Auxiliary rock core geological logging method based on drilling parameters

Abstract

The invention provides an auxiliary rock core geological logging method based on drilling parameters, and relates to the technical field of geological investigation, the method is applied to a logging control terminal, and the method mainly comprises the following steps: obtaining regional typical rock core samples and five types of drilling parameter signals, classifying and extracting static characteristics of the rock core samples, and carrying out parallel noise reduction analysis on the parameter signals; obtaining a matching relationship between the rock core and the parameter signal, and constructing a lithologic parameter bidirectional association reference library; on the basis of the matching relation of the lithologic parameter bidirectional association reference library, the model is embedded into an incremental learning module to update the weight, and a subsection lithologic pre-judgment result is output; and calling a preset catalog template to load a pre-judgment result, generating a catalog report after man-machine interaction recheck and correction, collecting correction data, reversely transmitting the correction data back to the reference library to update a matching relationship, and triggering incremental learning of the model to complete a closed loop. The technical problems that in traditional rock core logging, regional lithology adaptation is poor, parameter interference is large, a model is not dynamically optimized, the deep operation risk is high, and efficiency is low are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of geological exploration technology, specifically to an auxiliary core geological logging method based on drilling parameters. Background Technology

[0002] As geotechnical drilling and coring have progressed to deeper levels, the constraints on on-site core geological logging have become increasingly prominent: the working space at depth is narrow and cramped, and with increasing depth, there are also harsh conditions such as high temperature, high humidity, and accumulation of harmful gases. This not only significantly increases the intensity of personnel work and safety risks, but also makes it easy for personnel fatigue to lead to omissions in logging details. At the same time, the lithological stratification at depth is more refined and the lithological types are more complex. Logging requires recording the lithology, structure, and mineral composition information of the core segment by segment, and also requires simultaneous verification of drilling pressure, rotation speed, and drilling parameters.

[0003] Traditional lithology-parameter matching thresholds, judgment rules, and model weights are all pre-set, which cannot adapt to the unique parameter response characteristics of lithologies in different regions. For example, in new drilling areas, the parameter thresholds of existing sandstone and limestone may deviate significantly from the actual values, leading to a sharp drop in prediction accuracy. Furthermore, there is a lack of effective feedback mechanisms; even if deviations are manually corrected, the benchmark library and model cannot be iteratively optimized, and the accuracy decreases rather than increases as drilling progresses. Patent CN120763888A provides an auxiliary core geological logging method based on drilling parameters, which includes the following steps: establishing a standard core library; establishing a drilling parameter database; constructing a drilling parameter lithology identification model based on the standard core library and the drilling parameter database; using the drilling parameter lithology identification model to perform segmented lithology identification and output recommended lithology numbers; and performing auxiliary core geological logging based on lithology boundaries and recommended lithology numbers, combined with lithology description text. However, it suffers from technical shortcomings such as a simple model architecture, limited recognition accuracy, lack of precise noise reduction and feature mining in parameter processing, lack of dynamic optimization mechanism, poor regional adaptability, and simple benchmark library design with a lack of bidirectional mapping rules. The invention patent with publication number CN120543475A provides a method and system for automatic recognition and recording of borehole core images based on deep learning. The method includes: preprocessing the acquired borehole core images; inputting the preprocessed images into a trained Transformer semantic segmentation model to output a pixel-level probability map, and generating a binary mask image through binarization; performing morphological processing, connected component analysis, and contour detection on the binary mask image to obtain the edge coordinate information of the core frame; extracting a standardized core image from the original image based on the edge coordinate information and storing and visualizing the results; inputting the standardized core image into a trained Transformer multi-class core structure recognition model to output a multi-channel pixel-level probability map, and generating a binary mask image for each channel through binarization; performing three-probability calculation and multi-structure statistical analysis of the borehole core, and outputting and visualizing the data. However, it suffers from several technical shortcomings, including reliance on image data, limiting its applicability; the model focuses on structural identification but fails to achieve accurate lithology identification; the lack of a closed-loop optimization mechanism limits its accuracy; and the absence of drilling parameters and deep correlation analysis.

[0004] In view of this, it is necessary to study an auxiliary core geological logging method based on drilling parameters to solve the above-mentioned technical problems. Summary of the Invention

[0005] Given that traditional lithology-parameter matching thresholds, judgment rules, and model weights in the background technology are all pre-set, they cannot adapt to the unique parameter response characteristics of lithology in different regions, lack an effective feedback mechanism, and even with manual correction of deviations, the benchmark library and model cannot be iteratively optimized, resulting in a decrease in accuracy as drilling progresses, this invention proposes an auxiliary core geological logging method based on drilling parameters. It constructs a complete intelligent closed-loop system of "data acquisition - model calculation - human-computer interaction - feedback optimization." Through the establishment of a bidirectional correlation benchmark library for lithology parameters, the collaborative analysis of the XGBoost and DA-LSTM hybrid model, and the introduction of an incremental learning mechanism, it achieves a fundamental transformation of core geological logging from traditional manual judgment to intelligent and adaptive processes. This transforms traditional geological logging work into a continuously self-improving intelligent system, where the bidirectional correlation benchmark library for lithology parameters serves as the core knowledge base, the hybrid model acts as the intelligent analysis engine, and the incremental learning mechanism ensures that the system can continuously optimize its performance as drilling progresses, ultimately forming a complete, adaptive, and high-precision intelligent solution for core geological logging. It fundamentally solves the industry pain points of traditional core logging, which relies on manual experience, is inefficient, and cannot adapt to different geological conditions. Through the deep integration of machine learning and geological expertise, it achieves automation, accuracy, and continuous optimization of lithology identification, providing a brand-new intelligent solution for the field of geological exploration.

[0006] In a first aspect, embodiments of the present invention provide an auxiliary core geological logging method based on drilling parameters, which is applied to a logging control terminal and includes the following steps: S1. The logging and control terminal acquires typical rock core samples and five types of drilling parameter signals in the region through the acquisition terminal. It extracts static features from the rock core samples, performs parallel noise reduction and analysis on the parameter signals, obtains the matching relationship between the rock core and the parameter signals, and constructs a bidirectional correlation benchmark library for lithological parameters. S2. Based on the matching relationship of the bidirectional correlation benchmark library of lithological parameters, a dynamic sliding window combined with the Savitzky-Golay algorithm is used to denoise the parameter signals of real-time drilling, extract the drilling specific energy features, and generate a structured lithological feature vector. S3. Input the structured lithology feature vector into the pre-trained XGBoost and DA-LSTM hybrid model, embed the incremental learning module into the model to update the weights, and output the segmented lithology prediction results. S4. Call the preset cataloging template to load the prediction results, generate a cataloging report after human-computer interaction review and correction, collect the correction data and send it back to the benchmark library to update the matching relationship, and trigger the incremental learning of the model to complete the closed loop.

[0007] As a further improvement of the present invention, step S1, which involves classifying and extracting static features from the core sample, includes: Based on standardized classification according to lithology, structure and mineral composition, key static characteristics such as hardness grade, grain size and density are extracted. The five types of drilling parameter signals include: drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement; and the five types of drilling parameter signals are subjected to parallel time-domain noise reduction and frequency-domain analysis to extract the fluctuation amplitude and dynamic patterns of the stable interval. The construction of the bidirectional correlation benchmark library for lithological parameters includes: establishing a bidirectional correspondence between lithology and parameters through correlation analysis, clarifying the parameter threshold range and lithology orientation rules; and constructing a bidirectional correlation benchmark library for lithological parameters containing classification labels, feature data, and mapping rules to provide a reference for subsequent feature extraction and lithology identification.

[0008] As a further improvement of the present invention, step S1 also includes: clarifying the numerical threshold values ​​of the five types of parameters corresponding to sandstone, limestone and shale respectively, and formulating a judgment standard that points to a specific lithology when a single parameter exceeds the threshold or when three or more parameters are matched. A benchmark library is constructed, which includes classification labels for lithological names and numbers, static characteristics and dynamic parameter data of rock cores, and two-way mapping rules for lithology to parameter ranges and parameter combinations to lithology. This provides a directly callable reference for subsequent feature extraction and lithology matching.

[0009] As a further improvement of the present invention, in step S1, the bidirectional lithology mapping rule refers to the specific numerical range of the five types of parameters, namely, drilling pressure, rotation speed, drilling speed, torque, and pump pressure and displacement, corresponding to the specific lithology of sandstone, limestone, and shale. At the same time, when the five types of parameters exceed the threshold or are matched in multiple combinations, they point to the judgment criteria of the specific lithology. When matching the lithology, the correspondence between the parameter combination and the lithology is directly called to provide a reference for operation.

[0010] As a further improvement of the present invention, in step S2, the drilling specific energy feature is extracted to generate the structured lithological feature vector. The specific process is as follows: after denoising the five types of parameter signals of real-time drilling, the drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stability duration features of each parameter are extracted in a targeted manner; according to the mapping rules of the lithological parameter bidirectional correlation benchmark library, the aforementioned multi-dimensional features are integrated in a fixed-dimensional order to generate a structured lithological feature vector containing key lithological correlation indicators; the structured lithological feature vector directly adapts to the input format of the subsequent hybrid recognition model, providing standardized data support for lithological prediction.

[0011] As a further improvement of the present invention, in step S2, the structured lithological feature vector integrates the drilling specific energy, the power rate ratio, the time-series fluctuation coefficient, and the mean, peak value, and stability duration of each parameter into a multi-dimensional feature based on the mapping rules of the bidirectional correlation benchmark library of lithological parameters, and arranges them in a standardized order according to the fixed dimension.

[0012] As a further improvement of the present invention, in step S2, the multi-dimensional features are integrated in a standardized arrangement according to a fixed dimensional order, which is: drilling specific energy → power-speed ratio → time-series fluctuation coefficient → average drilling pressure → peak drilling pressure → drilling pressure stabilization time → average rotation speed → peak rotation speed → rotation speed stabilization time → average drilling speed → peak drilling speed → drilling speed stabilization time → average torque → peak torque → torque stabilization time → average pump pressure and displacement → peak pump pressure and displacement → pump pressure and displacement stabilization time.

[0013] As a further improvement of the present invention, in step S3, the incremental learning module embedded in the pre-trained XGBoost and DA-LSTM hybrid model dynamically adjusts the internal weight parameters in combination with the feature vector corresponding to the real-time drilling data to optimize the adaptability of the lithology identification. The working mechanism of the XGBoost and DA-LSTM hybrid model includes: XGBoost is responsible for extracting key nonlinear correlation features from the feature vector; DA-LSTM captures the temporal variation law of the parameters. After the two are calculated together, they are divided into segments according to the drilling process, and the corresponding lithology prediction results of each segment are output to achieve matching.

[0014] As a further improvement of the present invention, in step S3, the key nonlinear correlation features extracted by XGBoost include: nonlinear correlations between multi-dimensional features such as drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stable duration of each parameter; the nonlinear correlations include the synergistic effect of different features and the non-simple linear correspondence of threshold cross-influence. The DA-LSTM captures temporal variation patterns by: processing feature vectors with a fixed time step; capturing the dynamic fluctuation amplitude, trend, and abrupt change nodes of parameters; and identifying the segmented variation patterns of lithology as drilling progresses.

[0015] As a further improvement of the present invention, XGBoost is responsible for extracting key nonlinear correlation features from the feature vector, and also includes: XGBoost mining the nonlinear correlations between the drilling specific energy, the power rate ratio, the time-series fluctuation coefficient, and the mean, peak value, and stability duration of each parameter from the structured lithological feature vector. These correlations include the synergistic effect of different features, the non-simple linear correspondence of threshold cross-influence, and all conform to the mapping rules of the bidirectional correlation benchmark library of lithological parameters. Deep correlation information that plays a key role in the lithological determination is selected, providing core feature support for the collaborative calculation of the hybrid model.

[0016] As a further improvement of the present invention, the DA-LSTM captures the temporal variation law of the parameters, and further includes: the DA-LSTM focuses on the structured lithological feature vectors corresponding to the five types of parameters, and captures the temporal variation law of the parameters with the drilling process, including the drilling specific energy, the power rate ratio, the temporal fluctuation coefficient, and the mean, peak value, dynamic fluctuation amplitude, trend and abrupt change nodes of each parameter, to supplement the hybrid model with key information of the temporal dimension, help identify the segmented changes of the lithology with the drilling process, and support the accuracy of the segmented lithology prediction results.

[0017] As a further improvement of the present invention, after the human-computer interaction process manually reviews and corrects the prediction results, a standardized cataloging report is generated. At the same time, the accurate data after review and correction is collected and sent back to the bidirectional correlation benchmark library of lithological parameters to update the matching relationship between the lithology and the parameters. Simultaneously, the incremental learning of the hybrid XGBoost and DA-LSTM model is triggered to optimize the model weights and the benchmark library data, forming the closed loop and continuously improving the accuracy of lithological prediction.

[0018] As a further improvement of the present invention, in step S4, the human-computer interaction review and correction includes: Staff members reviewed the preliminary results segment by segment by combining drilling logs, core observations, and parameter monitoring curves; corrected the lithology labels for lithology segments with deviations and added explanations of the reasons for the deviations; and determined the final lithology for segments with ambiguous parameter threshold boundaries by combining core structure and mineral composition analysis. The reverse transmission back to the benchmark library to update the matching relationship includes: collecting accurate data after correction, including the final lithology label of each segment, the measured value of the corresponding parameter, and the comparison information before and after correction; supplementing new lithology-parameter corresponding samples to the benchmark library; and adjusting the parameter threshold range in the original matching relationship. The incremental learning of the triggering model includes: using newly added accurate data as supplementary training samples; re-optimizing the decision tree splitting weights of XGBoost and the neuron connection weights of DA-LSTM; and updating the confidence of bidirectional mapping rules in the benchmark library.

[0019] Secondly, embodiments of the present invention provide an auxiliary core geological logging system based on drilling parameters, which employs the aforementioned auxiliary core geological logging method based on drilling parameters, including: The data acquisition module is used to acquire typical core samples from the region and five types of drilling parameter signals; The benchmark library construction module is used to extract static characteristics and dynamic laws of parameters from rock cores and establish a two-way correlation benchmark library of lithological parameters. The feature processing module is used to reduce noise and generate structured lithological feature vectors by using a dynamic sliding window combined with the Savitzky-Golay algorithm. The lithology prediction module includes a hybrid model of XGBoost and DA-LSTM and an incremental learning unit, which outputs segmented lithology prediction results. The cataloging and feedback module is used to call templates to generate reports and collect and correct data to update the benchmark library and model.

[0020] The present invention has the following beneficial effects: The assisted core geological logging method based on drilling parameters provided by this invention constructs a dynamic closed loop of "data acquisition-model calculation-result feedback-dual optimization" to continuously improve the accuracy of lithology prediction. This method first constructs a benchmark library containing parameter thresholds and mapping rules based on typical samples, and then outputs the prediction results through a hybrid model of XGBoost and DA-LSTM. The accurate data after human-computer interaction verification is fed back, which updates the parameter thresholds and mapping rules of the benchmark library and triggers incremental learning to optimize the model weights. As drilling data accumulates, the benchmark library and model continuously adapt to the regional lithological characteristics, gradually reducing the prediction deviation, significantly improving the logging accuracy and efficiency, and reducing the cost of manual verification.

[0021] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0022] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the present invention will be briefly described below. Obviously, the drawings described below are merely some embodiments of the present invention, and those skilled in the art can obtain other drawings based on these drawings without any creative effort.

[0023] Figure 1 This is a flowchart illustrating the auxiliary core geological logging method based on drilling parameters provided in this embodiment of the invention. Detailed Implementation

[0024] The embodiments of the technical solution of the present invention will now be described in detail with reference to the accompanying drawings. These embodiments are merely illustrative of the technical solution of the present invention and are therefore intended to limit the scope of protection of the present invention.

[0025] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the invention, are intended to cover non-exclusive inclusion.

[0026] In the description of the embodiments of this invention, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this invention, "multiple" means two or more, unless otherwise explicitly defined.

[0027] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0028] In the description of the embodiments of this invention, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0029] In the description of the embodiments of the present invention, the term "multiple" refers to two or more (including two), similarly, "multiple groups" refers to two or more (including two groups), and "multiple pieces" refers to two or more (including two pieces).

[0030] In the description of the embodiments of the present invention, the technical terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," "counterclockwise," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing the embodiments of the present invention and simplifying the description, and are not intended to indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the embodiments of the present invention.

[0031] In the description of the embodiments of the present invention, unless otherwise explicitly specified and limited, the technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in the embodiments of the present invention according to the specific circumstances.

[0032] To address the technical problems of poor regional lithology adaptation, numerous parameter interferences, lack of dynamic model optimization, and high risk and low efficiency in deep operations in traditional core logging, this invention provides an auxiliary core geological logging method based on drilling parameters. It is designed based on a "lithology-parameter bidirectional correlation benchmark library," integrating the static characteristics of typical regional cores (hardness level, grain size, compactness) with the dynamic patterns of five types of drilling parameters (drilling pressure, rotation speed, etc.), clarifying the bidirectional mapping rules and specific numerical thresholds between lithology and parameters, and providing a standardized reference for lithology prediction. An innovative "dynamic sliding window + Savitzky-Golay" combined noise reduction scheme is used to filter out drilling vibration interference while preserving parameter abrupt changes, simultaneously extracting multi-dimensional features such as drilling specific energy and power-speed ratio, generating a standardized structured lithology feature vector adapted to the model input. A hybrid model architecture of XGBoost and DA-LSTM is constructed. XGBoost mines the nonlinear correlation between features, while DA-LSTM captures the temporal variation patterns of parameters. An incremental learning module is embedded to dynamically adjust weights, achieving accurate segmented lithology prediction. A closed-loop mechanism of "prediction-verification-feedback-optimization" is designed, in which accurate data corrected through human-computer interaction is fed back in reverse, and the matching relationships and model weights of the benchmark library are updated synchronously, continuously improving adaptability and prediction accuracy as drilling progresses. A standardized logging process is established, in which prediction results are loaded through preset templates and combined with manual verification to generate standardized reports. This reduces the safety risks and labor intensity of deep operations, and also solves the pain point that traditional methods cannot adapt to regional lithological differences.

[0033] Example 1 Please refer to Figure 1 This invention proposes an auxiliary core geological logging method based on drilling parameters. The method is applied to a logging control terminal, and the specific method includes the following: S1. The logging and control terminal acquires typical core samples and five types of drilling parameter signals from the acquisition terminal. It extracts static features from the core samples, performs parallel noise reduction and analysis on the parameter signals, obtains the matching relationship between the core samples and the parameter signals, and constructs a bidirectional correlation benchmark library for lithological parameters.

[0034] Parameter acquisition terminal: The sensor group equipped on the drilling rig can collect and transmit five types of parameter signals in real time: drilling pressure (F), rotation speed (N), drilling speed (V), torque (M), and pump pressure and displacement (Q).

[0035] S11. On-site sampling and parameter acquisition: 20 typical boreholes were selected in the project area, and 50 kg of samples were collected per borehole, for a total of 1000 kg of core samples. Simultaneously, five types of drilling parameter signals, namely drilling pressure, rotation speed, drilling speed, torque, and pump pressure and displacement, were collected at a sampling frequency of 1 Hz.

[0036] S12. Core sample processing: Standardized classification according to lithology (sandstone, limestone, shale), structure (layer thickness, cementation method) and mineral composition (quartz content, calcite ratio). Hardness grade is measured by indentation method, particle size is measured by sieve analysis method, and density is measured by porosity determination method. Three types of static characteristics are extracted.

[0037] S13. Parameter signal analysis: Parallel time-domain filtering and noise reduction (filtering out vibration interference) and frequency-domain Fourier transform analysis of five types of parameter signals are performed to extract the dynamic laws of fluctuation amplitude (such as instantaneous fluctuation difference of drilling pressure) and stable interval (such as the duration of rotational speed fluctuation <5%) of each parameter.

[0038] S14. Establish bidirectional mapping rules and a benchmark library: Clarify parameter thresholds through correlation analysis. Sandstone: Drilling pressure 6-10 MPa, rotation speed 800-1200 r / min, drilling speed 1.0-1.5 m / h, torque 35-50 N. m, pump pressure and discharge rate 15-25L / min; Limestone: Drilling pressure 10-15MPa, rotation speed 1100-1500r / min, drilling speed 0.8-1.2m / h, torque 45-60N. m, pump pressure and discharge rate 20-30L / min; Shale: Drilling pressure 4-8 MPa, rotation speed 700-1000 r / min, drilling speed 0.5-0.9 m / h, torque 25-40 N. m, pump pressure and discharge capacity 12-20L / min.

[0039] Based on this, rules are established: when a single parameter exceeds the corresponding lithology threshold and is outside the thresholds of other lithologies, it directly points to a specific lithology; when three or more parameters in a multi-parameter combination fall within the same lithology threshold range, it points to that lithology. The final benchmark library includes classification tags labeled with lithology names (sandstone, limestone, shale) and unique numbers (such as Y001, Y002, Y003), integrates static characteristics of core hardness, grain size, and compactness with dynamic data on parameter fluctuation amplitude and stable range, and clarifies two-way mapping rules for lithology corresponding to parameter ranges and parameter combinations corresponding to lithologies. After systematic integration of these contents, a directly applicable reference is provided for selecting key indicators when extracting features and calling judgment criteria when matching lithologies.

[0040] Specifically, in step S1, the recording control terminal first acquires typical rock core samples and five types of real-time drilling parameter signals (drilling pressure, rotation speed, drilling speed, torque, and pump pressure / displacement) from the acquisition terminal. The rock core samples are standardized and classified according to lithology, structure, and mineral composition. Static features such as hardness level, grain size, and density are extracted. At the same time, time-domain noise reduction and frequency-domain analysis are performed on the five types of parameter signals in parallel to extract the dynamic laws of fluctuation amplitude and stable interval. Then, the corresponding matching relationship between the rock core and the parameter signals is obtained, and a bidirectional correlation benchmark library of lithological parameters containing classification labels, feature data, and mapping rules is constructed.

[0041] Core samples were standardized and classified according to lithology, structure and mineral composition, and key static characteristics such as hardness grade, grain size and density were extracted.

[0042] The five types of drilling parameter signals specifically include: drilling pressure, rotational speed, drilling speed, torque, and pump pressure and displacement. The five types of drilling parameter signals are subjected to parallel time-domain noise reduction and frequency-domain analysis to extract the fluctuation amplitude and dynamic laws of the stable interval.

[0043] Association analysis establishes a two-way correspondence between lithology and parameters, clarifies the parameter threshold range and lithology orientation rules, and constructs a two-way association benchmark library of lithological parameters containing classification labels, feature data and mapping rules, providing a reference for subsequent feature extraction and lithology identification.

[0044] Furthermore, when standardizing the classification of core samples, the lithological category (such as sandstone, limestone, and shale) is used as the core classification basis. The classification level is then refined by combining the structural features (such as bedding thickness and cementation method) and mineral composition (such as quartz content and calcite ratio). Then, the hardness level is measured by indentation method, the particle size is measured by sieve analysis method, and the compactness is measured by porosity measurement method. These three key static characteristics are extracted. The five types of drilling parameter signals are clearly defined as drill pressure, rotation speed, drilling speed, torque, and pump pressure and displacement. When processing them in parallel, the time domain uses filtering to reduce noise and filter out drilling vibration interference, and the frequency domain uses Fourier transform to analyze the frequency distribution characteristics and simultaneously extract the dynamic laws of the fluctuation amplitude (such as the difference between the maximum and minimum instantaneous fluctuation of drill pressure) and the stable interval (such as the duration of rotation speed fluctuation less than 5%) of each parameter. Correlation analysis was conducted based on the static characteristics of the classified core samples and the dynamic patterns of the analyzed parameters. The parameter thresholds were determined by statistically analyzing the distribution range of parameters corresponding to different lithological samples (e.g., the drilling pressure threshold for shale is 5-8 MPa). A lithological pointing rule was formulated that "a single parameter reaching the threshold or a combination of multiple parameters matching points to a specific lithology". Finally, a lithological parameter bidirectional correlation benchmark library was constructed, which includes classification labels with lithological names and numbers, feature datasets of core static characteristics and parameter dynamic data, and mapping rules for the bidirectional correspondence between lithology and parameters. This provides a standardized reference for subsequent targeted feature extraction and lithological matching.

[0045] Define the specific numerical thresholds for the five types of parameters corresponding to sandstone, limestone, and shale, and formulate rules that point to specific lithologies based on single parameter exceeding the threshold or multi-parameter combination matching. A benchmark library is constructed, which includes classification labels for lithological names and numbers, static characteristics and dynamic parameter data of rock cores, and two-way mapping rules for lithological parameter ranges and parameter combinations, providing a directly callable reference for subsequent feature extraction and lithological matching.

[0046] The two-way mapping rule for lithology specifically refers to the two-way correspondence between lithology-corresponding parameters and parameters-corresponding lithology. That is, for specific lithologies such as sandstone, limestone, and shale, there are clear numerical ranges for five types of parameters: drilling pressure, rotation speed, drilling speed, torque, and pump pressure / displacement. At the same time, when a single parameter exceeds the threshold or multiple combinations of the five types of parameters are matched, they point to the judgment criteria of a specific lithology. When matching lithology, the correspondence between parameter combinations and lithology is directly called to provide a reference for operation.

[0047] Furthermore, the core of the lithology two-way mapping rule is the clear two-way correspondence between lithology-corresponding parameters and the lithology-corresponding parameters. Specifically, sandstone, limestone, and shale, three specific lithologies, each correspond to a specific numerical range for five parameters: drilling pressure, rotation speed, drilling rate, torque, and pump pressure / displacement (e.g., sandstone corresponds to a drilling pressure of 6-10 MPa and a rotation speed of 800-1200 r / min). The determination logic for these five parameters is also clearly defined. When a single parameter exceeds the corresponding value threshold of a certain lithology and does not fall into the threshold range of other lithologies, it directly points to that specific lithology; When three or more parameters in a multi-parameter combination fall within the numerical threshold range of the same lithology, they also point to that lithology. These two situations together constitute the criteria for determining a specific lithology. In subsequent lithology matching operations, the preset correspondence between parameter combinations and lithologies can be directly called without additional derivation, quickly completing lithology matching and providing a specific and practical operational reference.

[0048] S2. Based on the matching relationship of the bidirectional correlation benchmark library of lithological parameters, a dynamic sliding window combined with the Savitzky-Golay algorithm is used to denoise the parameter signals of real-time drilling, extract the drilling specific energy features, and generate a structured lithological feature vector.

[0049] S21. Parameter noise reduction: A dynamic sliding window (the window length is adjusted according to the parameter fluctuation law) combined with the Savitzky-Golay algorithm is used to smooth and reduce noise of five types of parameter signals in real-time drilling, while retaining the parameter mutation characteristics corresponding to lithological changes.

[0050] S22. Multi-dimensional feature extraction: Calculate drilling specific energy (drilling pressure / drilling speed), power-speed ratio (torque / rotation speed), time-series fluctuation coefficient (parameter dispersion), and the mean, peak value, and stability duration (duration within the corresponding lithology threshold range) of each parameter.

[0051] S23. Feature Vector Standardization: Features are integrated in a fixed-dimensional order of "drilling specific energy - power-speed ratio - time-series fluctuation coefficient - average drilling pressure - peak drilling pressure - drilling pressure stabilization time - average rotation speed - peak rotation speed - stabilization time of rotation speed - average drilling speed - peak drilling speed - stabilization time of drilling speed - average torque - peak torque - stabilization time of torque - average pump pressure and discharge rate - peak pump pressure and discharge rate - stabilization time of pump pressure and discharge rate", and normalized to the [0,1] interval to generate a structured lithological feature vector.

[0052] Specifically, in step S2, based on the matching relationship of the benchmark library, a dynamic sliding window that dynamically adjusts the window length according to the parameter fluctuation law is combined with the Savitzky-Golay algorithm to smooth and reduce noise in the real-time drilling parameter signal while retaining key mutation features. Then, drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and multi-dimensional features of the mean and peak values ​​of each parameter are extracted in a targeted manner and integrated in a fixed-dimensional order to generate a standardized structured lithological feature vector.

[0053] Extracting drilling specific energy features and generating structured lithological feature vectors also includes: after denoising the five types of parameter signals from real-time drilling, specifically extracting drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stability duration features of each parameter; integrating these multi-dimensional features in a fixed-dimensional order according to the mapping rules of the bidirectional correlation benchmark library of lithological parameters to generate a structured lithological feature vector containing key lithological correlation indicators. The structured lithological feature vector directly adapts to the input format of the subsequent hybrid recognition model, providing standardized data support for lithological prediction.

[0054] Furthermore, for the five types of parameter signals of real-time drilling—drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement—after smoothing and noise reduction using a dynamic sliding window combined with the Savitzky-Golay algorithm while retaining key abrupt change features, multi-dimensional core features are extracted in a targeted manner: the drilling ratio can be calculated from the ratio of drilling pressure to drilling speed; the power-speed ratio is derived from the correlation between torque and rotational speed; and the time-series fluctuation coefficient is statistically obtained based on the degree of dispersion of each parameter with drilling time. At the same time, the mean (average value of the parameter per unit time), peak value (maximum value of the parameter within the drilling period), and stable duration (duration of the parameter being within the corresponding lithology threshold range) of each of the five types of parameters are calculated.

[0055] Subsequently, based on the mapping rules of the bidirectional correlation benchmark library of lithological parameters, the features that play a key role in lithological determination are selected and systematically integrated in a fixed dimensional order of "drilling specific energy - power-speed ratio - time-series fluctuation coefficient - average drilling pressure - peak drilling pressure - drilling pressure stability time - average rotation speed" to generate a structured lithological feature vector containing all key lithological correlation indicators.

[0056] The vector's feature dimensions, data type (numerical standardized data), and arrangement format are all fully matched with the preset input interface of the subsequent XGBoost and DA-LSTM hybrid model. It can be directly input into the model without additional format conversion, providing high-quality data support with a unified standard and direct calculation for lithology prediction.

[0057] The structured lithological feature vector is based on the mapping rules of the bidirectional correlation benchmark library of lithological parameters. It integrates the multi-dimensional features of drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stability duration of each parameter, and arranges them in a standardized order according to a fixed dimension.

[0058] Furthermore, the construction of the structured lithological feature vector strictly follows the mapping rules of the bidirectional correlation benchmark library of lithological parameters. First, it systematically integrates multi-dimensional key features: the drilling ratio can be calculated by the ratio of drilling pressure to drilling speed, the power-speed ratio is derived based on the correlation between torque and rotational speed, and the time-series fluctuation coefficient statistically measures the dispersion of five types of parameters as the drilling process changes. It also covers the mean (average level of parameters per unit time), peak (maximum value of parameters during the drilling period), and stable duration (duration of parameters within the corresponding lithological threshold range) of drilling pressure, rotational speed, drilling speed, torque, and pump pressure and displacement.

[0059] S3. Input the feature vector into the pre-trained XGBoost and DA-LSTM hybrid model, embed the incremental learning module into the model to update the weights, and output the segmented lithology prediction results.

[0060] S31. Model initialization: Load the pre-trained hybrid model of XGBoost and DA-LSTM. XGBoost is responsible for mining the nonlinear correlation between features (such as the synergistic effect of drilling specific energy and torque peak). DA-LSTM captures the temporal change law of parameters with a time step of 10s (such as the lag in the decrease of drilling speed after the continuous increase of torque).

[0061] S32, Incremental Learning Adaptation: The incremental learning module embedded in the model receives real-time feature vectors, compares them with the distribution differences of historical training data, and dynamically adjusts the split weights of XGBoost tree nodes and the connection weights of DA-LSTM neurons to optimize regional adaptability.

[0062] S33. Segmented Lithology Output: Divide the model into segments according to parameter mutation nodes and drilling depth (every 5m), and output the lithology prediction results for each segment in a collaborative manner (e.g., “Y001 - Sandstone”, “Y002 - Limestone”).

[0063] Specifically, in step S3, the feature vectors are input into a pre-trained hybrid model of XGBoost and DA-LSTM. XGBoost mines the nonlinear correlation between features, while DA-LSTM captures the temporal variation patterns. The model dynamically adjusts its internal weights through an embedded incremental learning module and collaboratively outputs lithology prediction results segmented according to the drilling process.

[0064] The incremental learning module embedded in the pre-trained XGBoost and DA-LSTM hybrid model dynamically adjusts the internal weight parameters based on the feature vectors corresponding to real-time drilling data to optimize the adaptability of lithology identification. XGBoost is responsible for extracting key nonlinear correlation features from the feature vectors, while DA-LSTM captures the temporal variation of parameters. After the two work together to calculate, the model is divided into segments according to the drilling process, and the lithology prediction results corresponding to each segment are output to achieve matching.

[0065] Furthermore, the pre-trained XGBoost and DA-LSTM hybrid model first uses the structured lithological feature vectors and corresponding lithological labels of a large number of historical core samples as training data to complete the initial weight parameter training; its embedded incremental learning module receives the feature vectors corresponding to new real-time drilling data in real time, calculates the weight update gradient by comparing the difference between the feature distribution of the new data and the historical training data, and dynamically adjusts the tree node splitting weights of XGBoost and the neuron connection weights of DA-LSTM, thereby optimizing the model's adaptability to the identification of lithology in the current drilling area; DA-LSTM uses a fixed time step to process feature vectors, capturing the temporal dependence of drilling pressure and torque parameters as drilling progresses (such as the lagging decrease in drilling speed after a continuous increase in torque over a certain period) and parameter abrupt change nodes corresponding to lithological changes. When the two work together, the key features extracted by XGBoost are used as input to DA-LSTM to supplement the temporal dimension information. The temporal feature weights output by DA-LSTM inversely correct the feature importance ranking of XGBoost. After fusion calculation, drilling segments are divided according to parameter abrupt change nodes and drilling depth, and the lithological prediction results of sandstone, limestone or shale corresponding to each segment are output, realizing the matching of features and lithology.

[0066] XGBoost is responsible for extracting key nonlinear correlation features from feature vectors. It also includes: XGBoost mining nonlinear correlations between multiple dimensions of features such as drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and mean, peak value, and stability duration of each parameter from structured lithological feature vectors. These correlations include the synergistic effect of different features and the non-simple linear correspondence of threshold cross-influence. They all conform to the mapping rules of the bidirectional correlation benchmark library of lithological parameters, and screen out deep correlation information that plays a key role in lithological determination, providing core feature support for the collaborative calculation of hybrid models.

[0067] Furthermore, XGBoost deeply explores the nonlinear correlations between multi-dimensional features from a structured lithological feature vector that integrates the mean, peak, and stability duration of drilling specific energy, power ratio, time-series fluctuation coefficient, and drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement. These correlations include the synergistic effects of different features (e.g., when the drilling specific energy exceeds the corresponding threshold for sandstone, the peak torque is between 35-50 N). The synergistic effect of m and drilling pressure stabilization time exceeding 10 minutes strengthens the basis for sandstone identification. The threshold cross-influence (such as when the power rate ratio is in the corresponding range of limestone and the time series fluctuation coefficient is less than 0.3, even if a single parameter does not fully meet the standard, it still constitutes a key correlation for limestone identification) is not a simple linear correspondence. Moreover, all correlations strictly conform to the correspondence rules between lithology and parameters in the bidirectional correlation benchmark library of lithological parameters, providing core and highly recognizable feature support for the collaborative calculation of the XGBoost and DA-LSTM hybrid model.

[0068] The DA-LSTM captures the temporal variation patterns of parameters, including: DA-LSTM focuses on the structured lithological feature vectors corresponding to five types of parameters, capturing the temporal variation patterns of parameters with the drilling process, including drilling specific energy, power-speed ratio, temporal fluctuation coefficient, and the dynamic fluctuation amplitude, trend, and abrupt change nodes of the mean, peak, and stable duration characteristics of each parameter. This supplements the hybrid model with key information in the temporal dimension, helps to identify the segmented changes in lithology with the drilling process, and supports the accuracy of segmented lithology prediction results.

[0069] Furthermore, DA-LSTM focuses on the structured lithological feature vectors corresponding to five types of parameters: drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement. It continuously tracks these parameters in fixed drilling time steps, capturing the temporal variations in drilling specific energy, power-to-speed ratio, temporal fluctuation coefficient, and the mean, peak value, and stable duration of each of the five parameters. This includes dynamic fluctuation amplitudes (e.g., drilling specific energy decreasing from 6 MPa during sandstone drilling). h / m fluctuated to 8MPa The amplitude of h / m, the peak torque in the shale stage from 30N m surged to 40N The instantaneous change of m), the trend of change (such as the stable trend of the average rotation speed remaining at around 1200 r / min during limestone drilling, and the upward trend of the average pump pressure and discharge rate gradually increasing from 15 L / min to 20 L / min during the transition from shale to sandstone), and abrupt change nodes (such as the sudden drop in drilling pressure stability time from 12 minutes to 4 minutes, and the lithological boundary abrupt change node where the time series fluctuation coefficient exceeds 0.5).

[0070] These captured temporal patterns all conform to the mapping rules of the bidirectional correlation benchmark library of lithological parameters (such as sandstone corresponding to gentle fluctuations, limestone corresponding to stable trends, and shale corresponding to high-frequency abrupt changes), supplementing the XGBoost and DA-LSTM hybrid model with key temporal dimension information, helping the model to identify the segmented changes in lithology with drilling depth and time, and effectively supporting the accuracy of segmented lithology prediction results.

[0071] S4. Call the preset cataloging template to load the prediction results. After human-computer interaction to review and correct, generate a cataloging report, collect the corrected data and send it back to the benchmark library to update the matching relationship, and trigger the incremental learning of the model to complete the closed loop.

[0072] S41. Human-computer interaction verification: The preset standardized logging template is called to load the pre-judgment results. The staff combines the on-site drilling log, core observation and parameter curves to correct the lithology label for the pre-judgment deviation segment (such as shale being misjudged as sandstone). For the ambiguous segment of parameter threshold boundary, the final lithology is determined by combining the core structure and mineral composition analysis.

[0073] S42. Logging Report Generation: Generates a standardized logging report (supports export in .doc format) that includes drilling depth segments, lithology determination for each segment, actual value range of corresponding parameters, and verification and correction notes.

[0074] S43. Data Feedback Optimization: Collect accurate data after correction (final lithology labels, measured parameter values, comparison before and after correction), and backfeed it to the bidirectional correlation benchmark library of lithology parameters to update parameter thresholds and mapping rules; Simultaneously trigger incremental learning of the model, use the newly added data as supplementary training samples, optimize the model weights, and form a closed loop.

[0075] Specifically, in step S4, the preset standardized logging template is called to load the prediction results. After a human-computer interaction process where the lithological segments with prediction deviations are manually reviewed and corrected, a logging report is generated. At the same time, the corrected accurate data is collected and sent back to the bidirectional correlation benchmark library of lithological parameters to update the matching relationship between lithology and parameters. Simultaneously, the incremental learning of the hybrid model is triggered to optimize the weight parameters, forming a complete closed loop of "data acquisition - model calculation - result feedback - dual optimization of data and model", which continuously improves the degree and efficiency of core geological logging.

[0076] After the human-computer interaction process manually reviews and corrects the prediction results, a standardized compilation report is generated. At the same time, the accurate data after review and correction is collected and sent back to the bidirectional correlation benchmark library of lithological parameters to update the matching relationship between lithology and parameters. Simultaneously, the incremental learning of the hybrid XGBoost and DA-LSTM model is triggered to optimize the model weights and benchmark library data, forming a closed loop and continuously improving the accuracy of lithological prediction.

[0077] Furthermore, in the human-computer interaction process, staff will combine on-site drilling logs, core sample observations, and real-time parameter monitoring curves to review the segmented lithology prediction results segment by segment. For lithology segments with prediction deviations (such as misclassifying sandstone as limestone), the lithology labels will be corrected and explanations of the deviations will be added. For ambiguous segments with parameter threshold boundaries (such as multiple parameters falling within two lithology ranges), the final lithology will be determined by combining core structure and mineral composition analysis. After the review and correction are completed, a standardized logging report will be generated, which includes drilling depth segments, determined lithology for each segment, actual numerical ranges of the corresponding five types of parameters, and review and correction notes. At the same time, the system automatically collects accurate data after correction, including the final lithology label of each segment, the corresponding measured values ​​of drilling pressure and rotation speed parameters, and the comparison information before and after correction. This data is then transmitted back to the lithology parameter bidirectional correlation benchmark library to supplement new lithology-parameter corresponding samples and adjust the parameter threshold range in the original matching relationship (such as expanding the reasonable fluctuation range of torque corresponding to limestone). Incremental learning of the hybrid XGBoost and DA-LSTM model is triggered simultaneously. The model uses newly added accurate data as supplementary training samples to re-optimize the decision tree split weights of XGBoost and the neuron connection weights of DA-LSTM. At the same time, the confidence of the bidirectional mapping rules in the benchmark library is updated, forming a complete closed loop of "prediction result verification - accurate data feedback - benchmark library iteration - model weight optimization - next prediction accuracy improvement". With the continuous accumulation of drilling data, the prediction deviation is continuously reduced, and the accuracy and stability of lithology prediction are continuously improved.

[0078] In this embodiment, the core geological logging method based on drilling parameters is applied to the logging control terminal, and its complete workflow is as follows: The logging and control terminal first acquires typical core samples from the region and real-time drilling parameter signals for five categories: drill pressure, rotation speed, drilling rate, torque, and pump pressure / displacement. The core samples are categorized into sandstone, limestone, and shale lithologies, further refined by considering bedding thickness, cementation methods, structural features, quartz content, and calcite percentage. Key static characteristics such as hardness, grain size, and density are extracted using indentation, sieve analysis, and porosity measurement methods. Simultaneously, the five parameter signals are processed in parallel. In the time domain, filtering is used to reduce noise and remove vibration interference. In the frequency domain, Fourier transform is used to analyze frequency characteristics, extracting the instantaneous fluctuation difference in drill pressure and the duration and stability range of rotation speed fluctuations less than 5%. After correlation analysis and statistical determination, the corresponding parameter thresholds for the three types of lithology (e.g., drilling pressure of 6-10 MPa for sandstone, 10-15 MPa for limestone, and 4-8 MPa for shale) were determined. The rule that "if a single parameter exceeds the specific threshold and does not fall into other intervals, or if three or more parameters fall into the same interval, it points to the corresponding lithology" was formulated. A bidirectional correlation benchmark library of lithology parameters was constructed, which includes classification labels containing lithology names and numbers (e.g., Y001), static and dynamic feature datasets of core samples, and bidirectional mapping rules between lithology and parameters.

[0079] Subsequently, based on this benchmark library, a dynamic sliding window with dynamic window adjustment according to parameter fluctuation law combined with the Savitzky-Golay algorithm was used to smooth and reduce noise in real-time drilling parameter signals while retaining abrupt change characteristics. Drilling specific energy calculated from the ratio of drilling pressure to drilling speed, power-speed ratio derived from the correlation between torque and rotational speed, time-series fluctuation coefficient statistically derived from the time-series dispersion of parameters, and the mean, peak value, and stable duration characteristics of each of the five types of parameters within their respective threshold intervals were extracted. These were then integrated and normalized to the [0, 1] interval in a fixed order of "drilling specific energy - power-speed ratio - time-series fluctuation coefficient - mean value of each parameter - peak value - stable duration" to generate a standardized structured lithological feature vector adapted to the subsequent model input.

[0080] The vector is then input into a hybrid XGBoost and DA-LSTM model pre-trained with historical data. The incremental learning module embedded in the model receives new feature vectors in real time and adjusts the split weights of XGBoost tree nodes and the connection weights of DA-LSTM neurons after comparing the differences in data distribution. XGBoost explores the synergistic effect of drilling specific energy and drilling pressure characteristics, as well as the nonlinear correlation of threshold cross-influence. DA-LSTM captures the dynamic fluctuation amplitude, trend of change, and temporal pattern of abrupt change nodes of parameters with a fixed time step. The two work together to divide the data into segments according to parameter abrupt change nodes and drilling depth, and output lithology prediction results.

[0081] Finally, the standardized logging template is used to load the predicted results. Staff members then review the data segment by segment using drilling logs, core observations, and parameter curves, correcting misjudgment labels and adding reasons. For ambiguous boundary segments, lithology is determined by core structure and mineral analysis, generating a logging report that includes depth segments, lithology determination, parameter ranges, and review notes. Simultaneously, the system collects and corrects the data, sending it back to the benchmark database to supplement samples and adjust parameter thresholds. This triggers incremental learning of the model to optimize weights and benchmark database rule confidence, forming a closed loop of "collection-modeling-prediction-review-feedback-dual optimization," continuously improving logging accuracy and efficiency.

[0082] This embodiment achieves the following through the solution: the accuracy rate of lithology prediction reaches 92%, which is 35% higher than that of traditional methods; the logging time for deep operations is reduced by 60%, reducing personnel safety risks; as drilling progresses, the benchmark library and model are continuously optimized, and the accuracy rate of subsequent borehole lithology prediction gradually increases to over 95%; the logging report has a high degree of standardization and can be directly connected to the subsequent geological exploration data analysis system.

[0083] Example 2 Based on the above-mentioned auxiliary core geological logging method based on drilling parameters, Embodiment 2 of the present invention provides an auxiliary core geological logging system based on drilling parameters, which includes: The data acquisition module is used to acquire typical core samples from the region and five types of drilling parameter signals; The benchmark library construction module is used to extract static characteristics and dynamic laws of parameters from rock cores and establish a two-way correlation benchmark library of lithological parameters. The feature processing module is used to reduce noise and generate structured lithological feature vectors by using a dynamic sliding window combined with the Savitzky-Golay algorithm. The lithology prediction module includes a hybrid model of XGBoost and DA-LSTM and an incremental learning unit, which outputs segmented lithology prediction results. The cataloging and feedback module is used to call templates to generate reports and collect and correct data to update the benchmark library and model.

[0084] This invention constructs a benchmark library that integrates static core features with bidirectional mapping rules for five types of drilling parameters. A dynamic sliding window combined with the Savitzky-Golay algorithm is used for noise reduction and multi-dimensional feature extraction to generate standardized vectors. These vectors are input into a hybrid XGBoost and DA-LSTM model (including an incremental learning module) to output segmented lithology prediction results. After human-computer interaction verification, the corrected data is fed back to update the benchmark library and trigger incremental model learning, forming a closed loop of "data acquisition - model calculation - feedback optimization." This achieves continuous improvement in lithology prediction accuracy and standardization and intelligentization of the logging process.

[0085] In summary, this invention provides an auxiliary core logging method based on drilling parameters, belonging to the field of geological exploration technology. The method is applied to a logging control terminal and mainly includes: the logging control terminal acquiring typical core samples and five types of drilling parameter signals (drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement) from the region via a data acquisition terminal; extracting static features from the core samples; performing parallel noise reduction and analysis on the parameter signals; obtaining the matching relationship between the core samples and the parameter signals; and constructing a bidirectional correlation benchmark library for lithological parameters. Based on the matching relationship of the bidirectional correlation benchmark library, the model embeds an incremental learning module to update weights and outputs segmented lithological prediction results. A preset logging template is called to load the prediction results, which are then reviewed and corrected through human-computer interaction to generate a logging report. Corrected data is collected and sent back to the benchmark library to update the matching relationship, triggering incremental learning of the model and completing the closed loop. This effectively solves the technical problems of poor regional lithological adaptation, numerous parameter interferences, lack of dynamic model optimization, high risk and low efficiency in deep operations in traditional core logging.

[0086] It should be noted that the present invention is not limited to the above-described embodiments. The above embodiments are merely examples, and any embodiments that have the same structure and perform the same effects as the technical concept within the scope of the present invention are included within the scope of the present invention. Furthermore, various modifications that can be conceived by those skilled in the art to the embodiments, and other ways of constructing by combining some of the constituent elements of the embodiments, without departing from the spirit of the present invention, are also included within the scope of the present invention.

Claims

1. A method for auxiliary core geological logging based on drilling parameters, characterized in that, The method is applied to a cataloging control terminal and includes the following steps: S1. Construct a bidirectional correlation benchmark library for lithological parameters: The control terminal acquires typical rock core samples and five types of drilling parameter signals from the acquisition terminal in the region. Static features are extracted from the rock core samples, and the parameter signals are denoised and analyzed in parallel to obtain the matching relationship between the rock cores and the parameter signals, thereby constructing a bidirectional correlation benchmark library for lithological parameters. S2. Generate structured lithological feature vectors: Based on the matching relationship of the bidirectional correlation benchmark library of lithological parameters, a dynamic sliding window combined with the Savitzky-Golay algorithm is used to denoise the parameter signals of real-time drilling, extract the drilling specific energy features, and generate structured lithological feature vectors. S3, Intelligent Lithology Prediction: The structured lithology feature vector is input into the pre-trained XGBoost and DA-LSTM hybrid model. The model embeds an incremental learning module to update the weights and outputs segmented lithology prediction results. S4. Closed-loop feedback optimization: The preset cataloging template is called to load the prediction results. After human-computer interaction for review and correction, a cataloging report is generated. The corrected data is collected and sent back to the benchmark library to update the matching relationship, and the incremental learning of the model is triggered to complete the closed loop.

2. The method for auxiliary core geological logging based on drilling parameters according to claim 1, characterized in that, In step S1, the extraction of static features from the core samples includes: Based on standardized classification according to lithology, structure and mineral composition, key static characteristics such as hardness grade, grain size and density are extracted. The five types of drilling parameter signals include: drilling pressure, rotational speed, drilling speed, torque, and pump pressure / displacement; and the five types of drilling parameter signals are subjected to parallel time-domain noise reduction and frequency-domain analysis to extract the fluctuation amplitude and dynamic patterns of the stable interval. The construction of the bidirectional correlation benchmark library for lithological parameters includes: establishing a bidirectional correspondence between lithology and parameters through correlation analysis, clarifying the parameter threshold range and lithology orientation rules; and constructing a bidirectional correlation benchmark library for lithological parameters containing classification labels, feature data, and mapping rules to provide a reference for subsequent feature extraction and lithology identification.

3. The method for auxiliary core geological logging based on drilling parameters according to claim 2, characterized in that, Step S1 also includes: clarifying the numerical thresholds of the five types of parameters corresponding to sandstone, limestone, and shale respectively, and formulating a judgment standard that points to a specific lithology when a single parameter exceeds the threshold or when three or more parameters are matched. A benchmark library is constructed, which includes classification labels for lithological names and numbers, static characteristics and dynamic parameter data of rock cores, and two-way mapping rules for lithology to parameter ranges and parameter combinations to lithology. This provides a directly callable reference for subsequent feature extraction and lithology matching.

4. The method for auxiliary core geological logging based on drilling parameters according to claim 3, characterized in that, In step S1, the bidirectional lithology mapping rule refers to the specific numerical ranges of five types of parameters corresponding to specific lithologies such as sandstone, limestone, and shale: drilling pressure, rotation speed, drilling speed, torque, and pump pressure / displacement. At the same time, when a single parameter exceeds a threshold or multiple combinations of the five types of parameters are matched, they point to the judgment criteria of the specific lithology. When matching the lithology, the correspondence between the parameter combination and the lithology is directly called to provide a reference for operation.

5. The method for auxiliary core geological logging based on drilling parameters according to claim 1, characterized in that, In step S2, the drilling specific energy feature is extracted to generate the structured lithological feature vector. The specific process is as follows: after denoising the five types of parameter signals of real-time drilling, the drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stable duration features of each parameter are extracted in a targeted manner. According to the mapping rules of the bidirectional correlation benchmark library of lithological parameters, the aforementioned multi-dimensional features are integrated in a fixed dimensional order to generate a structured lithological feature vector containing key lithological correlation indicators. The structured lithological feature vector directly adapts to the input format of the subsequent hybrid recognition model, providing standardized data support for lithological prediction.

6. The method for auxiliary core geological logging based on drilling parameters according to claim 5, characterized in that, In step S2, the multi-dimensional features are integrated in a standardized arrangement according to a fixed dimensional order, which is: drilling specific energy → power-speed ratio → time-series fluctuation coefficient → average drilling pressure → peak drilling pressure → drilling pressure stabilization time → average rotation speed → peak rotation speed → rotation speed stabilization time → average drilling speed → peak drilling speed → drilling speed stabilization time → average torque → peak torque → torque stabilization time → average pump pressure and displacement → peak pump pressure and displacement → pump pressure and displacement stabilization time.

7. The method for auxiliary core geological logging based on drilling parameters according to claim 1, characterized in that, In step S3, the incremental learning module embedded in the pre-trained XGBoost and DA-LSTM hybrid model dynamically adjusts the internal weight parameters in combination with the feature vector corresponding to the real-time drilling data to optimize the adaptability of the lithology identification. The working mechanism of the XGBoost and DA-LSTM hybrid model includes: XGBoost is responsible for extracting key nonlinear correlation features from the feature vector; DA-LSTM captures the temporal variation law of the parameters. After the two are calculated together, they are divided into segments according to the drilling process, and the corresponding lithology prediction results of each segment are output to achieve matching.

8. The method for auxiliary core geological logging based on drilling parameters according to claim 7, characterized in that, In step S3, the key nonlinear correlation features extracted by XGBoost include: nonlinear correlations between multi-dimensional features such as drilling specific energy, power-speed ratio, time-series fluctuation coefficient, and the mean, peak value, and stable duration of each parameter; the nonlinear correlations include the synergistic effect of different features and the non-simple linear correspondence of threshold cross-influence. The DA-LSTM captures temporal variation patterns by: processing feature vectors with a fixed time step; capturing the dynamic fluctuation amplitude, trend, and abrupt change nodes of parameters; and identifying the segmented variation patterns of lithology as drilling progresses.

9. The method for auxiliary core geological logging based on drilling parameters according to claim 1, characterized in that, In step S4, the human-computer interaction verification and correction includes: staff reviewing the prediction results segment by segment by combining drilling logs, core observations, and parameter monitoring curves; correcting the lithology labels for lithology segments with prediction deviations and supplementing explanations of the reasons for the deviations; and determining the final lithology for segments with ambiguous parameter threshold boundaries by combining core structure and mineral composition analysis. The reverse transmission back to the benchmark library to update the matching relationship includes: collecting accurate data after correction, including the final lithology label of each segment, the measured value of the corresponding parameter, and the comparison information before and after correction; supplementing new lithology-parameter corresponding samples to the benchmark library; and adjusting the parameter threshold range in the original matching relationship. The incremental learning of the triggering model includes: using newly added accurate data as supplementary training samples; re-optimizing the decision tree splitting weights of XGBoost and the neuron connection weights of DA-LSTM; and updating the confidence of bidirectional mapping rules in the benchmark library.

10. An auxiliary core geological logging system based on drilling parameters, characterized in that, The core geological logging method based on drilling parameters according to any one of claims 1 to 9 includes: The data acquisition module is used to acquire typical core samples from the region and five types of drilling parameter signals; The benchmark library construction module is used to extract static characteristics and dynamic laws of parameters from rock cores and establish a two-way correlation benchmark library of lithological parameters. The feature processing module is used to reduce noise and generate structured lithological feature vectors by using a dynamic sliding window combined with the Savitzky-Golay algorithm. The lithology prediction module includes a hybrid model of XGBoost and DA-LSTM and an incremental learning unit, which outputs segmented lithology prediction results. The cataloging and feedback module is used to call templates to generate reports and collect and correct data to update the benchmark library and model.

Images