine similarity function, which indicates the number of data sources or perspectives (e.g., well logging data, core sample analysis data), Represents a single view or data source, including well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, geochemical data; represents a positive sample pair, represents a negative sample pair, Indicates the batch size; express, represents the degree matrix, Represents the feature representation matrix of all samples after passing through the encoder, Represents matrix transpose.

[0039] In summary, by leveraging the above components working together, the contrastive learning engine is able to learn powerful feature representations from unlabeled rock mass data in a self-supervised manner; in this way, the contrastive learning engine not only improves the quality of feature representation, but also enhances the performance of subsequent tasks (such as classification, regression, etc.).

[0040] like Figure 4 As shown in the figure, the predictive modeling unit includes an input adaptation layer, a feature enhancement and conversion layer, a prediction core module, a sharing layer, and an output layer connected in sequence. Among them, the input adaptation layer is responsible for receiving the feature representation generated by the contrast learning engine, and can be customized according to the specific geological data to adapt to different types of data formats. For example, well logging curves are usually one-dimensional time series or depth series data. For this type of data, one-dimensional convolution operations or recurrent neural networks (RNNs) can be used in the input adaptation layer to further extract features and adjust the feature dimensions to a size suitable for subsequent models. As two-dimensional spatial data, core images need to be processed using two-dimensional convolution operations. The input adaptation layer contains appropriate pooling layers and smoothing layers to ensure that image features are correctly captured and converted into a format suitable for the prediction model. Seismic reflection data involves complex three-dimensional structural information. 3D CNNs or other advanced technologies can be used to parse these data and convert them into effective feature representations. Assuming that a set of drilling surrounding rock data sets containing well logging curves and core images are being processed, for the well logging curves, the input adaptation layer will first extract local patterns through 1D CNN, and then use the fully connected layer to compress these patterns into fixed-length feature vectors and standardize them. For core images, 2D CNN is used to extract texture features, and after processing through the pooling layer and smoothing layer, a fixed-length feature vector is also generated. Finally, the two feature vectors are merged together and their dimensions are adjusted through appropriate linear transformations to match the requirements of the subsequent prediction model. This customized design not only improves the model's ability to process different types of geological data, but also enhances its pertinence and accuracy for drilling surrounding rock classification tasks.

[0041] Specifically, the feature enhancement and conversion layer introduces a self-attention mechanism to help the model focus on the features with the most predictive value and capture long-distance dependencies. Especially when dealing with complex geological phenomena, this mechanism can help the model better understand the long-term trends of the data.

[0042] Furthermore, the prediction core module adopts the Transformer architecture of the encoder-decoder structure combined with the self-attention mechanism. This architecture can process long time series or multidimensional data sets more efficiently while maintaining high performance. The encoder is responsible for compressing the input feature representation, removing noise, and obtaining a more concentrated implicit representation; the decoder reconstructs the output based on the implicit representation generated by the encoder, predicts future values ​​or fills in missing data.

[0043] Furthermore, common features between multiple prediction tasks can be learned through shared layers to promote knowledge transfer and generalization capabilities. In the classification of surrounding rocks while drilling, this allows the model to utilize different but related geological information for comprehensive analysis, thereby improving classification accuracy.

[0044] Furthermore, the output layer can output the corresponding prediction results according to different prediction tasks, for example, for numerical prediction, it outputs continuous values; for classification tasks, it outputs discrete category labels. In addition, uncertainty measures such as confidence intervals or probability distributions can also be output to better understand the quality of the prediction.

[0045] In summary, the self-supervised learning module introduces a contrastive learning engine, which enables the model to learn the feature representations that distinguish different types of drilling rock mass even without explicit labels, thereby helping to improve the model's understanding of complex and diverse drilling rock mass. When encountering new or rare types of drilling rock mass, the powerful feature representations learned by self-supervised learning can help the model adapt to new situations more quickly and make effective adjustments using limited new samples to achieve accurate classification. In addition, due to the use of self-supervision, the model can rely heavily on large-scale unlabeled data for pre-training, and then only a small amount of labeled data is needed for fine-tuning to achieve good performance, which can greatly reduce the cost and time of data annotation.

[0046] Optionally, the schematic diagram of the meta-learning module in the above initial model is as follows Figure 5As shown in Figure 1, the meta-learning module includes a fast adaptation mechanism, a task generator, and a memory bank. The fast adaptation mechanism quickly adjusts the model parameters by using a small amount of sample data, so that the model can accurately predict new types of surrounding rock while drilling. Specifically, in the fast adaptation process, known geological principles are used as prior knowledge to guide the update of model parameters. For example, different types of surrounding rocks have significant differences in physical properties (such as resistivity and density), and this information can be added as constraints to model training. According to the characteristics of the surrounding rock while drilling, the most critical features for distinguishing new types of surrounding rocks are selected and given higher weights. This can be achieved by analyzing the differential features between different types of surrounding rocks in historical data. A basic model is trained using a large amount of surrounding rock data accumulated previously, and when a new type of surrounding rock is encountered, it is fine-tuned based on this basic model. This method can quickly improve the model performance with a small number of samples. In addition, the learning rate can be dynamically adjusted according to the sample size. For a small sample data set of a new type of surrounding rock, a larger initial learning rate is used to speed up the convergence speed, and then the learning rate is gradually reduced to improve the accuracy. For example, when encountering a type of drilling rock that has never been seen before, the model can use a small number of samples of the drilling rock to quickly adjust and optimize its own parameters, thereby improving the classification accuracy of this type of drilling rock.

[0047] Among them, the task generator is used to randomly create a series of small-scale tasks, each task corresponding to the while-drilling surrounding rock classification problem under different types or conditions, so as to help the model adapt to new types of while-drilling surrounding rock more quickly.

[0048] Specifically, the workflow of the task generator includes the following aspects: Dataset partitioning: Divide the original dataset into multiple subsets, each representing a possible task. For example, in the problem of while-drilling surrounding rock classification, this may mean partitioning the data according to different while-drilling surrounding rock types, geological characteristics, or other relevant attributes.

[0049] Sample selection: Randomly extract a certain number of samples from the above subsets as the support set and query set. The support set is used to simulate the training data in the new task, while the query set is used to evaluate the performance of the model on this new task.

[0050] Task definition: Define a specific task for each selected support set-query set pair. For example, in a classification problem, the task might be to identify a specific type of surrounding rock while drilling, while in a regression problem, the task might be to predict certain geochemical parameters.

[0051] Introduction of task diversity: In order to ensure that the model can handle a variety of situations, the task generator will deliberately introduce some variation factors, such as adding noise, changing sample distribution, etc., to increase the challenge and diversity of the task.

[0052] Iterative update: As the model's learning progresses, the task generator can dynamically adjust the difficulty and complexity of the task based on the model's performance, such as adding more complex while-drilling surrounding rock patterns or reducing the amount of available training samples.

[0053] In summary, the memory bank is used to store various surrounding rock patterns and their solutions encountered in the past, so that relevant knowledge can be quickly retrieved and applied when facing similar new situations.

[0054] Optionally, the schematic diagram of the feature fusion and classification module of the above initial model is as follows Figure 6 As shown in the figure, the feature fusion and classification module includes a cross-domain attention network and a hierarchical decision tree, wherein the cross-domain attention network is used to make the model pay attention to the relationship between different data sources, so as to build a more comprehensive representation of the characteristics of the surrounding rock while drilling. The hierarchical decision tree is used to recursively classify the surrounding rock while drilling based on the importance of the features.

[0055] Specifically, the cross-domain attention network includes a multimodal input integrator, a cross-modal interaction layer, a geological semantic enhancement module, an attention aggregator, and a classification head, which are connected sequentially, as shown in Figure 7 As shown, the multimodal input integrator includes an encoder and a fusion layer. Different encoders can be used to extract features for different types of geological data. For example, for well logging curves, 1D convolutional neural networks (1D CNNs) are used for feature extraction because well logging data is essentially a one-dimensional time series or depth series data; for core images, long short-term memory networks (LSTMs) or 2D convolutional neural networks (2DCNNs) are used for feature extraction because core images provide information on rock structure and texture. Different types of rock features, such as cracks and particle size distribution, can be identified through 2D convolutional neural networks. The fusion layer uses bilinear pooling, and the specific methods include: Suppose there are two modes A and B, which are respectively encoded by their respective encoders to obtain feature vectors and ,in, and are the dimensions of modes A and B respectively.

[0056] Next, the feature vector and The conversion is performed through a bilinear mapping function, including: Calculate the eigenvector and The outer product of , the result of the calculation is a ; further, in order to reduce the number of parameters and computational complexity, the original feature vector is first projected into a lower-dimensional space. Let the projection matrix be and , the new feature vector is and ,in, k Represents the target dimension, and then calculates and The outer product of .

[0057] Since the matrix produced by the outer product may be very large, it needs to be vectorized and normalized to maintain numerical stability: first, the matrix G is flattened into a long vector v(G), and second, the flattened vector v(G) is normalized by the L2 norm, that is, .

[0058] Finally, a pooling operation is performed on the flattened and normalized vector, that is, a power root, such as a square root or a higher-order power root, is taken for each element in the normalized vector, and then L2 normalization is performed again.

[0059] After the above steps, a fixed-length vector can be output in the end. The vector contains the information interaction from two different modalities and can be used as the input of subsequent classifiers or sent to a deeper network layer together with other features for further processing.

[0060] In the classification of surrounding rocks while drilling in this application, bilinear pooling can help the model better understand the relationship between different geological data, such as the connection between well logging curves and core images, and how these connections affect the type of surrounding rocks while drilling. In this way, the model can not only learn useful patterns from a single modality, but also discover important cross-modal associations, thereby improving classification performance. In addition, since bilinear pooling introduces additional nonlinear transformations, it can increase the expressiveness of the model, making the model more robust and having better generalization capabilities.

[0061] Further, if Figure 7As shown in FIG. 1 , the cross-modal interaction layer includes a multimodal feature mapping layer, a feature interaction mechanism, and a dynamic context modeling component connected in sequence. The multimodal feature mapping layer uses an adaptive linear transformation technique to map data from different modalities (such as well logging curves, core images, etc.) to a common feature space through adaptive linear transformation. It can also automatically adjust the mapping parameters according to the specific characteristics of the input data. For example, when processing well logging curves, a specific weight adjustment strategy is applied considering its time series characteristics; for core images, an adjustment method suitable for image feature extraction is used (for example, a model pre-trained on a large image dataset (such as ImageNet) is used for feature extraction of core images through fine-tuning; or, considering that there may be important features of different scales in core images, a multi-scale feature extraction method can be designed, which means using receptive fields of different sizes to capture local and global information in the image, which helps to more comprehensively understand the content of the core image). The multimodal feature mapping layer can ensure that different modal features can interact effectively in the same dimension by using different processing strategies for different modal features.

[0062] Specifically, in the feature interaction mechanism, a bilinear pooling method based on the attention mechanism is introduced. This method can not only capture the complex relationship between different data sources, but also highlight the key features that are crucial to the analysis of geological phenomena. For example, when combining well logging data and core images, the model can automatically focus on key areas that may indicate changes in mineral composition or sudden changes in geological structure.

[0063] Specifically, in view of the time evolution characteristics of the surrounding rock data, the dynamic context modeling component can adopt GRU or LSTM, for example, to model long-term dependencies and local context information, so that the model can more effectively learn and predict geological phenomena that change over time.

[0064] Among them, by setting up a cross-modal interaction layer, not only can the model's ability to process different types of surrounding rock data be enhanced, but also the model's understanding depth and prediction accuracy of geological characteristics can be improved, thereby achieving more accurate surrounding rock data identification.

[0065] Further, if Figure 7As shown in the figure, the geological semantic enhancement module includes a geological term embedding layer, a context-aware module, and a knowledge distillation component. The geological term embedding layer is used to use a pre-trained language model specifically for the geological field (such as GeoBERT, which has been trained on a large number of geological documents and can capture semantic information specific to geology) to embed professional terms and concepts in the geological field into a low-dimensional vector space, so that the model can use this prior knowledge for learning and prediction. For example, for a specific mark in a well logging curve or a mineral component identified in a core image, the model can understand its meaning based on the existing geological terms and perform association analysis.

[0066] The context-aware module includes a geological background encoder, a conditional random field, a graph neural network (GNN), a self-attention mechanism, a geological rule constraint unit, and a dynamic context adapter. The geological background encoder is used to convert the geological background information of each sample (such as geological age, geographical location, etc.) into a vector representation that can be processed by the model. The conditional random field is used to model the sequential relationship between the surrounding rocks while drilling, especially when there is a clear hierarchical structure. For example, when the data presents a linear sequence (such as a well logging curve), a chained CRF can be used to capture the dependency between adjacent surrounding rocks while drilling. The graph neural network (GNN) is suitable for data that can be organized into a graph structure, such as the connection between geological bodies or the surrounding rock network while drilling. The self-attention mechanism enables the model to focus on the most critical part among a lot of information, so as to better understand and process geological data. The geological rule constraint unit is used to introduce rules and constraints in the geological field to ensure that the prediction results of the model follow the known geological laws. A set of rules based on geological common sense can be pre-defined and integrated into the model, such as sedimentary phase change laws, mineral paragenesis combinations, etc. The dynamic context adapter enables the model to dynamically adjust its parameters according to new or rare types of surrounding rock while drilling to adapt to the changing geological conditions. Specifically, a fast adaptation mechanism (such as the MAML algorithm) can be used to enable the model to quickly update its weights or parameters on a small number of samples to cope with new types of surrounding rock while drilling.

[0067] Further, if Figure 7As shown in the figure, the knowledge distillation component includes an input layer, a domain knowledge encoding layer, and a knowledge fusion layer connected in sequence. The input layer is used to receive data from multiple sources (well logging curves, core images, seismic reflection data, etc.) and domain knowledge (expert rules, knowledge graphs, literature mining results). The domain knowledge encoding layer includes an expert system module, a knowledge graph module, and a literature mining module. The expert system module is used to convert the professional knowledge of geologists into a rule set or decision tree form and embed it into the model as additional guidance information. The knowledge graph module uses GNNs to process the knowledge graph and extract the relationship between surrounding rock types. The literature mining module uses NLP technology to automatically extract important information about surrounding rock characteristics and classification from geological literature and convert it into feature vectors or label distributions for model learning. The knowledge fusion layer includes a weighted fusion unit and a dynamic adjustment unit. The weighted fusion unit can assign different weights according to the reliability of the knowledge source and integrate this knowledge into the learning process of the main model. The dynamic adjustment unit is used to dynamically adjust the weights of each knowledge source according to the performance of the model to optimize the overall performance of the model.

[0068] Specifically, the geological semantic enhancement module processes surrounding rock data as follows: First, the input layer collects multi-source data including well logging curves, core images, seismic reflection data, and domain knowledge from expert systems, knowledge graphs, and literature mining. Secondly, the geological term embedding layer is used to convert geological professional terms into vector representations, and the geological background encoder is used to convert background information into a form that can be processed by the model. Then, in the context-aware module, CRF is used to capture the dependencies between linear sequence data, GNN is used for feature extraction of graph structure data, the self-attention mechanism highlights key features, the geological rule constraint unit ensures that the prediction is in line with geological common sense, and the dynamic context adapter ensures that the model can quickly adapt to new surrounding rock types. Finally, in the knowledge distillation component, the weighted fusion unit integrates multi-source knowledge, and the dynamic adjustment unit continuously optimizes the weight of the knowledge source according to the model performance, ultimately improving the accuracy and reliability of surrounding rock data recognition.

[0069] Further, if Figure 7As shown in Figure 1, the attention aggregator includes an input feature mapping layer, a multi-head attention mechanism, a context-aware attention mechanism, a cross-modal interactive attention mechanism, and an attention fusion and compression layer connected in sequence. Among them, the input feature mapping layer is used to receive the feature representation of the geological semantic enhancement module and perform L2 normalization on each feature representation so that the scales of the features of different modalities are consistent to avoid some modalities dominating the attention mechanism due to their large numerical range. The multi-head self-attention mechanism creates multiple independent self-attention heads, each of which is used to capture different subspace information, thereby obtaining a richer feature representation, and then combines the results of all self-attention heads by weighted summation to form the final self-attention output, which helps the model understand the complex relationships in the surrounding rock data from multiple perspectives. The context-aware attention mechanism uses a geological context encoder (such as a GRU or LSTM-based model) to capture the changes in geological conditions during the drilling process, and provides this information as additional input to the attention mechanism, so that the attention mechanism can consider the specific geological environment of the sample and adjust the attention weight according to the context information provided by the geological context encoder. For example, higher attention is given to data at a specific depth of the surrounding rock while drilling, because this depth may correspond to a key geological interface or an important geological event. The cross-modal interactive attention mechanism promotes the effective exchange of information between different modalities through the interactive layer and captures the interdependence between them. In addition, graph neural networks (GNNs) can be used to model the complex relationship between different modal data. For example, a graph structure containing different types of geological data nodes is constructed, and the connection strength between nodes is learned through GNNs. In addition, the cross-modal interactive attention mechanism can also use feature selection algorithms to select the most representative features from the fused features, reduce redundant information, and improve model efficiency.

[0070] Specifically, the attention fusion and compression layer is used to effectively fuse the features of different data sources and compress them into a compact representation. Specifically, methods such as maximum pooling and average pooling can be used to reduce the feature dimension while retaining the most important feature information. For example, for the attention result of each modality, the maximum value or average value can be selected as the final feature representation. This application introduces a function expression of the attention fusion and compression layer, which is specifically expressed as follows: ,in: ; In the formula, represents the fused representation, Represents an activation function, such as ReLU or sigmoid; represents the gating unit, which is used to evaluate the importance of each modality and weight them according to their importance; represents the gating parameter, which controls the importance score of each modality; Indicates The weight coefficient of the mode; Indicates The feature conversion function of each modality is used to appropriately transform the attention weight of each modality to extract more useful information; represents the bias term; Represents the feature conversion function Parameter set; represents the scoring function used to evaluate the importance of each mode; Indicates The feature representation of the modality, Indicates the number of different modes.

[0071] Among them, the attention fusion and compression layer can obtain more comprehensive information by effectively fusing the features of different data sources, so that the model can comprehensively utilize this information to make more accurate predictions for surrounding rock classification and identification tasks; in addition, the attention fusion and compression layer also allows the model to dynamically adjust the weights of each modality according to the characteristics of the input data, thereby improving the adaptability and robustness of the model; further, the compressed feature representation can also reduce the computational burden of the model to improve the speed and stability of model training.

[0072] Specifically, the processing process of the attention aggregator for surrounding rock data is described as follows: First, the input feature mapping layer performs L2 normalization on the received surrounding rock data features of different modes to ensure scale consistency and avoid a certain mode dominating the subsequent analysis. Secondly, the multi-view feature extraction uses a multi-head self-attention mechanism to capture the features in the surrounding rock data from multiple angles to form a more comprehensive and rich feature representation. Thirdly, the context-aware attention mechanism uses a geological background encoder (such as a GRU / LSTM-based model) to capture geological changes during drilling, dynamically adjusts the attention weight according to the geological background, and ensures attention to key geological interfaces. Next, the cross-modal interactive attention mechanism uses GNNs to model the complex relationship between different types of geological data to achieve effective information communication, and applies feature selection algorithms to remove redundant information. Finally, the attention fusion and compression layer fuses and compresses the attention weights from multiple sources into a compact representation to provide support for the accurate classification of surrounding rock data.

[0073] In summary, based on the above components, the attention aggregator can not only enhance the model's ability to process different types of surrounding rock data, but also closely integrate with geological background information to improve the accuracy and reliability of the model's recognition of surrounding rock data.

[0074] Further, if Figure 7As shown in the figure, the classification head includes an input adaptation layer, a fully connected layer, a feature selection and compression layer, and an output layer connected in sequence, wherein the input adaptation layer is used to receive the compact feature representation from the attention aggregator and perform a linear transformation on the input features to adjust their dimensions to match the requirements of the subsequent layers. The fully connected layer is used to further compress the feature space to capture higher-level abstract features. The feature selection and compression layer is used to select the most representative features from the output of the fully connected layer to reduce redundant information. The output layer is used to output the classification results of the response according to the task requirements and pass it to the hierarchical decision tree for more detailed classification or regression analysis. Specifically, for multi-category classification tasks, the classification head can use the softmax activation function to output the probability distribution of each category; for binary classification tasks, the classification head can use the sigmoid activation function to output a value from 0 to 1, indicating the probability of belonging to a certain category.

[0075] The hierarchical decision tree is represented as follows: ; In the formula, represents a hierarchical decision tree model, represents the input feature matrix, represents the target label vector, represents the feature importance evaluation function, represents the process of selecting the optimal splitting feature and threshold, Represents the growth process of the tree, including recursive splitting and pruning, represents an ensemble learning method, such as a random forest or a gradient boosted decision tree, Represents the features obtained from the cross-domain attention network The weight of It represents the reduction of Gini impurity after segmentation using the feature, which is used to measure the contribution of the feature to the classification task.

[0076] Among them, the hierarchical decision tree shown above combines the weights and Gini impurity reduction output of the Cross-Domain Attention Network (CDAN) To calculate the importance score of the feature, ensure that the model can identify the most discriminative features. This adaptive method enables the model to dynamically adjust the weights of each feature during the training process, so as to better reflect their actual impact on the classification task. In addition, by selecting the feature and its threshold that maximizes the feature importance score at each node as the splitting criterion, the hierarchical decision tree can ensure that each split can separate samples of different categories to the greatest extent, thereby improving the classification accuracy of the surrounding rock while drilling. Furthermore, the hierarchical decision tree uses a cross-domain attention network to capture the relationship between different data sources, and can construct a more comprehensive representation of the surrounding rock while drilling features, which not only enhances the expressiveness of a single decision tree, but also provides an additional source of information for the entire model, enabling the model to handle more complex data patterns.

[0077] Furthermore, the output module of the above initial model is as follows: Figure 8 As shown in the figure, the output module includes a fully connected layer, an activation function (Softmax) and a classification decision layer. The fully connected layer includes 5 neurons, each neuron represents a type of surrounding rock while drilling, the activation function is used to output the probability distribution of the surrounding rock type while drilling, and the classification decision layer is used to select the category with the highest probability as the classification result according to the probability vector output by the activation function; the classification information output by the output module is specifically as follows: , ;in: , ; In the formula, Represents classification information, which is the category label output by the surrounding rock classification and identification model while drilling; represents the number of category labels, Represents the labels of each category obtained after multi-category confidence weighting and context-dependent classification steps The enhancement score, Represents the classification confidence score, indicating that the model is The confidence level is The maximum value among is taken as the classification confidence score; Represents a collection of adjacent categories The sum of all class probabilities in , For a neighboring category The probability of Indicates A set of adjacent categories related to a class, Represents the adjustment parameter used to balance the direct probability and related categories of impact; Represents another adjustment parameter to balance the original score and contextual information The weight of .

[0078] It should be noted that in the classification decision layer shown above, by considering the probability of adjacent categories , can introduce additional contextual information, so that the classification decision does not only rely on the probability of a single category, but also comprehensively considers the influence of multiple related categories, which helps the model capture more subtle pattern changes, thereby improving the accuracy and reliability of model classification. In addition, C It provides a quantitative indicator to indicate the model's confidence in the classification results and helps decision makers quickly judge the quality of the model's predictions, so that they can better assess potential risks.

[0079] S2, using the surrounding rock data to train the initial model; wherein, the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn the feature representations that distinguish different while-drilling surrounding rock categories when the surrounding rock data has no category labels.

[0080] S3, until the initial model meets the preset training end condition, the initial model that meets the training end condition is determined as the while-drilling surrounding rock classification and identification model.

[0081] The above-mentioned surrounding rock classification and identification model while drilling can be trained through the following steps: S21: collecting surrounding rock data of different types of surrounding rock while drilling, and dividing them into a training set and a validation set in proportion, for example, the division ratio may be 7:3, wherein the types of surrounding rock while drilling include, for example, sandstone, shale and limestone; S22: setting training parameters, for example, the training batch size is 32, the number of training rounds is 100, the learning rate is 0.001, and the while-drilling surrounding rock classification and recognition model is trained using the training set until the maximum number of training times is reached; S23: Use the validation set to validate the surrounding rock classification and identification model while drilling, and use accuracy, recall rate and F1 score as evaluation indicators to evaluate the performance of the model. When all indicators reach 0.9 or above, the model is validated. Otherwise, adjust the training parameters (for example, adjust the number of training rounds to 200 and the learning rate to 0.05) or expand the training samples to retrain the model until the model is validated.

[0082] S4, inputting the acquired while-drilling surrounding rock data to be classified into a while-drilling surrounding rock classification and identification model for processing, and obtaining classification information corresponding to the while-drilling surrounding rock data to be classified.

[0083] The following is a detailed description of the solution in this application through examples, and different sample data are collected on site, including: Sample 1: Well logging data: resistivity logging: 2.7 ohm·m, density logging: 2.66 g / cm 3 , Natural gamma ray logging: 32 API; Core sample analysis data: Porosity: 18%, mineral composition: quartz 82%, feldspar 13%, mica 5%; Seismic reflection data: reflection amplitude: 0.72, phase: 115°; Surrounding rock thickness and depth data: top plate burial depth: 1250 meters, layer thickness: 32 meters; Geochemical data: Carbon isotope ratio: δ 13 C = -26‰, oxygen isotope ratio: δ 18 O = -9‰, trace element concentrations: Sr = 42 ppm, Ba = 155 ppm.

[0084] Sample 2: Well logging data: resistivity logging: 0.9 ohm·m, density logging: 2.72 g / cm 3 , Natural gamma ray logging: 88 API; Core sample analysis data: Porosity: 6%, mineral composition: clay minerals 62%, quartz 28%, calcite 10%; Seismic reflection data: reflection amplitude: 0.31, phase: 265°; Surrounding rock thickness and depth data: top plate burial depth: 1520 meters, layer thickness: 16 meters; Geochemical data: Carbon isotope ratio: δ 13 C = -27‰, oxygen isotope ratio: δ 18 O = -11‰, trace element concentrations: Sr = 118 ppm, Ba = 295 ppm.

[0085] Sample 3: Well logging data: resistivity logging: 4.1 ohm·m, density logging: 2.72 g / cm 3 , Natural gamma ray logging: 52 API; Core sample analysis data: Porosity: 11%, mineral composition: calcite 94%, dolomite 6%; Seismic reflection data: reflection amplitude: 0.88, phase: 50°; Surrounding rock thickness and depth data: top plate burial depth: 1820 meters, layer thickness: 48 meters; Geochemical data: Carbon isotope ratio: δ 13 C = -4‰, oxygen isotope ratio: δ18 O = +3‰, trace element concentrations: Sr = 22 ppm, Ba = 85 ppm.

[0086] Based on the training model of surrounding rock classification and identification while drilling in this application, the prediction results shown in Table 1 can be obtained: Table 1 Sample No. Sandstone Probability Shale probability Limestone Probability Granite Probability Basalt Probability Final classification Sample 1 85% 10% 3% 1% 1% sandstone Sample 2 5% 85% 7% 1% 2% shale Sample 3 2% 5% 88% 3% 2% Limestone Based on Table 1, it can be seen that the drilling surrounding rock classification and identification model provided by this application calculates the probability of each sample belonging to a different drilling surrounding rock type by analyzing the input drilling surrounding rock characteristic data, and makes the final classification decision based on the maximum probability principle. After laboratory testing, the experimental test results of samples 1, 2, and 3 are consistent with the final prediction results given in this application, indicating that the technical solution provided by this application has significant advantages in improving the accuracy, consistency, and efficiency of drilling surrounding rock classification.

[0087] It should be noted that this application can achieve multiple effects and significant benefits by accurately identifying and classifying the surrounding rock while drilling, including: First, it can greatly shorten the time required for traditional manual analysis, making the exploration process more efficient. By quickly and accurately classifying the surrounding rock while drilling, it can provide immediate support for drilling decisions, reduce unnecessary drilling work, and increase the speed of resource development. Second, by using machine learning algorithms to process complex geological data, including multi-source information such as logging, core sample analysis, seismic reflection, thickness depth, and geochemistry, patterns and relationships that are difficult for the human eye to detect can be discovered, thereby achieving more accurate surrounding rock classification and improving the accuracy of understanding underground structures. Third, more efficient exploration means fewer ineffective drilling holes, which helps reduce interference with the natural environment. In addition, accurate classification results can help better plan mining activities, optimize resource allocation, and protect the ecological environment.

[0088] In summary, the method for classification and identification of surrounding rocks while drilling provided by this application not only innovates the traditional geological exploration methods, but also lays a solid foundation for the digital transformation of the resource exploration industry, and provides strong technical support for the realization of intelligent, green and efficient modern resource development.

[0089] Embodiment 2: The present application provides a device for classifying and identifying surrounding rocks while drilling, which is applied to a method for classifying and identifying surrounding rocks while drilling in Embodiment 1, such as Fig. 9 As shown, including: Model building module, which is used to collect surrounding rock data of different types of surrounding rock while drilling and build an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data; The model training module is used to train the initial model using the surrounding rock data; the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn the feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels; A training end module is used to determine the initial model that meets the preset training end condition as the surrounding rock classification and identification model while drilling until the initial model meets the preset training end condition; The classification and identification module is used to input the acquired drilling surrounding rock data to be classified into the drilling surrounding rock classification and identification model for processing, so as to obtain the classification information corresponding to the drilling surrounding rock data to be classified.

[0090] Embodiment 3: The embodiment of the present application provides a non-transitory computer-readable storage medium, characterized in that the non-transitory computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method in Embodiment 1.

[0091] Embodiment 4: The embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, and the method in Embodiment 1 is implemented when the processor executes the computer program.

[0092] The above specific implementation methods further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above are only specific implementation methods of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for classification and identification of surrounding rocks while drilling, characterized in that: The specific steps include: Collect surrounding rock data of different types of surrounding rock while drilling, and establish an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data; The initial model is trained using the surrounding rock data; wherein the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels; Until the initial model meets a preset training end condition, the initial model that meets the training end condition is determined as a surrounding rock classification and identification model while drilling; The acquired drilling-while-rock data to be classified is input into the drilling-while-rock classification and identification model for processing to obtain classification information corresponding to the drilling-while-rock data to be classified.

2. The method for classification and identification of surrounding rocks while drilling according to claim 1, characterized in that: The initial model includes an input and preprocessing module, a self-supervised learning module, a meta-learning module, a feature fusion and classification module and an output module. The preprocessing module includes a dynamic preprocessor, and the self-supervised learning module includes a contrastive learning engine.

3. A method for classification and identification of surrounding rocks while drilling according to claim 2, characterized in that: The input and preprocessing module also includes a multimodal sensor, wherein: The multimodal sensor is used to receive and integrate surrounding rock data or surrounding rock data to be classified while drilling. The output of the multimodal sensor is specifically: ； ， ； In the formula, represents the output of a multimodal perceptron; represents a nonlinear activation function; represents the weight matrix; is the bias vector; It is a multimodal feature representation of data, representing a single modality feature The result of weighted summation; Representing modal characteristics The importance weight of Indicates the total number of modes; Indicates The original input data of the modality; For the The encoder function corresponding to the modality; The dynamic preprocessor is used to select the optimal processing strategy according to the data characteristics and preprocess the integrated data. The preprocessing includes feature extraction processing and noise reduction processing. The output of the dynamic preprocessor is specifically: ； In the formula, Represents the output of the dynamic preprocessor; represents the original data matrix; represents a multimodal perceptron; is the adaptive enhancement parameter; represents an adaptive unit; represents the feature weight vector; represents feature selection and recombination; represents adversarial deep autoencoder; represents the nonlinear activation function in the autoencoder, Indicates normalization processing.

4. The method for classification and identification of surrounding rocks while drilling according to claim 2, characterized in that: The contrastive learning engine learns the feature representations that distinguish different types of surrounding rock while drilling without class labels in the surrounding rock data. This is achieved through the following steps: The input data is randomly and reasonably transformed to form a positive sample pair and a negative sample pair, wherein the positive sample pair includes a first enhanced sample and a second enhanced sample obtained by different transformation methods, and the negative sample pair includes a third sample and the first enhanced sample or the second enhanced sample, and the original data in the positive sample pair and the original data in the negative sample pair are surrounding rock data of different types of surrounding rock while drilling; The preset contrast loss function is used to calculate the contrast loss between positive sample pairs and negative sample pairs, so as to learn the feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels.

5. The method for classification and identification of surrounding rocks while drilling according to claim 4, characterized in that: The contrast loss function is specifically: ,in: ； ， ； In the formula, represents the contrast loss, represents a hyperparameter; represents the multi-view contrast loss, represents the local structure preserving term; represents the temperature parameter, represents the cosine similarity function, Indicates the number of data sources or perspectives, Represents a single view or data source, including well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, geochemical data; represents a positive sample pair, represents a negative sample pair, Indicates the batch size; express, represents the degree matrix, Represents the feature representation matrix of all samples after passing through the encoder, Represents matrix transpose.

6. The method for classification and identification of surrounding rocks while drilling according to claim 2, characterized in that: The feature fusion and classification module is used to effectively fuse the features of different data sources and compress them into a compact representation, specifically: ,in: ； In the formula, represents the fused representation, represents the activation function, represents a gating unit; represents the gating parameter; Indicates The weight coefficient of the mode; Indicates The characteristic transfer function of the mode; represents the bias term; Represents the feature conversion function Parameter set; represents the scoring function; Indicates The feature representation of the modality, Indicates the number of different modes.

7. The method for classification and identification of surrounding rocks while drilling according to claim 2, characterized in that: The output module is used to output the classification information, and the classification information is specifically: , ;in: ， ； In the formula, Represents classification information, which is the category label output by the surrounding rock classification and identification model while drilling; represents the number of category labels, Represents the labels of each category obtained after multi-category confidence weighting and context-dependent classification steps The enhancement score, represents the classification confidence score, Represents a collection of adjacent categories The sum of all class probabilities in , For a neighboring category The probability of Indicates A set of adjacent categories related to a class, Represents the adjustment parameter used to balance the direct probability and related categories of impact; Represents another adjustment parameter to balance the original score and contextual information The weight of .

8. A device for classifying and identifying surrounding rocks while drilling, applied to a method for classifying and identifying surrounding rocks while drilling according to any one of claims 1 to 7, characterized in that: include: A model building module is used to collect surrounding rock data of different types of surrounding rock while drilling, and to establish an initial model that introduces a dynamic preprocessor and a comparative learning engine. The surrounding rock data includes well logging data, core sample analysis data, seismic reflection data, surrounding rock thickness, depth data, and geochemical data; A model training module is used to train the initial model using the surrounding rock data; wherein the dynamic preprocessor is used to preprocess the surrounding rock data with an optimal processing strategy, and the comparative learning engine is used to learn feature representations that distinguish different types of surrounding rock while drilling when the surrounding rock data has no category labels; A training end module, used for determining the initial model satisfying a preset training end condition as a surrounding rock classification and identification model while drilling until the initial model satisfies the preset training end condition; The classification and identification module is used to input the acquired drilling surrounding rock data to be classified into the drilling surrounding rock classification and identification model for processing, so as to obtain classification information corresponding to the drilling surrounding rock data to be classified.

9. A non-transitory computer-readable storage medium, characterized in that: The non-transitory computer-readable storage medium stores computer instructions, which enable a computer to execute the method of any one of claims 1-7.

10. An electronic device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the method according to any one of claims 1 to 7 is implemented when the processor executes the computer program.

Images