Intelligent logging classification method and product integrating interpolation repair and multi-feature enhancement

Abstract

The invention provides an intelligent logging classification method and product fusing interpolation repair and multi-feature enhancement, relates to the technical field of logging, and aims to solve the problem of logging curve fragment missing caused by complex borehole environment or instrument fault in an interpolation repair mode. The characteristic parameters at the missing depth are accurately complemented, that is, fitting is carried out based on the adjacent second well logging depth and the corresponding characteristic parameters at each first well logging depth (the well logging depth corresponding to the missing segment of the well logging curve), and a complete well logging curve is recovered; meanwhile, the training set is constructed on the basis of the complemented feature parameters, so that training set defects caused by incomplete data are eliminated; furthermore, when the training set is used for training the basic model, the classification model can fully learn the corresponding relation between various characteristic parameters and the stratigraphic position, and the problems of incomplete model training and low prediction accuracy caused by incomplete logging curve in the traditional method are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of well logging technology, and more specifically to an intelligent well logging classification method and product that integrates interpolation repair and multi-feature enhancement. Background Technology

[0002] Well logging curves are the "geological CT" (Computed Tomography) of oil and gas exploration. Through physical characteristic parameters such as deep lateral resistivity (RD), shallow lateral resistivity (RS), natural gamma (GR), sonic transit time (AC), and 4-meter gradient resistivity (R4), these physical characteristic parameters can: quantitatively characterize formation lithology (such as sandstone, carbonate rock, and shale identification); identify fluid properties (distinguish between oil, gas, and water layers); assess reservoir properties (porosity and permeability calculation); and predict geological structures (fractures and fault zones).

[0003] In the training process of classification models used to determine formation locations (such as oil-bearing layers, water-bearing layers, mudstone layers, etc.) based on well logging data, the source of training data is well logging curves. The accuracy and completeness of the well logging curves directly determine the prediction accuracy of the classification model. However, in well logging operations, factors such as complex wellbore environments and instrument malfunctions often result in missing segments of well logging curves, leading to incomplete data in the training set, incomplete training of the classification model, and low prediction accuracy. Summary of the Invention

[0004] The purpose of this application is to provide an intelligent well logging classification method and product that integrates interpolation repair and multi-feature enhancement. It aims to complete the missing segments of well logging curves used to construct the training set of the classification model and add feature interaction enhancement to improve the accuracy of the classification model.

[0005] In a first aspect, embodiments of this application provide an intelligent well logging classification method that integrates interpolation repair and multi-feature enhancement, the method comprising:

[0006] Obtain multiple logging curves from at least one sample well. The logging curves include M logging depths and N characteristic parameters corresponding to each of the M logging depths. The characteristic parameters corresponding to each logging curve are different. M and N are both positive integers, and N is less than or equal to M.

[0007] For each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter is taken as the first logging depth corresponding to the logging curve.

[0008] For each first logging depth in each logging curve, multiple logging depths adjacent to the first logging depth are obtained from the logging curve, and these multiple logging depths adjacent to the first logging depth are used as multiple second logging depths; the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve are fitted to obtain the feature parameters corresponding to the first logging depth;

[0009] A training set is constructed based on the feature parameters corresponding to each first logging depth in each of the logging curves;

[0010] Labels are configured for the feature parameters included in each well logging curve in the training set, and the labels are used to indicate the formation category of the well logging depth corresponding to the feature parameters;

[0011] The basic model is trained using the training set to obtain the classification model, which is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

[0012] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth, the training sample including multiple feature parameters corresponding to the logging depth and the label, and the multiple logging curves including a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity.

[0013] Before training the base model using the training set to obtain the classification model, the method further includes:

[0014] Multiply the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first products;

[0015] For each of the first products, the first product is added to the training samples corresponding to the first product in the training set, wherein the logging depths included in the corresponding training samples are the same as the logging depths corresponding to the first product.

[0016] In some embodiments, the step of fitting multiple second logging depths and feature parameters corresponding to multiple second logging depths in the logging curve to obtain feature parameters corresponding to the first logging depth includes:

[0017] A K-degree polynomial function is constructed. The fixed parameters of the K-degree polynomial function are determined by fitting the K-degree polynomial function with multiple second logging depths and corresponding characteristic parameters of the multiple second logging depths, where K is greater than 1.

[0018] The first logging depth is input into the K-order polynomial function to obtain the logging characteristic parameters corresponding to the first logging depth output by the K-order polynomial function.

[0019] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth, and the training sample includes multiple feature parameters corresponding to the logging depth and the label;

[0020] The depth interval between any two logging depths among the plurality of logging depths is the first depth interval;

[0021] Before training the base model using the training set to obtain the classification model, the method further includes:

[0022] Multiple resampling logging depths are obtained, wherein the multiple resampling logging depths include the multiple logging depths, and the depth interval between any two resampling logging depths is less than the first depth interval;

[0023] The resampling logging depth other than the multiple logging depths is determined as the third logging depth;

[0024] The third logging depth is input into the Kth degree polynomial function to obtain the characteristic parameters corresponding to the third logging depth;

[0025] The label is added to the feature parameters corresponding to the third logging depth to obtain the resampling features;

[0026] The resampled features are added to the training set as new training samples.

[0027] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth, and the training sample includes multiple feature parameters corresponding to the logging depth and the label;

[0028] Before training the base model using the training set to obtain the classification model, the method further includes:

[0029] At least one of the plurality of logging curves is identified as the curve to be expanded, and each characteristic parameter of the curve to be expanded is identified as the parameter to be expanded.

[0030] For each parameter to be expanded, the parameter to be expanded is input into a configured M-degree polynomial function to obtain the expanded parameter, where M is greater than 1;

[0031] For each of the extended parameters, the extended parameter is added to the training sample corresponding to the extended parameter in the training set, where the corresponding training sample is the training sample where the parameter to be extended is located.

[0032] In some implementations, obtaining multiple logging curves from at least one sample well includes:

[0033] Obtain multiple raw curves from the at least one sample well. The raw curves include X logging depths and Y characteristic parameters corresponding to each of the X logging depths. The characteristic parameters corresponding to each raw curve are different. X and Y are both positive integers, and Y is less than or equal to X.

[0034] Evaluate the importance of the characteristic parameters of each of the original curves for distinguishing multiple stratigraphic horizons in the sample block;

[0035] Based on the importance of the characteristic parameters of each of the original curves, a plurality of logging curves are determined from the plurality of original curves.

[0036] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth, the training sample including multiple feature parameters corresponding to the logging depth and the label, and the multiple logging curves including a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity.

[0037] Before training the base model using the training set to obtain the classification model, the method further includes:

[0038] Determine the ratio between the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first ratios;

[0039] For each of the first ratios, the first ratio is added to the training samples corresponding to the first ratio in the training set, wherein the logging depths included in the corresponding training samples are the same as the logging depths corresponding to the first product.

[0040] Secondly, embodiments of this application provide an intelligent well logging classification device that integrates interpolation repair and multi-feature enhancement, the device comprising:

[0041] The acquisition module is used to acquire multiple logging curves of at least one sample well. The logging curves include M logging depths and N characteristic parameters corresponding to each of the M logging depths. The characteristic parameters corresponding to each logging curve are different. M and N are both positive integers, and N is less than or equal to M.

[0042] The first determining module is used to determine, for each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter is taken as the first logging depth corresponding to the logging curve.

[0043] The second determining module is used to obtain multiple logging depths adjacent to the first logging depth from the logging curve for each first logging depth in each logging curve, and to take the multiple logging depths adjacent to the first logging depth as multiple second logging depths; and to fit the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve to obtain the feature parameters corresponding to the first logging depth.

[0044] A construction module is used to construct a training set based on the feature parameters corresponding to each first logging depth in each of the logging curves;

[0045] The configuration module is used to configure labels for the feature parameters included in each logging curve in the training set, wherein the labels are used to indicate the formation category of the logging depth corresponding to the feature parameters;

[0046] The training module is used to train the base model using the training set to obtain the classification model. The classification model is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

[0047] Thirdly, embodiments of this application provide an electronic device, including:

[0048] The memory is configured to store instructions; and

[0049] The processor is configured to retrieve the instructions from the memory and, when executing the instructions, to implement the intelligent well logging classification method with fusion interpolation repair and multi-feature enhancement provided in the first aspect of the embodiments of this application.

[0050] Fourthly, embodiments of this application provide a machine-readable storage medium storing instructions that, when executed by a processor, cause the processor to implement the intelligent well logging classification method based on the first aspect of embodiments of this application, which combines interpolation repair and multi-feature enhancement.

[0051] In this embodiment, by using interpolation repair, the problem of missing logging curve segments caused by complex wellbore environments or instrument malfunctions can be addressed by accurately completing the feature parameters at the missing depth. That is, at each first logging depth (the logging depth corresponding to the missing logging curve segment), the complete logging curve is restored by fitting based on the adjacent second logging depth and its corresponding feature parameters. At the same time, by constructing a training set based on the completed feature parameters, the defects of the training set caused by incomplete data are eliminated. Furthermore, when the basic model is trained using this training set, the classification model can fully learn the correspondence between various feature parameters and formation horizons, thereby improving the model's accuracy in identifying formation horizons. This effectively solves the problems of incomplete model training and low prediction accuracy caused by incomplete logging curves in traditional methods. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the intelligent well logging classification method that integrates interpolation repair and multi-feature enhancement provided in the embodiments of this application.

[0053] Figure 2 This is a schematic diagram of the structure of the intelligent logging classification device that integrates interpolation repair and multi-feature enhancement provided in the embodiments of this application;

[0054] Figure 3 This is a schematic diagram of the structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0056] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0057] The following, in conjunction with the accompanying drawings, provides a detailed description of the intelligent well logging classification method, formation location category acquisition method, and product provided in this application, which integrates interpolation repair and multi-feature enhancement, through specific embodiments and application scenarios.

[0058] Please see Figure 1 This is a flowchart illustrating the intelligent well logging classification method that integrates interpolation repair and multi-feature enhancement provided in this application embodiment. Figure 1 As shown, the intelligent logging classification method that integrates interpolation repair and multi-feature enhancement includes the following steps S100 to S600.

[0059] Step S100: Obtain multiple logging curves from at least one sample well. The logging curves include M logging depths and N logging depths from the M logging depths, each with its own characteristic parameters. The characteristic parameters for each logging curve are different. M and N are both positive integers, and N is less than or equal to M.

[0060] Step S200: For each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter shall be taken as the first logging depth corresponding to the logging curve.

[0061] Step S300: For each first logging depth in each logging curve, obtain multiple logging depths adjacent to the first logging depth from the logging curve, and use the multiple logging depths adjacent to the first logging depth as multiple second logging depths; use the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve to fit, and obtain the feature parameters corresponding to the first logging depth.

[0062] Step S400: Construct a training set based on the feature parameters corresponding to each first logging depth in each logging curve;

[0063] Step S500: Configure labels for the feature parameters included in each logging curve in the training set. The labels are used to indicate the formation category of the logging depth corresponding to the feature parameters.

[0064] Step S600: Train the base model using the training set to obtain a classification model. The classification model is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

[0065] This application proposes an intelligent well logging classification method that integrates interpolation repair and multi-feature enhancement, including: acquiring multiple well logging curves from at least one sample well, wherein the well logging curves include M well logging depths and N feature parameters corresponding to the well logging depths; marking well logging depths without corresponding feature parameters as first well logging depths; extracting adjacent second well logging depths and their feature parameters for each first well logging depth and fitting them to reconstruct the missing feature parameters; constructing a training set based on the reconstructed complete feature parameters and configuring formation layer labels; and training a base model using the complete training set to obtain a classification model, which is used to output the target formation layer category for the target depth based on the acquired feature parameters, thus completing the formation layer classification process.

[0066] The well logging curves refer to the data set measured for different physical characteristics of at least one well. Specifically, this can be achieved using curve sets corresponding to parameters such as deep lateral resistivity, shallow lateral resistivity, and natural gamma ray, with each curve corresponding to a different formation characteristic dimension. The first well logging depth refers to the depth location in a specific well logging curve where a characteristic parameter is missing. This is achieved by identifying missing points through data integrity verification, used to locate the target location for data reconstruction. The second well logging depth refers to the known data points adjacent to the missing point. This can be achieved by selecting two known data points above and below the missing point, using the spatial distribution characteristics of adjacent data to support the calculation of the missing value. The fitting process refers to deriving the missing value based on the mathematical relationships of known data points. This can be achieved by using a quadratic polynomial function to fit curves to adjacent data points, leveraging the predictive power of the mathematical model to ensure reconstruction accuracy. Training set construction refers to integrating the complete characteristic parameters of all well logging curves. This can be achieved by establishing a mapping table between well logging depth and multi-dimensional features, providing structured data support for model training. Label configuration refers to annotating the formation attributes at each logging depth. This can be achieved using expert geological interpretations or core analysis data as supervisory signals, establishing a correspondence between feature parameters and formation categories. Basic model training involves establishing a mapping between features and formation categories through supervised learning. This can be achieved using convolutional neural networks to process multi-dimensional logging features, enabling the model to automatically classify formations based on input features.

[0067] Specifically, during the well logging data acquisition phase, well logging curves containing both complete and missing data are collected simultaneously. For each well logging curve with missing data, the missing point is first located as the first logging depth, and then adjacent second logging depths are selected above and below the missing point. A mathematical relationship model between depth and feature parameters is established by fitting the known feature parameters of the second logging depth to a predefined polynomial function. The missing point depth is then substituted into the fitted model to calculate the reconstructed feature parameter values. The complete dataset of all well logging curves is integrated into a training sample set containing multi-dimensional features and corresponding formation labels. A classification model is trained using a supervised learning algorithm, enabling the model to accurately predict formation categories based on the input feature parameters.

[0068] Compared with existing technologies, this approach accurately reconstructs missing values ​​through fitting, preserving the original data distribution characteristics while increasing the amount of effective training data. Compared with single interpolation methods, fitting can better capture the nonlinear variation of well logging parameters, significantly improving the accuracy of missing value reconstruction and providing a more complete and reliable data foundation for model training.

[0069] Through the above technical solution, this application effectively solves the problem of incomplete training sets caused by missing logging curve data. The missing value reconstruction method based on the principle of spatial correlation significantly improves the completeness of feature parameters, enabling the training set to comprehensively reflect the distribution patterns of formation features. A precise mathematical fitting model ensures the reliability of the reconstructed data and avoids interference from noisy data during model training. The final trained classification model possesses more accurate formation identification capabilities, providing reliable technical support for oil and gas exploration.

[0070] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth. The training samples include multiple feature parameters and labels corresponding to the logging depth. The multiple logging curves include a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity.

[0071] Before training the base model using the training set to obtain the classification model, the method also includes:

[0072] Multiply the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first products;

[0073] For each first product, the first product is added to the training set corresponding to the training sample. The logging depth included in the corresponding training sample is the same as the logging depth corresponding to the first product.

[0074] This application further proposes an operation step before training the basic model using the training set: multiplying the feature parameters of the same logging depth in the first logging curve corresponding to deep lateral resistivity and the second logging curve corresponding to shallow lateral resistivity to generate multiple first products, and adding each first product to the corresponding training sample.

[0075] The first logging curve refers to the logging data sequence obtained through deep lateral resistivity measurement. Specifically, it can be achieved by collecting resistivity values ​​at different depths within the wellbore using logging instruments, reflecting the deep resistivity characteristics of the formation. The second logging curve refers to the logging data sequence obtained through shallow lateral resistivity measurement. Specifically, it can be achieved by collecting data using resistivity measurement devices at different depths, characterizing the shallow resistivity characteristics of the formation. The multiplication operation refers to mathematically multiplying the deep lateral resistivity value and the shallow lateral resistivity value at the same depth point. This can be achieved using numerical multiplication calculations, generating a composite characteristic parameter representing the coordinated change of both values.

[0076] Specifically, before model training, for each logging depth point, the deep lateral resistivity parameter value in the first logging curve and the shallow lateral resistivity parameter value in the second logging curve are extracted. These two parameters are then multiplied to generate a new composite feature parameter. This product feature reflects the nonlinear relationship between deep and shallow resistivity; for example, when the deep resistivity is high and the shallow resistivity is low, the product result can indicate a specific formation fluid distribution pattern. The generated product feature is added as a new dimension to the original training samples, expanding the feature space of each sample to a combination of the original features and the product feature. This feature construction method strengthens the correlation between different logging parameters, enabling the model to learn the indicative pattern of resistivity parameter combinations on formation horizons. In cases where some logging data is missing, the introduction of the product feature can preserve the potential relationships between parameters, improving the model's ability to handle incomplete data through the complementary effects of features.

[0077] Compared to existing technologies, traditional methods typically use raw logging parameters directly as model input without considering the interactions between different parameters. Existing technologies often address missing features through interpolation or deletion, but fail to compensate for information loss by constructing correlated features. This proposed method actively constructs product features of resistivity parameters, revealing the coupling relationship of resistivity responses at different probe depths. This provides richer feature combination information, enabling the model to capture formation characteristic differences that cannot be represented by a single parameter.

[0078] Through the above technical solutions, this application effectively enhances the correlation between well logging curve features and improves the classification model's accuracy in identifying formation horizons. In the presence of missing data, the construction of product features preserves the cooperative variation patterns between parameters, improving the model's adaptability to incomplete data, thereby enhancing the robustness and prediction accuracy of the classification model in practical applications.

[0079] In some implementations, multiple second logging depths and characteristic parameters corresponding to the multiple second logging depths in the logging curve are used for fitting to obtain the characteristic parameters corresponding to the first logging depth, including:

[0080] Construct a K-degree polynomial function. The fixed parameters of the K-degree polynomial function are determined by fitting the K-degree polynomial function with multiple second logging depths and the corresponding characteristic parameters of the multiple second logging depths, where K is greater than 1.

[0081] Input the first logging depth into a K-order polynomial function to obtain the logging characteristic parameters corresponding to the first logging depth output by the K-order polynomial function.

[0082] This application further proposes to fit multiple second logging depths and the characteristic parameters corresponding to the multiple second logging depths in the logging curve to obtain the characteristic parameters corresponding to the first logging depth, including: constructing a K-order polynomial function, wherein the fixed parameters of the K-order polynomial function are determined by fitting the K-order polynomial function with multiple second logging depths and the characteristic parameters corresponding to the multiple second logging depths, where K is greater than 1; inputting the first logging depth into the K-order polynomial function to obtain the logging characteristic parameters corresponding to the first logging depth output by the K-order polynomial function.

[0083] Here, the K-order polynomial function refers to a nonlinear relationship model with logging depth as the independent variable and characteristic parameters as the dependent variable. Specifically, the polynomial coefficients can be determined using the least squares method, and its order K can be 3 or 4 to accommodate nonlinear variations in formation parameters. The second logging depth refers to the known data points adjacent to the missing data points. Specifically, it can be achieved by using two adjacent depth points above and below as input data, ensuring the continuity of the interpolation results through the spatial correlation of adjacent points. The function's fixed parameters refer to the combination of coefficients determined through fitting, which can be achieved by solving the normal equation system or using gradient descent optimization algorithms to ensure that the polynomial curve passes the constraints of the known data points.

[0084] Specifically, when a depth point in the well logging curve contains missing characteristic parameters, multiple known depth points adjacent to the missing point are selected as the data source for fitting. For example, two depth points above and below the missing point are selected, forming a sample set containing four known data points. Based on the depth values ​​and corresponding characteristic parameters of these sample points, a cubic polynomial function is constructed and its coefficients are solved. By substituting the depth value of the missing point into the polynomial function with determined coefficients, the predicted characteristic parameter value for that point is calculated. This process utilizes the vertical continuity of formation parameters and captures local variation trends through high-order polynomials, reflecting actual geological patterns better than simple linear interpolation.

[0085] Compared to existing technologies, traditional methods often use linear interpolation between adjacent points or moving averages to handle missing data, failing to accurately reflect the nonlinear variation characteristics of formation parameters. This proposed method, by constructing a high-order polynomial function, can effectively fit the complex vertical fluctuations of logging parameters, such as the abrupt changes in sonic transit time at the sandstone-mudstone interface. Compared to fixed-order interpolation methods, this method can automatically determine the optimal fitting curve based on the actual data distribution, avoiding subjective errors caused by manually setting the interpolation method.

[0086] Through the above technical solution, this application can accurately reconstruct missing data based on the local variation patterns of well logging parameters. The generated supplementary data conforms to the changing trends of formation physical characteristics, effectively improving the completeness and consistency of the training dataset. This method solves the problem of insufficient model training caused by missing data, providing high-quality training samples for subsequent classification models, thereby significantly improving the accuracy of formation horizon identification.

[0087] In some implementations, the training set includes multiple training samples, each corresponding to a logging depth, and the training samples include multiple feature parameters corresponding to the logging depth and labels;

[0088] The depth interval between any two logging depths in a plurality of logging depths is the first depth interval;

[0089] Before training the base model using the training set to obtain the classification model, the method also includes:

[0090] Multiple resampling logging depths are obtained, which include multiple logging depths, and the depth interval between any two resampling logging depths is less than the first depth interval.

[0091] The resampling depth other than the multiple resampling depths is determined as the third logging depth;

[0092] Input the third logging depth into a K-order polynomial function to obtain the characteristic parameters corresponding to the third logging depth;

[0093] Labels are added to the feature parameters corresponding to the third logging depth to obtain resampling features;

[0094] The resampled features are added to the training set as new training samples.

[0095] This application further proposes to obtain multiple resampled logging depths before training the basic model using the training set, wherein the multiple resampled logging depths include the original logging depths and the interval between any two resampled logging depths is smaller than the original interval; the newly added resampled logging depth is determined as the third logging depth; the feature parameters corresponding to the third logging depth are calculated using a K-order polynomial function; and the calculated results are labeled and added to the training set as new training samples.

[0096] Resampling logging depth refers to inserting higher-density virtual logging points within the original logging depth interval. This can be achieved using linear interpolation or proportional segmentation methods to generate a denser data distribution. Third logging depth refers to newly added virtual logging points through resampling, specifically defined as intermediate points inserted between adjacent original logging depths to fill in the original data gaps. A K-order polynomial function is a mathematical fitting model established based on the characteristic parameters of adjacent logging depths. This can be implemented using quadratic or cubic polynomial functions to ensure the mathematical continuity and geological consistency of the new data points. Label inheritance refers to labeling the formation category corresponding to the newly added logging depth with the category of the adjacent original logging depths. This can be achieved using nearest neighbor matching or weighted averaging methods to maintain the spatial correlation of formation physical properties.

[0097] Specifically, when the original logging depth intervals are large, a higher-density resampled logging depth set is constructed by inserting virtual logging points. For each newly added third logging depth, a pre-established K-order polynomial function is used to perform mathematical interpolation based on the feature parameters of adjacent original logging depths, generating feature parameters that conform to geological patterns. For example, when the original logging depth interval is 1 meter, a virtual point with a depth of 0.5 meters can be inserted between every two adjacent depth points. By combining the interpolated feature parameters with the corresponding formation category labels, new training samples are formed, increasing the data density of the training set. For example, a cubic polynomial function is used to fit the resistivity data of five adjacent original logging depths to calculate the resistivity value of the intermediate point. Thus, the model training can acquire richer feature distribution information, improving the underfitting problem caused by the sparsity of the original data.

[0098] Compared to existing technologies, traditional methods directly use raw well logging data for training, which makes it difficult to capture the continuous variation of formation characteristics when the well logging depth interval is too large. Common data augmentation methods in existing technologies often employ random noise addition or sample duplication, which may destroy the physical meaning of the well logging curves. This solution generates virtual well logging data through a deterministic interpolation method based on a mathematical model, maintaining the continuity of formation characteristics while avoiding interference from artificial noise. For example, compared to random interpolation methods, polynomial function interpolation can better reflect the nonlinear characteristics of resistivity variation with depth, ensuring the geological rationality of the newly added data.

[0099] By employing the aforementioned technical solution, this application effectively alleviates the problem of insufficient training samples caused by excessively large sampling intervals in well logging data. Through mathematical interpolation, virtual data points conforming to geological laws are generated, significantly improving the completeness and density of training data without increasing the actual cost of well logging operations. This solution improves the classification model's ability to capture continuous changes in formation characteristics, enhances the model's generalization performance under different geological conditions, and avoids the risk of overfitting caused by simple data amplification. For example, in carbonate formations, the virtual data generated by this method can accurately reflect the gradual change characteristics of resistivity in the dissolution zone, enabling the model to more accurately identify the reservoir-non-reservoir interface.

[0100] In some implementations, the training set includes multiple training samples, each corresponding to a logging depth, and the training samples include multiple feature parameters corresponding to the logging depth and labels;

[0101] Before training the base model using the training set to obtain the classification model, the method also includes:

[0102] At least one logging curve among multiple logging curves is identified as the curve to be expanded, and each characteristic parameter of the curve to be expanded is identified as the parameter to be expanded.

[0103] For each parameter to be expanded, input the parameter to be expanded into the configured M-degree polynomial function to obtain the expanded parameter, where M is greater than 1;

[0104] For each expansion parameter, the expansion parameter is added to the training set along with the corresponding training samples. The corresponding training samples are the training samples where the expansion parameter is located.

[0105] This application further proposes that after constructing the training set, at least one well logging curve among multiple well logging curves is identified as the curve to be expanded, and each feature parameter of the curve to be expanded is identified as the parameter to be expanded; for each parameter to be expanded, the parameter to be expanded is input into a configured M-degree polynomial function to obtain the expanded parameter, where M is greater than 1; for each expanded parameter, the expanded parameter is added to the training sample corresponding to the expanded parameter in the training set, and the corresponding training sample is the training sample where the expanded parameter is located.

[0106] Among them, the curve to be expanded refers to the logging curve whose characteristic parameters need to be enhanced. Specifically, it can be selected by evaluating the importance of the curve characteristics to the formation horizon classification, such as selecting deep lateral resistivity curves or natural gamma curves. This feature is used to specifically enhance the information content of key features and avoid ineffective expansion.

[0107] Among them, the M-degree polynomial function refers to a mathematical tool that generates extended parameters through nonlinear transformations. Specifically, it can be a quadratic or cubic polynomial function, for example, extending the original parameter x to xm. 2 x 3 This feature enhances the model's ability to recognize complex patterns by introducing higher-order terms to uncover nonlinear relationships between features.

[0108] The extended parameters refer to new feature parameters generated through polynomial transformations, such as converting the original resistivity parameter into square or cubic terms. This feature can increase the dimensionality of the training samples, enrich the data distribution, and alleviate the underfitting problem caused by data sparsity.

[0109] Specifically, before model training, curves that contribute significantly to formation horizon classification are selected from multiple well logging curves as curves to be expanded, such as deep lateral resistivity curves. For each feature parameter of this curve, it is input into a predefined quadratic polynomial function to generate combined features containing the original parameters and their squared terms. For example, if the original resistivity parameter is R, then after expansion, R and R' are obtained. 2 Two parameters are then used. The generated extended parameters are then added to the original training samples, making each sample contain more dimensional feature information. In this way, the model can learn the nonlinear correlations between different logging parameters, such as the quadratic trend of resistivity with depth, thereby more accurately identifying formation boundaries.

[0110] Compared to existing technologies, traditional methods typically only fill in missing data using linear interpolation or mean shift, without considering the potential nonlinear relationships between features. Our proposed solution, however, actively expands existing features using polynomial functions, not only increasing the amount of data but also enhancing the feature representation capabilities. For example, existing technologies may only retain the original resistivity parameters, while our solution introduces an additional squared term, enabling the model to capture the gradient changes in resistivity, thereby improving classification accuracy.

[0111] Through the above technical solution, this application can effectively expand the feature dimension of the training samples and solve the problem of insufficient model training caused by sparse logging curve data. The high-order features generated by polynomial transformation can reveal the complex correlation between different logging parameters, such as the interaction between resistivity and sonic transit time, thereby helping the classification model to more accurately classify formations such as oil-bearing layers and water-bearing layers.

[0112] In some implementations, multiple logging curves are obtained from at least one sample well, including:

[0113] Obtain multiple raw curves from at least one sample well. Each raw curve includes X logging depths and Y logging depths from the X logging depths, each with its own characteristic parameters. The characteristic parameters for each raw curve are different. X and Y are both positive integers, and Y is less than or equal to X.

[0114] Evaluate the importance of the characteristic parameters of each original curve in distinguishing multiple stratigraphic horizons in the sample block;

[0115] Multiple logging curves are determined from multiple original curves based on the importance of the characteristic parameters of each original curve.

[0116] This application further proposes to obtain multiple raw curves for a sample block, the raw curves including X logging depths and Y logging depths among the X logging depths, each raw curve having different characteristic parameters; to evaluate the importance of the characteristic parameters of each raw curve for distinguishing multiple formation horizons in the sample block; and to determine multiple logging curves from the multiple raw curves based on the importance of the characteristic parameters of each raw curve.

[0117] The original curves refer to the initial set of well logging data collected from the sample block. Each original curve contains X well logging depths, but only some depths have corresponding feature parameters. This can be achieved by using well logging instruments to collect data at multiple depth points, providing basic data for subsequent feature selection. Importance assessment refers to measuring the contribution of different feature parameters to formation classification using quantitative indicators. This can be achieved by using a random forest algorithm to calculate feature importance scores, used to identify physical parameters that play a key role in classification decisions. Well logging curve determination refers to selecting the original curves corresponding to key feature parameters from multiple original curves based on the importance assessment results. This can be achieved by setting an importance threshold or selecting the top K features by importance ranking, used to construct a high-value-density training dataset.

[0118] Specifically, after acquiring the raw logging data, the complete logging depth information of all raw curves is first preserved to ensure data integrity. Then, a random forest algorithm is used to evaluate the importance of the feature parameters of each raw curve, calculating the contribution of each feature parameter to distinguishing different formation layers. For example, deep lateral resistivity may be highly important for identifying oil-bearing layers, while natural gamma may contribute significantly to distinguishing mudstone layers. Based on the evaluation results, logging curves corresponding to feature parameters with importance exceeding a preset threshold are selected; for example, the top five most important feature parameters are retained, and the raw curves corresponding to these parameters are used as the logging curves for training. Thus, redundant or low-contribution features are automatically removed, effectively reducing data dimensionality.

[0119] Compared to existing technologies, traditional methods typically use all original curves directly for model training, failing to consider the varying contributions of different feature parameters to the classification results. This leads to low training efficiency and susceptibility to noise interference. This proposed solution addresses the data dimensionality expansion problem by employing a dynamic feature selection mechanism that removes redundant data while retaining core classification features.

[0120] Through the above technical solutions, this application can significantly reduce the computational complexity during the classification model training process, improve the model convergence speed, and enhance the model's ability to distinguish stratigraphic positions by focusing on key feature parameters, thus avoiding prediction bias caused by irrelevant features. The data screening process is optimized while ensuring the integrity of the original data, ensuring that the training set is both concise and has high information density.

[0121] In some implementations, the training set includes multiple training samples, each training sample corresponding to a logging depth. The training samples include multiple feature parameters and labels corresponding to the logging depth. The multiple logging curves include a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity.

[0122] Before training the base model using the training set to obtain the classification model, the method also includes:

[0123] Determine the ratio between the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first ratios;

[0124] For each first ratio, the first ratio is added to the training sample corresponding to the first ratio in the training set. The logging depth included in the corresponding training sample is the same as the logging depth corresponding to the first product.

[0125] This application further proposes to determine the ratio between the logging feature parameters corresponding to the same logging depth in the first logging curve and the second logging curve before training the basic model using the training set, thereby obtaining multiple first ratios, and adding each first ratio to the corresponding training sample.

[0126] The first logging curve refers to the set of logging data corresponding to deep lateral resistivity, which can be achieved by using a sequence of resistivity measurements collected by logging instruments, with each logging depth corresponding to a deep lateral resistivity value. The second logging curve refers to the set of logging data corresponding to shallow lateral resistivity, which can be achieved by using a sequence of values ​​obtained by resistivity measuring instruments at different detection depths. The first ratio refers to the ratio of characteristic parameters of deep lateral resistivity to shallow lateral resistivity at the same logging depth, which can be achieved by dividing the values ​​of the two logging curves at the same depth. This ratio reflects the relative change of formation resistivity at different detection depths.

[0127] Specifically, when data is missing at logging depths, directly using the original logging parameters may result in insufficient feature information in the training samples. By calculating the ratio of deep lateral resistivity to shallow lateral resistivity, new correlated feature parameters are generated. This ratio characteristic can characterize the vertical differences in formation resistivity; for example, in oil-bearing intervals, the ratio of deep to shallow resistivity typically exhibits a specific trend. Adding this ratio feature to the training samples allows the model to simultaneously learn the original parameters and their correlations, enhancing its ability to identify formation features. During data processing, the ratio of the two resistivity parameters is calculated simultaneously for each effective logging depth, and the calculation results are used as a new feature dimension, together with the original feature parameters, to form the training samples.

[0128] Compared to existing technologies, traditional methods typically process missing data from each logging curve individually, without considering the correlation characteristics between different physical parameters. Existing methods for missing data interpolation only restore single parameters, failing to supplement combined features reflecting formation characteristics. This approach introduces resistivity ratio features to construct correlation parameters across logging curves at the data level. This allows the model to capture the indicative effect of resistivity differences at different depths on formation horizons. This feature construction method overcomes the limitations of single-parameter interpolation.

[0129] Through the above technical solution, this application effectively solves the problem of insufficient training set features caused by missing well logging curve data. By generating deep and shallow resistivity ratio features, not only is the number of feature dimensions increased, but more importantly, combined features reflecting the correlation of formation properties are introduced. This feature expansion method enhances the model's ability to distinguish complex formation features. Especially when some original parameters are missing, the ratio features can provide additional discriminative information, thereby improving the accuracy of the classification model in predicting formation horizon categories.

[0130] Please see Figure 2 This is a schematic diagram of the structure of the intelligent well logging classification device that integrates interpolation repair and multi-feature enhancement provided in the embodiments of this application. A second aspect of the embodiments of this application provides an intelligent well logging classification device 10 that integrates interpolation repair and multi-feature enhancement. The device 10 includes:

[0131] The acquisition module 11 is used to acquire multiple logging curves of at least one sample well. The logging curves include M logging depths and N logging depths among the M logging depths, each corresponding to a characteristic parameter. The characteristic parameters corresponding to each logging curve are different. M and N are both positive integers, and N is less than or equal to M.

[0132] The first determining module 12 is used to determine, for each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter is taken as the first logging depth corresponding to the logging curve.

[0133] The second determining module 13 is used to obtain multiple logging depths adjacent to the first logging depth from the logging curve for each first logging depth in each logging curve, and to take the multiple logging depths adjacent to the first logging depth as multiple second logging depths; and to fit the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve to obtain the feature parameters corresponding to the first logging depth.

[0134] Module 14 is used to construct a training set based on the feature parameters corresponding to each first logging depth in each logging curve;

[0135] Configuration module 15 is used to configure labels for the feature parameters included in each logging curve in the training set. The labels are used to indicate the formation category of the logging depth corresponding to the feature parameters.

[0136] Training module 16 is used to train the base model using the training set to obtain a classification model. The classification model is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

[0137] The intelligent logging classification device 10 with fusion interpolation repair and multi-feature enhancement provided in the second aspect of this application can realize the various processes implemented in the above method embodiments and achieve the same beneficial effects. To avoid repetition, it will not be described again here.

[0138] Please see Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. A third aspect of this application provides an electronic device 1000, including a processor 1100 and a memory 1200. The memory 1200 stores machine-executable instructions that can be executed by the processor 1100. The processor 1100 can execute the machine-executable instructions to implement the above-mentioned intelligent well logging classification method of fusion interpolation repair and multi-feature enhancement.

[0139] The fourth aspect of this application provides a machine-readable storage medium storing instructions that, when executed by a processor, cause the processor to implement the aforementioned intelligent well logging classification method with fusion interpolation repair and multi-feature enhancement.

[0140] In some embodiments, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the intelligent well logging classification method based on the fusion interpolation repair and multi-feature enhancement described in the above embodiments.

[0141] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0142] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 The computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable apparatus for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0143] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0144] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0145] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include temporary computer-readable media (transistor silicon switching devices), such as modulated data signals and carrier waves.

[0146] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0147] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

[0148] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. An intelligent well logging classification method integrating interpolation repair and multi-feature enhancement, characterized in that, The method includes: Obtain multiple logging curves from at least one sample well. The logging curves include M logging depths and N characteristic parameters corresponding to each of the M logging depths. The characteristic parameters corresponding to each logging curve are different. M and N are both positive integers, and N is less than or equal to M. For each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter is taken as the first logging depth corresponding to the logging curve. For each first logging depth in each logging curve, multiple logging depths adjacent to the first logging depth are obtained from the logging curve, and these multiple logging depths adjacent to the first logging depth are used as multiple second logging depths; the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve are fitted to obtain the feature parameters corresponding to the first logging depth; A training set is constructed based on the feature parameters corresponding to each first logging depth in each of the logging curves; Labels are configured for the feature parameters included in each well logging curve in the training set, and the labels are used to indicate the formation category of the well logging depth corresponding to the feature parameters; The basic model is trained using the training set to obtain the classification model, which is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

2. The method according to claim 1, characterized in that, The training set includes multiple training samples, each training sample corresponding to a logging depth. The training sample includes multiple feature parameters corresponding to the logging depth and the label. The multiple logging curves include a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity. Before training the base model using the training set to obtain the classification model, the method further includes: Multiply the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first products; For each of the first products, the first product is added to the training samples corresponding to the first product in the training set, wherein the logging depths included in the corresponding training samples are the same as the logging depths corresponding to the first product.

3. The method according to claim 1, characterized in that, The step of fitting multiple second logging depths and feature parameters corresponding to multiple second logging depths in the logging curve to obtain feature parameters corresponding to the first logging depth includes: A K-degree polynomial function is constructed. The fixed parameters of the K-degree polynomial function are determined by fitting the K-degree polynomial function with multiple second logging depths and corresponding characteristic parameters of the multiple second logging depths, where K is greater than 1. The first logging depth is input into the K-order polynomial function to obtain the logging characteristic parameters corresponding to the first logging depth output by the K-order polynomial function.

4. The method according to claim 3, characterized in that, The training set includes multiple training samples, each training sample corresponding to a logging depth, and the training sample includes multiple feature parameters corresponding to the logging depth and the label; The depth interval between any two logging depths among the plurality of logging depths is the first depth interval; Before training the base model using the training set to obtain the classification model, the method further includes: Multiple resampling logging depths are obtained, wherein the multiple resampling logging depths include the multiple logging depths, and the depth interval between any two resampling logging depths is less than the first depth interval; The resampling logging depth other than the multiple logging depths is determined as the third logging depth; The third logging depth is input into the Kth degree polynomial function to obtain the characteristic parameters corresponding to the third logging depth; The label is added to the feature parameters corresponding to the third logging depth to obtain the resampling features; The resampled features are added to the training set as new training samples.

5. The method according to claim 1, characterized in that, The training set includes multiple training samples, each training sample corresponding to a logging depth, and the training sample includes multiple feature parameters corresponding to the logging depth and the label; Before training the base model using the training set to obtain the classification model, the method further includes: At least one of the plurality of logging curves is identified as the curve to be expanded, and each characteristic parameter of the curve to be expanded is identified as the parameter to be expanded. For each parameter to be expanded, the parameter to be expanded is input into a configured M-degree polynomial function to obtain the expanded parameter, where M is greater than 1; For each of the extended parameters, the extended parameter is added to the training sample corresponding to the extended parameter in the training set, where the corresponding training sample is the training sample where the parameter to be extended is located.

6. The method according to claim 1, characterized in that, The acquisition of multiple logging curves from at least one sample well includes: Obtain multiple raw curves from the at least one sample well. The raw curves include X logging depths and Y characteristic parameters corresponding to each of the X logging depths. The characteristic parameters corresponding to each raw curve are different. X and Y are both positive integers, and Y is less than or equal to X. Evaluate the importance of the characteristic parameters of each of the original curves for distinguishing multiple stratigraphic horizons in the sample block; Based on the importance of the characteristic parameters of each of the original curves, a plurality of logging curves are determined from the plurality of original curves.

7. The method according to claim 1, characterized in that, The training set includes multiple training samples, each training sample corresponding to a logging depth. The training sample includes multiple feature parameters corresponding to the logging depth and the label. The multiple logging curves include a first logging curve corresponding to deep lateral resistivity and a second logging curve corresponding to shallow lateral resistivity. Before training the base model using the training set to obtain the classification model, the method further includes: Determine the ratio between the logging characteristic parameters corresponding to the same logging depth in the first logging curve and the second logging curve to obtain multiple first ratios; For each of the first ratios, the first ratio is added to the training samples corresponding to the first ratio in the training set, wherein the logging depths included in the corresponding training samples are the same as the logging depths corresponding to the first product.

8. An intelligent well logging classification device integrating interpolation repair and multi-feature enhancement, characterized in that, The device includes: The acquisition module is used to acquire multiple logging curves of at least one sample well. The logging curves include M logging depths and N characteristic parameters corresponding to each of the M logging depths. The characteristic parameters corresponding to each logging curve are different. M and N are both positive integers, and N is less than or equal to M. The first determining module is used to determine, for each logging curve, if there is a logging depth in the logging curve that does not correspond to a characteristic parameter, then the logging depth that does not correspond to a characteristic parameter is taken as the first logging depth corresponding to the logging curve. The second determining module is used to obtain multiple logging depths adjacent to the first logging depth from the logging curve for each first logging depth in each logging curve, and to take the multiple logging depths adjacent to the first logging depth as multiple second logging depths; and to fit the multiple second logging depths and the feature parameters corresponding to the multiple second logging depths in the logging curve to obtain the feature parameters corresponding to the first logging depth. A construction module is used to construct a training set based on the feature parameters corresponding to each first logging depth in each of the logging curves; The configuration module is used to configure labels for the feature parameters included in each logging curve in the training set, wherein the labels are used to indicate the formation category of the logging depth corresponding to the feature parameters; The training module is used to train the base model using the training set to obtain the classification model. The classification model is used to output the target stratigraphic category corresponding to the target depth based on the feature parameters corresponding to the input target depth.

9. An electronic device, characterized in that, It includes a processor and a memory, the memory storing programs or instructions that can run on the processor, the programs or instructions being executed by the processor to implement the intelligent logging classification method of fusion interpolation repair and multi-feature enhancement as described in any one of claims 1-7.

10. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores instructions that, when executed by a processor, cause the processor to implement the intelligent logging classification method of fusion interpolation repair and multi-feature enhancement as described in any one of claims 1-7.

Images