Rock mechanical property characterization method and equipment based on indentation test and machine learning
By using indentation testing and machine learning methods, a rock mechanical parameter evaluation model was established. A handheld indentation tester was used to quickly obtain basic rock mechanical parameters without core sampling, which solved the problems of speed and accuracy in rock mechanical performance evaluation in existing technologies and met the safety requirements of deep resource mining.
Patent Information
- Application Number
- CN202511672897.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-14
- Publication Date
- 2026-02-13
AI Technical Summary
Existing methods for evaluating rock mechanical properties have long testing cycles, high costs, and are difficult to meet the need for rapid acquisition. In particular, in deep resource mining, rock cores are often caked up, making it difficult to prepare standard samples on-site, which affects engineering design and safety.
Based on indentation testing and machine learning methods, an integrated database and evaluation model are established by leveraging the high correlation between indentation index and basic rock mechanical properties. This allows for the rapid acquisition of basic rock mechanical parameters using a handheld indentation tester without the need for core sampling.
It enables rapid, accurate, and repeatable on-site assessment of rock mechanical properties, providing reliable predicted values and ranges to support mine safety construction and engineering design.
Smart Images
Figure CN121521608A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of rock mechanics technology, and in particular to a method and device for characterizing rock mechanical properties based on indentation testing and machine learning. Background Technology
[0002] Accurate assessment of rock mechanical properties is crucial for oil development, deep underground mining, deep surrounding rock stability analysis, and support design in underground space engineering. Commonly used rock mechanics testing methods include triaxial compression testing, uniaxial compression testing, Brazilian splitting test, and rock fracture testing. As a complex heterogeneous material, the mechanical properties of rock are influenced by various factors, such as the distribution of microcracks within the rock, mineral composition, and external loading conditions. During deep resource extraction, high stress concentration is significant, core fragmentation is severe, and obtaining standard samples in the field is difficult. Furthermore, these laboratory testing methods have long testing cycles, high costs, and are prone to parameter updates, often failing to meet the need for rapid acquisition of rock mechanical properties in practical engineering, thus affecting design and safety. Therefore, developing rapid, accurate, and repeatable methods for assessing rock mechanical properties is of great significance for rock mechanics engineering construction. Summary of the Invention
[0003] The purpose of this invention is to overcome the limitations and shortcomings of existing traditional rock mechanical performance evaluation methods and provide a rock mechanical property characterization scheme based on indentation testing and machine learning. Based on the high correlation between the indentation index and the basic rock mechanical properties, this scheme can quickly obtain the "indentation index" that can be mapped to multiple basic rock mechanical properties in a short period of time (tens of seconds) using a handheld indentation tester embedded with the corresponding model, without the need for core sampling. It also provides on-site predictive values and confidence intervals, achieving rapid, accurate, and repeatable on-site rock mechanical performance evaluation.
[0004] To achieve the above objectives, this invention provides a method for characterizing the mechanical properties of rocks based on indentation testing and machine learning, comprising the following steps:
[0005] S1, to conduct basic mechanical tests and indentation tests on rocks of various lithologies and types to obtain the basic mechanical parameters and indentation index of the rocks;
[0006] Basic mechanical parameters include uniaxial compressive strength Brazil splitting strength Elastic modulus and fracture toughness Indentation index includes the first peak indentation force of the indentation test. Maximum indentation force Standard indentation hardness Indentation modulus Energy consumption per unit displacement ,acting Specific energy ;
[0007] S2, Establish an integrated indentation-mechanical parameter database that includes basic rock mechanical parameters and indentation index;
[0008] S3. Correlation analysis is conducted between rock indentation index and basic mechanical parameters using correlation analysis functions. A rock basic mechanical parameter evaluation model based on indentation index is established through machine learning.
[0009] S4 is an integrated handheld indentation tester. The handheld indentation tester is embedded in the rock basic mechanical parameter evaluation model in S3 and has been calibrated. It can carry out indentation tests on exposed rocks in deep mine roadways, collect in-situ indentation indices of rocks, and directly output basic rock mechanical parameters based on the obtained indentation indices.
[0010] S5, on the exposed rock surface, uses an integrated handheld indentation tester to complete several indentation-unloading tests, acquires force-displacement curves in real time and automatically extracts in-situ indentation index, outputs basic rock mechanical parameters and prediction ranges in real time through the rock basic mechanical parameter evaluation model, generates a report and synchronizes it to the project database, thus achieving the characterization of rock mechanical properties.
[0011] Furthermore, the database established in S2 is an integrated database of indentation-mechanical parameters for rocks of various lithologies and types.
[0012] Furthermore, S3 includes the following sub-steps:
[0013] S31, aligns the indentation index and basic mechanical parameters of the same type and lithology in the indentation-mechanics integrated database, and performs data layering and cleaning at the same time;
[0014] S32, use the correlation analysis function to determine the strength and shape of the correlation between each indentation index and the target's basic mechanical parameters;
[0015] S33, the random forest algorithm is used to model the nonlinear relationship between the indentation index and the basic parameters of rock mechanics, and an evaluation model for the basic mechanical parameters of rock is obtained.
[0016] S34, use the validation set and independent test set to evaluate the model performance;
[0017] S35 introduces an interpretation method based on Shapley Additive Explanations theory to visualize the model output and realize quantitative correlation analysis between indentation index and basic rock mechanical parameters.
[0018] Furthermore, in S32, the Pearson correlation function is first used to analyze the linear correlation between the parameters, and then the Spearman correlation function is used to analyze the strong correlation between the parameters:
[0019]
[0020]
[0021] in, Indicates the indentation index, which is the independent variable. This represents the basic mechanical parameters of the dependent variable. This represents the Pearson correlation coefficient. Represents the Spearman correlation coefficient. This represents the number of sample pairs.
[0022] Furthermore, the random forest algorithm in S33 utilizes bootstrap sampling and random feature selection mechanisms to reduce the risk of overfitting and capture the nonlinear relationships and interaction effects between variables, as expressed by the following formula:
[0023]
[0024] in, Indicates the number of decision trees. It is the first The predicted results for each tree, It is the first The parameter set of the trees, This is a predicted value.
[0025] Furthermore, S33 uses Scikit-learn to develop a rock basic mechanical parameter evaluation model based on the random forest algorithm. The hyperparameters of the random forest in Scikit-learn include: number of decision trees, maximum tree depth, minimum number of sample splits, and feature selection method; the optimal parameter combination is determined through grid search or Bayesian optimization.
[0026] Furthermore, the performance evaluation metrics for the model in S34 include mean absolute percentage error. The root mean square error (RMSE) is calculated using the following formula:
[0027]
[0028] .
[0029] Furthermore, the SHAP summary plot in S35 shows: feature importance, the larger the average absolute value of the SHAP value, the more important the feature; direction of influence, positive SHAP values indicate improved mechanical properties, and negative values indicate decreased mechanical properties; degree of influence, the relationship between feature values and SHAP values reveals nonlinear effects.
[0030] Furthermore, in S5, representative rock cores are drilled from the field to conduct indoor standard tests. These cores are used as true values to verify the results of the integrated handheld field indentation tester, calculate error indices and interval coverage, and perform extrapolation verification across lithology / working conditions. Paired data from the field and indoor tests are fed back into the database for model calibration and updates. When the error of a certain layer exceeds the threshold or a systematic deviation occurs, incremental retraining and parameter recalibration are triggered.
[0031] The present invention also provides a rock mechanical property characterization device based on indentation testing and machine learning, which adopts the rock mechanical property characterization method based on indentation testing and machine learning as described above, including a body, a conical indenter and support disposed at the front end of the body, a drive unit disposed inside the body, and a control unit, a displacement sensor and a force sensor disposed inside the body.
[0032] The conical indenter is connected to the drive unit; the support is equipped with a three-point leveling structure to ensure that the axis of the conical indenter is aligned with the normal direction of the rock surface and to prevent slippage; the displacement sensor continuously measures the indentation amount of the conical indenter with micron-level resolution; the force sensor measures the loading reaction force in real time; a handle is also provided on the side of the machine body for stable gripping and force guidance; the machine body is also equipped with a display screen that can display the force-displacement curve and feature points in real time; the control unit can calculate and output the corresponding indentation index; the control unit integrates a rock basic mechanical parameter evaluation model.
[0033] The above-described solution of the present invention has the following beneficial effects:
[0034] The rock mechanical property characterization method and equipment based on indentation testing and machine learning provided by this invention utilizes indoor rock indentation tests and basic mechanical tests to obtain indentation index and basic mechanical parameters, establishes an integrated database of basic rock mechanical parameters and indentation index, and uses machine learning to perform correlation analysis on a large amount of data in the database to establish and verify a rock mechanical parameter evaluation model based on indentation index. Therefore, based on the high correlation between indentation index and basic rock mechanical parameters, without the need for core sampling, the device embedding the evaluation model can obtain an "indentation index" that can be mapped to multiple basic rock mechanical parameters during rapid indentation and unloading processes, and provide on-site usable predicted values and confidence intervals. This realizes rapid in-situ evaluation and predictive characterization of rock mechanical properties, which can be implemented in complex engineering sites. It provides data support for subsequent mine safety construction, stability evaluation, support method selection, and support technology parameter optimization to meet the technical requirements of safe and efficient mining resources.
[0035] Other beneficial effects of the present invention will be described in detail in the following detailed description section. Attached Figure Description
[0036] Figure 1 This is a flowchart of the method steps of the present invention;
[0037] Figure 2 This is a schematic diagram of the device of the present invention;
[0038] Figure 3 This is a flowchart of method S5 of the present invention.
[0039] [Explanation of Labels in the Attached Image]
[0040] 1-Body; 2-Conical pressure head; 3-Support; 4-Displacement sensor; 5-Force sensor; 6-Handle; 7-Display; 8-Control unit. Detailed Implementation
[0041] The following specific examples illustrate the implementation of this disclosure. Those skilled in the art can easily understand other advantages and effects of this disclosure from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this disclosure, and not all of them. This disclosure can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this disclosure. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.
[0042] It should be noted that various aspects of embodiments within the scope of the appended claims are described below. It will be apparent that the aspects described herein can be embodied in a wide variety of forms, and any particular structure and / or function described herein is merely illustrative. Based on this disclosure, those skilled in the art will understand that one aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number of aspects set forth herein can be used to implement the device and / or practice the method. Additionally, this device and / or method can be implemented using structures and / or functionalities other than one or more of the aspects set forth herein.
[0043] It should also be noted that the illustrations provided in the following embodiments are merely schematic representations of the basic concept of this disclosure. The illustrations only show components relevant to this disclosure and are not drawn according to the actual number, shape, and size of components in implementation. In actual implementation, the type, quantity, and proportion of each component can be arbitrarily changed, and the component layout may be more complex. Furthermore, specific details are provided in the following description to facilitate a thorough understanding of the examples. However, those skilled in the art will understand that the described aspects can be practiced without these specific details.
[0044] like Figure 1 As shown, embodiments of the present invention provide a rock mechanical property characterization method based on indentation testing and machine learning. In the current processes of underground metal mineral resource mining, tunnel excavation, and underground engineering construction, the basic mechanical properties of rocks are crucial for the safe mining of resources and engineering construction. However, with the increase in the depth of mineral resource mining, the geological environment of deep rock masses becomes increasingly complex. Existing indoor rock mechanics testing methods are difficult to quickly determine the basic mechanical properties of rocks under different geological conditions. Therefore, the proposed method provides a more economical, rapid, and accurate assessment of the basic mechanical properties of rocks under complex geological conditions. This method specifically includes the following sub-steps:
[0045] S1, conduct basic mechanical tests and indentation tests on rocks of various lithologies and types to obtain the basic mechanical parameters and indentation index of the rocks.
[0046] The laboratory tests can utilize various types and lithologies of rocks commonly found in practical engineering applications. Basic mechanical tests include uniaxial compression tests, Brazilian splitting tests, and fracture toughness tests. The fundamental mechanical parameters of the rocks that need to be obtained include uniaxial compressive strength. Brazil splitting strength Elastic modulus and fracture toughness The indentation index required for indentation testing includes the first peak indentation force. Maximum indentation force Standard indentation hardness Indentation modulus Energy consumption per unit displacement ,acting with specific energy The formulas for calculating each parameter are as follows:
[0047]
[0048]
[0049]
[0050]
[0051] In the calculation of uniaxial compressive strength and Brazilian splitting strength, , Both represent peak values. Indicates the cross-sectional area of the sample. Indicates the radius of the sample. Indicates the diameter of the sample. This indicates the sample thickness. In fracture toughness calculations, Indicates peak load. Indicates the length of the notch on the sample. Indicates the thickness of the sample. Indicates the length-to-diameter ratio of the crack. and span ratio The relevant functions, It can be calculated using the following formula:
[0052]
[0053] elastic modulus This represents the slope of the linear stress-strain segment during uniaxial compression of rock, and
[0054]
[0055] in, Indicates the amount of stress change. It represents the change in strain.
[0056] Indentation testing and basic mechanical testing can be conducted simultaneously, for example, by loading the specimen at a rate of 0.2 mm / min until failure. In this embodiment, taking a specimen size of 50 × 50 × 50 mm as an example, the first peak indentation force is calculated based on the indentation force-displacement curve during the specimen testing process. Maximum indentation force Standard indentation hardness Indentation modulus Energy consumption per unit displacement ,acting Specific energy It should be further explained that the indentation modulus Indicates the first peak value The slope of the straight segment of the preceding curve, based on the area of the indentation force-indentation displacement curve, can be used to calculate the work done during the indentation of the cutting tooth. Indicates the peak force of the first indentation. The work done by the cutting tooth on the rock This indicates the work done by the cutting edge on the rock when the sample finally fails. Indentation hardness index. Indicates the ease or difficulty of the cutting tooth intrusion, and the energy consumption per unit displacement. The specific energy indicates the ease or difficulty of initial rock failure. This represents the rock-breaking efficiency of the cutting teeth. Their calculation formulas are as follows:
[0057]
[0058]
[0059]
[0060]
[0061]
[0062] in, Indicates the maximum indentation force The corresponding indentation displacement (mm), Indicates indentation force The corresponding indentation displacement, Indicates the final displacement when the rock breaks. Less than or equal to 3mm It represents the final volume of rock failure under indentation force.
[0063] S2. Establish an integrated indentation-mechanical parameter database that includes basic rock mechanical parameters and indentation index.
[0064] In this step, the data obtained in S1 are classified and stored in the database according to the basic mechanical parameters and indentation index of the same type of rock, and an integrated indentation-mechanical parameter database for multi-lithology and multi-type rocks is established. The integrated indentation-mechanical parameters include the uniaxial compressive strength, Brazilian splitting strength, elastic modulus and fracture toughness mentioned in S1, as well as the first peak indentation force, maximum indentation force, standard indentation hardness, indentation modulus, energy consumption per unit displacement, work done and specific energy.
[0065] S3. Conduct correlation analysis between rock indentation index and basic mechanical parameters, and establish a rock basic mechanical parameter evaluation model based on indentation index through machine learning.
[0066] In this step, based on the database established in S2, correlation analysis is performed using correlation analysis functions. Simultaneously, an evaluation model is built based on the random forest algorithm, following a "first interpretable, then augmented" modeling strategy. Specifically, this may include the following sub-steps:
[0067] S31 aligns the indentation indices and basic mechanical parameters of the same type and lithology in the indentation-mechanical integrated database, while performing data layering and cleaning to reduce systematic contamination.
[0068] S32 uses correlation analysis functions to determine the strength and shape of the correlation between each indentation index and the target's basic mechanical parameters. First, the Pearson correlation function is used to analyze the linear correlation between the parameters, and then the Spearman correlation function is used to analyze the strength of the correlation. The specific formulas are as follows:
[0069]
[0070]
[0071] in, The independent variable is represented by the indentation index in this embodiment. This represents the dependent variable (in this embodiment, it represents the basic mechanical parameters). This represents the Pearson correlation coefficient. Represents the Spearman correlation coefficient. The number of sample pairs (i.e., the number of pairs that exist) (Group of independent and dependent variables).
[0072] S33 employs the random forest algorithm to model the nonlinear relationship between the indentation index and fundamental rock mechanics parameters, obtaining an evaluation model for these parameters. The random forest algorithm is an ensemble learning algorithm composed of multiple decision trees. It effectively reduces overfitting risk by utilizing bootstrapping (constructing diverse training sets) and random feature selection (constructing diverse tree structures), automatically capturing nonlinear relationships and interaction effects between variables. This can be expressed by the following formula:
[0073]
[0074] in, Indicates the number of decision trees. It is the first The predicted results for each tree, It is the first The parameter set of the trees, This is a predicted value.
[0075] In one specific implementation of this embodiment, Scikit-learn is used to develop a rock fundamental mechanical parameter evaluation model based on the random forest algorithm, including model construction and optimization. The hyperparameters of the random forest in Scikit-learn include: the number of decision trees (ensuring enough trees to stably capture complex nonlinear relationships), maximum tree depth (controlling model complexity and preventing overfitting to experimental noise), minimum number of sample splits (ensuring sufficient experimental data to support each mechanical law), and feature selection method (enhancing diversity among trees and improving model generalization ability). The optimal parameter combination is determined through grid search or Bayesian optimization to obtain a robust model that accurately describes the complex relationship between the indentation index and fundamental mechanical parameters without overfitting the data.
[0076] S34. The model performance is evaluated using a validation set and an independent test set. The evaluation metrics for model performance include mean absolute percentage error (MASE). The root mean square error (RMSE) and the root mean square error are calculated using the following formulas:
[0077]
[0078]
[0079] The sensitivity of a feature importance analysis model to different indentation parameters is used to assess the contribution of features: by analyzing the internal mechanism of the model, the relative importance of each indentation parameter to the prediction results is quantified; features with high importance scores indicate that they have a significant impact on the basic mechanical parameters of rocks. All indentation parameters are sorted from high to low importance scores to identify key parameters that play a dominant role in mechanical properties. The important parameters identified by machine learning are compared with rock mechanics theory to ensure that data-driven discovery is consistent with physical mechanisms.
[0080] Finally, extrapolation verification is performed using reserved samples to evaluate the model's transferability and robustness under new lithology or different experimental conditions. For example, samples of different rock types (not present in the training set) can be used to test the model's predictive ability, and the model's stability can be evaluated under varying experimental conditions (such as different loading rates and ambient temperatures). Based on the above results, it can be determined whether the model can be generalized to new geological conditions and experimental scenarios.
[0081] S35 introduces an interpretation method based on Shapley Additive Explanations (SHAP) theory to visualize the model output. For example, the SHAP summary plot can show: feature importance (the larger the average absolute value of the SHAP value, the more important the feature); direction of influence (positive SHAP values indicate improved mechanical properties, while negative values indicate decreased mechanical properties); and degree of influence (the relationship between feature values and SHAP values reveals nonlinear effects). Therefore, SHAP values can reveal the positive and negative influence direction and contribution of each indentation index on the prediction results, thereby achieving a quantitative correlation analysis between indentation indices and basic rock mechanical parameters.
[0082] S4, developed and calibrated an integrated handheld field indentation tester.
[0083] It should be noted that the integrated handheld field indentation tester can conduct field indentation tests on exposed rock in deep mine roadways, collect in-situ indentation indices, and directly output usable basic rock mechanics parameters based on the obtained indentation indices. Therefore, the integrated handheld field indentation tester needs to be calibrated.
[0084] At the same time, such as Figure 2 As shown, the front end of the integrated handheld indentation tester body 1 is equipped with a conical indenter 2. The conical indenter 2 is connected to the drive unit inside the body 1. Under the drive unit, the conical indenter 2 contacts the rock surface and is pressed in a controlled manner, converting the loading into local indentation fracturing. Simultaneously, the body 1 is also equipped with a support 3, which has a three-point leveling structure to ensure that the axis of the conical indenter 2 is aligned with the normal direction of the rock surface. It also has an anti-slip function to reduce slippage-induced tilting errors. The body 1 contains two types of sensing units: a displacement sensor 4 continuously measures the indentation amount of the conical indenter 2 with micron-level resolution; and a force sensor 5 (pressure sensor) measures the loading reaction force in real time. Both are sampled synchronously (sampling frequency...). High-quality force-displacement data is generated at kHz, thus producing a force-displacement curve. A handle 6 on the side of the fuselage 1 is used for stable grip and force guidance; it can be connected to the fuselage 1 using a buffer structure to absorb operational vibrations and improve reliability. The fuselage 1 is also equipped with a display 7, which can display the force-displacement curve and feature points (such as the first peak indentation force) in real time. The control unit 8 inside the fuselage 1 can calculate and output the corresponding indentation index. Obviously, the control unit 8 integrates the rock basic mechanical parameter evaluation model in S3, and completes the calibration of the rock basic mechanical parameter evaluation model through the steps of S1-S3.
[0085] The S5, on exposed rock surfaces in deep mine roadways or tunnels, uses an integrated handheld indentation testing instrument to perform several indentation-unloading tests, acquiring force-displacement curves in real time and automatically extracting the in-situ indentation index. Through an embedded, calibrated rock fundamental mechanical parameter evaluation model, it can output uniaxial compressive strength in real time. Brazil splitting strength Elastic modulus Fracture toughness The report includes basic rock mechanics parameters and a 95% prediction range, along with prompts indicating whether the data is reliable or requires retesting. This information is then generated and synchronized to the project database for rapid on-site decision-making regarding surrounding rock classification and support parameters. Additionally, if... Figure 3 As shown.
[0086] In addition, a small number of representative rock cores can be drilled from the field to conduct indoor standard tests. These cores can be used as "true values" to verify the results of the integrated handheld field indentation tester, calculate error indices and interval coverage, and perform extrapolation verification across lithology / working conditions. Paired data from the field and indoor tests are fed back into the integrated database, and the model is further calibrated and updated according to established rules (including coefficients, stratification, or mixed effects). When the error of a certain stratification exceeds the threshold or a systematic deviation occurs, incremental retraining and parameter recalibration are triggered, so that the model can be continuously optimized and become more accurate with use in practical applications.
[0087] As described above, the rock mechanical property characterization method based on indentation testing and machine learning provided in this embodiment, based on the high correlation between the indentation index and the basic mechanical parameters of rocks, acquires the "indentation index" that can be mapped to multiple basic mechanical parameters of rocks during rapid indentation and unloading processes using an integrated handheld field indentation testing instrument with an embedded evaluation model, without the need for core sampling. The method provides on-site usable predicted values and confidence intervals. This method achieves rapid in-situ assessment and predictive characterization of rock mechanical properties, can be implemented in complex engineering sites, and provides data support for subsequent mine safety construction, stability evaluation, support method selection, and support technology parameter optimization, thereby meeting the technical requirements for safe and efficient mining.
[0088] Based on the same inventive concept, this embodiment also provides a device, while as follows Figure 2As shown, the system includes a body 1, a conical indenter 2 and a support 3 located at the front end of the body 1, a drive unit located inside the body 1, a control unit 8, a displacement sensor 4, and a force sensor 5 (pressure sensor) also located inside the body 1. The conical indenter 2 is connected to the drive unit inside the body 1. Driven by the drive unit, the conical indenter 2 contacts the rock surface and is pressed in a controlled manner, converting the loading into localized indentation fracturing. The support 3 has a three-point leveling structure, which ensures that the axis of the conical indenter 2 is aligned with the normal direction of the rock surface, and also has an anti-slip function to reduce slippage-induced tilting errors. The displacement sensor 4 continuously measures the indentation amount of the conical indenter 2 with micron-level resolution; the force sensor 5 (pressure sensor) measures the loading reaction force in real time. Both sensors can sample synchronously (sampling frequency...). High-quality force-displacement data is generated at kHz, thus producing a force-displacement curve. A handle 6 is also provided on the side of the fuselage 1 for stable grip and force guidance. It can also be connected to the fuselage 1 using a buffer structure to absorb operational vibrations and improve reliability. The fuselage 1 also has a display 7, which can display the force-displacement curve and feature points (such as the first peak indentation force) in real time. The control unit 8 can calculate and output the corresponding indentation index. The control unit 8 integrates the rock basic mechanical parameter evaluation model in S3, and completes the calibration of the rock basic mechanical parameter evaluation model through the steps of S1-S3.
[0089] The device provided by this invention has the same inventive concept and beneficial effects as the aforementioned method, which will not be repeated here.
[0090] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0091] The above embodiments are merely illustrative of several implementation methods of this application, and their descriptions are relatively specific and detailed, but they should not be construed as limiting the scope of the application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.
Claims
1. A method for characterizing the mechanical properties of rocks based on indentation testing and machine learning, characterized in that, Includes the following steps: S1, to conduct basic mechanical tests and indentation tests on rocks of various lithologies and types to obtain the basic mechanical parameters and indentation index of the rocks; Basic mechanical parameters include uniaxial compressive strength Brazil splitting strength Elastic modulus and fracture toughness Indentation index includes the first peak indentation force of the indentation test. Maximum indentation force Standard indentation hardness Indentation modulus Energy consumption per unit displacement ,acting Specific energy ; S2, Establish an integrated indentation-mechanical parameter database that includes basic rock mechanical parameters and indentation index; S3. Correlation analysis is conducted between rock indentation index and basic mechanical parameters using correlation analysis functions. A rock basic mechanical parameter evaluation model based on indentation index is established through machine learning. S4 is an integrated handheld indentation tester. The handheld indentation tester is embedded in the rock basic mechanical parameter evaluation model in S3 and has been calibrated. It can carry out indentation tests on exposed rocks in deep mine roadways, collect in-situ indentation indices of rocks, and directly output basic rock mechanical parameters based on the obtained indentation indices. S5, on the exposed rock surface, uses an integrated handheld indentation tester to complete several indentation-unloading tests, acquires force-displacement curves in real time and automatically extracts in-situ indentation index, outputs basic rock mechanical parameters and prediction ranges in real time through the rock basic mechanical parameter evaluation model, generates a report and synchronizes it to the project database, thus achieving the characterization of rock mechanical properties.
2. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 1, characterized in that, The database established in S2 is an integrated database of indentation-mechanical parameters for multi-lithological and multi-type rocks.
3. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 1, characterized in that, S3 includes the following sub-steps: S31, aligns the indentation index and basic mechanical parameters of the same type and lithology in the indentation-mechanics integrated database, and performs data layering and cleaning at the same time; S32, use the correlation analysis function to determine the strength and shape of the correlation between each indentation index and the target's basic mechanical parameters; S33, the random forest algorithm is used to model the nonlinear relationship between the indentation index and the basic parameters of rock mechanics, and an evaluation model for the basic mechanical parameters of rock is obtained. S34, use the validation set and independent test set to evaluate the model performance; S35 introduces an interpretation method based on Shapley Additive Explanations theory to visualize the model output and realize quantitative correlation analysis between indentation index and basic rock mechanical parameters.
4. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 3, characterized in that, In S32, the Pearson correlation function is first used to analyze the linear correlation between the parameters, and then the Spearman correlation function is used to analyze the strong correlation between the parameters. in, Indicates the indentation index, which is the independent variable. This represents the basic mechanical parameters of the dependent variable. This represents the Pearson correlation coefficient. Represents the Spearman correlation coefficient. This represents the number of sample pairs.
5. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 4, characterized in that, The Random Forest algorithm in S33 utilizes bootstrap sampling and random feature selection mechanisms to reduce the risk of overfitting and capture the nonlinear relationships and interaction effects between variables, as expressed by the following formula: in, Indicates the number of decision trees. It is the first The predicted results for each tree, It is the first The parameter set of the trees, This is a predicted value.
6. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 5, characterized in that, S33 uses Scikit-learn to develop a rock basic mechanical parameter evaluation model based on the random forest algorithm. The hyperparameters of random forest in Scikit-learn include: number of decision trees, maximum tree depth, minimum number of sample splits, and feature selection method; the optimal parameter combination is determined through grid search or Bayesian optimization.
7. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 5, characterized in that, The performance evaluation metrics for the model in S34 include mean absolute percentage error. The root mean square error (RMSE) is calculated using the following formula: 。 8. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 7, characterized in that, The SHAP summary plot in S35 shows the importance of features: the larger the average absolute value of the SHAP values, the more important the feature. The direction of influence: positive SHAP values indicate improved mechanical properties, while negative values indicate decreased mechanical properties; the degree of influence: the relationship between eigenvalues and SHAP values reveals nonlinear effects.
9. The rock mechanical property characterization method based on indentation testing and machine learning according to claim 1, characterized in that, In S5, representative points are selected from the field to drill rock cores and conduct indoor standard tests. These are used as true values to verify the results of the integrated handheld field indentation tester, calculate error index and interval coverage, and perform extrapolation verification across lithology / working conditions. Paired data from the field and indoor environments are fed back into the database for model calibration and updates. When the error of a certain layer exceeds the threshold or a systematic deviation occurs, incremental retraining and parameter recalibration are triggered.
10. A rock mechanical property characterization device based on indentation testing and machine learning, employing the rock mechanical property characterization method based on indentation testing and machine learning as described in any one of claims 1-9, characterized in that, It includes a body, a conical pressure head and support disposed at the front end of the body, a drive unit disposed inside the body, and a control unit, a displacement sensor and a force sensor disposed inside the body; The conical indenter is connected to the drive unit; the support is equipped with a three-point leveling structure to ensure that the axis of the conical indenter is aligned with the normal direction of the rock surface and to prevent slippage; the displacement sensor continuously measures the indentation amount of the conical indenter with micron-level resolution; the force sensor measures the loading reaction force in real time; a handle is also provided on the side of the machine body for stable gripping and force guidance; the machine body is also equipped with a display screen that can display the force-displacement curve and feature points in real time; the control unit can calculate and output the corresponding indentation index; the control unit integrates a rock basic mechanical parameter evaluation model.