Rock stratum hardness identification method and related equipment
By acquiring rock stratum working condition characteristic data, determining target working condition characteristics, and constructing a multi-task recognition model, the problem of low accuracy in rock stratum hardness recognition was solved, achieving efficient and reliable rock stratum hardness recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- CENT SOUTH UNIV
- Filing Date
- 2026-02-10
- Publication Date
- 2026-04-17
AI Technical Summary
Existing methods for identifying rock hardness suffer from low accuracy. Traditional indoor tests are destructive, time-consuming, and costly. Furthermore, deep learning methods lack stability and robustness in feature selection and task collaboration mechanisms, making it difficult to achieve accurate rock hardness identification.
By acquiring working condition characteristic data of multiple target rock layers, the target working condition characteristics are determined. A multi-task recognition model is used to identify hardness and predict the specific work of breaking. A multi-task recognition model is constructed, and the final sample rock layers are selected through multiple constraints for training to improve the model performance.
It improves the accuracy and efficiency of rock layer hardness identification, reduces data size, enhances the correlation of feature data, improves the reliability and accuracy of prediction results, and adapts to different rock layer conditions.
Smart Images
Figure CN121880862A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of rock layer hardness identification technology, and in particular to a rock layer hardness identification method and related equipment. Background Technology
[0002] Rock hardness is a key mechanical indicator characterizing the ability of rock to resist external mechanical intrusion. It is closely related to equipment selection, parameter optimization, and construction efficiency prediction in engineering projects such as rock drilling, well drilling, and tunnel excavation. Accurate and real-time identification of rock hardness is of great significance for realizing intelligent geotechnical construction, improving energy utilization efficiency, and reducing engineering costs. At present, rock hardness identification methods include contact methods (indoor point load tests, microhardness tester measurements, etc.) and empirical engineering analogy methods. These traditional methods have the following limitations: (1) Indoor tests require on-site sampling and specimen preparation, which is highly destructive, time-consuming, costly, and difficult to achieve continuous drilling identification of strata; (2) Empirical methods usually rely on a single parameter and fail to fully consider the complex nonlinear dynamic relationship between multi-source working condition parameters and rock characteristics during rock drilling, resulting in poor adaptability and low reliability of the identification results in field applications.
[0003] With the development of sensor technology and big data acquisition systems, data-driven deep learning methods have provided a new approach for real-time identification of rock hardness using drilling data. However, existing deep learning-based rock hardness identification methods still have limitations: (1) Feature selection lacks stability. Existing feature selection methods ignore the high noise and non-stationary characteristics of the drilling data itself. The optimal features on different data subsets vary greatly, resulting in a lack of robustness in the model; (2) The task collaboration mechanism is simple. Existing multi-task learning frameworks usually only share parameters at the bottom layer and lack a deep interaction mechanism between tasks, which cannot adaptively adjust the degree of dependence between different tasks; (3) The utilization rate of semi-supervised learning is low. In scenarios where labels are scarce, traditional pseudo-labeling methods have difficulty distinguishing between "difficult samples" and "noisy samples," resulting in low accuracy in rock hardness identification. Summary of the Invention
[0004] This application provides a method and related equipment for identifying rock layer hardness, which can solve the problem of low accuracy in identifying rock layer hardness.
[0005] In a first aspect, embodiments of this application provide a method for identifying the hardness of rock strata, the method comprising: Acquire working condition characteristic data for multiple target rock strata; the working condition characteristic data includes the values of multiple working condition characteristics; Multiple target working condition features were identified from all working condition features; the correlation between the target working condition features and the rock layer hardness was greater than the correlation between all other working condition features and the rock layer hardness; For each target rock layer, based on the values of all target working condition characteristics of the target rock layer, a multi-task recognition model is used to identify the hardness of the target rock layer, and the predicted rock layer hardness and predicted breaking work are obtained. The multi-task recognition model is trained using the predicted rock hardness and predicted fracture specific work of all target rock layers to obtain the trained multi-task recognition model. The trained multi-task recognition model is used to identify the rock strata to be identified, and the hardness recognition result of the rock strata is obtained.
[0006] Optionally, multiple target operating condition characteristics may be determined from all operating condition characteristics, including: For each operating condition feature, based on the value of that operating condition feature, multiple correlation indices are calculated for that operating condition feature, and the overall correlation degree of the operating condition feature is calculated based on all correlation indices. Sort all comprehensive relevance scores from largest to smallest, and use the working condition features corresponding to the top few comprehensive relevance scores in the sorting results as the target working condition features.
[0007] Optionally, calculate the comprehensive correlation of the operating condition characteristics based on all correlation indices, including: Through the formula:
[0008] Calculate the first Comprehensive correlation of individual working condition characteristics ; in, For the first The average score of all correlation indices for each working condition characteristic. Indicates the first The stability variance of each operating condition characteristic This represents the stability penalty coefficient. For smoothing terms, To maximize stability variance, , This represents the total number of operating condition characteristics.
[0009] Optionally, based on the values of all target working condition characteristics of the target rock stratum, a multi-task recognition model is used to identify the hardness of the target rock stratum, obtaining the predicted rock stratum hardness and predicted fracture specific work, including: Using a multi-task recognition model, feature extraction and splicing are performed on the values of all target working condition features of the target rock stratum to obtain spliced features and generate the location code of the target rock stratum; Based on splicing features and location coding, the target rock layer is classified by hardness to obtain the predicted rock layer hardness; Based on splicing features and location coding, the specific energy of the target rock layer is predicted, and the predicted specific energy of the target rock layer is obtained.
[0010] Optionally, the multi-task recognition model is trained using the predicted rock hardness and predicted fracture specific energy of all target rock layers, resulting in a trained multi-task recognition model, including: Based on all predicted rock layer hardness and predicted fracture work, multiple final sample rock layers were determined from all target rock layers. A loss function is constructed based on all the final sample rock layers; The multi-task recognition model is trained using a loss function to obtain the trained multi-task recognition model.
[0011] Optionally, based on all predicted rock layer hardness and predicted fracturing work, multiple final sample rock layers are identified from all target rock layers, including: For each target rock layer, based on the predicted rock layer hardness and predicted fracture specific work, it is determined whether the target rock layer meets the multiple constraint conditions. If so, the target rock layer is used as a final sample rock layer.
[0012] Optional, multiple constraints include uncertainty constraints, physical scope constraints, time consistency constraints, and classification confidence constraints; The uncertainty constraint is: the prediction variance of the predicted rock hardness and the predicted breaking work is less than or equal to the preset variance threshold or the fluctuation standard deviation is less than or equal to the preset standard deviation threshold. The physical range constraint is: the predicted rock layer hardness is within the preset hardness range; The time consistency constraint is: the abrupt change values of the predicted rock layer hardness and the predicted fracture specific work are less than or equal to the preset abrupt change threshold; The classification confidence constraint is: the maximum probability of classification prediction is greater than or equal to the preset probability threshold.
[0013] Optionally, the loss function is:
[0014] in, This represents the value of the loss function. This represents the supervised training loss function constructed based on all final sample rock strata. This represents the unsupervised loss weighting coefficient. This represents the unsupervised training loss function constructed based on all final sample rock layers.
[0015] Secondly, embodiments of this application provide a rock stratum hardness identification device, comprising: The acquisition module is used to acquire working condition characteristic data of multiple target rock strata; the working condition characteristic data includes the values of multiple working condition characteristics. The determination module is used to identify multiple target working condition features from all working condition features; the correlation between the target working condition features and the rock layer hardness is greater than the correlation between all other working condition features and the rock layer hardness; The first identification module is used to identify the hardness of each target rock layer based on the values of all target working condition characteristics of the target rock layer, using a multi-task identification model to obtain the predicted rock layer hardness and the predicted breaking work. The training module is used to train the multi-task recognition model using the predicted rock hardness and predicted fracture work of all target rock layers, so as to obtain the trained multi-task recognition model. The second identification module is used to identify the rock layer to be identified using the trained multi-task identification model, and obtain the hardness identification result of the rock layer to be identified.
[0016] Thirdly, embodiments of this application provide a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aforementioned rock hardness identification method.
[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned rock hardness identification method.
[0018] The above-mentioned solution in this application has the following beneficial effects: In the embodiments of this application, by acquiring the working condition feature data of multiple target rock layers, then determining multiple target working condition features from all working condition features, and then, for each target rock layer, based on the values of all target working condition features of the target rock layer, using a multi-task recognition model to identify the hardness of the target rock layer, obtaining the predicted rock layer hardness and the predicted breaking work. Then, using the predicted rock layer hardness and the predicted breaking work of all target rock layers, the multi-task recognition model is trained to obtain the trained multi-task recognition model. Finally, the trained multi-task recognition model is used to identify the rock layer to be identified, and the hardness identification result of the rock layer to be identified is obtained. Specifically, the target working condition features were identified from the working condition characteristics, and the correlation information between the working condition features and the rock layer hardness was mined. This improved the correlation between feature data and rock layer hardness while reducing the data size, thereby improving the efficiency and accuracy of hardness prediction. Multi-task identification of rock layer hardness and predicted fracture specific work was performed, taking into account the fracture situation of the rock layer while identifying rock layer hardness, thus providing auxiliary explanation of rock layer hardness and improving the reliability of prediction results. Training the multi-task identification model can improve its performance, and using the high-performance model for hardness identification further improves the accuracy of rock layer hardness identification.
[0019] Other beneficial effects of this application will be described in detail in the following detailed description section. Attached Figure Description
[0020] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0021] Figure 1 This is a schematic flowchart of a rock layer hardness identification method provided in an embodiment of this application; Figure 2 A schematic diagram illustrating the comprehensive correlation of operating condition features provided in an embodiment of this application; Figure 3 This is a schematic diagram of a confusion matrix provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of a rock layer hardness identification device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of a terminal device provided in an embodiment of this application. Detailed Implementation
[0022] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.
[0023] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.
[0024] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.
[0025] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."
[0026] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.
[0027] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.
[0028] To address the issue of low accuracy in existing rock stratum hardness identification methods, this application provides an improved rock stratum hardness identification method. This method acquires working condition feature data of multiple target rock strata, then determines multiple target working condition features from all working condition features. For each target rock stratum, based on the values of all target working condition features, a multi-task identification model is used to identify the hardness of the target rock stratum, obtaining predicted rock stratum hardness and predicted breaking work. Then, the multi-task identification model is trained using the predicted rock stratum hardness and predicted breaking work of all target rock strata, resulting in a trained multi-task identification model. Finally, the trained multi-task identification model is used to identify the rock stratum to be identified, obtaining the hardness identification result of the rock stratum to be identified. Specifically, the target working condition features were identified from the working condition characteristics, and the correlation information between the working condition features and the rock layer hardness was mined. This improved the correlation between feature data and rock layer hardness while reducing the data size, thereby improving the efficiency and accuracy of hardness prediction. Multi-task identification of rock layer hardness and predicted fracture specific work was performed, taking into account the fracture situation of the rock layer while identifying rock layer hardness, thus providing auxiliary explanation of rock layer hardness and improving the reliability of prediction results. Training the multi-task identification model can improve its performance, and using the high-performance model for hardness identification further improves the accuracy of rock layer hardness identification.
[0029] The following is an exemplary description of the rock layer hardness identification method provided in this application.
[0030] like Figure 1 As shown, the rock stratum hardness identification method provided in this application includes the following steps: Step 11: Obtain working condition characteristic data of multiple target rock strata.
[0031] The above-mentioned working condition characteristic data includes the values of multiple working condition characteristics. Working condition characteristics are the characteristic parameters of the drilling rig when it is drilling the target rock formation, such as drilling speed-pressure ratio, mechanical specific energy, thermal efficiency index, energy consumption per unit footage, and hydraulic efficiency index.
[0032] In some embodiments of this application, raw parameters of the drilling rig (such as engine speed, engine load rate, hydraulic oil temperature, power head speed, impact pressure, slewing pressure, propulsion pressure, drilling speed, drilling depth, air pressure, etc.) can be collected by devices such as sensors, and then the values of multiple working condition characteristics can be calculated based on the raw parameters.
[0033] For example, the formula for calculating the drilling speed-pressure ratio is:
[0034] In the formula, PPR The drilling speed-pressure ratio, The drilling speed of the drilling rig. For rotational pressure, To push for pressure, To impact pressure.
[0035] The formula for calculating mechanical specific energy is:
[0036] In the formula, MSPE For mechanical specific energy, For engine load rate, Engine displacement. Engine speed, For rotational pressure, To impact pressure, The cross-sectional area of the drill bit. This refers to the drilling speed of the drilling rig.
[0037] The formula for calculating the thermal efficiency index is:
[0038] In the formula, The thermal efficiency index, For engine load rate, Engine displacement. Engine speed, For rotational pressure, To impact pressure, For fuel density, This refers to the engine's instantaneous fuel consumption. This refers to the calorific value of fuel oil.
[0039] The formula for calculating energy consumption per unit advance is:
[0040] In the formula, EC Energy consumption per unit of advance. For fuel density, This refers to the engine's instantaneous fuel consumption. This refers to the calorific value of fuel oil. This refers to the drilling speed of the drilling rig.
[0041] The formula for calculating the hydraulic efficiency index is:
[0042] In the formula, The hydraulic efficiency index. For rotational pressure, To push for pressure, To impact pressure, This refers to the engine oil pressure.
[0043] For example, after obtaining the above data, the data can be preprocessed using algorithms such as the Z-Score anomaly detection algorithm.
[0044] Step 12: Identify multiple target operating condition features from all operating condition features.
[0045] The correlation between the aforementioned target working condition characteristics and rock hardness is greater than the correlation between all other working condition characteristics (i.e., working condition characteristics other than the target working condition characteristics) and rock hardness.
[0046] In some embodiments of this application, the step of determining multiple target operating condition features from all operating condition features includes: The first step is to calculate multiple correlation indices for each operating condition feature based on its value, and then calculate the overall correlation of the operating condition feature based on all the correlation indices.
[0047] The aforementioned correlation indices include Pearson correlation coefficient, Spearman rank correlation coefficient, mutual information, and random forest feature importance indices.
[0048] Specifically, through the formula:
[0049] Calculate the first Comprehensive correlation of individual working condition characteristics .
[0050] in, For the first The average score of all correlation indices for each working condition characteristic. Indicates the first The stability variance of each operating condition characteristic This represents the stability penalty coefficient. For smoothing terms, To maximize stability variance, , This represents the total number of operating condition characteristics.
[0051] For example, the formula for calculating the Pearson correlation coefficient is as follows:
[0052] In the formula, The Pearson correlation coefficient is... , The working condition characteristic parameter variables (i.e., the values of the working condition characteristics and the actual rock layer hardness) have sample means as follows: , .
[0053] The formula for calculating the Spearman rank correlation coefficient is:
[0054] In the formula, The Spearman rank correlation coefficient. For variables rank, For variables rank, This represents the sample size.
[0055] The formula for calculating mutual information is:
[0056] In the formula, For discrete random variables and mutual information content For variables and The joint probability distribution, For variables Independent probability distributions, For variables The independent probability distribution of .
[0057] The importance of features in random forests is measured using a Gini impurity reduction-based method, and its calculation formula is as follows:
[0058] In the formula, Features Importance score The total number of decision trees in the random forest. For the first The set of all non-leaf nodes in a tree. For the node Features used The reduction in Gini impurity resulting from the division.
[0059] The second step is to sort all the comprehensive relevance scores from largest to smallest, and then take the working condition features corresponding to the top few comprehensive relevance scores in the sorting results as the target working condition features.
[0060] For example, the operating condition features corresponding to the top K comprehensive relevance scores are selected as the target operating condition features, where K is a preset value. The comprehensive relevance statistics of the operating condition features are as follows: Figure 2 As shown, Figure 2 The horizontal axis represents the Borda score (i.e., the overall relevance), and the vertical axis represents the operating condition characteristics.
[0061] It should be noted that, in order to improve the reliability of the target operating condition features, multiple rounds of resampling with replacement can be performed, and then the total score of the features in all resampling can be calculated. The target operating condition features can then be determined through the sorting process described above.
[0062] Step 13: For each target rock layer, based on the values of all target working condition characteristics of the target rock layer, use a multi-task recognition model to identify the hardness of the target rock layer and obtain the predicted rock layer hardness and predicted breaking specific work.
[0063] The predicted rock layer hardness mentioned above is the predicted value of the rock layer hardness of the target rock layer, and the predicted breaking energy is the predicted value of the breaking energy of the target rock layer (used to evaluate the ease of rock breaking).
[0064] In some embodiments of this application, the steps of using a multi-task recognition model to identify the hardness of the target rock layer based on the values of all target working condition characteristics of the target rock layer, and obtaining the predicted rock layer hardness and predicted fracture specific work, include: The first step is to use a multi-task recognition model to extract and stitch together the values of all target working condition features of the target rock stratum to obtain stitched features and generate the location code of the target rock stratum.
[0065] For example, the convolutional pyramid algorithm can be used to extract and concatenate the values of all target working condition features of the target rock layer to obtain concatenated features. A gating mechanism can then be used to generate the location code of the target rock layer. For instance, a set of parallel dilated convolutional layers can be used to extract features, with the dilation rate of each branch set to... It also includes a 1x1 convolutional branch to capture instantaneous features. The outputs of all branches are concatenated and fused after causal cropping, as shown in the following formula:
[0066]
[0067] in, Indicates splicing characteristics, A vector representing the values of all target operating condition characteristics.
[0068] Introduce actual borehole depth value Embedded vector And through a gating mechanism and standard sinusoidal position encoding Adaptive fusion to generate depth-aware location codes :
[0069]
[0070] in, This indicates element-wise multiplication.
[0071] The second step is to classify the target rock layer by hardness based on splicing features and location coding to obtain the predicted rock layer hardness.
[0072] For example, a support vector machine or similar model, pre-trained with rock hardness sample data, can be used to process the spliced features and positional codes to obtain the predicted rock hardness. For instance, the spliced features and positional codes are concatenated, and then the spliced result is encoded using a transformer encoder or similar method. The encoded result is projected into hardness classification features and fracture ratio regression features. Then, algorithms such as attention mechanisms and activation functions are used to enhance and fuse the hardness classification features and fracture ratio regression features to obtain enhanced hardness classification features. Finally, a support vector machine or similar model pre-trained with rock hardness sample data is used to process the enhanced hardness classification features to obtain the predicted rock hardness.
[0073] The third step is to predict the specific fracturing energy of the target rock layer based on the splicing features and location coding, and obtain the predicted specific fracturing energy of the target rock layer.
[0074] For example, a pre-trained support vector machine or similar model based on the fracturing strength training data can be used to calculate the splicing features and location encoding to obtain the predicted fracturing strength of the target rock layer. For instance, the fracturing strength regression features can be input into a pre-trained support vector machine or fully connected layer model based on the fracturing strength training data for processing to obtain the predicted fracturing strength and the corresponding uncertainty.
[0075] It is understandable that the above process is the computation process of the multi-task recognition model.
[0076] It should be noted that in multi-task recognition models, an uncertainty gating mechanism can be introduced to establish bidirectional interaction between two tasks. The regression branch includes an internal uncertainty estimator, which outputs the log-variance. Used to calculate the gating coefficient Controlling the weights of regression features on the classification task:
[0077]
[0078] in, It is the Sigmoid activation function. Feature representation for hardness classification task. This represents the feature representation of the broken work regression task.
[0079] Introduce learnable class weight factors before the hardness classification task head. This is used to reweight the rock layers in the loss function to alleviate the problem of uneven rock hardness distribution.
[0080] in, This represents the total number of hardness categories. For real labels, To predict probabilities for the model, It is correlated with the inverse of the number of samples in each category and is fine-tuned during training.
[0081] Step 14: Use the predicted rock hardness and predicted fracture work of all target rock layers to train the multi-task recognition model and obtain the trained multi-task recognition model.
[0082] In some embodiments of this application, the steps of training the multi-task recognition model using the predicted rock hardness and predicted fracture specific work of all target rock layers to obtain the trained multi-task recognition model include: The first step is to identify multiple final sample rock layers from all target rock layers based on all predicted rock layer hardness and predicted fracture specific work.
[0083] Specifically, for each target rock layer, based on the predicted rock layer hardness and predicted fracture specific work, it is determined whether the target rock layer meets the multiple constraint conditions. If so, the target rock layer is used as a final sample rock layer.
[0084] The aforementioned multiple constraints include uncertainty constraints, physical scope constraints, time consistency constraints, and classification confidence constraints.
[0085] The uncertainty constraint is that the predicted variance of the predicted rock hardness and the predicted breaking work is less than or equal to the preset variance threshold or the fluctuation standard deviation is less than or equal to the preset standard deviation threshold.
[0086] The physical range constraint is: the predicted rock layer hardness is within the preset hardness range.
[0087] The time consistency constraint is that the abrupt changes in the predicted rock hardness and predicted fracture work are less than or equal to the preset abrupt change threshold.
[0088] The classification confidence constraint is: the maximum probability of classification prediction is greater than or equal to the preset probability threshold.
[0089] The second step is to construct a loss function based on all the final sample rock layers.
[0090] Specifically, the loss function is:
[0091] in, This represents the value of the loss function. This represents the supervised training loss function constructed based on all final sample rock strata. Indicates as training rounds Dynamically changing unsupervised loss weight coefficients This represents the unsupervised training loss function constructed based on all final sample rock layers. With training rounds To avoid variance collapse, the unsupervised regression loss is linearly increased and uses mean squared error (MSE).
[0092] in, This represents the number of iterations in the supervised training phase. This represents the number of iterations in the semi-supervised training. This is the maximum weight hyperparameter for unlabeled data.
[0093] The third step is to train the multi-task recognition model using the loss function to obtain the trained multi-task recognition model.
[0094] For example, it is determined whether the value of the loss function is less than the preset loss function value. If so, the multi-task recognition model is used as the trained multi-task recognition model. Otherwise, the parameters of the multi-task recognition model are updated, and the steps of using the multi-task recognition model to identify the hardness of the target rock layer and predict the rock layer hardness and the specific work of breakage are returned.
[0095] The performance of the model is evaluated using a validation set. For classification tasks, the evaluation metrics are accuracy, precision, recall, and F1 score. For regression tasks, the evaluation metrics are root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (COP). , Coefficient of Determination).
[0096] The hyperparameters used during training are shown in Table 1: Table 1
[0097] The classification performance of the trained model on the test set is as follows: Figure 3 As shown in the confusion matrix, Figure 3 Each row represents the actual category, each column represents the predicted category, and the values on the diagonal indicate the proportion of correctly classified samples. For example, the accuracy for the "medium-hard" rock category reached 92.63%, indicating that the model has high reliability in identifying high-hardness rock layers. Meanwhile, the model's accuracy, precision, recall, and F1 score for rock layer hardness classification on the test set were 0.9104, 0.9196, 0.9104, and 0.9134, respectively; the RMSE for predicted breaking power was 1.3464 kW·h / m³. 3 MAE is 0.5522 kW·h / m 3 The R² value reached 0.9245, indicating that the model has a strong feature learning ability for identifying rock layer hardness.
[0098] Step 15: Use the trained multi-task recognition model to identify the rock layer to be identified and obtain the hardness recognition result of the rock layer to be identified.
[0099] The rock strata to be identified are those that require hardness identification. The hardness identification results are used to describe the hardness-related data of the rock strata to be identified, such as hardness and specific work of breaking.
[0100] Specifically, the values of multiple target working condition features of the rock stratum to be identified are obtained. Using the trained multi-task recognition model, feature extraction and splicing are performed on the values of all target working condition features to obtain spliced features and generate position codes. Hardness classification is performed based on spliced features and position codes to obtain the predicted rock stratum hardness. The specific energy of the target rock stratum is predicted based on spliced features and position codes to obtain the predicted specific energy of the rock stratum to be identified. The predicted rock stratum hardness and predicted specific energy of the rock stratum are used as the hardness identification results of the rock stratum to be identified.
[0101] It is worth mentioning that the target working condition features were identified from the working condition characteristics, and the correlation information between the working condition features and the rock layer hardness was mined. This improved the correlation between the feature data and the rock layer hardness while reducing the data size, thereby improving the efficiency and accuracy of hardness prediction. Multi-task identification of rock layer hardness and predicted fracture specific work was performed. While identifying rock layer hardness, the fracture situation of the rock layer was considered, which provided auxiliary explanation of rock layer hardness and improved the reliability of the prediction results. Training the multi-task identification model can improve its performance. Using the high-performance model for hardness identification further improves the accuracy of rock layer hardness identification.
[0102] Furthermore, addressing the high noise and non-stationarity of drilling data, this application constructs a feature optimization mechanism. By comprehensively evaluating the mean and variance of feature importance, it effectively eliminates "pseudo-strong features" that perform well under specific distributions but are unstable overall, thus selecting the most discriminative and robust feature subset. Secondly, it utilizes a dilated causal convolutional pyramid to capture multi-scale temporal patterns and introduces an uncertainty-aware task collaboration mechanism. This mechanism dynamically adjusts the influence weight of the prediction uncertainty of the fracture-to-work regression task on the hardness classification task, significantly enhancing the model's adaptability under lithological abrupt changes and marginal geological conditions. Finally, to address the scarcity of labeled samples, a semi-supervised collaborative training strategy based on uncertainty filtering is designed. This strategy filters high-quality pseudo-labels through multiple constraints of physical range, statistical variance, and classification confidence, and effectively utilizes massive amounts of unlabeled data through dynamic weight scheduling, preventing model collapse. In summary, this approach achieves end-to-end enhancement from feature engineering to model optimization, providing high-precision and highly reliable data support for intelligent construction in geotechnical engineering.
[0103] The rock layer hardness identification device provided in this application will be described exemplarily below.
[0104] like Figure 4 As shown, this application embodiment provides a rock layer hardness identification device, the rock layer hardness identification device 400 including: The acquisition module 401 is used to acquire working condition characteristic data of multiple target rock strata; the working condition characteristic data includes the values of multiple working condition characteristics. The determination module 402 is used to determine multiple target working condition features from all working condition features; the correlation between the target working condition features and the rock layer hardness is greater than the correlation between all other working condition features and the rock layer hardness; The first identification module 403 is used to identify the hardness of each target rock layer based on the values of all target working condition characteristics of the target rock layer using a multi-task identification model, and obtain the predicted rock layer hardness and the predicted breaking work. Training module 404 is used to train the multi-task recognition model using the predicted rock layer hardness and predicted fracture specific work of all target rock layers, so as to obtain the trained multi-task recognition model. The second identification module 405 is used to identify the rock layer to be identified using the trained multi-task identification model, and obtain the hardness identification result of the rock layer to be identified.
[0105] It should be noted that the information interaction and execution process between the above-mentioned devices / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.
[0106] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.
[0107] like Figure 5 As shown, an embodiment of this application provides a terminal device, wherein the terminal device D10 of this embodiment includes: at least one processor D100 ( Figure 5 The diagram shows only one processor, a memory D101, and a computer program D102 stored in the memory D101 and executable on the at least one processor D100, wherein the processor D100 executes the computer program D102 to implement the steps in any of the above method embodiments.
[0108] Specifically, when the processor D100 executes the computer program D102, it acquires working condition feature data of multiple target rock layers, then determines multiple target working condition features from all working condition features, and then, for each target rock layer, based on the values of all target working condition features of the target rock layer, uses a multi-task recognition model to identify the hardness of the target rock layer, and obtains the predicted rock layer hardness and predicted breaking specific energy. Then, it uses the predicted rock layer hardness and predicted breaking specific energy of all target rock layers to train the multi-task recognition model, and obtains the trained multi-task recognition model. Finally, it uses the trained multi-task recognition model to identify the rock layer to be identified, and obtains the hardness identification result of the rock layer to be identified. Specifically, the target working condition features were identified from the working condition characteristics, and the correlation information between the working condition features and the rock layer hardness was mined. This improved the correlation between feature data and rock layer hardness while reducing the data size, thereby improving the efficiency and accuracy of hardness prediction. Multi-task identification of rock layer hardness and predicted fracture specific work was performed, taking into account the fracture situation of the rock layer while identifying rock layer hardness, thus providing auxiliary explanation of rock layer hardness and improving the reliability of prediction results. Training the multi-task identification model can improve its performance, and using the high-performance model for hardness identification further improves the accuracy of rock layer hardness identification.
[0109] The processor D100 can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.
[0110] In some embodiments, the memory D101 may be an internal storage unit of the terminal device D10, such as a hard disk or memory of the terminal device D10. In other embodiments, the memory D101 may be an external storage device of the terminal device D10, such as a plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, etc., equipped on the terminal device D10. Furthermore, the memory D101 may include both internal and external storage units of the terminal device D10. The memory D101 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory D101 can also be used to temporarily store data that has been output or will be output.
[0111] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.
[0112] This application provides a computer program product that, when run on a terminal device, enables the terminal device to implement the steps described in the various method embodiments above.
[0113] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to the rock hardness identification method device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, such as a USB flash drive, a portable hard drive, a magnetic disk, or an optical disk.
[0114] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.
[0115] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0116] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention.
Claims
1. A method for identifying the hardness of rock strata, characterized in that, include: Acquire working condition characteristic data of multiple target rock strata; the working condition characteristic data includes the values of multiple working condition characteristics; Multiple target working condition features are identified from all working condition features; the correlation between the target working condition features and the rock layer hardness is greater than the correlation between all other working condition features and the rock layer hardness; For each target rock layer, based on the values of all target working condition characteristics of the target rock layer, a multi-task recognition model is used to identify the hardness of the target rock layer, and the predicted rock layer hardness and predicted breaking work are obtained. The multi-task recognition model is trained using the predicted rock hardness and predicted fracture specific work of all target rock layers to obtain the trained multi-task recognition model. The trained multi-task recognition model is used to identify the rock layer to be identified, and the hardness recognition result of the rock layer to be identified is obtained.
2. The rock stratum hardness identification method according to claim 1, characterized in that, The process of determining multiple target operating condition features from all operating condition features includes: For each operating condition feature, based on the value of the operating condition feature, multiple correlation indices of the operating condition feature are calculated, and the comprehensive correlation of the operating condition feature is calculated based on all correlation indices. Sort all comprehensive relevance scores from largest to smallest, and use the working condition features corresponding to the top few comprehensive relevance scores in the sorting results as the target working condition features.
3. The rock stratum hardness identification method according to claim 2, characterized in that, The calculation of the comprehensive correlation of the operating condition characteristics based on all correlation indices includes: Through the formula: Calculate the first Comprehensive correlation of individual working condition characteristics ; in, For the first The average score of all correlation indices for each working condition characteristic. Indicates the first The stability variance of each operating condition characteristic This represents the stability penalty coefficient. For smoothing terms, To maximize stability variance, , This represents the total number of operating condition characteristics.
4. The rock stratum hardness identification method according to claim 1, characterized in that, Based on the values of all target working condition characteristics of the target rock stratum, a multi-task recognition model is used to identify the hardness of the target rock stratum, resulting in predicted rock stratum hardness and predicted fracture specific work, including: Using the multi-task recognition model, feature extraction and splicing are performed on the values of all target working condition features of the target rock stratum to obtain spliced features and generate the location code of the target rock stratum; Based on the splicing features and the location encoding, the target rock layer is classified by hardness to obtain the predicted rock layer hardness. Based on the splicing features and the location encoding, the specific energy of the target rock layer is predicted to obtain the predicted specific energy of the target rock layer.
5. The rock stratum hardness identification method according to claim 1, characterized in that, The multi-task recognition model is trained using the predicted rock hardness and predicted fracture specific energy of all target rock layers to obtain the trained multi-task recognition model, including: Based on all predicted rock layer hardness and predicted fracture work, multiple final sample rock layers were determined from all target rock layers. A loss function is constructed based on all the final sample rock layers; The multi-task recognition model is trained using the loss function to obtain the trained multi-task recognition model.
6. The rock stratum hardness identification method according to claim 5, characterized in that, Based on all predicted rock layer hardness and predicted fracturing specific work, multiple final sample rock layers are determined from all target rock layers, including: For each target rock layer, based on the predicted rock layer hardness and predicted fracture specific work, it is determined whether the target rock layer meets multiple constraint conditions. If so, the target rock layer is taken as a final sample rock layer.
7. The rock stratum hardness identification method according to claim 6, characterized in that, The multiple constraints include uncertainty constraints, physical scope constraints, time consistency constraints, and classification confidence constraints. The uncertainty constraint is: the prediction variance of the predicted rock hardness and the predicted breaking work is less than or equal to a preset variance threshold or the fluctuation standard deviation is less than or equal to a preset standard deviation threshold. The physical range constraint is: the predicted rock layer hardness is within a preset hardness range; The time consistency constraint is that the abrupt changes in the predicted rock hardness and predicted breaking work are less than or equal to a preset abrupt change threshold. The classification confidence constraint is: the maximum probability of classification prediction is greater than or equal to a preset probability threshold.
8. The method for identifying rock strata hardness according to claim 5, characterized in that, The loss function is: in, This represents the value of the loss function. This represents the supervised training loss function constructed based on all final sample rock strata. This represents the unsupervised loss weighting coefficient. This represents the unsupervised training loss function constructed based on all final sample rock layers.
9. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the rock layer hardness identification method as described in any one of claims 1 to 8.
10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the rock layer hardness identification method as described in any one of claims 1 to 8.